[music] So, every time you ask it something, it opens file after file and burns tokens finding what it already had. Layer one is raw transcripts, notes, voice messages land there and stay [music] exactly as they are. Layer two is a wiki and one nightly job at 3:00 in the morning is the only thing allowed to write to it. It reads what is new [music] in raw, works out what it means, and files it on the page it belongs on. Then, a routing table, one row [music] for every kind of work I do.

Der Modell Hardwarestandard kurz MHS, eine gemeinsame Sprache mit der ein Agent Laborgeräte und Fertigungsmaschinen bedienen kann. Mikroskope, Pipettierroboter, Roboterarme, in einem Fall die Laser, mit denen ein Quantenrechner eingestellt wird. Eine MHS-Datei, die festlegt, wie schnell ein Roboterar am fahren darf und in welchem Winkel, erfüllt genauso eine Sicherheitsfunktion. Wer diese Datei schreibt, schreibt in Europa womöglich ein reguliertes Bauteil. Wir reden hier gerade noch über Software, aber ein Fehler in einer Textdatei bewegt hier ein Stück Metall.

Hermes geht spaßer mit dem Gedächtnis um und zwar ab sofort von allein. Bisher hat er dabei einen dicken Brocken der alten Unterhaltung wörtlich mitgeschleppt in jede neue Runde. In den Messungen von NOS bleiben bei einer sehr langen Sitzung rund 49 000 Token übrig, statt rund 162 000. Bei gleicher Arbeit, also grob ein Drittel der Last. Wer Hermes den ganzen Tag laufen lässt, merkt das an der Rechnung.

You see, now Hermes can already use tools and completes tasks, but the context about your work is scattered across calls, notes, files and messages. If you have it or if you can provide all the necessary files, you can just provide the files here to your Hermes agent and tell to build up your second brain based on that documents. All right, it's finished and we can see all the files that it edited and what the source nodes were used, the size, etc. Actually useful thing you can do is just to copy the URL of this video, go into your own Hermes agents or some LM and tell it to guide you step-by-step how to install all of this. Just make sure to keep your raw files because when you delete your wiki, you can recreate it with your raw files.

Fehler 1: Du baust was der Kunde will, statt den wahren Engpass zu lösen. Um den Engpass zu finden, stellst du eine Frage. Dein Ziel durch KI sind 12 Termine und 10 Stunden Ersparnis für den Vertrieb. Wenn der Kunde den ROE anzweifelt, legst du dieses Dashboard auf den Tisch. Ein Mitarbeiter kostet 40 pro Stunde und verbringt 10 Stunden pro Woche mit Tickets.

Únete a la Cofradía IA y ahorra 70 y domina el código IA en 30 días o te devolvemos tu dinero 9 GRADOS, 9 SELLOS: subes de nivel y desbloqueas módulos secretos, créditos API, reuniones 1-a-1 conmigo, la Cofradía GRATIS de por vida y hasta volverte socio. Cada grado, un sello en tu perfil Dale NEURONAS PREMIUM a tu IA instala SlashStack Otras Guías IA Gratis en mi TELEGRAM Quieres un PLAN que FUNCIONE? yo te lo diseño CORREO: alfa appsclavitud.com SÍGUEME EN: Youtube: Facebook: Instagram: Tiktok: El modelo es el alma. :::::::::::::::::::::: :::::::::::::::::::::: :::::::::::::::::::::: :::::::::::::::::::::: :::::::::::::::::::::: usa mi código repo de deepseek harness: :::::::::::::::::::::: :::::::::::::::::::::: :::::::::::::::::::::: :::::::::::::::::::::: :::::::::::::::::::::: 00:00 El modelo no es el agente 01:00 Qué es deepseek harness 02:41 Todo es un plugin 03:20 Cada corrida tiene rastro 04:10 Los 4 modos 05:38 pestaña de trayectoria 06:54 ideas de plugins 08:20 Veredicto de una semana 10:00 crea cualquier plugin 10:45 qué es CORDIS? 11:20 casos de uso 13:10 integra otros modelos en Deepseek Harness 14:12 Deepseek Harness Hermes Agent

I built an app that replaced 3 SaaS tools I was paying thousands of dollars for in just 3 prompts inside ChatGPT. codex agents ChatGPT_Partner aiagents ChatGPT Sites comes with built-in database, hosting, authentication, storage, and custom domains. And you can share your app with your whole team. Plus how ChatGPT Work fits between ChatGPT and Codex, using skills like Scrape Creators to pull data from any social platform, and editing your app with annotations. TIMESTAMPS: Chapters: 0:00 Intro 1:04 Inside ChatGPT Work 3:26 Background on ChatGPT Work 3:32 Creating Our First Prompt to Use in Sites 6:09 Editing the App 8:16 How the Scraping Skill Works 10:19 Site Settings and Database 12:08 Wrap Up More about me rileybrown.xyz Agent Native Resources (free): Sponsorships: riley rakugomedia.com

This is my Discord where I talk to my Hermes bot, so I can say, "Hermes bot, draft me a spec to add on to my Pomodoro app the ability to switch skins." So, here's my little desktop Pomodoro app. Like should I consider switching around my times for when I do certain tasks, or take more breaks?" And it can actually tell me like what the pattern looks like. Like should I like take a break right now or should I like keep pushing through it, right? So, if you're interested in something like that, where I'll actually explain how to actually build up these functionalities and multi-agent systems, do comment Hermes workshop because I actually do monitor all the comments and if I see a lot of interest, then I will consider it. And again, please let me know in the comments, just write like Hermes workshop or something like that, and that will let me know how much interest there is in doing a Hermes workshop, and maybe we'll do one.

A profile is a separate agent setup with its own config, its own memory, its own skills, and its own chat history. Each bot has its own role, its own model, its own memory, its own skills, and its own avatar. So, one bot can be your researcher, one your writer, one your coder, and they can all run on different models. Hermes through the bundled gateway adapter, a direct HTTP custom runtime for your own orchestrator stack or a built-in demo gateway, a mock backend with demo agents. And grab the prompts so your agents get clear instructions from day one, which is the difference between an agent team that helps and one that just makes noise.

But, what if I told you that you can build a really great AI-powered second brain that is completely free, runs on your own computer, and keeps all your notes as plain markdown files that you own forever? Sources go into a sources folder, permanent notes go into a notes folder, and daily notes go into a daily folder. Because the agent can search and read all your files, you can just ask it stuff like "What have I noted down about transformer architectures?" or "Summarize everything in my vault about my thesis topic and tell me where the gaps are." It goes through your notes, pulls the relevant pieces together, and gives you an answer based on your own knowledge with references to the actual notes. You can just tell the agent, "Go through my inbox folder, file everything into the right place, fix any broken links, and give me a list of orphan notes that aren't linked to anything." And it'll go and clean your whole vault up. If Anti-Gravity dies tomorrow, or Google nerfs the free tier again, which, let's be honest, they might, your notes are still just markdown files, and you can point literally any other AI tool at the same folder.

But if you want to truly unlock the power of Hermes, you need to know what Hermes can see, what it can remember, and what you're allowing it to do. But if you want to see how I structure something called a second brain, which is a way that you can take all of your files, your documents, transcripts, all of that stuff, and distill down the key insights that your agent can then tap into, I've done a full video on that and how to set it up, all of that stuff. Now you can click manage profiles here and this is where you can add additional profiles and each one of these as you can see comes with a soul.md file. So yeah, you can see here we need to create an application, give it the permissions, and we can then go into the setup, the gateway setup, Hermes gateway setup in your terminal, and you can add in the configuration keys and get all of that set up. Now, like in Claude Code, you can have different approval settings where you either manually accept all of the edits and changes that the agent wants to make or you can put it into auto mode or have your own separate rules according to the different types of tasks and tools that it calls whether it's going to ask you for permission or not.

Boek een call: In deze complete Nederlandse Hermes-cursus bouw je vanaf nul een eigen AI-team dat onderzoek doet, strategie bepaalt, werk uitvoert en het eindresultaat controleert. Daarna geef je Hermes de juiste context over jou en je bedrijf, bouw je vijf gespecialiseerde AI-operators en zet je een succesvolle workflow om in een herbruikbare skill. Je ziet hoe je Hermes gebruikt voor deep research, funnels, brandbooks, websites, advertentievarianten, e-mailflows en het omzetten van lange YouTube-video s in bruikbare skills en agents. Na deze cursus heb je geen losse chatbot, maar een eigen operatorsysteem met: - Een orchestrator die opdrachten verdeelt - Een researcher die onderzoek uitvoert - Een strategist die de richting bepaalt - Een builder die het werk maakt - Een reviewer die alles controleert - Een second brain met jouw bedrijfskennis en werkwijze - Herbruikbare skills voor terugkerende processen - Toegang tot Hermes vanaf je telefoon via Telegram Benodigdheden - Een Mac of geschikte computer - Hermes Agent - Een OpenRouter-account en API-key - Telegram als je Hermes vanaf je telefoon wilt bedienen Bestanden en resources - Hermes download: - GitHub-repository met skills en bestanden: - AI Operator-programma masterclass: Boek een call: Gebruik altijd afzonderlijke API-keys, deel tokens nooit publiek en maskeer gevoelige gegevens voordat je screenshots of video s publiceert. Hoofdstukken 0:00 Intro Bouw je eigen AI-team 3:35 Hermes installeren 4:06 Hermes starten en OpenRouter koppelen 4:54 API-key, model en Hermes Doctor 6:18 Hermes jouw bedrijf laten begrijpen 6:34 Bedrijfscontext en AI-profiel bouwen 11:42 Je eerste herbruikbare skill installeren 12:56 Vijf gespecialiseerde AI-operators bouwen 16:05 De researcher en strategist aan het werk 16:24 Builder en reviewer voltooien de workflow 17:06 Je agents afzonderlijk bekijken 17:18 Het eindresultaat van het AI-team 18:27 Van geslaagde workflow naar skill 19:02 De Operator Team Workflow hergebruiken 20:46 Telegram en de Hermes Gateway instellen 22:20 Hermes verbinden met je telefoon 23:14 Hermes bedienen via Telegram 24:51 Van cursus naar AI Operator 27:45 Deep research en je funnel bouwen 33:32 Research omzetten in marketingassets 34:24 Een brandbook en Visual DNA maken 36:30 Je funnel analyseren en verbeteren 38:03 Websiteframework, assets en second brain 42:40 Het complete websiteresultaat 45:17 YouTube-video s omzetten in skills en agents 52:45 Een complete advertentiebatch genereren 53:11 On-brand e-mailflows en Mission Control 53:51 Mission Control, Obsidian en de complete afsluiting HermesAI AIAgents Automatisering

I installed more skills, added more tools, gave it more memory, and every single time my agent got slower, heavier, and I trusted it less. So, in [music] this video, I'm going to show you all three upgrades in the exact order they have to happen, the one line I now put in every skill so it cannot lie to me, and a 30-second check you can run on your own agent right now. What your agent is worth is not how much work it attempts. If you only trust half of what your agent does, doubling its workload does not give you an employee, it gives you a second job. It cannot send, it cannot write files, it cannot spend anything, not because I told it not to, because it does not have those tools.

Mit bis zu vier parallelen Codex-Agenten, je eigene Kopie der Software, waren es anderthalb Wochen Aufwand über zwei Kalenderwochen und rund 12.000 Dollar. Jetzt 75 Prozent, über 60 Prozent entschieden dagegen, nur noch 4 Prozent entschieden dafür. HERMES-FUNK, HERMES CLOUD IST OFFIZIELL: Elon Musk schrieb, sein Agent habe einen eigenen Rechner in der Cloud. Nous Research antwortete, Hermes Agent habe das auch, man sei nur nicht daran gefesselt, und es koste ein paar Dollar im Monat statt 200. KAPITEL: 0:00 Kalt-Start: 5 Jahre in 2 Wochen 0:46 Asana: Codex räumt 5 Jahre Arbeit weg 2:25 Aufstand gegen Rechenzentren 4:09 Anthropics geheimes Model 2 6:02 Die menschliche Minute 6:33 GitHub-Ecke: 381 Token auf einer RTX 3090 8:40 Hermes-Funk: Hermes Cloud ist offiziell 10:37 Warum wir in der Community Hermes nutzen 11:25 Commander-Logbuch 12:52 Schlagzeilen-Runde 13:48 Jetzt auf Spotify Ausblick QUELLEN: Asana Codex: Rechenzentrums-Umfrage: Model 2: Elliptische Kurve Rang 30: GitHub-Fund: Hermes Cloud: Qwen-Nachmessung (AIME 2026): Börsengang: KI News heute ist Deutschlands täglicher KI-Podcast von AIIANER.

And then you can rename this to your agent and you can make things you can make things bigger by double pinching and just kind of zooming in like that, kind of like when you zoom in on a picture on your phone, okay? And then you can also turn things 3D by just holding down like this and spinning it around, you know, you can spin it around like that and make it bigger like that. So, I have a folder here called props, so I'll open that and then I'll go down here to uh Let's see, I have this hologram folder down here and these are all 3D images down here and I can pick whatever 3D image I want to pull up on the screen just like that and it renders in 3D. And then I can make it bigger, I can small make it smaller, I can manipulate it in 3D space just like that. And they render like this and you can like I said, you can change the color of all these and then these can be manipulated in 3D space as well.

One AI writes for you, one AI checks the work, one AI makes it better, all at the same time, all while you do something else. Instead of one chatbot in one tab, you get a whole team in one place. With an agent operating system, I get one group chat with every model together. One agent writes, one agent edits, one agent plays judge and scores it. Pick one workflow, just one, one that generate leads.

One bot's notes stay with that one bot. Now, if you're watching this and thinking, "Okay, but how do I actually set up a bot the right way?" This is exactly what we go through inside the AI Profit Boardroom. Right now, we've got walkthroughs on setting up Hermes Desktop, building your first bot, and wiring your bots to talk to each other without it all turning into a mess. And because the bots talk to each other, my research bot can hand its findings straight over to my writing bot. Bot mode is one of the first tools that puts that whole idea on your own desktop for free in a way a normal person can actually use.

In this video, I'll cover everything Hermes new bot mode, show you how to use it, and let you know if you should be ripping out Grobbot for Hermes bot. And so you can either log into your own personal accounts if that's something you want, or you can use agent mail, which is what I've been doing, which is just a free email service for these bots, and it creates its own accounts for everything, which gives really good separation. So, when you have one single massive agent you talk to, like most people do with Hermes or Open Claw, you're including your entire agent's history, all the skills you use, all the tools you use, all the MCPs you use with every single prompt. But when you use bot mode, whether it's Grockbot or Hermes bot, you are splitting up responsibilities from one bot to many, many bots. So if you have one bot that does a hundred things rather than a 100 bots that do one thing, those hundred bots that do one thing will do that one thing way better because it's going to have way less context with every single prompt you send.

Your research agent runs one model, your writing agent runs a different one, your inbox agent runs a small, fast, local model that lives entirely on your own machine. It's a complete agent team pre-configured, pre-named, with personality files already written for each agent, models already assigned, routines already set up. Hermes gives you every agent with its own brain, its own memory, its own schedule, talking to every other agent, all running on your machine, under your control, using the models you choose. So, if you want to skip the setup and get straight to running your own agent team, the full agent OS zip file is inside the AI Profit Boardroom, ready to download and install today. On top of that, you get the full 30-day roadmap for getting the most out of your Hermes setup, four live coaching calls every week where you can bring your exact agent configuration and get feedback in real time, and daily tutorials as new features ship.

Now, here you see that uh from this AI Researcher, I asked this Code Reviewer bot uh I have sent the message, and uh here you see that Code Reviewer also acknowledged it, and uh here you see, message from AI Researcher, and uh message replied to the AI Researcher. So this is a AI researcher team, this is a code reviewer team, and you can set up multiple teams like one team for debugging, another agent is for the sales, and another agent for the marketing, okay? Okay, so if I just copy it, and I can ask it that want to create a front-end, or you can say back-end. Now, what I can do, I can actually um message from this code reviewer to the back end dev. So, here you see uh back end dev developer back end dev bot got the message from the code reviewer.

And just me and my husband, we're just having a meet around, and I just asked them, "Do you have any Evelyn bags?" Because I've been after an Evelyn bag for the longest time. I think I might have one." Voila, here it is and she just brought this out and I said, "Oh my goodness." Literally for about 7 years, 6 years or something, I'd been looking for the for this one and never managed to get my hands on one and she just said, "Yeah, we have one." So, yeah, she became my sales associate that day and I did ask her whether she would take me on as a client and she said, "Yes, of course." Now, at this point, I did have a wish list at the London Bond Street store cuz that's where I used to shop. He said, there's black with gold, black with palladium, there's gold, and I can't remember, I think it was with palladium or there was a red and I cannot remember what the red was, but as soon as he said there was a black with gold, I was like I was saying, that's the one. But, I couldn't say no, because these things do not come along that often, and I just I just love this bag so very So, this is the Constance 18. So, as we were talking, I also said, "Oh, by the way, is the Evelyn 23 is it out now?" Because it kept get getting delayed and she said, "Yes, it is." And I said, "Oh, I don't know if one will come along, but I would definitely be interested." And as we were chatting, she was showing me some things in the display cabinets and she said, "Do you like that color?" And it was a beautiful gray and I said, "Oh, yes, I love gray." And she said, "That cuz it's quite a light gray." And I said, "Yeah, I I like a light gray.

Bisher hieß es in Hermes: ein Chat, eine Aufgabe, fertig. Mit dem Bot Mode werden aus deinen Agentenprofilen benannte Bots, jeder mit eigenem Job, eigenem Gedächtnis, eigenen Skills und eigenem Modell. Und sie können miteinander reden. Die ganze Folge gibt es hier auf dem Kanal. KINews Hermes KIAgenten KI Shorts

Because the agent-to-agent communication between the different agents is so good, you can just rely on talking to just one bot and allowing them to distribute the work, to delegate the work. Also, by the way, a majority of people I see using Grok Bot just like make the names of each of their bot like content, email, coding, right? The reason why like this is better than having like one mono agent that just handles everything is Bill just has context just around my technical work, just around my computers, how things are set up in my projects. People aren't as engaged after a few days after they join, Dusty will message them and say, "Hey, here's a few ideas based on my research about you of things you can build." And Dusty just keeps people engaged, keeps things moving, posts new AI news in the community every day. Slate manages Reed, so I don't go to Reed and say, "Hey, do this, do that." I just tell Slate, "Hey, have Reed try different things out and see if there's different business opportunities online." Then Slate just watches Reed do all those things all day.

Also immer wenn du eine Sitzung aufmachst, sie beendest, einen Prompt abfeuerst, einen Tool Call machst, wird agentmemory benachrichtigt und kann diese Information dann selbst auswerten agentmemory weiß dabei genau, in welcher Sitzung welche Informationen stehen. Man muss nicht immer wieder dran denken, dass agentmemory angeworfen werden muss und kann einfach so agieren, wie man es immer tut. Was wir jetzt zusätzlich bekommen, ist der agentmemory Viewer, der die gesamten Sitzungen von Claude Code und Codex oder anderen Harnesses zusammenfasst und uns wichtige Informationen verschafft, wie zum Beispiel die Token Savings, die bei mir 48 % beziehungsweise mehr als 40 US-Dollar umfassen. Teilweise hat man das Problem, dass dieser Graph 'n bisschen veraltet ist und das kann man dann lösen, indem man hier unten rechts einmal auf Rebuild Graph klickt. Claude's Auto Memory kann die Regel sauber überschreiben, aber nur agentmemory kann beide Arbeitsschritte erfassen und ihre Herkunft nachvollziehen.

Solche mathematischen, akademischen Gespräche sind jetzt natürlich eher Quatsch, aber man kann hier natürlich ganz konkret an Softwarepten arbeiten und das Spannende ist, es gibt hier eben auch Projekte und wir haben hier einfach ein komplettes GitHub in Bass implementiert. Dass also jetzt mein Arbeitskollege nicht einfach irgendwelche wundersamen Codeänderungen produziert und ich frage mich dann, ist das jetzt mit KI gemacht, wie hat er das gepromptet, wie viel hat der Mensch da gerade untersucht oder nicht, sondern dass wir stattdessen wirklich gemeinsam mit den KI arbeiten können. Wir drücken einmal Shift Tab, um in den Auto Mode zu wechseln und dann sagen wir mal: "Moin Clord, ich habe hier auf dem VPS eine Buzz Relay installiert in Docker und ich hätte das gerne, dass du mal ein Cloud Code Agent einrichtest. Jetzt haben wir also schon mal mich selbst, wir haben Fis, der auf meinem Laptop läuft und wir haben Cloud Code, der auf dem Server läuft, auf dem auch das Bass Relay installiert ist. Oder wenn der Agent jetzt noch kein Nost PUBG hätte, dann würde ich sagen: "Hey, installier mal bitte Nosterrichter alles ein und dann sende mir deinen PUBG." Weil es tatsächlich so ein bisschen fittelig war, diesen Agenten zu verbinden, habe ich hier einen kleinen Notion Guide am Ende auch den Agenten schreiben lassen.

So, I have my little agent buddy here and I can ask uh can you summarize this repo? So, in this video, I'm going to show you how you can activate this HUD mode, how you use it, and we're going to try it out on a bunch of different apps, and try to really put it to the test to see how well it actually works. You wouldn't necessarily need to use it all the time, but just having it, as Brooklyn described, as an agent buddy to try to ask questions about what you're seeing on the screen, is kind of useful. Uh so, you do need to tell it, okay, I've changed apps or what am I on now or what do you see now? If you're working mainly in here on some larger coding project, but you have questions, you can use the a background feature on this, but if you don't want to do that, um you want something a little bit more um full session, then you could see See, see we're getting our answer here.

See how to build an accessible mobile library of standard operating procedures, safety protocols, and FAQs. Connecteam Knowledge Base centralizes company guidelines, workflows, and protocols so employees always have key operational resources right in their pockets. Chapters: 00:00 Introduction to Connecteam Knowledge Base 00:30 Structuring categories and adding SOPs 01:10 In-app search for instant information access 01:40 Managing permissions and knowledge updates Learn more: Sign up for free: Book a demo:

Die können Input liefern, aber der Owner der Agents, der kann sich um das ganze Overhead kümmern und da sehe ich jetzt jetzt Bass und wie du jetzt quasi Hermes z.B. Und wenn ihr jetzt Hermes hier hinzufügen wollt, müsst ihr quasi unter Settings gehen und habt den hier Agents und jetzt erstmal gar nicht verfügbar. So, das bringt euch jetzt nicht näher, weil jetzt wäre meine Intuition, okay, was hat mich jetzt dahineleitet, dass ich denn downloade oder ich muss es jetzt installieren, was nicht ganz falsch ist, aber die bessere Erklärung finde ich eigentlich hier, wenn ihr unter Docs geht und dann Integration und dann habt ihr hier die Bassintegration. ähm kann ich gerne mal noch mal als gesondertes Video machen, ähm wie ihr das quasi auf den Remote Server macht und da macht ihr dann aber auch Hermes einrichten. Und jetzt habt ihr Hermes, könnt ihr quasi unten einrichten, dann als Default Harness Hermes Agent, das jedes Mal mit den neuen Agent einstellt, der quasi dem Hannes von dem Hermes läuft, das heißt übernimmt die Skills, den Kontext und die Tools und ja heißt ihr seid jetzt quasi hier Agents könnt quasi ein Agent Team anlegen, wo verschiedene Systemproms drin sind, die auf den gleichen Hermes Agent Harness gehen und könnt jetzt hier sagen Bobby könnt jetzt hier ähm helpful full oder hilfreicher Assistent.

This is Hermes Agent's response to the Grok Bot released 2 days ago where we said that Grok Bot has a bigger advantage over Hermes Agent in terms of computer and browser use cuz each Grok Bot has its own cloud computer. All right, so that took about 5 minutes and as you can see, there is a change in the UI on the top left, you can see the bots panel and on the right side, you can see the cron jobs list as well for your bots. All right, so even with that very simple description, uh Hermes Agent is smart enough to know what the requirements are and gives a formal description of what its roles, what its responsibilities, so it can browse the web, run code, delegate tasks. But, if you have been working with your Hermes Agent on the desktop app, what you can do is you can, uh, instead of creating a new bot from scratch, you can actually go to advanced and clone from the profile, the existing profiles that you have. You can see that there's a lot of value here for using bot mode in the Hermes desktop app because you have one interface to communicate with all of your Hermes agent rather than you having to enter a separate session to communicate with them one by one.

Plus, you get the agent operating system where you can plug in Hermes alongside Claude, open Claude, free Claude code all inside one dashboard. You can export the file, you can send it over, they import it, they add their own keys, and in seconds they're running the same agent you spent weeks perfecting. For a small agency, that could mean, for example, the agent one person built for client research becomes the agent everyone uses the same day. Credentials get stripped automatically, so the file is safe to hand to anyone, and you can back up your agent, move it to a machine that day that maybe it runs day and night, maybe it runs on a VPS, and you can hand it to your team, or you can share it with your community. You get a prompt library you can pull from to build your agent's skills faster, and you get the agentic operating system, the dashboard, where your Hermes plugs in right next to your Claude, your open Claude, and your free Claude code with shared memory across all of them.

Now, inside the AI Profit Boardroom, we've got the full Agent OS as a ready-to-install zip file, Astros, Oracle, Apollo, the video agent, the full memory system, all set up and working together. You go back to Astros, you find your topic, you click video agent, it takes you to your video director, you plug in the topic and the angle, you hit write the script, and then it generates a video avatar script, B-roll, all in one place. Astros sits inside the Agent OS next to Hermes Oracle, which monitors Twitter and finds the latest trending news, then lets you draft social content or SEO content in one click and Hermes Apollo, the voice agent, where you can talk to your agent, chat with it, and use mixture of agents for higher quality outputs than a single model alone. With Astros, the monitoring is automatic, the ranking is automatic, the angle generation is automatic, and the content creation is one click, and every step gets logged to your memory, so your whole system gets smarter over time. You can get the full Agent OS, including Astros, Oracle, Apollo, the video agent, and the Obsidian Memory Vault as a complete zip file you install directly.

Instead of dumping every folder on the desk to find one page, your agent just opens the right drawer, grabs what it needs, and closes it again. If you run a business, you could feed your onboarding process into Hermes agent, and every task from then on follows your exact process without you explaining it again to every new hire or every new agent you spin up. On top of that, you get a 30-day roadmap that walks you through setup step by step, four live coaching calls a week where you can bring your exact build and get help on it directly, and daily tutorials showing you how to turn documents like this one into permanent skills your agent actually uses every day. Inside the AI Profit Boardroom, you get this exact system plus a 30-day road map that walks you through setting it up piece by piece, plus four live coaching calls a week where you can bring your exact setup and get help on it directly from people who use it daily. Daily tutorials show you how to keep adding new skills like the ones Hermes builds with /learn straight into your Agent OS so nothing you learn ever gets stuck sitting in one place.

Then install Unthropics front-hand design skill. This gives Claude a clear direction, typography, colors, and spaceships. Now, Claude can search for real companies and pull that code into your project. Making sure that your website doesn t lock like AI Slop. If you want to state the cursing edge of AI, hit subscribe for more content like this.

It asks the AI to analyze the problem, before jumping straight into it. Detern your raw request into a complete feature plan. So you can actually agree on what you re building before AI starts building it. So it stops looking for APIs that existed two years ago. If you want to state the cutting edge of AI, hit subscribe for more content like this.

By the end of this, you'll know every major command, every unique feature nothing else has, how it compares to Hermes agent, and exactly how to run prime agent safely because this tool definitely needs safety rules, and the story about why explains exactly how that works. You get the HNOS agent operating system where Prime agent plugs in alongside your Claude, your Hermes, your Open Claude, all sharing one memory of tools for video SEO agents and AI advertisers already wired in. You get the zip file, the 30-day implementation road map, a full video tutorial, and daily updates as we ship new versions, daily step-by-step tutorials on exactly what's in this course, the command formula, building scores, designing checks, four coaching calls every week where you share your screen, show your Prime agent setup, and get it fixed live. So, you got three workers running at the same time, one manager assembling the results, one command from you, and three sub agent details worth noting because each one shows how deliberately this was built. A self-improving agent gets better at whatever actually gets rewarded, not at what you meant, which means how you check the work now matters more than ever has in the history of these tools, which is why you want to make sure that you have a very clear bar of what the work and done looks like.

Again, we've got a closer look, a bit of a breakdown for flavors, which is cool, and you can just see the basics of this, and I I think this is, you know, based on the prompts I gave it, pretty cool, but you can see exactly the kind of effect that this design style has. Like, if you say, "Find out what works about this, but I want to change ABC, or look at these four things as inspiration, and adapt You can now have these assets fully built using Claude in this new system. And by the way, if this sounds like I'm speaking Spanish, I'll put a link down below for the full Claude code masterclass that goes through all this stuff, design systems, building Claude, learning how to use Claude, all its features, websites, memory systems, everything you could possibly imagine, including all the stuff that I just don't share on YouTube. So for example, if I come off this and show you, if I just come down here and I click on Jack design, what this is, as you can see, is basically images of me with some descriptions. So if I'm going to get my consistent thumbnails and if I just want to change one section, I just literally select Jack design or it could be Jack thumbnail style one or Jack sky image or whatever I want.

It was an idea that Andrej Karpathy, who now works at Anthropic, he put this idea out earlier this year and it blew up because this was a way of taking all of your raw data and pulling out insights so that the AI doesn't have to search through all of your raw transcripts, it can actually just go to the wiki and find the key insights and the summaries that's actually going to help with whatever it is that you're working on and building. So, what we're going to set up, if you haven't guessed already, is we're going to set up a nightly run so that anything that gets added into the raw folder is going to then get compiled overnight. I'm not going to answer these questions cuz I'm going to ask Hermes to just answer them for me cuz he's got all the information about me, but if you don't have anything set up, then you can answer the questions. You can regenerate the wiki from those, but that's where everything's going to be stashed and overnight, which is what we're going to set up in a second, it's going to do the processing every single day. So, you can see we're giving each folder an agents.md file, which is going to tell Claude Code or whatever harness it is that you're using, it's going to tell it what's in that folder and how to work with that information.

So, right here, you can see that Klaus, my agent right here, is inside of my school community, logged in as me, and it's going through, and it's liking posts that I wanted it to like. I'm going to create a new agent real quick, and I'm going to call this one sort of like my media agent, and I'm going to connect this agent to Slack, and so that it can basically notify me when there's things going on in the Slack channels that have to do with sponsorships or new videos or new opportunities, anything to do with the media side that I deal with in Slack with a few different teams. Uh maybe you should just check in with Money to see if he has any information about, you know, what's going on with our Anthropic reel or just what's been going on with that." So, what this should do is it should basically tell Klaus that, you know, "I'm just an executive assistant, but I can see that Money is a Slack media agent. And what else is cool is if I go to my conversation with Money, it says, "Hey, by the way, Klaus asked me about this." So, they're not going to be able to keep things secret, they're going to snitch on each other if they're, you know, being weird, but that's how that works and I think that that is a super cool feature. For me, I'm going to try to set up all of the Hermes agents that I have right in here and I think it's going to feel more cohesive because they can talk to each other more easily than the way I have them setting up talking to each other in Hermes.

If Claude keeps ignoring your instructions, your Claude.md might be the problem. If you use Claude every day, that file probably has too much context in it. On used skills, MCP servers and plugins, then it shows you the proposed fixes before it changes anything. Review the plan and then just remove what you don t use. Then if you want to state the cutting edge of AI, hit subscribe for more content like this.

Now, to access this, just come over to the cloud website, click on design in the bottom left-hand corner, and you can literally create your design system by here under design systems, and then you can create a new design system. Think of this as having the world's best designer, and he has at the moment what is known as a terrible memory, something like Forrest Gump {slash} Memento style, and what we're going to do here is give him a codified blueprint, typography, graphics, everything so they understand exactly your specific design style, and you'll get something that looks like this. I said, "Hey, I love this presentation, turn it into a design system." So, action one, take any design assets that you've got, give it to Claude, and it will turn it for you into a full design system. So, effectively, what we're going to do is give Claude a reference design of, "Hey, I want it to look like this or be this," give it a criteria, and it's going to spin up multiple critics who will ruthlessly challenge Claude's design, and it will go round in a loop, hence the called the gauntlet loop, as Matt showed, or the design loop, until it reaches its end state, meaning that you have a last back and forth, and you don't need to keep basically coming back and forth as much, and the first design is way better. And let's just put it to the test and say, "Hey, can you actually, if I just gave you the URL, I'm not even going to tell you anything, I'm just going to give you the URL, can you actually design this?" And look at this.

Because Moxo turns a business process into a live workflow. The AI handle of the routine work while your team handles the judgment. Even people outside your company can complete their task without even creating an account. So instead of chasing files and approvals and updates across so many different tools, you can see what s moving, what s stuck, and exactly who needs to act on it. If you want to state the cutting edge of AI, hit subscribe for more content like this.

I used to have five windows open, one for Claude code, one for Codex, one for Hermes, five windows open and Telegram, web browser, desktop interface, command line interface, different agents on different interfaces with no context shared between them. It's like combining Slack and Discord together but designed for humans and AI agents to truly work together. So, my colleagues, Claude code, multiple Hermes agents can all work together. So, let's say I bring in uh one agent and I can actually talk to the agent. >> And so, you can Buzz is not for everyone, but it may be the clearest preview of what an AI first company looks like.

My main agent, Luffy, is talking to two other agents and asking them to reply. So, I can have one controlling agent managing agents on my local machine and different machines and all talking together. I'm going to call it social media, and then I'm going to create a channel, and then I can add or create agents. I'm going to add in I'm going to add in Robin at and now that I've added Robin, I can start talking to her. So, hey Robin, I say, "Hey, give me an update on my social media stats." You'll see that this signifies that they read it.

And it still doesn't solve the main issue, which is the context in Hermes still doesn't get carried over to other AI agents Buzz is the first tool that I've used that does It's like combining Slack and Discord together But designed for humans and AI agents to truly work together So my colleagues, Claude Code, multiple Hermes agents can all work together sharing the same project history and handing off stuff to each other in this video, I'm gonna show you How Cloud Code cross-checks code from Codex How I have two Hermes agents, one on VPS and one on my local hosted machine talk to each other And start a meeting inside Buzz with my human colleagues, with AI agents chiming in and turning them into actions immediately And the moment that blew my mind was when I asked Hermes to do something for me. And you can check for any updates So last but not least, on the top right, you can see who are the channel members, and you can even bring up your terminal, and you can do things with your command line interface if needed And then you can also start a huddle and chat with your teammates with voice if that's what you wish. He can bring in his agents, I can bring in my agents, and I can also set up a channel with my wife and have an agent that deals with my family life, updates my wife, and I can give the agent the calendar of my kids, And The agent can help organize my family activities together. So let's show you an example of agent talking to agents So I have two Hermes agents, one on a VPS and one on another machine that's ho-hosted locally, and I'm gonna make them chat to each other So what I'm gonna do is I'm gonna get my main agent to coordinate this I'm telling it to compile a list and there are some jobs that are useless and not used anymore And so I've given a command and basically, like my two Hermes are all running cron jobs, and I wanted to come back with a list of what they're doing, make sure they're not duplicated because, you know, you-- a lot of you are probably running OpenClaw Hermes and like multiple instances of it, and some things might be duplicated. But if it's not something super intensive, which is 95% of all the other stuff I do, I'm gonna go back to Buzz And the main reason that I'm moving to Buzz is that I realized I was spending so much time opening different windows, copying and pasting context And explaining the same thing again and again in Hermes, Codex, Cloud Code, I was using all of them now they can work all together, including my human teammates, who I can give them access to my agents One of the moments that really sold me was that I asked Claude Code to do something, But it was missing some files and credentials.

Get the Agent OS Hermes Agent Masterclass Want to make money and save time with AI? Join here: Video notes links to the tools Get a FREE AI Course Community 1,000 AI Agents Get a FREE AI SEO Strategy Session Get 200 Free AI SEO Prompts Get out SEO link building book here Turn Any Book into an AI Skill: Hermes AI Agent Update Learn how the new Hermes AI update allows your agents to master any book or PDF using a single command. Discover how the Book Brain Engine creates permanent knowledge assets that are 50x more efficient than traditional context windows. 00:00 - Intro: AI Agents Can Now Learn Books 01:18 - The Problem: AI Forgetfulness 01:58 - How the Book Brain Engine Works 02:38 - 50x Better Token Efficiency 05:51 - Business Use Cases for AI Skills 07:45 - Why You Need Skill Stacking 10:27 - How to Build Your AI Specialist

Claude Code와 Codex는 잘 돌아가는데, 정작 결과와 내가 남긴 피드백은 툴마다 흩어져서 그때 왜 그렇게 판단했지 를 다시 찾기 어려웠던 적 있으신가요? 이번 영상에서는 Block이 공개한 오픈소스 AI 협업 도구 Buzz에 이미 쓰고 계신 Claude Code와 Codex를 그대로 붙여서, 결과도 피드백도 판단 근거도 한 채널 한 스레드에 모으는 방법을 정리했습니다. 여러분의 업무 시스템에 Buzz를 도입해보고 싶으신가요? 댓글로 편하게 의견을 남겨주세요 타임라인 0:00 AI 작업 결과가 흩어지는 문제 1:20 Slack과 Buzz의 차이 3:18 방과 모델 실행 위치 구분 5:30 Buzz 커뮤니티 만들기 8:20 Claude Code Codex 에이전트 설정 12:58 Buzz 작업공간과 컨텍스트 14:51 해석담당 반박담당 실습 19:43 로컬 모델 사용하는 방법 20:56 사용자 초대 21:47 아웃트로 (Buzz의 한계와 운영 조건) 영상 상세 가이드 Buzz 앱 다운로드 Buzz 공식 GitHub Buzz 공식 소개 Buzz ClaudeCode Codex AI에이전트 업무자동화

And to do that, you can say, "I'd like you to go ahead and summarize all the things today and just write for me a handover document that I can take over now into claude code and pass on to another agent." Literally give it that prompt and Hermy's agent will go grab the skill and write you a fully detailed handover document. So to do that, we come over to Hermes and he can say, "Hey, I want you to go ahead and use my teach me skill to teach me the best way to grow an Instagram account." And just like that, it's access the skill. You can connect it to Telegram super easily and then you'll have your entire Hermes agent running on your phone and you can use it whenever you want to fully host it and you can come back and control it in your actual panel. But one of the limitations of Telegram is the fact you can do one chat at a time, which is why I really like this agentic OS because I can have one chat here and say, "Hey, explain to me the um last 30 days skill." for example, send that bad boy off. So you can see uh here for example we may have something like sweet spot references station and sharp or we can go uh full sort of like um if you think about like the odyssey level of information if you want to then fable 5 looks at those three questions those three responses and then gives you an answer based on that and it's so so cool and to activate it all you need to do is come down and do forward slash and just type MOA and it just runs it through your mixture of agents model present and just say hey I want to know how to double my net profit please tell me how and the Cool thing about this is it's also token saving because it brings over the cash in a really cool way.

All unsere Unterhaltungen liegen natürlich auf dem Slack Server, aber kann man KI Agenten in Slack hinzufügen, Agenten wie Clod, Open AI, Hermes oder irgendeine andere externe Plattform oder einen Agenten, damit sie an den Diskussionen teilnehmen und Aufgaben mit uns ausführen. Ich habe natürlich einen Mac, also kann ich einfach nun ja den Download der Anwendung anfordern und überprüfen, ob sie auf meinen Rechner heruntergeladen wurde. Es ist also sehr, sehr wichtig, einen VPS zu haben, einen sicheren externen Server und darüber hinaus brauchen wir einen Server, der etwas leistungsfähiger ist. Also gut, ich werde jetzt einfach einen weiteren Agenten erstellen. Also dieses Mal der Agent, wir werden also versuchen einen neuen Agenten zu erstellen.

Why you might want a dedicated memory provider instead of using built-in memory, the different memory options and why I picked the two we're setting up today, how to install an unofficial lightweight memory layer, and how to install an official heavyweight memory engine. And like we can see, it can be used for Claude code, for Codex, it can be used for Cursor, it can be used for any agent that you want, and we can have all of them connect to the same memory bank if we want to. And one thing I want to mention just very quickly is that if you're running Hermes and you have your terminal execution back end running in Docker, we can't just give Hermes the task of setting up Mnemosyne for us because Hermes would try and run its terminal, and it would run inside of a Docker container here, and that's not what we need because we need our memory system to survive outside of the terminal back end Docker container. Okay, so I'm now on the hindsight profile and if we go over to memory and context again, we can see here that I have persistent memory and user profile still turned on and the built-in memory provider is set. And now if we go over to Docker Desktop, which by the way this has to be running, I have a whole video that goes deeper into how to get your Hermes agent and now something like hindsight containerized inside of Docker, but we can go here and we can see that we have our server is now running.

Мой телеграм и гайд по установке и регистрации Buzz здесь: Три дня назад я поставил себе Buzz, новый мессенджер для ИИ-агентов от Джека Дорси, создателя Твиттера. Внутри него Claude Code и Codex общаются друг с другом напрямую, без моего участия, и к концу работы у меня уже готовый результат. В этом видео разбираю, что такое Buzz и чем он отличается от других инструментов, и прохожусь по четырём главным фишкам: агенты разговаривают между собой прямо в канале, автоматизации по расписанию, встроенное хранение кода вместо Гитхаба и общие вычислительные мощности на всю команду. Ссылка на Buzz: Плейлист про Hermes Все мои соц. сети Таймкоды: 0:00 Вступление 0:40 Что такое Buzz 2:04 Установка и приватные ключи 2:43 Подключение своих агентов 4:02 Создание каналов для агентов 5:56 Агенты работают без тебя 8:24 Workflows и автоматизация 11:16 Свой Git-релей вместо GitHub 11:51 Share Compute 13:28 Звонки агентам и минусы Buzz buzz buzzai hinkok

So, that's why in this video, we're going to go over exactly what all those skills are, how we can be able to use those skills here to building applications with the highest accuracy, and also how does it work behind the scene, and how are they different compared to skills that we have already talked about on this channel, like G stack, Sue Powers, GSD. Now, unlike any other traditional code review skills that we have seen before, where it's just only checking for bugs in the same sessions that it wrote it, this time is actually going to be very different because we're actually going to start a fresh context window here and following a code review checklist that Matt Pocock has created, walk through the names step-by-step to make sure that the code here is fully checked. Uh the last one here you can see is called a data clumps means that if there's any many data types here that are being referenced by many functions, then maybe we should group those data types into a one same type that's much more easier to reference by many functions here, okay? There could be one of the misconception here is that people think that this is actually just putting all those code, all those smaller function here, stuff it into one single file, which cause like 10,000 lines of code inside of one single file. Now, of course, if you want to go in a further deep dive on all the other skills that I have talked about, as well as some deeper use case on how we can be using those skills in a practical situations, you can also check out our ASS builder in our school community in lesson four here, we actually going to dive in deep onto what those skills are and how we can apply that into building real applications.

Also du kannst es verbinden mit Cloud Code, mit Cloud Desktop, sprich mit Cloud Cowork oder eben mit Cloud Chat in der Desktop Version, aber auch mit Cloud AI, also mit der Webvversion, mit Chat GPT, Chat GPT Work, Codex, Antigravity, viele weitere bis hin zu Open Code. Und hier kannst du jetzt eben ein eigenes Plugin unter Anführungszeichen hinzufügen, also eben einen Connektor auf gut Deutsch und dafür gibst du irgendeinen Namen, also Testwebseite, also auch hier habe ich es schon eing gerichtet. Aber jetzt grundsätzlich mal der Unterschied zwischen der kostenlosen Version und der Pro Version, denn wie gesagt, es gibt verschiedenste Möglichkeiten und auch andere MCP Lösungen, um deine Webseite, um WordPress mit der KI deiner Wahl zu verbinden. Und genau dazu möchte ich jetzt noch was sagen, also dass der dass die Firma hinter Novier diese Warnung hier um zur eigenen Absicherung hier hinschreibt, ist völlig verständlich, denn Novier ist extrem mächtig, denn mit Nova kann deine KI quasi alles in WordPress oder auf dieser WordPress Webseite machen, was eben ein echter Mensch mit Admin Zugriff auf dieser Webseite auch machen könnte. Nicht nur was die Sicherheit per se betrifft, sondern sie haben ein automatisches Crash Recovery, das bei mir noch nie zum Einsatz kam, bei Nova und ich habe das jetzt auf ja mindestens 15, 16 verschiedenen Webseiten im Einsatz gehabt für unterschiedlichste Aufgaben, noch nie auch nur irgendwas kaputt gemacht hat, aber es gäbe eine Crash Recovery, die quasi sofort alles zurücksetzt, falls doch mal irgendetwas passieren sollte auf den Zustand von davor, einen Safe Mode noch noch dazu so ein 30 Sekunden Limit, dass spezielle Skripts nicht unendlich lang in einem Loop laufen können und die Seite deswegen nicht mehr aufrufbar ist etc.

Now we can see our three team members and let's actually just send a message like, "Hey!" Alright, Fizz just responded saying, "Hi Leon, I'm Fizz. Of course, we've got Claude Code and Codex but the system will also detect any other harnesses that we might have like Hermes Agent, Kimmy Code, OpenCode, Goose which is kind of Buzz's own platform, Cursor, Grock, Omi Pi, Open Claw and we can even add a custom harness. And you know what, just to keep things interesting, let's add one more agent and let's call this one Walter and let's say your name is Walter. What you can do is install open code on your machine and then you can use open code to run models through Olama or Alum Studio. If we wanted to see more details as to what Fizz is up to, we can click on view activity and here we can see all Fizz's outputs.

VOTE EM MIM NO PRÊMIO iBEST: HOSPEDAGEM VPS NA HOSTINGER Use o CUPOM: ELIRIGOBELIAI (10 OFF) Se inscreva para receber novidades: Link do buzz: Você precisa assistir estes vídeos Crie seu app: App de jogo: Curso N8N: Agente Whatsapp: Me segue no insta: elirigobeli

Aber was sehr interessant ist, nach mehreren Monaten Arbeit, musik Tests und Experimenten gibt es bei Hermes natürlich Dinge, die ich mit Clote nicht machen kann und genau deshalb wird Hermes heute zu einem Trend, ist sehr gefragt und wird vor allem zu einem Werkzeug für Unternehmen. Wir werden also versuchen, den Unterschied zwischen Claude und Hermes ein wenig zu verstehen, was die Stärken der einzelnen Tools sind, um die Wahl besser zu verstehen und zu wissen, wie man sie einsetzt, wann man Clot und wann man Hermes verwendet, insbesondere im Unternehmenskontext. Das kann ich mit Cloud nicht machen, weil Hermes irgendwo versucht, mich basierend auf den verschiedenen Kontexten, die ich mit ihm habe, zu verstehen. Was Hermes mich also kostet, ich lasse Hermes auf einem VPS laufen und ich werde Ihnen gleich erklären, warum ich es auf einem VPS laufen lasse. Heutzutage kostet ein Server nicht viel und besonders für ein Unternehmen empfehle ich Ihnen tatsächlich einen Onlinees Server zu nehmen, denn nun ja, bei einem musik Server gibt es den KVM1, den hier für Anfänger und voila, er kostet nur 5,49 Cent pro Monat nicht teuer.

) Buzz (coupon : GOBUZZ) : Documentation : Accès à mes 36 formations (coupon : FIRAS30) : J ai entièrement testé Buzz, une solution collaborative capable de remplacer Slack tout en intégrant directement vos agents IA dans votre équipe. Dans cette vidéo, je vous montre comment installer Buzz sur un serveur VPS, le connecter à votre nom de domaine et configurer une connexion sécurisée avec des clés SSH publiques et privées. Vous découvrirez également comment créer des canaux de discussion, ajouter de nouveaux agents IA et communiquer avec eux directement depuis Buzz. Buzz peut devenir un véritable espace de travail centralisé pour gérer vos conversations, vos équipes et vos agents d intelligence artificielle, tout en gardant le contrôle de vos données grâce à une installation auto-hébergée. SOMMAIRE : 00:00 - Introduction : découvrez Buzz, l alternative à Slack 02:52 - Installer Buzz facilement sur un serveur VPS 06:56 - Comprendre la différence entre clé publique et clé privée 08:06 - Générer et configurer vos clés SSH publique et privée 12:23 - Connecter Buzz à votre nom de domaine 15:36 - Créer votre premier canal de discussion sur Buzz 16:28 - Communiquer avec vos agents IA directement dans Buzz 20:15 - Ajouter un nouvel agent IA à votre équipe 22:54 - Tester votre nouvel agent IA dans Buzz 24:08 - Télécharger mes 36 formations avec ce code promotionnel Buzz AgentsIA

Ein KI-Agent braucht kein Plugin und keine Vektordatenbank er kann Dateien lesen und schreiben. Genau darauf setzen zwei Konzepte auf: Andrej Karpathys LLM-Wiki und Googles Open Knowledge Format. In diesem Video baue ich beides live in Obsidian auf, zusammen mit Claude Code. Du siehst die drei Schichten RAW, Wiki und CLAUDE.md, wie der Agent aus Rohquellen verknüpfte Wiki-Seiten macht und wo das Ganze an seine Grenzen stößt. KI Strategie Call: Meine kostenlosen Templates: Instagram: LinkedIn: KAPITEL 0:00 - Obsidian als Wissensdatenbank für KI-Agenten 0:42 - Karpathys LLM-Wiki: Warum RAG nicht reicht 2:59 - Die drei Schichten und Googles OKF-Standard 5:03 - Setup live: Vault, Terminal-Plugin, Claude Code 7:46 - Wissen einlesen und die ehrlichen Nachteile Links Ressourcen Andrej Karpathy, LLM-Wiki (Gist): Google Open Knowledge Format, Spezifikation: Google Cloud Blog zu OKF: Obsidian Download: Obsidian ClaudeCode SecondBrain LLMWiki OKF Karpathy Wissensmanagement KIAgent PKM KIAutomatisierung

So these docs, not only can you use them to learn my skills better, you can also use them just to learn about AI coding and understand how things work at a deep level. Before we had model invoked skills and user invoked skills and the benefit of user invoked skills which we can see inside the main skilled MD file is these are hidden from the agents context window until you invoke them. However, this has a horrible failure mode, which is when you get to the end of a grilling session and most of the hard stuff is already done, you just have a bunch of easy questions, one per turn, that you're just basically saying, "Yeah, that sounds good. So you can use it for writing skills, you can use it crucially for taking your agents.mmd pulling it out into skills to stop that horrible frontloading. And so actually having a document that you can just pull out, give to someone, they can comment on it on Google Docs or whatever is actually really useful.

Trag dich hier in den Newsletter ein für mehr AI-Hacks, Agent-Systeme Automations-Frameworks: BUZZ: Bass: Jack Dorseys neues Open-Source-Framework bringt Menschen und Agents in einen gemeinsamen Workspace Ich habe live gesehen, wie zwei Agents sich selbstständig eine Geschichte ausgedacht und ohne mein Eingreifen miteinander kommuniziert haben. In diesem Video zeige ich dir, was Bass ist, wie du es installierst und selbst hostest und warum es kein OpenClaw- oder Hermes-Killer ist, sondern genau die Schicht, die zwischen Menschen und Agents bisher gefehlt hat. Kapitelübersicht: 00:00 Live-Demo: Zwei Agents kommunizieren autonom 02:15 Was ist BUZZ? Self-Hosted Community 07:06 Agents erstellen und konfigurieren 10:22 Compute Sharing: Eigene Hardware fürs Team bereitstellen Was du lernst: Was BUZZ ist und warum es kein OpenClaw- oder Hermes-Killer ist, sondern die fehlende Schicht zwischen Menschen und Agents Wie du in wenigen Minuten eine eigene Community aufsetzt gehostet oder self-hosted auf deinem eigenen Server Wie du einen Agent erstellst, ihm eine Instruction gibst und ihn mit Claude Code, Codex oder OpenClaw verbindest Warum die Anzahl paralleler Tasks pro Agent entscheidet, ob du wirklich eine Workforce aufbaust Wie du einzeln oder in Gruppenchats mit deinen Agents kommunizierst, ohne kompliziertes Setup Wie Compute Sharing funktioniert und warum dein Team von einer geteilten lokalen Maschine profitiert Warum Self-Hosting auf einem deutschen Server dir volle Datensicherheit für deine Chatverläufe gibt Fazit: Bass ist kein Ersatz für OpenClaw, Hermes oder Cloud Code es ist die Schnittstelle, die bisher gefehlt hat. Du bekommst: ein Open-Source-Tool, das Menschen und Agents in einem gemeinsamen Workspace zusammenbringt die Möglichkeit, dein gesamtes Team mit den passenden Agents auszustatten, ohne dass sich jeder selbst um das Setup kümmern muss volle Kontrolle über deine Daten durch Self-Hosting Zugriff auf beliebig viele parallele Agents, die proaktiv mitdenken und selbst Ideen einbringen Wer jetzt einsteigt, baut sich gerade die Workforce auf, die in ein paar Monaten Standard sein wird.

We put a Claude agent and a GPT agent in a channel and told them to work together on many tasks, and we went through everything else it can do. Now, coordinating a group of agents where they can talk to each other isn't a new thing entirely, and Claude code already does it with something called agent teams. Buzz lets you connect the Claude Code and Codex subscriptions you're already paying for, along with Goose, which is Dorsey's own coding agent, and then you can have all of them work together on the same thing. So if you've got a few Mac minis or any other computers sitting around, you can join them together and the combined power hosts one big model and you can share that whole setup with your team so everyone's working off the same one. If you want agents arguing with each other, so one of them catches what the others missed, Claude Code has agent teams.

This is the Harold release and a lot of the highlights here center around their new and improved voice mode where it's a lot more hands-off now and you can communicate with your Hermes a lot more naturally like you would with another human. And as you can see, while we're recording this right now, this voice conversation here, it's picking up what I'm saying right now, which we find very inconvenient, you know, if you're already on your computer. Or maybe if you're outside and you have access to your Hermes desktop app on your phone, this is also another way for you to work with your Hermes agent. And lastly, there's also Hermes import agent command, which can migrate a Claude code or code X CLI setup into Hermes, which saves you a lot of time. The bottom line is the new and improved voice mode really gives you an additional way to communicate with your Hermes, but I wouldn't say it's a primary way to communicate with Hermes.

Learn Buzz Deploy Buzz: A couple of weeks ago I told you my Buzz wasn t on Block s servers, and I never actually showed you how. I paste one public key into Railway, click deploy, and about thirty seconds later I ve got my own Buzz relay running: Postgres, Redis, storage and the relay itself, with no terminal, no SSH and no Linux anywhere. 0:00 How to Self Host Buzz AI (Jack Dorsey Setup) 0:30 What is a Buzz AI Relay? (Buzz App vs Server) 1:34 Buzz AI Security Private Access Control 2:02 How Much Does It Cost to Self Host Buzz AI? 2:49 How to Deploy a Buzz AI Relay on Railway (Tutorial) 3:57 How to Connect Buzz Desktop App to Your Server 4:39 How to Add a Custom Domain to Buzz AI Relay 5:09 How to Make Your Buzz AI Relay Private 5:46 How to Invite Users Team Members to Buzz AI 6:37 Private Buzz AI Workspace AI Agent Showcase 7:08 Is Self Hosting Buzz AI Worth It?

My multi-agent team Claude, GPT, and Hermes agents spread across my PC, my DGX Spark, and a VPS now researches and works together in one shared Buzz workspace, and this is how I set it up: Run agents? Agent Wikis has free, curated LLM wikis on a ton of AI topics plus a Pro tier ( 9.99 mo) for super-sized wikis with way more depth: Sign up for my FREE weekly newsletter, where I spill my unfiltered thoughts on the latest AI news, cool research, and projects I m building: A follow-up to my Buzz beginner s guide: now that Nous Research has shipped a Hermes Agent integration, I walk through getting Hermes running in Buzz through the native gateway (my recommended path over the desktop ACP), including the Buzz CLI prerequisite that quietly stops agents from ever replying. I create a new agent, run the gateway setup wizard, wire up the Nostr keys, and use a skill I wrote to get the CLI installed and the whole thing connected. Then I show the real payoff three Hermes agents on three different machines researching, cross-checking each other s work, and collaborating in a shared repo. Resources: Buzz: Hermes Agent: My buzz-skills repo: Timestamps: 0:00 - A multi-agent team working together in Buzz 1:03 - Three ways to run Hermes Agent in Buzz 3:35 - The Buzz CLI prerequisite (the 1 gotcha) 4:46 - Creating an agent the gateway setup wizard 8:45 - Installing the Buzz CLI with my skill 10:24 - Admiral goes live: memory across machines 11:37 - Getting three agents to work together 13:28 - Wrap-up the buzz-skills repo Buzz HermesAgent NousResearch AIAgents MultiAgent Nostr AITools OpenSource AgentWorkflow

Ahora podemos tener un equipo completo de humanos y agentes de AI coordinados para trabajar de forma eficiente y al mismo nivel en nuestros proyectos Prueba Skywork desde: y consigue un descuento con el cupón Alejavi20 Curso intensivo IA: Curso de Make: Curso de Gemini: Curso de ChatGPT: Enlaces: - Buzz: - Guía Buzz con IA GRATIS: Conoce mi academia: 00:00 Equipo de agentes 01:15 Buzz AI 02:47 Caso de uso 1 06:49 Equipo de trabajo 08:20 Caso de uso 2 13:00 Resultado coordinado 16:42 Caso de uso 3 21:35 Caso de uso 4 25:22 Caso de uso 5 28:27 Caso de uso 6 Suscríbete a mi newsletter: Sígueme en mi Instagram (subo contenidos más resumidos): Te gustó el vídeo? Compártelo, suscríbete y déjame un me gusta, me ayudarás muchísimo Colaboramos? Si eres empresa, te dejo por aquí mi email: alejavi.colaboraciones gmail.com

Zum Deep-Dive-Video: Komm in den Fokus-Club: Ständige Rückfragen im Team kosten wertvolle Arbeitszeit. Jede kleine Unterbrechung durch wiederkehrende Fragen wie Kampagnenfreigaben oder Onboarding-Templates reißt dich aus deinem Arbeitsfluss. In diesem Video zeige ich dir, wie du Wissensmanagement mit Hilfe von Gemini Notebook (ehemals Notebook LM) im Team etablierst, um wiederkehrende Fragen zu reduzieren. Wenn du deine Prozesse optimieren willst, abonniere den Kanal für wöchentliche Tipps zum Thema Zeitmanagement und schreib in die Kommentare, welche Frage dein Team am häufigsten stellt. 0:00 Warum dein Team dich ständig unterbricht 1:13 Warum ChatGPT hier nicht die Lösung ist 3:10 Der neue Reflex bei Teamfragen 5:37 In 10 Minuten zum eigenen KI-Gatekeeper

Get the Agent OS Buzz Setup Get a FREE AI SEO Strategy Session Want to make money and save time with AI? Join here: Video notes links to the tools Get a FREE AI Course Community 1,000 AI Agents Get 200 Free AI SEO Prompts Get out SEO link building book here Rank 1 on Google with FREE AI SEO: 4 Buzz Workflows (Jack Dorsey s New App) The video demonstrates how to use Buzz, Jack Dorsey s new free app, to run AI-powered SEO workflows that function like an always-on SEO team. The presenter shows four workflows that worked largely on the first try: creating and deploying a formatted blog post directly to a website, generating keyword opportunities by pulling from an Obsidian memory vault and Google Search Console data (including trending topics captured via Agent OS workflows), prospecting verified link-building outreach leads and saving them to a file with the option to email via Google Workspace APIs, and building and deploying a full website to a test Netlify domain with five blog posts. The script also covers configuring multiple agents (e.g., Hermes Agent, Claude) inside Buzz, using free models APIs, team collaboration, and resources via the AI Profit Boardroom and a free SEO strategy session at goldie.agency. 00:00 Buzz AI SEO Promise 00:41 Buzz Setup and Channels 01:26 Auto Blog Post Deployment 02:27 Keyword Research via Memory 04:32 Link Building Outreach 05:37 Build a Website Instantly 06:59 Configure Agents and Models 08:38 Mobile and Gateway Integration 09:15 Why Buzz Beats Alternatives 11:38 Workflow Tips and Wrap Up 12:18 Community Training and Offer

See how the InfraNodus Obsidian plugin  helps you visualize connections in your notes and extract deeper insights from your knowledge base. This video covers how the InfraNodus Obsidian plugin goes beyond native tools to map your ideas. By comparing the InfraNodus GraphView against the native Obsidian GraphView, you can decide which visualization method fits your workflow better. This breakdown is designed for users who need to organize unstructured information and want practical ways to see how their notes relate to each other without manual effort. Timecodes: 0:00 Native Obsidian graph view vs InfraNodus 1:32 Finding gaps in your thinking 3:32 How the InfraNodus graph view works 3:52 Main difference: wikilinks only vs concept co-occurrences 6:32 What that allows you to do is to build a text network with graph metrics 7:27 Visualizing wikilinks only (skipping concepts) 10:00 Making InfraNodus work the same as Obsidian graph view 11:34 Navigating across different pages using the graph 12:06 Why adding concepts is better for LLMs 12:50 Generating AI insights from inside the graph itself 14:05 So you can take elements from the graph and integrate them into your AI workflow 14:15 Using an MCP server for graph view 15:09 Conclusion

Link to Resources: Presented by Zapier: Andrew Warner and Matt Van Horn break down this week s top GitHub repos for AI agents, coding workflows, browser automation, and developer productivity. Andrew Warner is joined by Matt Van Horn to review the most interesting AI and developer tools trending on GitHub. They discuss Buzz , Paperclip , Orca , Hermes Agent , ElgoLite , Alibaba s Open Code Review , productivity skills for Claude and Codex, browser automation, AI-powered code reviews, and tools that help developers build faster with coding agents. They also explore open-source collaboration, contributing to GitHub projects with AI, turning books into reusable skills, and why terminal-first AI workflows are becoming the preferred way to build software. Repos featured: - Buzz - Paperclip - Orca - Hermes Agent - ElgoLite - Open Code Review - Greptile - Zapier MCP - I Have ADHD - Pascal Editor - Book to Skill - JCode - Matt Pocock Skills - Compound Engineering - T3 Code - Claude Code - Codex - World Monitor 00:00 - GitHub Roundup: This week s biggest AI and developer repositories.

I m going live to show how I create an AI agent in Buzz that responds to anyone who posts in a channel... We ll cover the agent draft, system prompt, channel setup, and how the harness routes every message so the bot can act. As a fun side demo, the agent we ll ship is a profile pic roaster but the real topic is building always on channel agents for human and agent collaboration on Buzz (via Nostr). 0:00 Collaborative AI agents in Buzz 6:44 How the Roast agent works 11:51 Editing your Buzz profile photo 17:10 Connecting local models to Buzz 26:41 Structuring Buzz relays for business 32:18 Security and privacy on Nostr 36:46 Vision for remote Buzz agents 42:28 Writing custom instructions for agents 49:19 Using Canvas for shared knowledge 56:34 Advanced environment variables for agents 1:04:11 Explaining Buzz ACP subscribe kinds 1:09:53 Triggering agents on channel joins 1:19:44 Troubleshooting Buzz community connections 1:29:51 Building automated trading agents 1:34:59 Generating YouTube thumbnails with Grok 1:39:26 Creating Nano Banana image agents 1:47:23 Orchestrating multi agent workflows 1:54:01 Testing multi agent image generation 1:59:38 Installing and testing Buzz locally

But, you can invite people in and they can work alongside you, your Open Claw, your Hermes, your other agents, and they can all discuss with each other and have these kind of adversarial conversations to get way better results. You don't need to glue it together with MCP servers to GitHub and this and that, all in one place, and that opens it up even more with the fact now that you can bring human beings in here, and your humans, your AI agents can all work together to write code together, and ship code, and commit it to the local projects. Here, everyone works alongside each other, so you can see who's doing what, who's building what, who's thinking what, whether it's your AI agents, your humans, and Buzz, they're the same thing, and it makes building code so much nicer. Instead of having to write skills where Claude code calls Codex or Codex calls Claude code, you just have it all in one place where you can just tag Claude and Claude can say, "Hey, check this out. But I'm able to connect all these computers, run a bunch of models on them, and then everyone in here can not only talk and use this free compute that's being hosted on my side, but what's going to be really amazing is everyone in your Buzz communities that's sharing compute can actually team up together and build apps together, multiplayer app building.

The idea behind Hyperagent is that you build agents that go off and do the work, and every agent you make gets its own computer in the cloud, so it can go and do things out in the world instead of [music] just describing them. Then there's tools, where I switch on the ones Scout needs, web search and documents, so it can save what it finds, and every agent comes with a real browser built in, so it can open a competitor's pricing page and read it itself. And I can hand that same skill to any other agent on the team, so they all judge things by one standard instead of each doing its own thing. And one agent doing that on its own is the part that got me, but it isn't where this actually points, because the moment you've got one running itself, the obvious move is a second and a third, and that's when it stops being a tool and starts being a team. So, under Scout, I've got an inbox agent that clears the overnight pile and hands me the things that actually need a human, and a content agent that turns anything I hand it into posts ready to go out, each with its own computer, skills, and memory.

So first now we are going to use uh Gmail trigger because whenever we get a new email it s going to trigger it s going to activated and I m going to explain it uh uh very fast like uh I m not going to um go in depth for every nodes because like whatever I used in this workflow that s what I m going to explain and I just wanted to see this node uh okay uh you know that how we can connect the credentials to the Google account. Now we got five PDFs and these are all the five PDFs we got from the email and this is how we need to set it up first node and the second node we are going to use was if node let me take that one and I just wanted to explain and this node was uh let me check it once I just wanted to take this kind of object. So you you can set it up you know very well about this Google sheets was like a very kind of simple thing that I m going to use I can use choose my invoice a track record then I m going to choose the invoices tab I m going to just add one filter here that was invoice number basically invoice number had I m going to drag and drop from the previous field that s it and basically I just wanted to uh tell you something about this node basically why we wanted to implement to here only to check whether we have any duplicates in the Google sheet so we just given the condition the condition was going to work on the based on the invoice number now let me execute this step and it s given me data that means there is no any existed row number row number means a kind of invoice number and one more thing uh I m going to use here that uh if If node was like uh let me take that uh if node and I we just need to put something here. Let me I m not deleting at last we are going to check it all the things how it s one now work finally now I m not going to delete anything but to continue we need to delete that I just want to delete that 1 the first one I m going to delete row this row now let me execute once now it will to the first again row okay let me check how it s working basically how things was working here. Let me go back and do the things again properly from the agent we need to put invoice date and this is the invoice date and coming to this where is that so received date so received date was of course yeah this one is received date got it now we are almost done uh I just wanted to uh I just deleted the all the data now we are going to check how it s going to work I m going to execute this step.

আপন যদ ক ল উড এআই (Claude AI)-এর এই সমস ত প ইড মড ল দ ই বছর র জন য ব ন ম ল য ব যবহ র করত চ ন , অথব চ য টজ প ট (ChatGPT)-এর প ইড মড লগ ল দ ই বছর র জন য ব ন ম ল য ব যবহ র করত চ ন, ত হল এই ভ ড ওত আপন ক স ব গতম এই ভ ড ওত আম আপন ক শ খ ব ক ভ ব ক ল উড এআই (Claude AI)-এর এই সমস ত প ইড মড ল দ ই বছর র জন য ব ন ম ল য ব যবহ র কর য য এর জন য আপন র গ টহ ব স ট ড ন ট ড ভ লপ র প য ক (GitHub Student Developer Pack) প রয জন হব এই ভ ড ওত আম ইত মধ য ই ব য খ য কর ছ ক ভ ব গ টহ ব স ট ড ন ট ড ভ লপ র প য ক (GitHub Student Developer Pack) প ওয য য আপন যদ এখনও এই ভ ড ওট ন দ খ থ ক ন, ত হল প রথম এট দ খ ন পর আপন ব ষয ট ব ঝত প রব ন এর জন য, আপন র ব র উজ র খ ল ন এব গ টহ ব স ট ড ন ট ড ভ লপ র প য ক (GitHub Student Developer Pack) ল খ স র চ কর ন ত রপর, এই ল ঙ ক ক ল ক কর ন, এব আপন এই প জট ত চল আসব ন এরপর, আপন ন শন (Notion) খ জ প ব ন ন শন (Notion)-এর ম ধ যম আপন ক ল উড এব চ য টজ প ট উভয মড লই ব ন ম ল য ব যবহ র করত প রব ন যদ ও আম ইত মধ য ই ব য খ য কর ছ ক ভ ব আপন ন শন এআই (Notion Ai)-এর ব জন স প ল য ন সম প র ণ ব ন ম ল য প ত প র ন আপন এই ভ ড ওট দ খ ত শ খত প র ন আম এখন আপন ক শ খ ব ক ভ ব এর প ল স প ল য নট দ ই বছর র জন য ব ন ম ল য প ওয য য আপন র প ল য নট কতদ ন সক র য থ কব , স দ ক এখ ন ব শ ষভ ব মন য গ দ ন আপন র এড ক শন প ল য নট যতদ ন সক র য থ কব , এই প ল য নট ও আপন র ইম ল ঠ ক ন য ততদ ন সক র য থ কব ক ভ ব দ খব ন? আপন র এড ক শন স ব ধ ট কতদ ন সক র য থ কব ? এখ ন ক ল ক কর ন ত রপর এখ ন ক ল ক কর ন Your Benefits Application-এ ক ল ক কর ন এরপর, আপন Expires in almost two years ল খ ট দ খত প ব ন এর ম ন হল আপন র অ য ক উন টট প র য দ ই বছর স থ য হব স ই অন য য , আপন র Notion অ য ক উন টট ও দ ই বছর স থ য হব এর স ব ধ হল আপন র চ য ট হ স ট র ইত য দ ম ছ য ব ন আপন র অ য ক উন টট একট ম ত র WorkSpace থ ক দ ই বছর ধর চ ল থ কব এট দ ব কর র আগ , আপন ক Notion Ai-ত য ত হব আপন র ব র উজ র এইভ ব Notion Ai ল খ স র চ কর ন ত রপর আপন এট ত ক ল ক করব ন এব ত রপর Get Notion Free-ত ক ল ক করব ন এরপর, আপন ক আপন র ইম ল ঠ ক ন , অর থ ৎ আপন র ইম ল ঠ ক ন ট প রণ করত হব আপন য ক ন ইম ল ঠ ক ন দ ত প র ন আম আপন র GitHub অ য ক উন ট র স থ ত র কর ইম ল ঠ ক ন ট ব যবহ র কর র পর মর শ দ চ ছ এট ই শ র য ত রপর, Continue-ত ক ল ক কর ন আপন র ইম ল ঠ ক ন য একট ক ড প ঠ ন হব ওই ক ডট এখ ন ল খ ন এট আপন র ক ষ ত র ও ক জ করব এট কপ কর এখ ন প স ট কর ন এব ত রপর Continue ক ল ক কর ন এরপর, আপন র ন ম ল খ ন মন র খব ন, আপন ক আপন র GitHub অ য ক উন ট র ন ম র মত ই একই ন ম ল খত হব ত রপর, Continue ক ল ক কর ন এরপর, For Personal Life ন র ব চন কর ন ত রপর, Continue এব ত রপর Escape ক ল ক কর ন অথব , আপন চ ইল আপন র ড স কটপ Notion অ য প ল ক শনট ড উনল ড করত প র ন এভ ব আপন র Notion অ য ক উন ট ত র হয য ব এব আপন প রস ত ত হয য ব ন এখ ন , আপন মড লগ ল দ খত প ব ন তব , আপন এই সমস ত মড ল ব ন ম ল য অ য ক স স প ব ন ক ন ত ক ছ স ম বদ ধত আছ আপন খ ব ব শ ক ছ অ য ক স স করত প রব ন ন আপন শ ধ ম ত র একট স ম ত সময র জন য এট অ য ক স স করত প রব ন আপন র যদ আরও ব শ ক ছ র প রয জন হয , ত হল আপন ক প র প ল য ন ব ব জন স প ল য ন স বস ক র ইব করত হব এখন, প ল স প ল য নট সক র য করত , আপন ক আপন র GitHub অ য ক উন ট য ত হব ত রপর, এই ল ঙ ক ক ল ক কর ন: Notion ওয বস ইট Get Redirected Access ত রপর, এভ ব ন চ স ক রল কর ন ওয র কস প স র ন মট খ জ ন যদ আপন র অ য ক উন টট এভ ব লগ ইন কর থ ক , ত হল ওয র কস প স র ন মট স বয ক র য ভ ব প রদর শ ত হব এরপর, Get Verified With GitHub -এ ক ল ক কর ন অন ম দন কর র পর, আপন র ওয র কস প সট এড ক শন প ল স প ল য ন আপগ র ড হয য ব এর ম ন হল , আপন র প ল স প ল য নট এড ক শন প ল স প ল য ন র মত ই হব ক ন প র থক য থ কব ন এট পর ক ষ করত , এখ ন ক ল ক কর ন এব আপন এড ক শন প ল স প ল য নট দ খত প ব ন এরপর, আপন এখ ন থ ক এই মড লগ ল ব যবহ র করত প রব ন আপন আপন র কন ট ন ট অন সন ধ ন করত এব আরও অন ক ক ছ করত প রব ন আপন র প রয জন য অন য য ক ন ক জও করত প রব ন তব , আপন এখ ন থ ক ক ড -এর মত ব শ ষ ক ছ করত প রব ন ন ত ই, আম ক ড কর র পর মর শ দ ব ন যদ আপন র ক ড -এর প রয জন হয , তব আপন Kiro ব যবহ র করত প র ন, অথব আপন ক ChatGPT-এর প ল স প ল য ন ব ব জন স প ল য ন ন ত হব যদ আপন র কন ট ন ট গব ষণ , স ক র প ট ল খ ন , ব ক ড ছ ড সবচ য কঠ ন ক জ, অর থ ৎ গব ষণ , কর ন র প রয জন হয , তব আপন ত এখ ন থ ক কর য ন ত প র ন এত আপন র ক ন সমস য হব ন এছ ড ও, আপন র অ য ক উন ট র ম য দ দ ই বছর থ কব এট আপন র ড ট ও স রক ষ ত র খব এছ ড , আপন র যদ ক ন প রশ ন থ ক , আপন র ক ন অস ব ধ হল , কম ন ট স কশন ন র দ ব ধ য আম ক জ জ ঞ স করত প র ন এতদ র দ খ র জন য আপন ক অস খ য ধন যব দ

Dieses Video von "Claude AI models escape during testing | Microsoft’s best day" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

Ne und ähm und da geht s eher um Konversationen, um Chat-Dialoge, ne, also wenn ihr selbst an euer Verhalten euch vielleicht den Claw, Chat-Pubility, Perplexity erinnert, ihr, ihr schreibt ja mit der KI und daraus entstehen eben viel längere Suchanfragen, sehr viel mehr Kontext, und dadurch eben auch ganz neue Faktoren, wie dann der Content von euch empfohlen wird. Der Traffic, der dann aber entsteht, ist sehr viel wertvoller, weil der Nutzer kommt sehr viel gebildeter und aktivierter auf eure Webseite und seht ihr auch gleich, Traffic, der über die Sprachmodelle kommt, der scheint euch schon sehr, sehr viel mehr zu vertrauen und sehr viel höhere Conversion Rates auch zu haben, weil er sich eben vorher schon informiert hat und vielleicht Vergleiche mit den Konkurrenten gemacht hat oder ähnliches. Also ich zeige euch jetzt nicht, wie man Zielgruppen anlegt, weil ich hoffe, wenn ich davon ausgehe, dass einige von euch das schon gemacht haben, ist es ganz leicht, ihr könnt sogar einfach nur eure URL von eurer Webseite hinterlegen und dann machen wir euch schon einen Vorschlag über das Zielgruppenprofil. Jetzt siehst du, der Chatflash überlegt jetzt, liest jetzt den Artikel, das ist wirklich ein Agent, der jetzt hier arbeitet und schreibt jetzt quasi die FAQ. Also da muss ich auch erstmal gucken, so würde ich das vielleicht jetzt nicht auf unserem Blog posten, aber es scheint halt so diese Struktur und wirklich die Key-Aussagen ganz leicht zitierbar zu machen für Sprachmodelle, ist eben ein sehr, sehr starker Ranking-Faktor für diese Artikel.

Also, diese Kompetenz hier war tatsächlich eher auf die Cloud ausgerichtet, aber keine Sorge, wenn ich einfach nur diese URL nehme und zu Hermes gehe, äh, also hier sage ich ihm, kennst du diese Kompetenz und dann füge ich den Link ein. Also daher, aber was sehr interessant ist, schaut, er kann dir ganz einfach Schritt für Schritt helfen, diese Konfiguration durchzuführen und deshalb muss man wirklich gut verstehen, wie Hermes die Kontrolle über eine Software übernimmt. Man muss ihm nicht den Login und das Passwort geben, aber man muss ihm den MCP einen MCP Zugang geben. Heutzutage kann man tatsächlich mehrere Umgebungen erstellen, also mehrere Profile auf Hermes und er wird sich einfach dein Profil vorstellen und das ermöglicht es genau zu verstehen, wer Sie sind und dadurch wird er in der Lage sein, ihnen perfekt zu antworten. Also dem einen werde ich diese Aufgabe geben, dem zweiten werde ich diese Aufgabe geben, der dritte Agent wird diese Aufgabe ausführen und deshalb hat er tatsächlich diese Fähigkeit, wie Sie hier sehen können, diese Agenten parallel arbeiten zu sehen, wobei jeder auf diese bestimmte Aufgabe spezialisiert ist.

Ich habe das irgendwie gespürt, dass da etwas auf uns zukommt und ich habe ja schon früher immer Podcasts gehört während dem Autofahren über das Thema KI auch mit dieser ganzen progressiven Entwicklung Brin stecken, wo das losging mit Chat GPT habe ich natürlich von Anfang an immer etas Versuche unternommen. Man ist ja dann immer irgendwie in einem Tunnel oder man man baut nur Tools mit Python etc., aber es gibt im mehr und das ganze Thema mit der Automation und eine etc., Ja, das musst du mal irgendwo sehen. Also all, die mich davon abgehalten haben, gab es eigentlich nicht und ich habe schon mega lange keine Weiterbildung mehr gemacht und ich dachte jetzt es wird wieder mal Zeit, denn etwas zu schenken und dann äh habe ich recherchiert und bin auf das gestoßen und ich dachte komm das das geht ja nicht lange und ich war dann sehr positive Branch das war sehr viel Stoff und mega spannend war ja auch diese ganze Klickerei mit drin und so wie wie der Aufbau ist oder wie man diese diese verschiedenen Tools in Kombination anwenden kann auch für einen Telefonagenten. eines war super und wenn man das mal gesehen hat, dann hat man es drin und dann ist man später schneller als wenn man alles irgendwo nachlesen muss oder über die KI das irgendwie herausfinden muss, wie das geht. Was sehr wichtig ist, ist das, dass wir in Zukunft musik voll drauf setzen, die menschlichen Kontakte zu pflegen mit mit Kunden, mit Serviceeverträgen, um zu erreichen, dass die Leute gerne mit uns als Person zusammenarbeiten, weil die die Produkte werden irgendwann nicht mehr so gut vergleichbar, weil ja alle mit Kriararbeiten und alle das extrem pushen, geht es künftig um den Mensch und das ist das Hauptziel.

You can just write a quick little JavaScript snippet right inside the workflow to manipulate the data exactly how you need it. And the source material lists a few specific triggers like a cron node which is just time based like telling the system to wake up every Monday at 9:00 a.m. If I just pipe that raw unpredictable chaos directly into chat GPT inside my workflow, won t the AI just completely break? No, it s just a massive wall of text with lots of ums and uh a s, which is why nan takes that raw transcript and immediately passes it down the line into a chat GPT node. Okay, for the non-developers listening, how does a Python node using pandas actually process data differently than me just looking at a spreadsheet?

Anthropic, which is the maker of Claude, one of the most popular AI products with consumers and enterprises, has put out a blog post this morning saying that it has walked over its evaluations that it s been doing on its AI models, various models that it has, some that have been publicly released, some that it s just testing, and found that over tens of thousands of evaluation runs, so it s doing these tests internally, that in at least three instances, the models had gained access to the internet and then gone on to gain access to another company s systems via the internet as part of completing its challenges. Yeah, so these models that these companies have are capable of a lot of things and and the companies look at them and say, We ve got to figure out exactly what we re dealing with here. So, in the process of of testing them, where they do these challenges, including this one known as capture the flag, where they say, Hey, AI model, I want you to go and get this thing. Unfortunately, it seems that they actually were connected to the internet, that a third-party provider that was essentially providing the software, you know, the almost like the the playground where this was happening, had left it open to the internet, and then as a result, these AI models, which in some cases knew that they weren t supposed to go out on the internet and and do these kinds of, you know, unauthorized access, decided, Well, we were told this wasn t the internet, so clearly what s happening is we actually just have all this in our playground, and so we went and did these things thinking we were doing the right thing, but not actually knowing that in reality they were out there, you know, wreaking havoc. How come it these things are happening, these incidents where you re going out in the open, you re affecting other companies, potentially causing real damage, and you don t even know it s happening till after the fact?

The scariest thing about an AI answering your phone isn t the robot voice. So, imagine it quotes some guy 500 bucks on a five grand job. Write down the eight questions your phone gets every single week. How do you, you know, how do quotes work? My first client answered exactly eight questions and that became his receptionist.

Basically, if you have a brand, if you have a website, you could actually go through and create a design system by simply just clicking on create design system, you could create by connecting this to Figma or GitHub or uploading slides or assets, or you could actually create this using Claude Code, which is going to allow you to have the best fidelity. For example, we could say I run a Shopify e-commerce business, I manage inventory, do all of this, and then this actually goes through and finds all of the different Claude connectors that you need to make sure you are connecting to Claude because if we actually come over to Claude, we come into customize, we come into connectors, we could see that this is actually a huge pain for us to try to go through. If we come back over here, click on customize, this is going to be skills, we come under connectors, and we come over here and click on browse, this is incredibly difficult for us to actually go through, but llmhelper will literally just tell you exactly what ones you should connect based on what tools you use and what your role is. If we come over here, for example, we could do it for a video or we could literally send this directly to Quad Code, to Canva, or we could connect more destinations if we click right here, and we could actually automatically send this to all of these different placements, which is incredible. If we come over here, recent activity, we could see our side effects, we could see our weight, we could come over here and see our schedule, we could see our progress, we could see if there s actually a community here.

Jetzt werde ich also einfach den Zugang zu meinen Websites zu meinem Hosting gewähren, denn heutzutage kann Cloud den Domainnamen hinzufügen, meine Website hosten, sicherstellen, dass die Seite einwandfrei funktioniert und sogar das E-Mail Marketing direkt auf meinem Hostinger Server durchführen. Also, ich bitte jetzt Atlas Cloud, damit Cloud zu Atlas Cloud geht. Also lassen wir ihm jetzt die nötige Zeit, damit er das Bild holen und an Atlas Cloud senden kann. Also voila, eine kleine Demonstration von 5 Sekunden und ich kann ihn auch bitten, 10 Sekunden zu machen für weniger als ein und das ist wirklich sehr, sehr interessant. Also kopiere ich einfach hier ist also meine Anfrage.

Also nicht nur Videos, die man nur schaut, um ein Thema zu lernen, sondern wirklich ein interaktives Video, wo man etwas anklicken kann, wo man Fragen beantworten muss, wo man irgendwie Punkte sammeln kann, weil es in vielen Bereichen sehr nützlich ist. für jegliche Firmen, die intern Schulungen durchführen, sei es zum Thema Datenschutz, Compliance, KI oder auch Onboardings oder auch für Coaches oder Creatorn, um in ihren Kursmodulen eben nicht nur stumpf Videos abzuspielen, sondern wirklich so ein bisschen Interaktivität reinzubekommen und dass es für den Teilnehmer auch einfach attraktiver ist, spannender ist, da auch mehr hängen bleibt oder auch für Berater und Agenturen, die das als Dienstleistung verkaufen wollen, denn solche Schulungen zu erstellen war bislang wirklich extrem teuer bei irgendwelchen E-Learning-Agenturen z.B., aber mittlerweile kann man das mit KI deutlich günstiger umsetzen. Daraufhin hat Cloud dann erstmal ein Kurrikulum erstellt mit den gesamten Lerninhalten dieser Schulung, also, was kommt vor, was sind die Lernziele, welche Level gibt es hier dann mit Übersicht, dann die einzelnen Level im Detail, denn du willst natürlich erstmal durchlesen, was hier überhaupt jetzt konzipiert wird, bevor das ganze generiert wird, denn die Generierungen, die wir uns gleich anschauen, die kosten natürlich auch etwas Geld und da willst du natürlich erstmal schauen, dass das ganze hier inhaltlich stimmt. Von diesem kleinen KI Roboter und auch den Menschen hier, der dann im Video immer konsistent gezeigt wird, dann auch die Umgebung, dann erstellt er die Stimmen, also die Erzählerstimme, die dann in den Videos vorkommt, dann generiert er die Videos nach und nach, so wie wir es gerade gesehen haben und am Ende baut er das ganze dann mit HTML zusammen zu einer interaktiven Schulung, denn Sprachmodelle sind extrem gut darin mittlerweile mit HTML zu arbeiten, weil sie auf extrem viel HTML Inhalten trainiert wurden und in der Kombination mit Tools wie Hyperframes, die dann noch mal anleiten, wie sie konkrete Animationen generieren sollen, entstehen dann super Videos, wie wir gerade gesehen haben. Falls jetzt mal ein Video, das du generiert hast, nicht gepasst hat von Hixfil z.B., dann musst du das neu generieren, aber wie gesagt, wir haben jetzt hier auch so viele verschiedene Elemente, dass man eben nicht nur gebunden ist an diesen KI-generierten Videos, sondern die sind dann nur vereinzelt vorhanden und der Großteil sind eben die HTML-generierten Inhalte.

One downside of MCP is that you would define these tools that you re giving to your model in the definition for those tools, which could run hundreds of lines of instructions, imagine doing that for 20 different tools. So if your skill includes scripts that you ve written for to run deterministically, you can have that a lot of times your skills aren t going to have that, but it s sort of like the superpower if you choose to equip your skill scripts and other tools. Like oftentimes you re going to have both MCP and Skills, and maybe the Skills are going to be telling your agent how to use the MCP or the different tools that are available via MCP. And then we also came up with now Skills, because I think one of the things we realized was that oftentimes people don t necessarily need external tools, but what they need is some way to capture a prompt and to reuse that prompt over and over again or to like, share the prompt. It s just absolutely wild like how much utility you can get out from just telling a machine like, this is what I want you to do.

Ja, das hat vielleicht sogar  eher als Verschwörungstheorie angefangen, aber das ist eine Theorie, die heutzutage sicherlich  noch viel mehr Richtigkeit hat, nämlich, dass ein Großteil des sogenannten Traffics, also  die angeklickten Webseiten, dass die vielleicht gar nicht mehr von Menschen angeklickt werden,  sondern von Maschinen und dass die Inhalte, die man dann sieht, eben auch nicht mehr von  Menschen kreiert wurden, sondern von Maschinen. Und  jetzt die nächste Stufe wird dann der synthetische Inhalt sein, in dem eigentlich nur noch  computer generierte Inhalte im Internet rumliegen und wenn es dann ganz düster wird, wird  der dann auch nur noch von Computern gesehen. KI generierte Inhalte fluten natürlich auch  die sozialen Medien und wir stellen fest, dass soziale Netzwerke inzwischen  eigentlich gar nicht mehr her werden, wenn man sich diese KI generierten Inhalte,  diesen Slop, der halt sehr sehr schnell und sehr sehr einfach zu produzieren ist, von  dem es dann dementsprechend extrem viel gibt, gegen den dann Createrin eigentlich auch gar  nicht mehr ankommen. Ich habe auch gemerkt, dass gerade in dem Bereich sehr viele Creator aus dem Kereich  entstehen, was für große Brands einfacher ist für Kooperation und da merkt man das schon natürlich,  dass man gerade bei solchen Sachen super schwierig rankommen kann und diese Woche herstell Influencer  im Modebereich sind längst musik Realität. Die eine, die weiterhin sehr viel Müll abbekommt in Social Media und auch mit den Suchmaschinen  nicht mehr sehr viel sehr gute Inhalte finden wird und die andere, die bereit ist und auch in  der Lage ist für guten Inhalt zu bezahlen.

Erste KI Videos vom neuen Halo Modell überzeugen und der GPT Work kann jetzt mit euren eingeloggten Accounts arbeiten, sofern ihr das möchtet. Sofern YouTube uns nicht wieder einen Strich durch die Rechnung macht und unsere Links nicht funktionieren, wir kennen diesen Bug oder diesen Fehler, können ihn teilweise auch reproduzieren und es gibt tatsächlich da nichts aktuell, was wir daran ändern können. Laut Anthropic kann jeder mit einem solchen Link den freigegebenen Stand sehen, inklusive aller Nachrichten und Artefacts, die vor der Freigabe erstellt wurden. Die wichtigste Erkenntnis ist deshalb mit einem Linkteil kann bei KI Diensten praktisch eher einer Veröffentlichung als einer privaten Weitergabe Cloud Freigaben genutzt hat, kann unter Einstellung Privacy und Share Chats prüfen, welche Inhalte noch öffentlich sind und die entsprechenden Links dort wieder deaktivieren. Open AI verschenkt damit also nicht einfach 100.000 normale Abos, sondern baut ein gezieltes Netzwerk für den wissenschaftlichen Einsatz seiner Modelle auf.

Dieses Video von "Make AI Videos FREE🔥 Long (15 Mins) + Realistic + Dialogues✨AI Video kaise banaye🔥AI Video Generator" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

So, if you need to set up some configuration, or you need to go out and talk to someone, and actually go and run an errand, then Wayfinder can figure that out for you, as well. One really cool thing about Wayfinder is the way that it establishes blocking relationships between tickets, because some decisions can only be made once other decisions are made. I literally just called to spec on the Wayfinder map, and it pulled in this enormous document with basically all of the decisions that have been pulled from the Wayfinder map into this GitHub issue. So, instead of doing Grill with Docs and then doing to spec into tickets, you're spending a lot more time in Wayfinder creating this enormous map, and then taking that map, turning it to spec, turning it to tickets, and then implementing each ticket, and then running code review at the end. If you kind of already know the way to your destination, then there's no need to use Wayfinder because you can just path your way there in a single session and just figure it out.

So, let s start here, inside Claude chat, and work our way through each layer one by one. Instead of every tool needing a custom one-off integration with Claude, a tool can expose an MCP server and Claude can communicate with it through a standard protocol. What I want to show you is two capabilities that only really make sense in Co-work and that completely changed how I think about the Claude ecosystem because Co-work is in Claude, but bigger. What I want to show you is one thing, how the same ecosystem skills plugins connectors MCP, extends into development because this is where Claude stops being an assistant and starts being infrastructure. Verdict, if you write code or work with developers, Claude code plus the community MCP catalog is where this ecosystem stops being a chat tool and starts being real infrastructure.

Join here: Video notes links to the tools Get a FREE AI Course Community 1,000 AI Agents Get a FREE AI SEO Strategy Session Get 200 Free AI SEO Prompts Get out SEO link building book here Hermes Agent Buzz Integration: 3 Ways to Connect (Desktop, Relay Bridge, Gateway) Setup Walkthrough Hermes Agent can now run Buzz, an AI-powered Slack-like app where agents (and humans) work in channels, DMs, threads, reactions, and scheduled tasks. The script explains three ways to run Buzz with Hermes: Buzz Desktop auto-discovery, a relay bridge that provides a hosted identity in channels, and connecting through the Hermes Gateway to use Buzz as a full external platform. It shows the setup steps: download Buzz from buzz.xyz, update Hermes, ensure Hermes ACP is set up, then run Hermes gateway setup, select Buzz, add the relay community URL, optionally leave channel UUIDs empty to join all channels, allow community members to talk to the agent, and restart the gateway. It also covers setting Hermes Agent as the default provider, choosing models (including free options), controlling Buzz locally, and promoting the AI Profit Boardroom community and trainings. 00:00 Buzz Meets Hermes 00:08 Three Connection Options 00:34 Install Buzz and Add Runtimes 00:56 Fix Auto Discovery and Update 02:12 Gateway Setup Walkthrough 03:04 Set Hermes as Default Provider 03:20 Use Free Models in Buzz 03:42 Control Buzz Locally and Remotely 04:35 Why Use Buzz Anyway 04:44 Wrap Up and Next Steps 04:56 Join AI Profit Boardroom 05:15 Community Training and Support 05:40 Coaching Calls and Networking 05:53 Final Thanks and Links

त जव ब ह ब ल क ल भ नह और आज म त म ह स ट प ब य स ट प द ख न व ल ह क क स त म भ अपन फ न स ऐस ह ई क व ल ट ड क य म ट र व ड य बन सकत ह व भ ब ल क ल फ र म स ट प वन, ज सम हम एआई क ह ल प स क स व यरल ड क य म ट र आइड य ज ल न ह स ट प ट म हम उन आइड य ज स फ ल ड क य म ट र स क र प ट बन ए ग फ र उस स क र प ट क हम फ र ट ल स स व इस ओवर म कन वर ट कर ग च थ स ट प म हम ड क य म ट र स ट इल इम ज स जनर ट कर ग और फ र उन इम ज स क एक स क र ट एआई व ड य जनर टर क य ज करक व ड य म कन वर ट कर ग और ल स ट म इन व ड य स क क स मर ज करन ह मतलब क व ड य एड ट ग करक भ बत ऊ ग त व ड य क ध य न स आख र तक द खन क य क अगर क ई स ट प आपस म स ह गय त आपस ड क य म ट र व ड य नह बन ग चल ए फ र ब न बकव स क व ड य क स ट र ट करत ह त द स त सबस पहल हम च ह ए हम र ड क य म ट र व ड य क ल ए आइड य ज और उसक ल ए हम आ ज ए ग च ट ज प ट पर अब हम एक ऐस प रम प ट च ह ए ज स स ड करत ह च ट ज प ट हम व यरल ड क य म ट र व ड य आइड य ज द द त उसक ल ए म प स ट कर ग यह व ल प र ट ज सम ल ख ह म झ ह स ट र कल ड क य म ट र व ड य स बन न ह म झ 10 इ टर स ट ग और प वरफ ल ड क य म ट र ट प क आइड य ज द ज नम सस प स, स ग त म स ट र , श क ग इव ट स य अनट ल ड स ट र ह और ज ऑड य स क इ ग ज कर सक हर आईड य क स थ एक श र ट ह क भ द अब हम इस च ट ज प ट क स ड कर द ग और च ट ज प ट हम 10 ड क य म ट र ट प क और स थ म उनक श र ट ह क भ ल खकर द द ग इसन क फ व यरल ट प क ज स जल य व ल ब ग, भगत स ह क अस बल ब म ब ग क प ल न, न त ज स भ ष च द र ब स क म त क रहस य द द ए ह अब अगर आपक इनम स क ई भ आईड य अच छ नह लगत ह त हम इस ब ल ग क 10 और आईड य ज द और यह हम 10 नए व यरल आइड य ज ल ख कर द ग अब इनम स आपक ज भ आईड य अच छ लग , ज स म झ यह इ ड य प क स त न प र ट शन व ल आईड य अच छ लग त म इस क छ इस तरह स क प कर ल ग और प र प ब क स म प स ट कर ग यह प र प ट प र प ट म ल ख ह म झ ट प क पर 2 म नट क एक सस प सफ ल और इ ग ज ग ह द ड क य म ट र स क र प ट ल ख कर द स ग त श र आत एक स ट र ग ह क स ह ज त र त क य र य स ट क र एट कर और स ट र व य अर क ए ड तक इ ग ज रख स क र प ट स न ब य स न ह फ क ट स एक य र ट ह और ए ड म क ई क ल ज ग ल इन य स ट ए न ह अब यह ज ट प क ल ख ह , हम इसक जगह पर अपन आईड य क ल ख ग ज भ आईड य हम अच छ लग ह , उस यह पर प स ट कर ल ग म 2 म नट क स क र प ट ल खव रह ह आप च ह त द क जगह 10, 20, 50 क तन भ म नट क स क र प ट ल खव सकत ह और द स त यह सब क छ आप अपन म ब इल पर भ कर सकत ह चल ए फ र अब हम इस प र प ट क च ट ज प ट क स ड कर द त ह और च ट ज प ट न अपन क म करन च ल कर द य ह और इसन हम 2 म नट क स क र प ट ल खकर द द ह ज सम ट टल स क स स न स ह स न वन ज सम ह क ह और स थ म स क र प ट भ ह अब हम व इस ओवर क ल ए स क र प ट जनर ट करन ह त हम इस यह व ल प र प ट द ग द स त , आपक प र प ट स क ट शन नह ल न ह क य क म स र प र प ट स क ल क ड स क र प शन म प र व इड कर द ग अब हम इस व इस ओवर व ल प र प क च ट ज प ट क स ड कर द ग और यह हम स पल अपन स क र प ट क ल ख कर द ग ज स हम ड यर क ट क प करक अपन व इस ओवर जनर ट कर सकत ह हम इस स क र प ट क क प कर ल ग और स ग त चल ग फ र म व इस ओवर बन न हम आ ज ए ग Google AI स ट ड य क व बस इट पर और इसक ल क भ आपक ड स क र प शन और ड क य म ट फ इल म म ल ज एग व बस इट पर आन क ब द आपक क छ इस तरह क इ टरफ स द खन क म ल ग जह आपक क छ ऑलर ड बन बन ई ट पल ट स म ल ज ए ग स ग त हम इस पहल ट पल ट द एवर ड अस स ट ट पर क ल क कर ग और न क स ट प ज ओपन ह ज एग अब यह पर आपक क छ स ग त स क श स द ख रह ह ग ज स पहल स न जह हम अपन स न य व इस ओवर क क य पर पस ह व ल खन स ग त ह द सर ह स पल क न ट क स ट जह हम अपन व इस क ड ट ल स ल खन ह क व इस स ग त क स ह न च ह ए फ र ह स प कर क ऑप शन इस पर क ल क कर ग त स प कर स ग त क स ट ग स ओपन ह ज ए ग और यह पर क य स ट ग रखन ह वह सब म आपक थ ड द र म बत ऊ ग और यह स हम अलग-अलग व इस स क स न सकत ह और क स एक आव ज क अपन व इस ओवर क ल ए स ल क ट कर सकत ह द स त , स न व ल ऑप शन म और स पल क न ट क स ट म क य ल खन ह , यह सब स ग त भ आपक ड क य म ट फ इल म म ल ज एग बस इन ह एक-एक करक क प करन ह ज स म यह स न व ल ट क स ट क क प कर ल ग और Google AI स ट ड य पर आकर इस स न व ल ब क स म प स ट कर द ग ठ क ह ? इस तरह स स पल क न ट क स ट व ल ट क स ट क भ हम क प कर ल ग और इस स पल क न ट क स ट व ल ब क स म इस प स ट कर ल ग न क स ट अब हम इस स प कर क स ट ग स म ज ए ग और हम च ज स करन ह इस ड यर क टर स स ग त न ट म ड क य म ट र व इस ओवर क ल ए हम स ट इल क ए प थ ट क रखन ह और ब क क स भ स ट ग म क ई च ज नह करन ह स ग त उसक न च ह व इस स त आप इन ह स नकर द ख सकत ह ल क न म र ह स ब स यह ल पटस क आव ज ड क य म ट र क ल ए एकदम परफ क ट ह न क स ट अब हम अपन स क र प ट क क प कर ग और इस ब क स म प स ट कर ल ग स क र प ट क प स ट करन क ब द न च यह ज रन ल ख ह , हम इस पर क ल क करन ह और हम र व इस ओवर जनर ट ह न च ल ह ज एग क छ ह द र म हम र व इस ओवर जनर ट ह ज एग चल ए आपक स न त ह क इसक आव ज क तन कड क और स ग त र यल स ट क ह स त बर 1947 प क स त न क तरफ स एक ट र न स ग त अम तसर र लव स ट शन पर आकर र क ल क न अज ब ब त यह थ क ट र न स एक भ य त र हम इस ड उनल ड बटन पर क ल क करक अपन व इस ओवर क ड उनल ड कर ल ग और स ग त द स त , आपक इस एड ट करन क भ जर रत नह ह बह त स ल ग बत त ह क ऑड य एड टर म ज कर व इस क एड ट करन ह त ह , ल क न इस व इस ओवर क आपक एड ट करन क भ जर रत नह ह त द स त , स ग त हम र ड क य म ट र व ड य क व इस ओवर र ड ह और अब हम च ह ए व ज अल स य न क स ग त इम ज स हम व ज अल स क ल ए यह प र प ट फ र क क प कर ग और स म च ट म च ट ज प ट पर प र प ट ब क स म प स ट कर ल ग और इस स ड कर द ग च ट ज प ट हम व इस ओवर क हर एक ल इन क ल ए इम ज स क प र प स ल खकर द द ग और द स त , इसन म र 2 म नट क ड क य म ट र व ड य क ल ए 37 इम ज स क ड ट ल ड प र ट स ल ख कर द ए ह अब हम जनर ट कर ग इम ज स और उसक ल ए हम चल ग Google फ ल क व बस इट पर फ र न च यह ज न य प र ज क ट ल ख ह , हम इस पर क ल क करक एक नय प र ज क ट क र एट कर ल न ह अब आपक प स द तर क ह इम ज स क र एट स ग त करन क पहल ह क म न अल ज सम अपन इम ज स क प र प ट स क क छ इस तरह स क प करन ह और Google फ ल पर प र प ट ब क स म प स ट करक फ र इम ज स क स ट ग स क क छ इस तरह स अपन र क व यरम ट क ह स ब स स ट कर ल न ह और इम ज स क जनर ट कर ल न ह क छ ह द र म हम र इम ज स बनकर आ ज ए ग ज स क आप स क र न पर द ख सकत ह अब यह त थ घ स प ट म हनत व ल तर क और द सर तर क ह एक क ल क म स र इम ज स क बनव न त हम च ट ज प ट पर आकर अपन स र इम ज स क प र प ट स क न च क प बटन पर क ल क करक क प कर ल ग फ र Google फ ल पर आकर यह ज ब क स म आपक एज ट ल ख द ख रह ह हम इस पर क ल क करन ह उसक ब द ब क स क र इट स इड म ह एज ट क स ट ग स ज स पर क ल क कर ग त एक प पअप ओपन ह ज एग पहल ऑप शन ह क फर म ब फ र जनर ट ग त हम इस न वर पर ट क कर ल ग और ब क क स ट ग स क ऐस ह रहन द ग ठ क ह ? त अब बस अपन इम ज क स र प र ट स क ब क स म प स ट कर ल न ह और आख र म ल खन ह जनर ट ऑल द इम ज य ज ग द अबव प र ट स मतलब क हमन ऐस ब ल ह क ऊपर द ए गए स र प र ट स इम ज स जनर ट करक द अब हम इस स ड कर द ग और एज ट अपन क म प लग ज एग और क छ ह स क ड स म यह हम र स र इम ज स क जनर ट करक द द ग ज स क आप स क र न पर द ख ह सकत ह मतलब क स र फ एक स 2 म नट क अ दर इसन म र स र इम ज स क जनर ट कर द य ह ब क इम ज स क क व ल ट आप द ख ह सकत ह ब ल क ल स न म ट क व ज अल स ह मतलब क अब ब र-ब र इम ज स बन न क झ झट ह खत म बस अपन स र प र ट स क क प कर , प स ट कर और इम ज स क जनर ट करव ल अब हम क छ नह करन ह बस अपन इम ज स क एक-एक करक क छ इस तरह स इम ज क ऊपर कर सर ल ज कर थ र ड ट स पर क ल क करन ह और इस ड उनल ड पर क ल क करक इन ह ड उनल ड कर ल न ह त द स त , हम र इम ज स भ जनर ट ह गई ह और अब हम र अगल स ट प ह इन इम ज स क व ड य म कन वर ट करन और ज स क र ट AI व ड य जनर टर ज सक हम य ज करन व ल ह वह ह wips. AI यह म ट AI क नई व बस इट ह और ज AI व ड य म डल ह व इनक नय म डल ह ज सक क व ल ट भ क फ श नद र ह हम इस ल ग इन बटन पर क ल क करक अपन Instagram य Facebook स ल ग न कर ल ग ल ग इन करन क ब द हम क छ इस तरह क इ टरफ़ स द खन क म ल ग हम यह ज ग ट स ट र ट ड ल ख ह , इस पर क ल क कर ग और एक नय प र ज क ट क र एट ह ज एग अब आपक न च एक ब क स द ख रह ह ग ज सम पहल ऑप शन म व ड य ल ख ह आप च ह त इम ज स भ क र एट कर सकत ह ल क न व फ लह ल 9 16 एस प क ट र श य म ह बनत ह आप अपन क स भ व ड य क ल प स ग भ कर सकत ह ल क न हम व ड य क र एट करन ह त हम व ड य क स ल क ट कर ग फ र ह फर स ट फ र म ल स ट फ र म क ऑप शन जह हम अपन इम ज क अपल ड करक उस व ड य म कन वर ट कर सकत ह ल स ट एडव स ड क ऑप शन ह जह स हम अपन व ड य क क व ल ट क 720p स ग त म जनर ट कर सकत ह चल ए फ र हम अपन स र इम ज स क यह ऊपर यह ज अपल ड स ग त म ड य ल ख ह , यह स अपल ड कर ग हम एक ब र म 12 इम ज स अपल ड कर सकत ह त इधर क ल क करक अपन इम ज स क अपल ड कर ल त ह य क न स हम र इम ज स अपल ड ह न च ल ह गई ह और क छ ह द र म यह अपल ड ह ज ए ग द स त , अब हम र प स इम ज क व ड य म कन वर ट करन क द तर क ह पहल त यह क हम न च फर स ट फ र म और ल स ट फ र म पर क ल क करक अपन इम ज क अपल ड कर ल और प र प ट ब क स म अपन प र ट ल खकर स ग त व ड य क जनर ट कर ल द सर तर क ह हम अपन इम ज पर क ल क कर ग फ र र इट स इड म आपक द ऑप शन द खन क म ल ग पहल ह ऑट एन म ट और द सर ह म न अल एन म ट सबस पहल हम ऑट एन म ट पर क ल क कर ग त यह ख द ह अपन स ब न क स प र प ट क इम ज क व ड य म कन वर ट कर द ग और र जल ट आप स क र न पर द ख ह सकत ह द सर हम म न अल एन म ट पर क ल क कर ग त यह हमस प र प ट ल खन क कह ग ल क न हम र प स त इम ज स व ड य बन न क प र प ट स ह ह नह त हम चल ग च ट ज प ट पर और इस ब ल ग क इन सभ क ल ए इम ज स व ड य प र ट स ल ख कर द और इस स ड कर द ग स ड करत ह च ट ज प ट हम अपन स र इम ज स क इम ज स व ड य प र ट ल खकर द द ग चल ए फ र हम अपन पहल इम ज क प र प ट क क प कर ल त ह और व इब स. पर आकर इस प र प ट ब क स म प स ट करक व ड य क जनर ट कर ल ग हम र व ड य जनर ट ह न च ल ह गई ह और क छ ह द र म यह बनकर आ ज एग ब क यह रह र जल ट आपक स मन एक स न म ट क श ट अम तसर क र लव स ट शन क द स त , म आपक स र इम ज स क व ड य म कन वर ट करक त नह द ख सकत , ल क न एक-द क और करक द ख द त ह ज स यह व ल इम ज स पल च ट ज प ट स प र ट क क प करन ह और यह पर ब क स म प स ट करक व ड य क जनर ट कर ल न ह फ र क छ इस तरह स त सर इम ज क भ व ड य म कन वर ट कर ग ब क व ड य आप स क र न पर द ख ह सकत ह क र जल ट त बह त ह श नद र बनकर आय ह यह क छ व ड य स और ह ज म न इस प र स स क फ ल करक बन ए ह त आप समझ गए ह ग क क स इम ज स क व ड य म कन वर ट करन ह ल क न द स त एक प र ब लम ह क जब भ आप क स भ व ड य क यह ऊपर स ड उनल ड बटन पर क ल क करक ड उनल ड कर ग त व ड य पर आएग व टरम र क त हम स ग त यह स क स भ व ड य क ड उनल ड नह करन ह हम क य कर ग क इस व ड य पर र इट क ल क कर ग और ऑप श स म सबस न च इ स प क ट पर क ल क कर ग र इट स इड म एक प पअप ओपन ह ज एग जह आपक द ख ग यह व ड य एसआरएसस त उसक आग ज ब ल कलर म एक ल क ह उस हम क छ इस तरह स क ल क करक क प कर ल न ह फ र द सर ट ब ओपन करक हम स ग त उस ल क क ओपन करन ह और अब आप द ख सकत ह क हम र व ड य म क ई भ व टरम र क नह ह मतलब क व ड य स व टरम र क हट स ग त च क ह अब व ड य क न च र इट स इड म यह ज थ र ड ट स द ख रह ह , हम यह पर क ल क करक व ड य क ड उनल ड कर ल न ह त क छ इस तरह स म न आपक व टरम र क क समस य क भ स ल व कर द य ह त इस ब त पर व ड य क ल इक जर र कर द न और एक ब र जब आपक स र व ड य स ड उनल ड ह ज ए त हम इन ह क स भ व ड य एड टर म ज स क म क पकट क य ज कर रह ह क पकट क ओपन कर ल ग और फ र न य व ड य पर क ल क करक हम अपन स र व ड य स क स ल क ट करक इ प र ट कर ल ग व ड य स इ प र ट करन क ब द हम न च ऑप श स म स ऑड य पर क ल क कर ग फ र स उ ड स पर क ल क कर ग उसक स ग त ब द इस फ ल डर आइकन पर क ल क कर ग और ल स ट म ड व इस पर क ल क कर ग अब यह सर च ब र म अपन व इस ओवर क न म सर च करक हम उस क छ इस तरह स ग त स उस पर क ल क करक और इस प लस स इन पर क ल क करक ऐड कर ल ग अब स पल स प र स स ह अपन स र व ड य स क व इस ओवर क अक र ड गल अर ज करन ह व ड य ल ब ह त उस कट कर ल न ह और व ड य छ ट ह त उसक स प ड क कम करक उस ल ब कर ल न ह म न अपन स र व ड य स क अर ज कर ल य ह और स र व ड य क अर ज करन क ब द हम एक अच छ स ब कग र उ ड म य ज क ऐड कर ल ग और यह व ल ब कग र उ ड म य ज क आपक म र Telegram च नल पर म ल ज एग ब कग र उ ड म य ज क क व ल य म क जर र कम कर ल न ह त क हम र व इस ओवर अच छ स स न ई द सक इतन करन क ब द हम ऊपर स एआई य एट पर क ल क करक अपन व ड य क एक सप र ट स ट ग स स ग त क म क स मम पर स ट कर ल ग और एक सप र ट बटन पर क ल क करक हम इस एक सप र ट कर ल ग और द स त हम र ड क य म ट र व ड य बनकर त य र ह ज एग त यह त थ प र प र स स क क स हम ब ल क ल फ र म ब न एक भ प स लग ए ड क य म ट र व ड य बन सकत ह आई ह प आप ल ग क व ड य पस द आय ह ग त व ड य क ल इक जर र कर द न और च नल पर नए ह त सब सक र इब भ कर ल न ब क अगर आपक क स स ट ट क र क टर क स थ व ड य बन न ह त यह ज स क र न पर आपक व ड य द ख रह ह इस पर क ल क करक स ख सकत

Anything you can kind of think of that an agent that you're used to, any other agent can do, it can do that here. So it's a great um solution to build uh web apps with um agents because you get things like you know authentication uh one command deployments so these things usually take hundreds of lines of code to do and with Wasp it it's all glued together and it's all available to you or an agent in one command one CLI command or a couple lines of code so you're like freeing up your agent to do more important work than this boilerplate mundane coding tasks and things like that. I can reply to it and then I can ask one of the agents like yo um you know what's what are some I asked here give me some topics to tweet about today something safe based on what I'm doing something in the middle and something wild and creative right so like I'm getting it to help me come up with tw uh tweet ideas things like that and um uh the workflow is basically you can get it to do stuff based on that when this comes in it triggers a a um workflow that tells me, hey, new tweet stats just landed and then I know, oh, cool. So whenever there's a message posted right now in that one channel >> so only if it says urgent for example or whatever you want to screen for you could then add a step what's the step that then happens where's the where's the action on the trigger >> right here you have a filter expression so I can filter for it gives you an example here contains text and let's say you're saying >> the the example I have is urgent if I use the word urgent >> then this triggers off it's not just if messages just use But let's let's click on add a step then and see what happens next. You can delegate you can tell um another thing I do in in this where it's like you can delegate multiple like a complex task to multiple agents at once.

And then Codex found something and it was like, "Okay, we should do this." And then Claude searched the internet, searched the old contract that we've signed earlier, and then like came up with a counterpoint and it was just so fun to see them and then they settled on this like final agreement and I've just found that Buzz does a really good job of like allowing these agents to like concisely discuss and then it'll create the thing and then the thread just stops. So therefore if you only have like if you only have Codex, you can allow it to spin up like multiple sessions and work in parallel and you can set that somewhere in the settings like say uh five max or something like that or you want to let it go crazy, you can do 10, 15, 20, you know. It's like one QR code, one scan, now I have it all lined up and so I can just go into my coding channel and I can very easily at mention Claude code and I can say, I'm just going to say hi guys and they will respond and it's like very they're like perfectly in sync, which is really cool. And like the point that I hope we get to is like I could just fire up an agent with Fable 6 uh and or it's like Fable 6 on continuous mode and I could just say, "Hey, like for these hours I need you to constantly check on certain things um and like here's the end goal every day for this thing and just trust that it'll get done correctly." Cuz that will just give me just [clears throat] an insane amount of leverage. >> Yeah, um so I'm from what I gather most of the audience is focused on like entrepreneurship and and stuff like that, but these these guys are great at coding and making apps and now with these kind of apps that are very open, you can build stuff that like that you can build really bespoke tools that integrate with Buzz and you know, close this con close the loop on the context that you're kind of like shuffling around everywhere and um um I'm part of a team and we we're working on a full-stack TypeScript framework that works really well with agents.

Und unter meinem letzten Video, nämlich das Tutorial zum zweiten Gehirn, war das eines der meist diskutierten Themen, nämlich nicht Obsidien, sondern es war genau diese Frage, wie findet die KI eigentlich das Richtige? Und in diesem Video zeige ich dir, was RAG eigentlich bedeutet, welche verschiedenen grundlegenden Wege es gibt, Wissen in die KI zu bekommen und wie du entscheidest, welche Architektur zu deinem Wissen passt und welche konkrete Kette hinter der Antwort von gerade eben steckt, das löse ich auch später auf, denn dann kannst du es selber beurteilen. Frage 1, ein exakter Begriff FCP XML, ein Schnittdateiformat, das genauso in meinen Notizen steht. Die Wiki Schicht holt also den einfachsten Weg von allen zurück ins Spiel, nämlich vollständig alles laden und wächst das Wiki irgendwann selbst zu groß, wird eben darauf wiedergesucht. Und andersherum, ein riesiger Support bestand, tausende Tickets, täglich neue, viel Material, aber ein Wiki müsste permanent herunter kompilieren.

Ja, dieses Jahr seit seit langer Zeit ausnahmsweise mal ohne uns äh und gar nicht mal wegen dem gleichen Grund oder weil der eine nicht geht und der andere nicht geht, sondern alles andere Gründe privater Natur. Also das betrifft dann aber tatsächlich glaube eher diejenigen, die Agents schreiben oder eben im größeren Umfang äh äh Anwendung schreiben der Enduser, also je nachdem in welcher Rolle ihr unterwegs seid, als diejenigen, die vielleicht Copilot in Microsoft 365 nutzen, werden jetzt nicht zukünftig oder zumindest ist es noch nicht angekündigt. Also, man muss ja mal vorsichtig sein beim Bding, aber den den Webw aus Grenzen, das wäre jetzt nicht mein oder wahrscheinlich unser beider Ansatz nicht, sondern eben tatsächlich zu was passt nicht in unsere Firmenpolcy rein, so wie du schon sagtest, weil es einfach veraltete Quellen sind oder Fake News, ist eben ein Portal bekannt für Fake News oder irgendwie sowas in der Richtung ähm äh was man hier ausschließen würde und sagen würde, nee, das das wollen wir nicht, dass es von Copilot herangezogen wird fürs Backgrounding. Ich glaub, wir hatten einfach so viel mit Copilot zu tun, dass wir so unwichtige Themen wie Planner Support in Private und Share Channels haben wir einfach, aber wir sind ja rechtzeitig dabei diese Themen aufzuarbeiten, denn anstelle, dass es jetzt dann rauskommt und wir im Juli quasi den Release haben, kommt s erst im August, late August, um es genau zu sagen, also so genau Microsoft das halt mag, aber grundsätzlich die gute Nachricht ist, es steht explizit mit mit dass die Planner Integration für die Planner Tabs in Private Channel und in Share Channel mit reinkommt. Ich blende jetzt gerade noch mal in groß ein äh auch einen Suspected Threats äh wo eben möglicherweise Scam oder andere Geschichten von Microsoft detiktiert wird, weil eben die PN als spricht die die ZPUI wieder das nah legt, weil irgendwie dort nicht Microsoft.com steht, sondern Microsoft und dann eine 0ft.com, also was eben ausschaut wie Microsoft, aber keine Microsoft Domain ist.

If you are automating your business in 2026, you are likely choosing between three dominant platforms: n8n, Cloud Code, and Taskade Genesis. If you use the code writing path with Cloud Code, the AI generates a code base using an item potency key. Claude code offers infinite flexibility, but it shifts the entire burden of managing raw infrastructure, observability, and token scaling onto the user. To use a managed path, you have to surrender low-level architectural tweaking and raw code ownership to the platform. For the developer, Claude Code wins, prioritizing a high capability ceiling and code review.

What's cool this, obviously, if you want to, if you want to actually create multiple profiles inside Hermes agent, you can just ask it and tell it this new thing. If not, you can actually get the ideas, and this is what I do, and I just forward them to Hermes agent, and then Hermes saves them, and then I can ask them any questions that I want to. The very first thing that you need to do is send a prompt to Hermes agent to say, "Hey, I want you to interview me about everything that I do on a daily basis, and I want you to tell me ways that you can help me save time and take admin off my shoulders." Once you've done that, you're going to want to go ahead and connect it to what I would call the core three, your calendar, your email, and even your calls. For example, if I'm chatting with Hermes agent, I might come down and say, "Hey, what was the What is the title of the next appointment that I have inside my calendar?" But, the coolest thing that it does is actually drop me a daily brief, a daily insight digest every single morning of all the calls I have coming up that day, all the emails that are unread. I can come over to a new chat and I can say something like, "Hey there, I want you to go ahead and in fact let's go ahead and use Kimmy K3, shall we?" Let's go down and find Kimmy K3.

Sag Chat GPT oder Cloud einfach, er soll dir Fragen stellen zu deinem vergangenen Schuljahr, beantworte die ausführlich einfach mit Sprachnachricht, so wie ich das mache und dann hast du auch schon ein Ergebnis, mit dem du arbeiten kannst oder du benutzt die KI erstmal für einen Monat und machst dann die Auswertung. Also den Promp findest du in der Videobeschreibung und hier wird jetzt analysiert, ja, wo sind denn die Zeit und Energiefresser und auch zur Auswahl stehen dann Lösungen, also natürlich einen wieder verwendbaren Einzelpromt. Wir legen ein Projekt an mit Kontext und das könnte ich jetzt eben ihm sagen, ja, ich möchte dieses Projekt und dann erstellt er mir eine Datei oder hier sagt er mir sogar einen Cloud Skill, den könnte ich jetzt eben auch bauen lassen. Ich lasse wieder einen Prompt schreiben für einen Skill, den ich dann in Cloud anleg oder einfach einen Einzelpromt, den ich jetzt erstellen lassen könnte und dann in meiner App in der Klassenflow App einfach hinterlegen könnte. Also, ich habe das mal eben schnell gemacht und jetzt siehst du schon hier sogar mit Login könnte ich mich jetzt einloggen in mein persönliches Lehrer Cockpit.

Wenn ich also Cloud Opus 4.8 vergleiche, liegt es bei 85, also sehr, sehr nah dran. Also 30 Tage lang kannst du den Server testen, einfach Cloudcode darauf installieren und alle Anwendungen machen, die du möchtest, um sicherzugehen, dass es wirklich viel besser ist als der klassische Cloud Code. Also Olammer ist jetzt installiert, wird gerade ausgeführt und Sie werden sehen, dass wir jetzt eine kleine Konfiguration vornehmen, um Cloud mit Olama zu installieren und auch das Modell zu installieren, das wir ausgewählt haben. Also hier in diesem Terminal, wenn ich mir die Kursunterlagen anschaue, die ich euch tatsächlich in der Beschreibung dieses Videos hinterlegt habe, dann müsst ihr einfach nur klicken, um diesen gesamten Code zu kopieren. Hier auf dem VPS, wenn ich ihn installiere, wenn ich den VPS bei Hostinger erstelle, Hostinger gibt mir tatsächlich, wie Sie hier sehen werden, einen Code, also ein Passwort.

I m Sam Tucker from the Big Mac and you may have heard recently that the artificial intelligence inside Claude s Cowork app for Mac OS is able to escape the app containing it and gain full access to all the files on your computer. Anyway, this virtual machine is inside your Mac and the Claude AI agent is just a normal user in there with no special privileges. But also in that virtual machine is another process called Coworked and Coworked is more powerful with root access and their job is to share and sprinkle just some of your Mac files with the agent, you know, to help you summarize a document or whatever else you re too lazy to do yourself. However, researchers found a flaw in this Linux music virtual machine thing that when exploited could give the Claude AI root access as well, meaning that all your Mac files could then be mounted and the AI could read or write or do whatever it wanted to them. But in the meantime, you could make sure you re using Claude s cloud execution mode instead of local processing, but realistically, you should be pretty safe as the Claude agent only went rogue after receiving a single instruction from the researchers that discovered the exploit.

Auf der einen Seite ist es so, dass so viele Leute da draußen auch in den Kommentaren unter diesem Video auf Reddit und überall schreiben: Hey, ich habe Informatik studiert oder ich habe eine Ausbildung zum Fachinformatiker gemacht und ich finde keinen Job und sogar Leute mit Berufserfahrung sagen, dass sie keinen Job finden. Wenn du die heutigen Tools, die es gibt, um mit KI Software zu entwickeln, nicht einsetzen kannst und nicht verwendest, dann bist du auf dem Arbeitsmarkt in den meisten Stellen einfach nicht mehr so gefragt. KI Automation Engineer, das ist ein Begriff, der klingt sperrig, ich weiß, aber lass mich dir zeigen, was der eigentlich macht, denn dann verstehst du auch, warum Firmen gerade so viel Geld dafür auf den Tisch legen. Also, wenn wir das Ganze noch einmal auf einen einzigen Satz zusammenfassen, was ein KI Automation Engineer macht, dann ist es so, du gehst letztendlich in eine Firma rein, du schaust dir an, wie die Leute arbeiten, setzt dich daneben, verstehst den Prozess und überlegst, wie dieser Prozess effizienter gestaltet werden kann und wie hier sehr viel Zeit durch Automatisierung gespart werden kann. Und wenn du dich interessierst für unsere neue Weiterbildung zum KI Automation Engineer, wo wir dir genau das beibringen, was du brauchst, dann schau gerne mal auf den ersten Link ganz oben in die YouTube Beschreibung.

My coworkers on the next video aren t human they re AI agents I m planning and researching alongside, in the same shared workspace as my human teammates, thanks to a new open-source app from Block called Buzz. Agent Wikis has free, curated LLM wikis on a ton of AI topics plus a Pro tier ( 9.99 mo) for super-sized wikis with way more depth: Sign up for my FREE weekly newsletter, where I spill my unfiltered thoughts on the latest AI news, cool research, and projects I m building: Buzz just announced by Jack Dorsey and Block is an open-source, agent-first take on Slack where agents are real members of the room with their own keypairs and their own signed entries in one shared event log, not bots bolted onto a human account. I break down the core design idea (one identity, one append-only log, built on the Nostr protocol and why it s explicitly not a blockchain product), then install the desktop app, create an identity, and wire up agents across Claude Code and Codex runtimes. From there I set up agent memory and workspaces, talk to agents in DMs and threads, look at the shared-compute mesh for pooling GPUs across a group, spin up my own community, and link the mobile app. Resources: Buzz: Timestamps: 0:00 - My coworkers are AI agents (the concept) 0:41 - What Buzz is: an open-source, agent-first Slack 2:27 - One identity, one event log (and how Nostr fits) 6:00 - Installing Buzz creating an identity 7:21 - Setting up agents and runtimes 10:53 - Shared compute: pooling GPUs across a group 12:22 - Agent memory, workspaces talking in threads 15:32 - Creating my own community the mobile app 18:41 - First-look verdict Buzz JackDorsey Block Nostr AIAgents OpenSource Slack AITools AgentWorkspace

I put Claude Code and Cursor in the same chat and had them build together. Buzz is Block s new free open source workspace where AI agents live in channels like teammates. Get Buzz free: Clicky, 50 off: (LUKAS-PRO) BYQ component library: MagicPath Infinite Canvas: Join the CreatorNTWRK Discord: Join BuilderzGym on Skool: Normally a multi tool project means copy pasting context between Claude Code, Cursor and three browser tabs. In this video I connect Claude Code, Cursor, Mobbin, Higgsfield and BYQ inside one Buzz workspace and have them redesign a full blog page together, then push the result straight into a MagicPath canvas. What we cover: - Downloading Buzz and creating your first community - The Buzz interface compared to Slack - Creating agents from scratch and choosing Claude Code, Cursor, Codex or Kimi - Agent permissions and who can talk to each agent - Building agent teams so multiple agents work on one project - Connecting a new MCP and forcing an agent to refresh it - Wiring MagicPath in as a canvas for agent generated designs - A full multi agent build: Cursor reads the codebase, Claude Code redesigns with BYQ components, Higgsfield fills in the images - Exporting agents with or without their memories Timestamps: 0:00 What is Buzz 1:34 Setup UI tour 3:50 Creating your first agents 6:25 Why this beats juggling separate tools 8:05 Agent teams in action 12:11 Adding MCPs (BYQ MagicPath) 14:56 The full workflow: redesigning a blog page 19:33 Final thoughts Tools used: Buzz: BYQ MCP: MagicPath: Clicky: (LUKAS-PRO) Mobbin: Higgsfield: vidIQ: Convex: Framer: Let s connect: X: LinkedIn: claudecode cursor aiagents buzz mcp designengineer aiworkflow opensource

In this video, I build and test the complete workflow to see whether an AI agent can: Understand my brand voice Generate content ideas Write platform-specific posts Create social media designs Build a content calendar Schedule posts automatically Reduce the amount of manual work required I followed the framework demonstrated in the reference video, rebuilt the workflow myself and tested it with a real product. Tools used Claude Brand voice files AI content-generation workflows Social media scheduling tools Design automation Platform-specific content instructions Original workflow inspiration This build was inspired by the social media automation workflow demonstrated in this video: Full credit to the original creator for sharing the framework. Who this video is for This video will be useful for: Solo founders SaaS builders Content creators Digital marketers Freelancers Professionals building a personal brand Anyone trying to automate social media using AI Subscribe for more practical AI builds, agent workflows and real-world automation experiments. Tools used Claude Brand voice files AI content-generation workflows Social media scheduling tools Design automation Platform-specific content instructions Original workflow inspiration This build was inspired by the social media automation workflow demonstrated in this video: Full credit to the original creator for sharing the framework. Who this video is for This video will be useful for: Solo founders SaaS builders Content creators Digital marketers Freelancers Professionals building a personal brand Anyone trying to automate social media using AI Subscribe for more practical AI builds, agent workflows and real-world automation experiments.

Man kann jede Datei untersuchen , erstellen und den Quellcode exportieren oder das Spiel kompilieren, wann immer man möchte. Man kann so viel iterieren, wie man möchte, Dinge hinzufügen , Level bearbeiten und vieles mehr. Man kann hier Dinge erstellen, und wenn man mehr Freiheit möchte, kann man Dinge direkt in Kodot bearbeiten. Man braucht viel mehr Kontrolle, und für vollwertige 3D-Spiele in Unreal Engine- Qualität benötigt man die Unreal Engine. Und wenn man etwas Großartiges schafft, kann man es weiterentwickeln und in guten Händen weitermachen oder später versuchen, etwas Ähnliches mit der Unreal Engine oder Unity zu realisieren .

Die Datei .env sorgt dafür, dass Ihre Zugangsdaten nicht im eigentlichen Code landen, sodass wir den Code auf GitHub hochladen können, ohne die Geheimnisse mitzuübertragen. Der Vorgang ist im Wesentlichen folgender: Folgen Sie den Schritten, die Claude Ihnen gibt, oder den Schritten, die ich Ihnen hier gleich erkläre, damit nichts schiefgehen kann. Sollte Claude melden, dass die GitHub CLI nicht installiert ist, folgen Sie zunächst den Installationsanweisungen und führen Sie dann den Authentifizierungsbefehl aus. Nach wenigen Minuten ist die Bereitstellung abgeschlossen, und Hostinger zeigt Ihnen hier einen Screenshot der fertigen Website an, damit Sie überprüfen können, ob alles korrekt aussieht, bevor Sie die Website öffnen. Meine Datei ist leer, aber falls du Probleme hattest, könntest du den Inhalt des Protokolls kopieren und ihn Claude mit folgendem Text senden: Mein Hostinger-Deployment wurde abgeschlossen, aber die Live-Website funktioniert nicht.

Weil diese Strukturierung von Ideen, von Konzepten und Gedanken relativ, also es ist ja ein Key Skill, den man damals in der Beratung immer haben müsste, ist, wie kriegt man das denn jetzt den Leuten auf Slides irgendwie sinnvoll dargestellt, dass sie eine Storyline haben und so weiter und so fort? Bei uns ist das so, mittlerweile hat unser CDO hat jetzt, weil wir jetzt nur noch HTML Files erstellen und das ist halt irgendwie doof, wir schickst ein HTML File und dann hast Du halt in Slack irgendwie die HTML Vorschau. Hat da jetzt quasi son son Viewer gebaut, wo wir dann halt einfach über diesen Viewer einfach die die reingeben und dann den Link stellen, dass nicht jede einzelne Seite immer irgendwie publishen muss sozusagen, dass es ein bisschen leichter ist einfach für die Kommunikation. Hier und der Punkt, der hier noch mit drinsteht mit den mit den Prozessen, eine Prozessdoku ist halt immer ein und ich will jetzt, wie oft werden Prozesse nicht dokumentiert, weil s halt einfach so ein ist, diese die Schaubilder, das irgendwie alles irgendwie sauber zu dokumentieren. Ich hab jetzt viel mehr Zeit natürlich, mich ja meine Mitarbeiter zu kümmern und zwar die wirklich wertstiftenden Themen als Manager als irgendwie, keine Ahnung, ich korrigier dir da eine E-Mail oder sag dir noch mal zum zehnten Mal, wie man Konzepte irgendwie grundsätzlich aufbaut oder Admin arbeitet.

Check out Junie Use code CODEKING for 30 OFF (first 300 users): In this video, I ll be showing you how to configure different models and agents in Buzz, including how to use Kimi K3 as your main coding model, set up per-agent model configurations, and connect free API options like NVIDIA Build and OpenRouter through the Goose harness. -- Key Takeaways: Buzz lets you configure different models for different agents using harnesses like Claude Code, Codex, and Goose. NVIDIA Build offers free OpenAI-compatible API access to many preview models, including GLM, Nemotron, Kimi, DeepSeek, Qwen, and more. OpenRouter also provides free models that can be connected to Buzz through Goose, including Nemotron 3 Ultra and coding-focused models. Overall, Buzz s per-agent model configuration makes it practical to build a multi-model AI team while keeping costs low.

>> That's a that's such a huge deal because I feel like a lot of us are ping-ponging between different models and harnesses and stuff like that, and it's tiresome, you know, when you're when you're starting from scratch, um it does feel like and then I need to understand, okay, what is Saul good at, too, you know, like and what are the contexts for this particular model to get the most out of it. So like my team um like how much we're tweeting, how good what are what are our top tweets, all that stuff all that information was being um uh collected in this in this app, this full-stack app, and then I was like, "Oh, well, um instead of just building it with Buzz, why don't I get Buzz to also access the app, get that information, then we can get like, you know, daily reviews here, daily digests, and um we can also like kind of um brainstorm on what's working uh in our Twitter strategy for marketing, what's not working, get it to um find like links between the the popular uh tweets, and you know, brainstorm, come up with new stuff." So, that's exactly what I did. >> One thing that I really like about this kind of uh you know, this close we were talking about this closed circle of like context um that I'm getting back from my app so that tweet dashboard app is that like here it sends me the tweets every day and the the stats like the top tweets and my stats and impressions and I can just ask uh the agent like um let's make sure I'm I'm replying there and I'll just say like uh what can like what is the common thread between my top tweets? I can chat about it here, and then um you know, we're we can chat with colleagues about like what you know, what are some more tweets we want to push out, what's a a different marketing angle we could take, and um uh that's what I really like about this uh here. So, like I can honestly imagine a big open community of let's say, I don't know why I keep gravitating towards graphic designers, but like I say designers and uh you know, you need a really cool logo and AI gen image gen isn't really doing it for you, and you ask some people to like, you know, spit out some ideas and and some random person in in some, you know, graphic design community uh on Buzz gives you like an awesome logo, and you just, you know, tip them.

Für einziges Produkt muss man mehr Lieferant anschreiben, man muss die vergleichen, man muss schauen, wer hat die besten Preise bzw. Oftmals muss man mehrere Tage warten, bis man das richtige Angebot gefunden hat und erst dann, sollte man einen Paliferanten gefunden hat, beginnt die richtige Arbeit. Anstatt alles manuell zu machen, überlasse ich heute das Sourcing Tool Kit von XW Work komplett den Ablauf fürs Sourcing bzw. Nachdem ich selbst getestet habe, werde ich euch sagen, wie viel Zeit ich dadurch erspart habe und ob ich es für meine ganzen eigenen E-Commerce Brands selbst nutzen würde, warum Sourcing so viel Zeit kostet. Du fragst das Produkt an, fragst nach dem Preis, fragst nach den Maßinheiten, nach der Größe, nach der Qualität, nach den Verpackungsoptionen, aber auch z.B.

or some AI automations and she doesn t really have much experience doing it, but she just cold DM people on LinkedIn and she was able to make like more than 50k the last month selling these kind of automations and she didn t really have that much experience or credentials, she just chose a market that had a painful problem and because there s not a lot of people selling the same thing, people are willing to talk to her. So when you use my claw scale, it s going to give you the landscape, right, of okay, like what does your offer triangle look like, what you cover, what to sell, who to sell it to, what to charge, and then it goes more depth into like, okay, like this is the things that you can sell. They re all unprofitable, but they re still spending millions of dollars on like paying influencers, creators, running ads, and the model is like, hey, like VCs, the people that back these companies, they know that only a small percentage will survive, and that one winner will pay for all the losses. At the end of the day, you need to understand what s their pain that they want to get away from, what s their desire, and how can you get them from where they are to where they actually want to go. So even though you have an idea of an offer like let s say it s oh 3K per month for some LinkedIn posts the way you sell it is you don t sell the post you just take all the problems he has reverse it into solutions and you just literally mirror his problems once you do that the right way the person will naturally feel like wow this guy actually gets it like how do I work with you right it s as simple as that you are a mirror to their problems and you re showing them you have a solution now if you want to see the full live role play I actually do it on the free 3-day live training that I have coming up.

In diesem Video zeige ich Azubis wie du die beste Pflege-KI als Wissensdatenbank nutzt: Pflegeplan schreiben nach DIN 13004 Medikamente Schritt-für-Schritt erklärt Hygiene Infektionsprävention aktuell Für Prüfung lernen mit KI Prompts Für Pflege Azubis 1., 2., 3. Azubi KI-Prompts Wissensdatenbank: slimli.de kipflegBERATUNG Mein Azubi Discord für Fragen 0:00 Warum Ausbildung schwer ist 1:00 KI Wissensdatenbank einrichten 2:30 Pflegeplan Beispiel 4:00 Medikamente lernen 5:30 Prüfung Hacks 6:30 Fazit für Azubis Hier erfährst du, wie du KI in der Pflege sinnvoll nutzen kannst, um Zeit zu sparen, Stress zu reduzieren und den Kopf wieder frei zu bekommen egal ob auf Station oder bei dir zuhause am Küchentisch. Hier gibt es kein IT-Fachchinesisch, sondern echte Lösungen von Kollege zu Kollege. Lass uns die Technik nutzen, damit wieder mehr Zeit für das bleibt, was wirklich zählt: Der Mensch und DU. Pflegeausbildung KIinderPflege DigitalisierungInDerPflege Wissensdatenbank Pflegeazubi SmartNursing LernenMitKI EdTech Praxisanleitung PflegeCommunity

An agent sends you a question to a model along with a list of tools and the model is allowed to say, Hold on, I need to use one of those first. The other two are your choice, which is worth knowing cuz N s own documentation page for this node says, and I m quoting, you must connect at least one tool sub node to an AI agent node. Two of them run on your own machine and cost nothing to call, which is what we re using, and Lemonade, which is AMD s local server, and I haven t tested that one, so all I m telling you is that it s there. http col localhost col11434 that s three separate ideas http col is just how a web address starts localhost means this computer right here not the internet and 1 1434 is the port number that lama listens on every part of that stays on your machine to know before you move on if that default gives you a connection error N s own troubleshooting notes say to swap the word local host for 127.0.0.1. Uh I agent tool call nan workflow tool code tool and HTTP request tool which is the big one cuz that lets your agent call any API on the internet.

Five new things ChatGPT can do that you  probably didn t know about and why I switched to ChatGPT work. Number two, click the new work toggle, then click the plus icon and select  a folder with all your project files. Once it s connected, ask ChatGPT summarize my most important emails from the last week and  you ll get a summary of what matters most. Number  four, ask ChatGPT to send you that same summary every Monday morning. Number five, I ll ask ChatGPT to organize my messy desktop.

Then we run into a context problem where Claude is always trying to use the MCP tools to fetch for context on what s going on inside of the workflows. So, what I m going to do is just take a screenshot of this, send it over to the AI assistant, and ask it to give me some sample content as well as to make it check if the workflow is actually using these columns. It also gave me a heads up on images here where it identified that if I didn t use an image, I would get an error with this workflow and probably I could have it work on that and make it so that the images are optional and if there s no image, it will still make the post. Now, let s add one more step to this process, which is a Telegram AI agent that should help me with ideas and eventually take those ideas and add them as content to the Notion database. As you can see, after a few attempts, it just decided to replace the Notion tool with an API fetch, which may or may not have been the right decision here.

Bevor wir damit loslegen, das Video ist nicht gesponsort und ich mache diesen Grundkurs einfach nur, damit du Chat Work für deine Aufgaben nutzen kannst. Wenn du in der Cloud arbeitest, kann Chat GPT nicht auf Dateien auf deinem Computer zugreifen und deshalb wählen wir hier Onut und über Projekte auswählen kannst du ein bestehendes Projekt, was du vorher angelegt hast, wählen oder neues Projekt starten. Hier gibst du dem Projekt jetzt einen Namen und kannst da drunter Ordner auf deinem Computer hinzufügen, die Chat GBT lesen und bearbeiten kann. Ich habe auf meinem Computer ein Ordner erstellt, den ich Instagram Analyse nenne und jetzt starte ich eine neue Aufgabe in Chat GBT Work. Chgp Work arbeitet jetzt auf deinem Computer mit deinen lokalen Dateien und deshalb habe ich persönlich in meinem Chat GPT in den Einstellungen von meinem Content unter Datenkontrolle diese Einstellung hier deaktiviert, das Modell für alle verbessern.

Now, next step is, now I will add the Google Sheet trigger node. Now that we have explored a technical use case connecting Google Sheets and Gmail, let s move on to a workflow that s even more approachable, perfect for anyone who wants automation without writing a single line of code. Next, we will clean and format the data using a Set node, and finally, send a personalized confirmation email using the Gmail node. I will also show you how to test the webhook using a simple POST request, and we will verify that everything follows smoothly from data capture to final email notification. And now here you can see that I have created a Google Sheet called a subscribers demo, and I have the fields timestamp, name, email, and source.

Join my FREE Creator Magic community on Buzz Buzz Mobile is here. Buzz Mobile just launched on the App Store and Google Play and there was not a single video about it. I pair my phone to my desktop with a QR code, tour the whole mobile app, and then take my AI team to the beach to prove the point. 0:00 Intro: Jack Dorsey s Buzz Mobile App 0:27 What is Buzz AI by Block? 1:10 Desktop Setup Creator Magic Community 1:50 How to Download Buzz App Pair Mobile 3:04 Creating Custom AI Agents Mobile Chat Demo 4:11 Remote AI Agent Control Custom Emojis 5:12 Multimodal Vision Test: Sending Photos Videos 6:33 Multiple Communities, Local LLMs Themes 7:35 Vibe Coding an HTML Game on Mobile 9:54 Desktop Harnesses: OpenClaw Hermes Agent Support 10:16 Current Limitations: Review How to Join Buzz: Buzz on GitHub: Bring Your Own Harness PR: 𝕏 Me on X: Buzz iOS App: Buzz Android App:

The creator went from handling two projects per month generating around 10,000 to handling music five or six projects per month generating between 25 to music 30,000 revenue per hour increased because well development time was no longer the bottleneck. Ryan Dozer proved music the digital product angle by packaging around 20 skills into a 99 bundle covering categories like SEO, blog writing, anti-slop editing, and music YouTube thumbnail design. Some people use Cloud Code to build the wrapper around the API, while others music connect the API to no code tools like Zapier and automate entire workflows without touching the underlying code. Claudetlab.net NET documents a three-phase progression where operators start around 30 to 500 per month during the content phase, grow music to roughly 1,00 to 1,500 per month while providing knowledgebased services, and eventually reach 2,000 to 3,000 per month after a year of consistent operation. music It s one of the fastest way to turn an idea into something people can download, interact with, and pay for which it music makes it very accessible starting point for someone who wants to, I don t know, make their first dollar with AI.

A task management app only becomes useful once users can actually create tasks, add details, and see those tasks saved inside the app. Add a dashboard at the top of the app showing four summary cards: total tasks, completed tasks, tasks in progress, Below the card, add a progress bar showing the overall percentage of tasks completed. So, for our task management app, that just makes the workflow a lot faster, because now the user can describe what they want in plain language, and then the assistant can just understand the message, figure out the task action, and then apply it inside of the app. And this is where the task manager starts becoming more convenient, because now users can still use the normal interface when they want to, but now they also have a faster option for creating, updating, or checking tasks through conversation. So, for this upgrade, the agent should be able to create new tasks, update existing ones, mark tasks as complete, answer questions about the task list, suggest related tasks, and also help with related work through conversation.

Auf dem Papier versprechen Chat GPT Work und Claude Co-Work deshalb fast genau das Gleiche. Damit beide wirklich dieselben Voraussetzungen haben, habe ich zwei identische Testordner erstellt: einen für Chat GPT Work und einen für Claude Co-Work. Damit gewinnt ChatGPT Work für mich den zweiten Test deutlich. Diesmal wollen wir prüfen, wie gut ChatGPT Work und Cloud Co-Work eine kreative Aufgabe über eine externe Integration umsetzen. Damit gewinnt ChatGPT Work für mich den dritten Test ganz ganz ganz knapp.

And then finally because 63 of you people find AI tools expensive, If you watch this entire video, you will also be able to make a good looking AI film. Because by keeping things simple in the beginning, You will be able to focus on what type of prompts give better results or what difference does it make in the output by removing some add certain keywords Credits and time will also be less for small videos. We can choose the camera, lens, focal length and the aperture Just like we do for real videos and photos So let s do something crazy by selecting ARRI Alexa Fish eye lens, ultra wide angle For those who don t know, the face gets completely distorted in fish eye close-ups, And as you can see, it was able to generate that Let s try something new Side profile with a Persian cat in a swimming pool We ll go with a spherical lens, 85mm, f 2 And look at this. So again, I got the prompt from GPT, I asked the first one to pick subject and environment, second one to pick camera angle, added some customizations like camera should be near the keyboard and not show much background. But if you don t want to put this shot list, references, back and forth from GPT and time in iterations, if you want to save credits, then you can write a simple prompt for the first image, turn on the storyboard feature and then click on generate.

You can kick off a cloud agent like this, and it will start up a new development environment in the cloud that has all of its libraries, all of its context, all of its development information needed so I can work independently. And so you might be asking yourself right now, well, if it's just like tmux, can I get all my tmux bindings from tmux into herder? So I can write this, I can quit, and then I can just do herder server reload dash config. I can just hit yes or whatever, yes again, and then if I close Herder by typing prefix Q, if I look for the process of Claude, I can see Claude is actually still running right now. I can just say, "Make a new pane in Herder to um write "Hello world." Now, Cursor actually loads the Herder skill, and it knows how to make a new pane in Herder.

Join here: Video notes links to the tools Get a FREE AI Course Community 1,000 AI Agents Get a FREE AI SEO Strategy Session Get 200 Free AI SEO Prompts Get out SEO link building book here Jack Dorsey s Buzz: Free Slack-Like Workspace for AI Agent Teams (Codex Claude CLI) The script introduces Buzz, a newly launched app from Jack Dorsey and Block that looks like Slack but is AI-powered and free to use with multiple employees and teams of AI agents. Buzz runs on Codex CLI or Claude Code CLI, letting users leverage existing subscriptions without using APIs or paying extra. The presenter demos creating channels and tagging agents to generate improved image thumbnails, then shows how Buzz can reuse prior Claude Codex context and skills (including WordPress posting and Google Search Console analysis) to generate and publish a fully formatted SEO blog post without logging into WordPress. Buzz supports separate workflow channels, custom agents, combining models (including an open source agent called Goose), local context like an Obsidian vault, invites for human teammates, optional additional models, and monitoring active workflows. 00:00 Buzz Launch Overview 00:29 Thumbnail Agent Demo 01:35 Channels and Workflows 01:54 Getting Access to Buzz 02:08 Humans Plus Skills Context 03:22 WordPress Blog Automation 04:26 Build and Manage Agents 05:25 Clean Up and Rename Channels 05:51 Obsidian and SEO Keywords 07:47 Team Invites and Compute 08:29 Models Experiments Mobile 09:18 Wrap Up and Training Offer

Now, I ll paste this reference image into ChatGPT and ask it to generate a prompt that describes this lighting and visual style. Once we have the prompt, you can use it with any AI image generator, such as Nano Banana or ChatGPT s image generation, to recreate the same look with your own photo. Use the provided reference, remove the background visuals, place all the content on top of the existing scroll animation, and make sure the scroll animation continues to work properly. I also like these two sections, so I ll take screenshots, paste them into Antigravity, and use this prompt. Whether it s an about page, contact page, blog, or services page, simply provide a reference image and a clear prompt, and let AI do the heavy lifting.

In addition to connecting your apps, right down  here, you can also upload existing files, or you   could even create new documents directly in Second  Brain. Instead of searching through emails, documents,  or chat messages individually, you simply ask your   Second Brain a question, and it pulls together  the relevant information for you. Second Brain pulls together   my emails, calendar, meeting notes, documents, and  anything else related to Elizabeth. Now  that the meeting's over, let's see what Second   Brain captured. Now that  my meeting has become part of my Second Brain,   I can ask questions about it just like I can with  my emails, documents, and everything else I've   connected.

Und das finde ich auch irgendwie super super cool, weil ich erinnere mich noch Sabrina, wir saßen mal, ich glaube 4er Stunden im Studio und haben einfach nur gebrainstormt über unsere Strategie und haben das einfach aufgezeichnet, also haben das einfach mit quasi wie so einem Diktiergerät mäßig aufgezeichnet, was wir gesprochen haben und letztendlich haben wir das dann alles von Claude eben zusammenfassen lassen und der hat dann auch gesagt, ja, darüber habt ihr irgendwie noch gar nicht gesprochen, da habt ihr noch nicht so drüber nachgedacht, also auch so das Krit kritisch hinterfragt und das finde ich halt super cool. Also neben dieser ganz normalen Canva AI, ja, wo du auch theoretisch einfach reinschreiben kannst und gemeinsam Sachen entwickeln kannst, gibt es halt eben diese Tools und da gibt es halt super super coole Sachen, die einfach super viel Zeit sparen. Also, kann man echt mal ausprobieren, weil ich weiß, die meisten oder sehr, sehr viele von unseren Zuhörern, die arbeiten auch mit Can und haben da wahrscheinlich sogar auch ein Pro Abo und da kann man sich da einfach mal so ein bisschen ja, bisschen kreativ werden und einfach mal austesten, wenn man vielleicht noch gar keinen Zugang zu Cloud hat, dann kann man so auch coole Freebes z.B. Aber meiner Meinung nach ist es alles in deinem Abo schon mit inkludiert, weil bei den anderen AI Tools in Canva hast du ja auch kein Limit, also dass du nur fünf mal ein Background irgendwie entfernen kannst oder sowas und sind einfach Tools, die damit drin sind, weil ich glaube, weil was ich auch geil finde bei Ken ist ja dieses neue Carousel Studio und das muss man aber wirklich sehr gut prompten, muss man sagen. komplett von A bis Z habe ich das jetzt auch nicht alles verwendet, aber wirklich, wenn es rund um Videogenerierung geht, dann halte ich mich immer dort auf, weil es einfach diesen Vorteil bietet, dass du dich nicht auf einziges Tool beschränken musst, sondern du hast alles an einem Ort und kannst dann für jeden einzelnen Fall, für jedes einzelne Video selber entscheiden, welches Tool nutze ich, was mache ich damit und wie viel verbrauche ich auch mit dem einzelnen Tool, ne?

So, in today's video, we'll take a look at Buzz, see how it works, and we'll test it out by making a joint human plus AI agent workspace and run some fun experiments on it. So, basically, I want Hackerman to analyze both codebases cooked up by Dinesh and Gilfoyle, and determine which one is the most secure one, which in turn will let us know which model is the superior one in this case, Gemini 3.6 Flash or GLM 5.2. And another interesting thing to note here is that Buzz puts all your projects in the .buzz directory, so that's where all the work done by the agents live. For example, you might encounter situations where if you're using one of these Open Router models for your agents, they might end up searching the inner Buzz docs to see how to execute certain commands, like adding others to channels or similar actions. But while testing it, I also came across situations where I exceeded my token limits because the problem with these multi-agent workflows is that they keep messaging each other and piling requests upon each other until you max out your tokens.

So und angenommen, du möchtest jetzt eine neue Website erstellen oder deine bestehende Website bearbeiten, aber irgendwie bekommt es Fable nicht ganz hin, das was du dir vorstellst auch so abzubilden, aber dir fällt es auch schwer, das Ganze manuell auszuschreiben, denn eine Design Vorstellung in Text umzwandeln ist echt nicht immer einfach. diese hier, dann kann ich jetzt hier von einen Screenshot machen und den einfach nur ein Fable geben und der wird dann genau analysieren, okay, was für Elemente hat die Seite, wie muss es aussehen und kann sich dementsprechend orientieren. Es gibt auch die Seite Motion Sites und hier kannst du sogar nicht nur mit Screenshots arbeiten, sondern du kannst dir sogar den kompletten Prompt kopieren und den einfach an Fable geben und damit wird dann exakt diese Seite hier nachgebaut. Damit sparst du dir dein Fable Kontingent und dann im Anschluss kannst du die Seite mit Fable gerne weiter bearbeiten und das Design auch noch weiter anpassen und den Inhalt und so weiter, denn du willst jetzt auch nicht nur stumpf die Seite kopieren und das so lassen, das kommt auch nicht gut. Ich hinterlasse den Skill auch in der Videobeschreibung und du kannst den extrem einfach installieren, indem du einfach hier diesen Command kopierst oder auch einfach den Browser Link und einfach nur zu Cloud gehst und sagst, er soll den Skill installieren.

Firstly, let s test out the image generator by clicking here on image and typing in a simple prompt like we music can say realistic image of a dog playing in a park. Now do keep in mind that the AI platforms will progressively keep music getting better and at the end I ll show you the ultimate AI platform music that lets you use almost every single AI model out there for completely free. Now, let s test out the video generating model because if you click right here on create video, you can see you can type in a simple prompt to test this out. So firstly, before we test out the video generator, let s try generating images with it by clicking right here on create music image. All you have to do is click right here on generate and in music no time you should have this beautiful image right here.

A lot of people are asking, well, what is the primary difference between Chat GPT Work and Chat GPT Chat? Now, if you don t know what Chat GPT sites is, I have a video about it right here, but Chat GPT sites is so much more than what I was able to explain in that video or in this one. So, number one, web research, deep research, file analysis, image generation, create downloadable documents spreadsheets presentations charts, PDFs, use connected apps, taking actions through supported connected apps, schedule reminders and monitoring, multi-step reasoning, agentic website tasks writing coding analysis brainstorming, working inside ChatGPT projects, voice conversations, especially with the new GPT live model, producing multiple deliverables. So, even if Claude, Grok, or some other platform is your primary AI, you should still at least have the 20 ChatGPT plan because you can use ChatGPT messages to get so much of the grunt work done up front and then hand that off to Claude code or whatever other platform you re using to take the project across the finish line. Whereas, ChatGPT is cloud-based, but everything that work can do just about, ChatGPT can do as well besides run your local desktop apps, use your built-in browser, read your file system, and obviously create ChatGPT sites.

It can also just work through a research question without producing a specific output, but it s able to take those multiple steps throughout Here s the bit that s confusing a lot of people right now. With any apps connected, it can still access the web, files you upload, and project context, but it is helpful for getting the latest from these apps as you work and so that it comes into ChatGPT and it s part of the workflow rather than, again, having to copy and paste. ChatGPT work is really good tool for actually going, searching the web, searching different websites, seeing what s out there, doing the analysis, and analyzing what s best depending on the criteria you give us. I actually think the understanding the delivery fee is really great part of ChatGPT work, so that you re making a good decision rather than kind of just guessing based on the prices that you might see in a regular chat. So, basically the audit I m going to run is just one on a subscriptions I have, and and what I wanted to do is build a spreadsheet with the tool name, the plan, how much it s costing me, basically calculate the totals, and I ve told it because it s going into my emails do not reply forward or archive any or delete any email.

How this is going to send an alert to Google Sheet or the Slack channel so that our DevOps engineer is aware of that server is actually down and they need to take an action. So, I want to use server status report because I have already created one Google sheet that is called server status report. You can also go go to Google sheet and just create one Google sheet and just give the name to server status report or anything that you want. So, you need to first authenticate to Slack, then only you will see multiple options to select which channel, which message, where to send and everything. You actually do your work in back end, it every time runs every 5 minutes, 10 minute, and whenever there is issue encountered, it will trigger an alert and send a message to the Slack channel, so that your DevOps engineer are aware there is an issue with this particular server, and they need to work upon it.

Ich habe das mit Cloud verbunden und Cloud speichert sich da immer musik alles rein und da liegen super viele Informationen über mich drinnen, die ich vorher Claud gegeben habe und wo ich Cloud gesagt habe: Hey, bitte pack die alle systematisch in Obsidien rein, damit du immer weißt, was los ist. Da ist beispielsweise drin, wer ich bin, welche Firmen ich betreibe, was die Firmen genau machen, welche Umsätze die Firmen machen, welche Probleme in der Firma herrschen, welche Leute dabei sind, welche Freelancer wir haben, zack, zack, zack, einfach alles, jeglichen Kontext und jedes Mal, wenn eine Session durch ist, erweitert sich diese Datenbank. Jetzt kann man einfach einen prompt eingeben, dann bekommt man den Inhalt, den kann man dann noch abändern und wenn der dann passt, werden die Slides literally in Minuten erstellt. Wenn die KI sehr gemein zu dir ist oder sehr ehrlich zu dir ist, dann ist das nicht gut für Business und viele Leute musik können das nicht handeln und gehen deswegen aus dem Abo raus. Und immer wenn ich meiner KI jetzt beispielsweise sage: Hey, mach mir bitte ein Bild musik von dem und dem in diesem Stil, im Improvement Movement Stil, weil so heißt die Community, musik dann weiß die KI exakt, was zu tun ist und ich kriege immer wieder super Ergebnisse, mit denen ich direkt zufrieden bin.

Want to make money and save time with AI? Join here: Video notes links to the tools Get a FREE AI Course Community 1,000 AI Agents Get a FREE AI SEO Strategy Session Get 200 Free AI SEO Prompts Get out SEO link building book here

Now, if you do want to master AI tools and you want to learn how to build profitable SAS apps and websites, AI agents and mobile apps with AI, I ve created a complete master class that shows you exactly how to do it all step by step. It explains how users will enter their information, what they ll see when they open up the app, and the fact that every user should have their own account. Right now, every meal is being saved into one daily list, and that works for a basic tracker, but it can get messy once users start logging multiple meals throughout the day. It just makes the daily view easier to scan and it also makes the app just feel closer to how people actually track their meals. So for a website, start by defining the sections that you want, generate the first version, then publish once everything looks right, and then say for applications, begin with the core functionality, make sure that the foundation works, and then add new features one layer at a time before finishing with a final polish.

This app also lets you choose from a library of custom alarm sounds that go well beyond anything on your default clock app. You can stack multiple challenges into one alarm sequence and the alarm does not stop until you complete all of them. By the time I finished the last one, my brain has no real choice but to be awake. The app also learns over time, which sounds like marketing copy, but turned out to be real. Before we close out, I want to drop three bonus apps that did not fit cleanly into a time of day slot.

You see, big creators have hours of podcasts, YouTube videos, and live streams just sitting there, and they know that short clips, like the ones you might see on TikTok or Instagram, are the easiest way to reach new fans. But to be fair, most people don t spend enough time learning what performs best and how to properly use the AI tools to their full potential. Step one is to pick an AI page builder and learn it by building a sample page for an imaginary offer so that you ve got proof of what you can actually do. If you want to learn exactly how to find qualified businesses and get these businesses to pay you for any service, including this one, I m actually running a free live online training very soon. Step one, pick one type of local business like a dentist, salons, or plumbers.

Roboter für musik diese aufeinander folgenden Arbeitsschritte werden inzwischen von mehreren chinesischen musik Unternehmen angeboten, darunter Bright Dream Robotics und Fang Robotics. musik Nach Angaben von Wienerberger haben 12 dieser Roboter bis April musik 2026 bereits rund 40.000 km Mauerwerk in mehreren europäischen Ländern errichtet. Das Unternehmen musik spricht außerdem von einer Genauigkeit bei Position und Bohrtiefe von 99,97 musik und von einer Arbeitsgeschwindigkeit, die bis musik zu zehn mal höher sein soll als bei der herkömmlichen Methode. Ein chinesischer Fliesenleger musik Roboter trägt selbstständig den Kleber auf, positioniert jede einzelne Fliese präzise musik und verlegt bis zu 18 m² pro Stunde. Ein Roboter des singapurischen Unternehmens Fabrika AI erkennt selbstständig musik die Fugen zwischen den Fliesen, füllt sie mit Fugenmasse und entfernt anschließend das musik überschüssige Material.

And the part most people get wrong isn t the tool, it s the one step that decides whether your agent is reliable or garbage. In the AI agent node, you attach tools to the agent itself. Trigger fires, agent reads, agent writes, email sends. Every multi-step agent burns through what these platforms call operations or tasks and they drain faster than you d expect. Open an 8 N, drop in a schedule trigger, add the AI agent node, connect your model.

Give your Buzz agents hands: Join my community and share our compute: I have been building on Buzz all week and the one thing my agents could not do was touch anything outside the app, so I gave them hands. In this video I spin up a private channel, switch on Workflows, and build two agents: Wizard, who researches me and writes the voice guide onto the channel canvas, and Merlin, who writes LinkedIn posts and nothing else. Then I wire two Zaps, so every draft Merlin writes lands in a Zapier Table, and a single thumbs up publishes it straight to LinkedIn with the receipt posted back into the channel. Every message and reaction on Buzz is a signed event, so only my key can release a post, and the approval sits in the audit log right next to the thing it authorised. Try Buzz here: 0:00 Next Gen AI Agents vs OpenClaw Hermes 0:39 How to Enable Experimental Workflows in Buzz 1:23 Creating a Research AI Agent for Brand Voice 3:00 Building a Dedicated LinkedIn Writer Agent 4:12 Buzz Shared Compute Free Local AI Models 5:41 Setting Up Buzz Workflows Zapier Webhooks 6:37 Capture AI Post Drafts with Zapier Tables Agent Signatures 8:34 One Click Emoji Thumbs Up Approval Workflow 9:51 How to Auto Publish AI Drafts to LinkedIn via Zapier 11:12 Live Demo: 1 Click Thumbs Up to Live LinkedIn Post 12:40 How Peer to Peer Local AI Compute Sharing Works 13:41 Why Signed Autonomous Teammates Are Replacing AI Assistants

Un peu comme si pour construire notre automatisation et bien on devait aller à le roi Merlin ou euh ou par exemple pour construire une salle de bain et ben il faut regarder qu est-ce qui est compatible avec quoi c est long, c est chronophage, ça prend du temps et surtout bah on doit se casser la tête un petit peu et c était un peu la la manière de faire d avant en fait pour automatiser des choses. Déjà premièrement, c est parce que mon besoin était d automatiser un processus qui était répétable, qui était toujours le même et que j avais besoin qui tourne sans que je sois là et pas que il y ait un assistant, une assistante qui vient cliquer à chaque fois pour donner accès à des gens de ce qu ils ont acheté. Et bien sachez que vous allez avoir un premier effet wow assez rapidement mais que derrière il va falloir monter en compétence sur ce que c est qu une base de données, ce que c est qu une API, ce que c est qu un webhook, comment fonctionne un front-end, un backend, une queue et cetera. Maintenant, si vous avez une application qui a une logique 100 custom et vous pouvez gérer le déploiement, c est que vous savez utiliser des outils comme Docploy Versel desquels on a déjà parlé sur cette chaîne, mais qu en plus vous avez du front end, du back end, vous avez une app compliquée, et bien faut aller sur Cloud Code. Je vous incite même après cette vidéo à aller tester une des applications que je vous ai cité que vous ne connaissez peut-être pas parce que sur YouTube, c est bien, c est du Netflix, vous avez l impression d apprendre mais la réalité se passe sur votre clavier.

In this video, I ll be telling you about Buzz, the new open-source workspace from Jack Dorsey s Block, where humans and AI agents work together as equal teammates. Buzz combines team chat, AI agents, code collaboration, cryptographic identity, and a unified event log, all built on top of the decentralized Nostr protocol. -- Key Takeaways: Block has launched Buzz, an open-source workspace where humans and AI agents collaborate as equal team members. Buzz supports Claude Code, OpenAI Codex, and Goose through the Agent Client Protocol, making it model- and agent-agnostic. Overall, Buzz is an exciting early look at how team collaboration could work when AI agents become first-class teammates.

Once a skill is installed, it sits in your agent's memory and can be triggered when prompting the agent to do a specific task, or you can just trigger it yourself by telling the agent to use the specific skill. One skill and your agent can operate your entire business stack. Inspired by the famous O my Claude skill, which has 36,000 GitHub stars, it turns one agent session into a coordinated multi-agent workflow, and it decomposes it into subtasks, assigns specialist agents or external CLI workers like Codex, Gemini, Cursor, runs them in parallel or staged pipelines, then verifies the result instead of stopping at a half-done answer. Each one, every other skill on this list makes your agent better at things it could already do. To install, just tell your agent help me install Agent Reach and and the link in the chat.

It's often hard when you're planning out a spec to figure out exactly how things should look, and how things should look, how they should behave under certain circumstances, making sure that you've road tested all of this idea in kind of semi working code, that means you need a higher fidelity. So, the agent is going to ask me some questions, probably through a grilling session, and then I'm going to say, "Okay, yeah, this sounds good." But then, as soon as we get anything that I think, "Oh, I really need to see this in action. But when you need to raise the fidelity of the discussion, making a cheap, rough, concrete artifact to react to, an outline, a rough take a stab at UI logic code via the prototype skill, links the prototype as an asset, use when how should it look or how should it behave is is key question. So, I'm actually going to compact and say, "We're going to do some more QA on this." Compacting at this point makes sense to me because we kind of need to retain all of the information about the prototype, need to retain all of the design decisions that went into it, but we just need to give it some feedback. Front-end definitely benefits a lot from prototypes because how should it look, what should it look like is a really key question that's really hard to answer during the discussion phase.

This is the official Hermes Agent website, and if you want, you can read through this and get an understanding of what Hermes Agent is all about. You can access it through all sorts of channels, like Discord, Telegram, WhatsApp, email, really whatever you want, you can hook it into your agent. So if you teach your Hermes Agent a certain workflow that you go through every morning where you log into a system, you pull certain stats, you then add those figures or numbers to some kind of spreadsheet, and then you email that off to somewhere else, you can teach that workflow to your Hermes Agent, and that agent will create a new skill that will use going forward. We can view all of the cron jobs and those are all of those little scheduled jobs that the agent can set up and all of the little automations. So, from the skills tab, we can add skills, we can disable or enable skills, we can go to plugins and of course view all of the installed plugins, disable them, everything that you would expect to do with the agent on your local machine.

So if you've ever like left an agent running and then you know you're worried about what it's going to do when it's running in the background, well basically you can switch it on by default so that every um basically flagged command gets approved by you manually or you can switch on full auto. So if the agent wants to run a command, there's a smart gate where a second AI assesses it or you can it can run or it can ask you as well. We actually have a system for this where you can see here if we go over to the chat section, we can have all of our different agent profiles set up and then we can autodelegate from the main Hermes agent to those different profiles using this new system that just dropped. you can customize it way more and also with it being open source you can run it as uh free models inside the AI agent and then also the harness itself Hermes agent itself is free and open source so you know this is a really really powerful model that is free to use as well and also some people say well I don't need sub agents but honestly the sooner you learn how to use sub agents with AI the faster you're going to move and I'll show you some examples in a minute you also might think okay this stuff is too technical for the average person but if you look at the ARI profit boardroom and our testimonials and reviews. And also when you're running a a board like this where you're for example creating SEO content like you see in this blog post or a full video as you can see here when you're doing that sort of stuff the main thing here is that you don't need to approve every single flag command because you've got that second AI now that can just assess it and go yeah it's fine no problem.

Sie haben sehr, sehr viel darüber gesprochen, dass das System jetzt sehr, sehr leistungsstark geworden ist. Und das ist eine sehr, sehr wichtige Aufgabe, Transparenz und vor allem wird es für mich zu einer Box, in der ich das System perfekt nutzen und genau verstehen kann, wie es gerade denkt. Er bittet mich also den Ordner anzugeben und das ist sehr interessant, weil er sich tatsächlich nur dann stoppt, wenn es notwendig ist, da es eine Verwirrung darüber gibt, welchen Dateityp ich löschen möchte. Und jetzt lädt er also die sogenannten Skills und jetzt gibt er mir das genaue Log dieser drei Agenten, damit ich einfach die Transkriptionen sehen kann. Also ehrlich gesagt habe ich nur ein einziges Profil, das mein Hauptprofil ist, da ich der einzige bin, der an meiner Hermesinstallation arbeitet, also habe ich nur einen Telegram Account.

But now, what they've implemented, and this is also very interesting, is that instead of waiting, as you see here, for 30 seconds while the system is reflecting, but we can't see what it's doing. We can read exactly how it's processing, and this takes Hermes from being a black box to a truly transparent one, where I can actually see the tools and how they're working. And now, it's actually going to give me the content, so you'll understand that today, in fact, the sub-agents are really something that strengthens the Hermes agent. So, as a result, the concept of sub-agents is really being taken very seriously by the Hermes team, and every time I see an update, I notice they're working a lot on the sub-agents, and that's very interesting. Now, there are other tasks, well, other features which I think aren't very interesting, uh but still it's important to mention them to be aware of, you know, the updates to Hermes.

Today, I'm walking you through the brand new Hermes agent update, what changed and how I run Hermes inside my agent OS. Hermes is one of the agents living inside my Agent OS, which means every speed gain in this release shows up in the work I do every single day. There are Hermes walkthroughs covering the exact setup I'm describing here, live coaching calls where you can bring your own install and get it unstuck, and the prompts I use to run these agents. And look, the honest truth about Hermes and Agent OS is that the install is where most people store keys in the wrong place, the gateway not connecting, approvals set too tight or too loose, agents that work once and then sit there doing nothing. You get my full Agent OS zip file ready to install, the 30-day roadmap that takes you from empty dashboard to running agents, the Hermes walkthroughs, the prompts, and live coaching calls where you can share your screen and get it fixed on the spot.

Today, I'm walking you through what changed in Hermes' agent, why it matters more than it sounds, and how I run Hermes' agent inside my agent OS. Inside my agent OS, I run Hermes' agent as the writer on my SEO station, and I use that to draft a full content series answering the questions AI Profit Boardroom members ask me most. There are walkthroughs covering how I wire Hermes agent into the Agent OS, how I set up profiles for different models, and how the memory layer is structured. Fanning out 10 missions only works if you trust what's happening while you're not looking, and that's what live sub-agent transcripts fix. You get my full Agent OS zip file ready to install, the 30-day roadmap so you know the order to do it in, the prompts I use for the profiles, tutorials on wiring Hermes agent into the stations, and live coaching calls four times a week where you can share your screen and get unstuck in minutes instead of days.

Most AI agents forget everything the moment a session ends forcing teams to rebuild context from zero on every run. LLM Wikis take a different approach to agent memory: instead of just storing information, the agent reads, reconciles, and rewrites its own knowledge over time, much like a person maintaining a living wiki. Join Izma Aziz for a live, hands-on build using LangGraph and Deep Agents watching an LLM Wiki come together step by step, with the reasoning behind each design decision explained as it happens. You ll learn: What an LLM Wiki is and why persistent, growing memory matters for agents How it s different from RAG, file search, and standard chat history The full lifecycle: ingesting, organizing, querying, and updating knowledge How to watch a working LLM Wiki get built live, not just described in slides The design tradeoffs and limitations to expect in production Why the cost of forgetting compounds as agents take on longer, more autonomous tasks

By the end of this video, I'll have a full AI team in there, and I'll show you how to do it too, including a chief of staff agent that runs all the other agents for me, and also stick around to the end, because I'll show you how you can join my community and work alongside my AI agents. so now just like Slack I can add a channel member and let's find the Sea Swim agent perfect managed by me on my own harness that is now added to the channel. So now Sea Swim is here and I can say tell me what it's like today and also bring me up to speed on what's happened in this channel so far. Again, I can click down here and view the activity and see Claude Code literally running my agent right now. Now it doesn't stop there because I can actually click into Sea Swim and click message to drop a direct message to Sea Swim anywhere better than Paphos to swim in Cyprus today.

Most people run AI like one little chatbot, one chat box, one job at a time. I can make images all from one place, one command, a whole army. We show you step-bystep how to plug in Kim K3 for free, four live coaching calls every week where you can share your screen and ask about your own setup, daily tutorials, a prompt library, and a member map so you can meet other people building agents near you. Now, you might say, "Won't 10 agents just make 10 piles of junk?" Fair worry, but no, because they all share one brain, one fault, your voice, your standards. And one more thing, if you want the full process, all the SOPs, and 100 plus AI use cases just like this one, come join the AI success lab.

Es bringt nichts, wenn ihr euch in der Theorie irgendein Wissen aneignet, überlegt, wie so etwas aussehen könnte, was ihr da alles tolles machen könnt und am Ende des Tages macht ihr gar nichts, weil ihr euch noch damit beschäftigt, ja, wie das denn aussehen könnte. Das ist alles sehr vorbildlich, nur die befinden sich jetzt in so einem Teufelskreis, weil die sehr, sehr viele Einfälle haben, was sie nun jetzt alles noch damit machen können und die überschlagen sich dann völlig, die haben einen Backlock ohne Ende und morgen kommt wieder eine andere Sache. Aber der Punkt ist wirklich, mit dem, was ihr habt, mit diesem Werkzeug aktuell, könnt ihr es wirklich schaffen schon so viel umzusetzen, dass ihr gar nicht äh ja erstmal auf irgendwas noch warten müsst oder irgendwelche ähm KI Lösung euch anschauen müsst. Ja, und dann muss man natürlich auch die Zeit haben, wenn etwas fertig ist und nicht läuft, dass man das dann auch korrigiert, weil das bringt euch auch nichts, wenn ihr etwas zum Laufen bringt, aber es am Ende gar nicht läuft, weil es irgendwo irgendwelche Fehler gibt oder irgendwelche Komplikationen und euch das dann zu komplex ist oder ja, ihr auch nicht so wirklich Lust habt, das ist ja auch zerben, gar keine Frage, dass ihr dann sagt, ah, ich gehe jetzt erstmal wieder das nächste Thema an. Auch wenn ihr da keine Lust drauf habt, ist auch nicht jederms Sache, sich da reinzufuchsen oder auch diese ganzen Fehler, die man da bekommt, da ist nicht jeder für gemacht und das ist auch überhaupt nicht schlimm.

Am Ende zeige ich dir den fertigen und wie er in Sekunden einen Report schreibt, für den du sonst eine Stunde brauchst. Also lässt du NotebookLM erst die relevanten Fakten aus seiner Wissensbasis herausziehen, sauber mit Quellenangabe und gibst diese Fakten dann an Gemini, das daraus den fertigen Text baut. Ich gehe hier in Gemini in meinen Gem, den ich angelegt habe und da ich die Vorarbeit bereits in Notebook LM gemacht habe, brauche ich in Gemini einfach nur mein Notebook. Notebook LM zieht aus meinen Quellen mit Fundstelle und der Gemini Gem baut den fertigen Wochenbericht in unserem Format, den wir quasi haben wollen. Kurz zu Datenschutz: Weil dein KI-Agent mit Firmenwissen arbeitet, für alles mit Personenbezug, Kundennamen, Mitarbeiterdaten nutzt du NotebookLM bitte nur in der Google Workspace Business Version, nicht die kostenlose Version.

This starts building the web UI which then launches here in the background as you can see in my browser and we can now interact with Hermes directly through this which is a lot easier to do so simply click on chat on the left menu and here we've got another version of that terminal but this also doubles as our chat interface so we can start working with Hermes right in here now let's take a look at this dashboard on the right hand side here we can create new chat sessions and this allows us to see those sessions on the right hand side as history as well as create them. This again isn't probably something you'll need to access very much at all, but it is useful for the Hermes agent itself to have access to its own logs so it can identify what's going right, what's going wrong, and make its own updates. Right now, this is a pretty simple job, but I can work through this to do things like check all my emails, check my calendar, and maybe even check things like Slack and just give me a brief of everything I should be aware of for the day ahead, which can be incredibly useful when you set it up quite well. You can turn these on or off and you've got a few predefined ones like this one here for claw design which is currently turned on and I've just turned that off but we can turn that back on and take a quick look at it. This step here will set up in the background for a couple of minutes and finally take you to this dashboard here where you can click over here to access your Hermes agent.

But instead of just reading the headlines and seeing what's good, I also want to just key in like kind of the recent updates overall and how Hermes has improved over time and what we are using to get the max out of Hermes. Um, we don't really run these on the desktop versions or whatnot because I do believe that a 24-hour agent is the most powerful regardless of like, you know, you don't have to use it on your computer, you can just use it on chat. one is um not requiring the API key, but also of course they have a smart filtering system that they released like here now you can just like uh I wanted to show you guys the video but of this but the smart approvals just mean that you can save your passwords and do what you need to do. I still want to use Kimmy independently of uh Hermes like I use the Kimmy model uh Kimmy K3 and I use codeex model. I I've never really thought of seeing this day because yet again uh we run a company and we do have uh assistants to help us with our work but more and more of that work is like even the scheduling is becoming uh a Kimmy task rather than an assistant task and it's saving a lot of our assistance work for them to work more on company filing company registration or whatever whatever like workload they have to do.

Wenn ich ihn Hermes gebe, wird Hermes mir genau sagen, was er für mich tun wird und wie er mir tatsächlich im Alltag helfen kann. Genauso, ich bin tatsächlich ein wenig zu Cloud zurückgekehrt und versuche einfach äh ihm die Quellen zu nennen, die ich bevorzuge, wenn ich recherchiere und diese Recherchen zusammenfasse und anschließend gibt er mir tatsächlich die drei Quellen GitHub, da ich also sehr viel arbeite mit den Tools und den Updates, die auf dem Tool auf GitHub sind. Er kann uns wirklich dabei helfen, die Idee zu nehmen und sie durch den gesamten Prozess zu führen, den wir ihm gegeben haben, den wir ihm gegeben haben. alles was mit Briefings zu tun hat, denn dadurch gibt er mir die Informationen, die ich berücksichtigen muss, die tatsächliche Konnektivität meiner Infrastruktur zu testen, denn die Infrastruktur hostet Anwendungen und ich finde es wirklich sehr interessant, dass er das für mich macht und auch was den VPS und die Kommentare angeht, kann er tatsächlich die Kommentare herausfiltern und mir sagen, welche technischen Kommentare interessant sind, welche davon sind Level 1 Kommentare, also von Kunden und welche Antworten bereit sind, verschickt zu werden. Also hat er den Prompt für den morgendlichen Punkt erstellt, den ich auf Hermes starten muss, damit Herr Mess die Arbeit erledigt und mir das Ergebnis zurückgibt.

Now, instead of actually typing it because of AI psychosis and I'm incapable of like writing stuff anymore, what I like to do is go to a chatbot like Claude, for example, and then just write like, "Give me a brief description about what you know about me that I can give to Hermes agent." So, yes, it will just tell me about myself, which I can just copy that and I'll write, "Here is a bit about myself. Click open folder as vault, and you can just pick it wherever you want, just create a file, call it like Hermes vault, or whatever, and then just click open, and then you have your vault. So, what I can say here is, "Use Obsidian skill." Probably don't you actually need to like specify Obsidian skill, but I'm just going to do that just to make sure it's like very specific, "to grab the PRD for the food tracker and come up with a step-by-step plan for how I can implement the MVP version in 1 hour." Great. If you are super baller and you're like totally cool just paying for like Anthropic Claude models or like other models through news for whatever and you're like willing to spend like hundreds of hundreds dollars, that's totally cool, okay? I'm going to put on screen now some of the things that I have my multi-agent system set up, more advanced things and skills and just like things that I'm doing right now that you can do, you can unlock set up your Hermes agent properly.

Wenn ich mir also mein Schema anschaue, werde ich alles verbinden, also alles blockieren, um Clod die Möglichkeit zu geben, auf allen sozialen Netzwerken zu veröffentlichen, dank dieser Erweiterung. Also, wenn ich jetzt hier zurückkomme, klicke ich auf synchronisieren und dann lädt er alle Fähigkeiten und das Knohow, also auf die Art und Weise, wie Cloud heute tatsächlich den Inhalt hinzufügt und ihn mit unseren sozialen Netzwerken verbindet. So, ich kann also den Link und das Programm sehen, genauso wie ich es einfach neu planen oder es so lassen kann. Also hat er hinzugefügt genau, also die notwendigen Tage, damit es hier ist und das ist wirklich sehr, sehr interessant. Also jetzt hier wird er den Beitrag verfassen, das Bild generieren, den Beitrag überprüfen und mir dann das Bild zeigen.

jetzt eben so ein Faceless YouTube Channel und ich werde mich jetzt einfach mal strickt an diesen Guide hier halten und auch die Proms nutzen, die jetzt hier angegeben wurden, damit wir einfach schauen, ob das wirklich so klappt, wie die es sagen. Also habe ich einfach nur den Link von diesem Kanal kopiert und habe diesen minimalen Prompt hier geschrieben, dass er den Channel einfach nur analysieren soll. Das heißt, hier kann ich mir jetzt das gesamte Videos auch einfach durchlesen und danach habe ich ihm lediglich gesagt, dass er ein fünfminütiges Video jetzt erstellen soll mit dem Videogenerierungsmodell CDE 2.0 null in HD Auflösung und das Ganze soll dann am Ende auf einem Faceless YouTube Channel landen, damit ihr auch so ein bisschen den Kontext kennt. Und klar, wenn man jetzt mal die Videos von Primal Space anschaut, dann sieht man natürlich trotzdem noch einen Qualitätsunterschied, aber wir haben gerade in einem ziemlich ähnlichen Stil ein komplettes Video hinbekommen mit Script, Visuals, Sprecherstimme und Schnitt in einer halben Stunde und ich habe dafür buchstäblich nichts gemacht, außer zwei Proms zu schreiben. Das klingt erstmal viel, ist es aber überhaupt nicht, denn es braucht eigentlich nur ein Video, das wirklich sehr gut performt und du hast direkt 4000 Stunden Watchtime drin, weil es ist ja insgesamt gezählt von all den Leuten, die das Video schauen.

Wie wird aus gesammeltem Wissen ein System, dem ich vertrauen kann, statt ein wachsender Müllhaufen? Und wenn ihr dieses Video komplett durchschaut und ihr die Informationen, die ich euch mitteile, alle versteht, könnt ihr danach dieses System komplett nachbauen. Bevor auch nur eine einzige Zeile Code entstanden ist, haben wir aufgeschrieben, was das System können muss, um die Frage vom Anfang zu beantworten, was es braucht, damit ich jede Information wiederfinde und damit ich dem Bestand auch vertrauen kann. Jedes andere Teil der Architektur ist einfach nur eine abgeleitete Sicht auf diese Dateien und kann jederzeit weggeworfen und neu gebaut werden, ohne dass ein einziges Stück Wissen verloren geht. Wenn ihr tiefer verstehen wollt, wie Token und der Tokenizer funktionieren, dazu habe ich auch ein separates Video, das blende ich hier ein und der Link ist in der Beschreibung.

Entertainment ist ein riesen US Case, natürlich auch die ganzen anderen Use Cases in der Industrie und der Engine war jetzt für mich auch bis dato eine der beeindruckendsten Unternehmen und vor allem auch gar nicht mal nur das Thema Sicherheit, sondern auch wie menschlich sich diese Objekte hier bewegen und diese Humanoiden, also der Gang und die ganzen Bewegungen, ja, die sind menschlicher als manch ein Mensch und das ist definitiv wirklich höchst gerade beeindruckend, was Engine hier baut. Weise das Arbeiten mit Agentic Coding, ja, die Zukunft ist und nicht mehr das Arbeiten über einfache Chatbot Apps und das zeigt sich eben in den Nutzerzahlen, welche nun rasant in die Höhe schnellen seitens Open AI, wie du hier siehst, von 1 Million auf 9 Millionen Nutzer innerhalb der letzten 5 Monate und ein Großteil davon, also die letzten 3 Millionen User laut Angaben von Open AI innerhalb von 3 Tagen dann dazu gewonnen und das bestätigt ja noch einmals mehr das, was ich jetzt hier seit Monaten predige, das Agenticoding, der Nummer Skill auch für Nichtentwickler ist, also wirklich für jeden, der heute am Rechner arbeitet und genau dazu wird es auch in Zukunft noch mal eine Wiederholung unseres Liveentic Coding Workshops geben. Corporate LLM ist ja der Kategoriebegriff für solche sicheren Unternehmens KI Arbeitsbereiche, den wir jetzt seit über einem Jahr geprägt haben und Relation Flow, das ist eben unser Corporate LM, welches unter anderem der Marcel Fehl dafüren vorantreibt und da uns eben die Community wichtig ist und auch deine Use Cases, deine Ideen, deine Vorschläge, um das immer besser zu machen, hast du jetzt eben die Möglichkeit, Teil der Relation Flow Community zu werden und lass es mich unbedingt auch in den Kommentaren wissen, ob ich noch mehr solche realen Einblicke und Tests hier innerhalb der KI News oder auch in anderen anderen Videos interessieren. Aber äh dieses Jahr bin ich wirklich überrascht, wie viele Unternehmen sagen, nein, jetzt erst recht integrieren wir KI und wir nutzen es sogar als Vorteil, als Chance, dass wir sagen, wir starten jetzt mit der KI Integration und nutzen eben diese Gelegenheit, wo unsere Wettbewerber teilweise in irgendwelchen Betriebsferien herumhängen, weil sie gar nicht verstehen, was mit KI gerade auf den Markt zurollt. Und da bin ich selbst extrem überrascht, wie viele Anfragen wir gerade haben und wie viele Unternehmen sich jetzt dafür einfach entscheiden, KI zu integrieren, ja, um am Wettbewerb vorbeizuziehen und diesen bedeutenden Zeitpunkt der Menschheitsgeschichte wirklich für sich zu nutzen und fürs eigene Unternehmen zu nutzen und nicht in ein paar Jahren ja beschämt und bedröppelt zurückzuschauen und selbst noch zu sagen, okay, ich wusste von allem, aber ich habe einfach nicht gehandelt und da freue ich mich natürlich ganz besonders, dass wir hier bei Everlast mittlerweile eine solch starke Community an Menschen haben, die sagen, wir nutzen K halt wirklich Gewinn bringt und spielen nicht einfach nur damit herum oder ignorieren es gar und um da jetzt noch mal ein bisschen tiefer einzutauchen, da habe ich die hier schon Video zu aufgenommen, also rund um das Thema Coding, ein ausführlicher Deep von null.

Und genau da brauchen wir Cowork, denn ich erinnere daran, hier im klassischen Chat, wenn du hier einen sehr, sehr langen Prompt schickst, dann wird das Ergebnis selbst, wenn du mit Fab 5 oder einem anderen leistungsstarken System arbeitest, immer viel Zeit in Anspruch nehmen und manchmal muss er einen Teil verarbeiten, um zum nächsten übergehen zu können, sodass das Ergebnis mittelmäßig ist. Jedes Mal, wenn ich im selben Projekt einen neuen Chat starte, denn ich kann in diesem Projekt mehrere Chats, mehrere Sitzungen starten, bleiben die Anweisungen also gleich. sage, dass ich einen neuen Ordner auf dem Computer erstellen kann, hierherkommen kann und einen neuen Ordner anlegen kann. Also, um das Cowork zusammenzufassen, ich kann damit online arbeiten in der Cloudvion im Browser, genauso wie ich auch offline damit arbeiten kann, also in einer lokalen Version, was sehr interessant ist, weil alle Dateien innerhalb des Ordners erstellt werden. Das ist eine kleine Zusammenfassung, um euch den Unterschied zu erklären zwischen dem schnellen Arbeiten mit Cowork hier für eine sofortige Aufgabe, die auf eurem Computer ausgeführt werden soll in einem bestimmten Ordner mit Ausgaben sowie der Möglichkeit, ein Projekt zu erstellen und innerhalb dieses Projekts mehrere Sitzungen zu eröffnen, um immer im gleichen Kontext und mit den globalen Anweisungen an einem bestimmten Projekt zu arbeiten.

LLM Wiki: The Next Evolution of AI Knowledge Architecture Most AI systems today rely on Retrieval-Augmented Generation (RAG), where the model retrieves relevant document chunks every time you ask a question. Instead of retrieving raw documents for every query, it compiles knowledge once during ingestion into a persistent, structured, and cross-linked knowledge base. Key Highlights - Persistent AI-maintained knowledge base - Automatically cross-linked pages - Better multi-source knowledge synthesis - Long-term memory for AI agents - Reduced repeated retrievals - Human-readable and AI-friendly Markdown - Ideal for enterprise knowledge, research, software engineering, and personal knowledge management Rather than replacing RAG, LLM Wiki complements it. A growing trend is to use RAG over the compiled wiki, combining persistent knowledge with real-time retrieval for more reliable and context-rich AI systems. As AI agents become more autonomous, persistent knowledge architectures like LLM Wiki could play a key role in enabling continuous learning instead of starting from scratch with every interaction.

Obsidian Skills (42,361 , by the CEO of Obsidian himself) gives your AI agent hands inside your notes. I plugged them together on my own machine, with real commands and real numbers and this video literally maps itself on screen while you watch. FREE PDF THE ATLAS SETUP Every command from this video, the full loop diagram, the first prompts to run, and the honest caveats. CHAPTERS 0:00 Two constellations (89,956 42,361 ) 1:19 Karpathy s raw folder: the pile problem 2:17 Tool 1 Graphify: the map (god nodes, honesty tags) 3:44 I ran it on nanoGPT the real output 4:23 71.5 fewer tokens (their benchmark) 4:52 THE BRIDGE: graphify export obsidian (92 real notes) 5:34 Tool 2 Obsidian Skills: the hands 6:44 The atlas flags an unclassified node 7:17 The loop: a second brain that maintains itself 9:07 The 15-minute setup (4 commands) 10:31 The finished atlas this video, as a graph STEP-BY-STEP BEGINNER PLAYBOOKS (paid, on Gumroad) Claude Code for Complete Beginners ( 14.97) Codex for Complete Beginners ( 9.97) Claude Cowork: Sales Playbook ( 14.97) Claude Certified Architect Prep ( 29) EVERY SOURCE (check us) Graphify (MIT): Obsidian Skills (MIT): Community combo repo by Lucas Rosati: Obsidian CLI docs: JSON Canvas open spec: Defuddle: Real receipts: we ran graphify on karpathy nanoGPT on our own machine (73 nodes 76 edges 19 communities 92-note vault export). FOLLOW HYPERAUTOMATION LABS ALL THREE YouTube: subscribe, it s free, and it tells the algorithm to send you more Facebook: the Supporter button keeps the lights on Instagram: graphify obsidian secondbrain claudecode knowledgegraph aiagents pkm ai

jetzt Gemini Notebooks für dich interessant ist, dann schreib mir doch mal in die Kommentare, ob du ein eigenes Video dazu sehen willst, indem ich mal einen Deep Dive mache. Eine weitere Neuerung von Gemini Notebooks ist, dass man jetzt direkt über die Gemini App auf dem Handy auf seine eigenen Notizbücher zugreifen kann und auch neue erstellen kann. Da bin ich mal gespannt drauf, wie das funktionieren soll, ob man quasi immer, wenn man etwas sucht, das zu einem Notizbuch hinzufügen kann und so immer mehr Kontext in der Suche liefert. Aber schauen wir mal, was sich Google dabei denkt, wenn es dann auch rauskommt." Ich habe jetzt auch direkt mal in der Gemini App nachgesehen und hier gibt es schon die neue Funktion Notebooks erstellen, aber witzigerweise habe ich einen Google Workspace Account, also einen Pro Account. Man kann bei dem Tool Quellen hochladen und dabei ist das Besondere, dass man eben nicht nur PDFs, Websites und so weiter hochladen kann, sondern auch YouTube Videos.

Also, die geplanten Aufgaben sind wirklich die große Stärke von Hermes und ich wollte heute in diesem Video eigentlich den Lebenszyklus der Aufgaben erklären und vor allem zeigen, wie man sie einrichtet und programmiert, damit sie genau unseren Bedürfnissen entsprechen. Und ihr werdet sehen, das macht wirklich einen riesigen Unterschied zwischen dem Hermes, der diese Dateien nicht definiert hat und dem Hermes, dem man die Informationen gibt. Das ist also die Aufgabe und hier sage ich ihm, das ist die Aufgabe und ich bitte ihn tatsächlich nach Neuigkeiten zu suchen, also jeden Tag zu KI Themen und zur Automatisierung. Also hier hier werden wir also die Anfrage starten und sehen, ob er tatsächlich in der Lage ist, mir das zu erstellen, also den Schädel, die geplante Aufgabe. Jetzt werden wir versuchen, dieses kleine nicht wirklich ein Problem, aber an dieser Stelle müssen wir ein paar Aktualisierungen vornehmen, damit im Docker Compose File hier, also in den Basisinformationen, immer ein Server mit diesem Gateway darin läuft, der ständig aktiv ist.

Google s Open Knowledge Format (OKF) formalizes the LLM-wiki pattern into a portable way to represent agent-ready knowledge. This short explains the key distinction: an LLM Wiki is a workflow for continuously maintaining a living, linked knowledge base; OKF is the interoperable Markdown-and-YAML format that lets that knowledge travel between agents, tools, and teams. openknowledgeformat okf llmwiki aiagents agentmemory knowledgegraph rag aiengineering

Egal, ob du SEO, Geo, AI, SEO für Kundenprojekte durchführst oder die Sichtbarkeit von deiner eigenen Webseite, von deinem eigenen Webprojekt steigern möchtest, heutzutage kannst du mit Cloud große Teile des SEO komplett automatisieren und dabei über 90 der kompletten Arbeitszeit einsparen. Und jetzt sage ich Cloud: Okay, jetzt wo du weißt was hier alles nicht passt, würde ich dich bitten, dass du mir das direkt in der Webseite reparierst mit den Mitteln, die dir zur Verfügung stehen. So, Clud hat jetzt auch noch eine schöne und detaillierte Präsentation erstellt und vielleicht kannst du dich noch an den Anfang des Videos erinnern, wo ich gesagt habe, mit Cloud kannst du deinen SEO automatisieren und dabei bis zu bzw. Jedenfalls sobald du SE Ranking mit Cloud verbunden hast, kann Cloud auf echte SEO Daten und Metriken zugreifen, mit echten Daten arbeiten, deine Webseite analysieren, die Konkurrenz analysieren, jegliche Webseite analysieren. Du könntest theoretisch eine komplette WordPress Webseite include erstellen lassen über Novamira nur durch die Eingabe von ein paar Prompts oder Seiten updaten, neue Seiten hinzufügen oder wie ich es jetzt eben im Video auch gezeigt habe technische SEO Fehler beheben, denn Nova Mira kann auch super mit SEO Plugins in WordPress arbeiten, Josto Rankmaph, aber auch mit anderen Plugins mit Advanced Custom Fields Einstellungen optimieren.

Und ihr werdet sehen, das macht wirklich einen riesigen Unterschied zwischen dem Hermes, bei dem diese Dateien nicht definiert sind und dem Hermes, dem man diese Informationen gibt. Das heißt, ich kann einen Buchhalter erstellen, ich kann einen Entwickler erstellen, ich kann einen Designer erstellen, ich kann jemanden erstellen, der sich um die Strategie kümmert und jeder arbeitet in einer separaten Umgebung. Und genau hier kann man Hermes wirklich einsetzen, damit es ein sehr mächtiges Werkzeug wird, denn im Profil kann ich Unteragenten erstellen. Das bedeutet, dass ich in der Buchhaltung auch mehrere Personen erstellen kann, die verschiedene Aufgaben bearbeiten und genau dadurch kann ich ein vollständiges System haben, das wirklich sehr effizient funktioniert. Also, ich werde Ihnen jetzt den Bereich der Unteragenten zeigen, damit Sie ein wenig den Unterschied verstehen und sehen, wie man zwischen diesen beiden Methoden wechseln kann.

It is technically a client-server architecture, which means if you start something in your herdr session, let's say here I'm starting a PyCoding agent session, and I'm asking it, "What can you do for me?" I can technically quit this session here, and when I restart it, I'm back to where I was, and as you see, it finished its work. It works like this because herdr is using a client-server infrastructure, and when you run herdr, you're connecting with a client to a server that's also automatically started if it's not already running, and your sessions then technically run on that server and that server keeps on running even if you close that herdr pane, if you detach from that session. With this plug-in, for example, I have a new shortcut which opens this file viewer, and now I can see the work my agent did, and I also get highlights about changes, so uncommitted changes, and I can dive into those and take a look and see what changed and then give my agent feedback and so on. And I like that extensibility, that I can make herdr exactly the tool I want it to be, just as I can make Py, for example, as a coding agent, exactly the coding agent I want it to be. Alternatively, I can run it from my MacBook here with herdr dash, dash remote and then the identifier or the IP of my VPS, and then herdr will behind the scenes connect to it via SSH and will still run herdr on the VPS, but for example, with my key binding and settings from the MacBook, which is nice, lets me bring together both worlds.

Set your channel s spec once AI Master runs it Two of Karpathy s projects have been everywhere lately LLM Wiki and AutoResearch. They look like two different tools. They re actually the same method applied twice. This video breaks down the method itself and how it changes the way you think about working with AI, period. Timestamps: 00:00 Intro: two Karpathy projects 00:39 The Spec: writing a recipe, not a wish 04:41 The Verifier: checking output automatically 09:56 The Knowledge Base: how the system gets smarter 12:15 Inside LLM Wiki 13:46 Inside AutoResearch 15:08 Same method, different ratios If you use AI daily and want your results to compound instead of staying a lottery every time, this one s worth watching.

Wenn wir jetzt hier das Projekt einmal entfernen und dann auf neues Projekt und auf vorhandenen Ordner verwenden gehen, dann können wir jetzt hier diesen Ordner Rechnungen auswählen und haben jetzt auch dazu das entsprechende Projekt angelegt. So und um dir einmal zu demonstrieren, wie gut jetzt Chat Team mit diesen Dateien und dieses Zugriff hat arbeiten kann, habe ich jetzt in diesem Ordner Rechnungen verschiedene Dateien angelegt. Schauen wir uns das Ganze einmal an und wir sehen, der Ordner sieht auf jeden Fall schon mal sehr viel sauberer aus als vorher und wir sehen, wir haben jetzt hier auf jeden Fall alle Rechnungen sauber nach Monat und Jahr sortiert. Und was jetzt eben dazu kommt und das ist das Schöne an ChatGBT Work, wir können jetzt hier einfach in den Ordner schauen, wo auch unser Feedback drin war und wir sehen, ChatGBT hat jetzt hier ein Ordner Outputs erstellt und hier sehen wir jetzt auch direkt unsere Präsentation als PowerPoint und die können wir selbstverständlich auch direkt mit der App öffnen. Im dritten Schritt empfehle ich dir also dich mal mit diesen ganzen Plugins von ChatGBT Work vertraut zu machen, einmal zu durchsuchen und zu schauen, ob es Apps gibt, mit denen sich ChatGBT verbinden kann, die in deine täglichen Arbeitsabläufe passen.

Ohne zentrale Wissensdatenbank kann dein Unternehmen bei Ausfällen ins Stocken geraten. Bewirb dich für dein KI-Pilotprojekt: Eine zentrale Wissensdatenbank sichert den reibungslosen Betrieb, selbst wenn Schlüsselpersonen ausfallen. So bleibt dein Unternehmen flexibel und effizient, während du Zeit und Ressourcen sparst. Entdecke, wie ein Logistikunternehmen seine Effizienz steigern konnte und wie auch du davon profitieren kannst. Mehr von tiles: - Website: - LinkedIn: - Instagram: - Facebook: - YouTube: Wissensdatenbank Prozessautomatisierung Effizienz B2B KI

Also, das ist ja eigentlich schon eine irsinnige Vorstellung und die gab es aber immer wieder in der Menschheitsgeschichte, das berühmte Infinite Monkey Theorem, das sagt er im Kern, ich kann einfach Millionen von Affen den ganzen Tag auf der Tastatur herumhämmern lassen von Emil Borell, der hat das 1913 beschrieben und dann habe ich irgendwann die gesamte Bibliothek Nationale zusammen, also die Nationalbibliothek Frankreichs, einfach nur durch reinen Zufall, indem ich Affen ja den ganzen Tag auf einer Tastatur herumhämmern lasse. Und das führt dann dazu, dass man immer mehr und mehr Regeln und immer mehr und mehr Wissen äh diesen KI-Agenten zur Verfügung stellt und im besten Fall dann einfach immer sagt, wenn ein Fehler gemacht wurde, merkt ihr das jetzt, speicher dann eine Markdown File ab und irgendwie wird die KI ja dann schon schlauer werden und beim nächsten Mal das Wissen äh richtigerweise wiederfinden. All das liegt an einem zentralen Ort und darauf können dann wiederum meine KI Agenten zugreifen, meine Apps greifen auf diese zentrale Single Source of Truth, auf dieses Hem zu und der Name impliziert es schon, das ist ein äh lebendes äh System, also die Ontologie eures Unternehmens, wenn man so möchte und kein statisches Konstrukt, auch alle Workflows, alle Prozesse basieren auf diesem Hem, also auf diesem Company Brain ein paar ganz reale Beispiele, weil wir das bei uns schon umgesetzt haben und laufend immer erweitern, z.B. Ja, also wenn unsere Abteilungsleiter im Marketing, im Vertrieb sagen, ich brauche jetzt mal ein eigenes Dashboard, ich muss einen Überblick bekommen über meine Vertriebs KPPI, dann können sie einfach zugreifen auch als normale Mitarbeiter auf dieses Single Source of Truth und sich ein eigenes Dashboard bauen oder auf die bereits vorhandenen Dashboards zugreifen und müssen sich nicht erst die Daten aus irgendwelchen Excel Tabellen oder Markdown Files zuliefern lassen. genauso diese ganzen ähm Agent Hness Konstrukte, die könnt ihr genauso ignorieren und das was wir bauen mit Corporate LM, das ist im Wesentlichen genau diese Vision, also so ein company Brain zu schaffen, dass es eben nicht jetzt jeder selbst tun muss, sondern dass ich ein System habe, indem ich mein Company Brain anlegen kann.

Community: Karpathys Github Gist: Googles Open Knowledge Format: Ist das LLM-Wiki die bessere Alternative zu klassischem RAG? In diesem Video erfährst du, wie das Konzept von Andrej Karpathy funktioniert, wie du mit einem KI-Agenten, Markdown und Obsidian deine eigene Wissensdatenbank aufbaust und welche Rolle Googles Open Knowledge Format dabei spielt. Außerdem vergleichen wir die Stärken und Schwächen von LLM-Wikis und RAG-Systemen und klären, für welche Anwendungsfälle sich welcher Ansatz eignet. 00:00 Intro 00:36 Was das LLM-Wiki lösen will 03:14 Googles Open Knowledge Format 04:25 Tutorial: Eigenes LLM-Wiki bauen 12:50 Anwendungsfälle für das LLM-Wiki 13:59 Kritik am LLM-Wiki 16:00 Mein Fazit zum LLM-Wiki KI LLM RAG Obsidian llmwiki

Und ihr werdet sehen, das macht wirklich einen riesigen Unterschied zwischen dem Hermes, bei dem diese Dateien nicht definiert sind und dem Hermes, dem man diese Informationen gibt. Also sage ich ihm ganz einfach, ich werde diese Information auswählen und ihm sagen, dass er diesen Teil aktivieren soll und dass er einfach selbst diese Recherche durchführen soll, die mir diese Tabelle ausgibt. Also irgendwo haben sie Hermes für ihre eigene Maschine, für ihr eigenes Unternehmen verwendet und irgendwann haben sie tatsächlich die Möglichkeit einfach auf Hermes zu gehen und Hermes darum zu bitten, ihnen das zu erstellen, was man nennt die Skills, die Kompetenzen. Also, es ist eine einfache Erstellung und wir machen es einmal, wir bearbeiten es nur einmal und danach kann man sehr schnell weitermachen, denn es ist eine Datei, die jeder herunterladen kann, sobald er bezahlt hat. Also diese Datei, ich denke, es wird ein paar Minuten dauern, bis ich das Ergebnis bekomme, aber ich habe viel mit den Kompetenzen gearbeitet, ich habe andere Kompetenzen getestet und ich habe sehr viel ausprobiert.

Das ist jedoch ein Irrtum, denn heutzutage, wenn man Telegram mit Hermes verbindet, erhält man ganz einfach vollständigen Zugriff auf Hermes. Und ich wollte euch in diesem Video einfach zeigen, welche Vorteile es tatsächlich hat, wenn man mit Telegram auf Hermes arbeitet und vor allem, wie man es korrekt über die Weboberfläche von Hermes verbindet. Und ihr werdet sehen, das macht wirklich einen riesigen Unterschied zwischen dem Hermes, der diese Dateien nicht definiert hat und dem Hermes, dem man die Informationen gibt. Und selbst wenn ich also nicht vor dem Computer sitze, sondern im Fitnessstudio bin und Hermes etwas fragen möchte, öffne ich einfach Telegram und schreibe ihm direkt. Ihr wisst ja, bei Telegram reicht es einfach Telegram.org org einzugeben und ihr könnt Telegram ganz einfach auf dem iPhone, auf Android oder auf jedem beliebigen Handy herunterladen.

Denn wenn wir sie in verschiedenen Bereichen einsetzen und genau deshalb ist Hermes heute nicht einfach nur ein Tool, dass man installiert, um es zu testen, nein, wir integrieren es ins Unternehmen und versuchen es zu nutzen. Entweder hat man die Möglichkeit direkt mit Hermes zu arbeiten, das werde ich Ihnen in diesem Video demonstrieren, um Hermes selbst zu jemandem zu machen, der lernt und ihnen hilft, Kompetenzen zu entwickeln. Und ihr werdet wirklich sehen, das macht einen riesigen Unterschied zwischen dem Hermes, bei dem diese Dateien nicht definiert sind und dem Hermes, dem man die Informationen gibt. Ganz einfach, das ist ein Dokument, indem wir Hermes Mess zeigen, wie man etwas macht, obwohl es stimmt, dass Hermes Mess er kann sie kennen, um Aufgaben auszuführen, die ich ihm auftrage, da er mit einem LM Cloud oder anderen verbunden werden kann. Das ist ein kleines Szenario, das ich mir ausgedacht habe, und deshalb, wenn ich dann, wenn man sich hier befindet, dort wo man ihn bittet, etwas zu erstellen, ist es so, als würde ich in der ersten Methode Herr Mess einfach bitten, mir die Fähigkeit zu erstellen.

Second Brain mit Obsidian und Claude Code: Ich habe mir eine lokale Web-App gebaut, die mein gesamtes Wissen als interaktive Karte zeigt fast 2.000 Notizen, über 4.000 Dateien, durchsuchbar in Sekunden. In diesem Video zeige ich dir mein komplettes Second Brain System gebaut mit Claude Code (Fable 5), ohne selbst eine Zeile Code zu schreiben. Das siehst du im Video: - Warum der Obsidian Graph am Haarball-Problem scheitert - Cloud-Ansicht: Themen-Cluster mit Struktur statt Punkte-Chaos - Suche mit Kameraflug: Notizen, Wikilinks und PDFs direkt im Graph lesen - Ring-Ansicht: aktive KI-Skills, Memory, Routinen, Plugins und Connectors auf einen Blick - Globus-Ansicht: das gesamte Wissen als drehbare Kugel - LLM Wiki von Karpathy: Claude pflegt mein Wissen selbst liest Quellen, verlinkt Seiten und markiert Widersprüche, statt sie zu überschreiben Alles läuft lokal auf meinem Rechner: kein Abo, keine Cloud, meine Daten bleiben bei mir. Kostenlose Ressourcen: Claude Masterclass (kompletter Kurs, kostenfrei): Prompting Guide als Gratis-Download: Video: Prompting für die aktuellen KI-Modelle: Kapitel: 00:00 Mein gesamtes Wissen auf einer Karte 00:27 Das Problem mit dem Obsidian Graph View 01:41 Was ist mein Second Brain? 02:39 Cloud-Ansicht: Struktur statt Haarball 03:30 Suche und Inspector: Finden und Lesen ohne Umweg 04:46 Ring-Ansicht: das komplette System auf einen Blick 06:09 Globus und KI-Wiki 06:55 Vorher gegen Nachher 07:57 Wie du das nachbaust Wenn du sehen willst, wie ich das System mit Claude Code gebaut habe von der ersten Idee bis zur fertigen App schreib Second Brain in die Kommentare.

Wenn du Hermes noch nicht installiert hast, ist das das beste Video, das du heute sehen wirst, denn wir werden Hermes wirklich auf einfache Weise nutzen, seine Komponenten verstehen und lernen, wie man diese Komponenten tatsächlich bedient. Und es ist sehr wichtig, ihre Definition und ihren Zweck ein wenig zu verstehen, denn heute kann Herr Mess uns ganz einfach ermöglichen, mehrere Profile zu erstellen und jedes Profil hat automatisch seine eigenen Dateien wie diese. Also ich erinnere am Ende daran, dass all das, also Identität, Gedächtnis, Sitzungen, Fähigkeiten, man sie in einem einzigen Profil finden kann. Es ist also so, als würde ich ihm das Gedächtnis auffrischen oder ich füge Daten oder Informationen ein, die ich anderswo erstellt habe, damit er hier im Inneren damit weitermachen kann. hierher und ich sage ihm: "Hör zu, ich möchte, dass du dir diese Informationen merkst und sie in deinen Speicher aufnimmst." Also werde ich hier einfach diesen ganzen Block kopieren und ihn darum bitten.

Your AI second brain has a problem: your agent can open anyone s LLM wiki, but it can t find anything in it - because there s been no standard for how these wikis are built. Google s new Open Knowledge Format fixes that with just markdown, folders, and one required field, and in this video I show you how it works and how to adopt it with literally one prompt. Karpathy s LLM wiki- Google s OKF blog - Google s OKF github - Google s OKF spec.md file - ai

Und genau deshalb installieren viele Leute Hermes, probieren es aus und nach ein oder zwei Wochen sagen sie: "Hör zu, ich arbeite lieber mit Chat GPT oder Cloud, das ist einfacher." Und genau das ist der Fehler, denn wenn du die Funktionen nicht einstellst und den Unterschied zwischen diesen Funktionen nicht verstehst, ist es ganz normal, dass Hermes einfach nur ein gewöhnliches LM bleibt. Also, was ich jetzt machen werde, ich werde euch hier einfach alles erklären, euch die verschiedenen Funktionen zeigen, den Unterschied, wie man sie auswählt, warum man sie auswählt und in welchem Kontext, vor allem im beruflichen Kontext. Als erstes werde ich euch jetzt die Bedeutung dieser drei Dateien erklären und ich werde einfach direkt Prompts in mein Hermes eingeben, um ein bisschen zu sehen, wie es reagiert. Also Hermes, wenn du Hermes startest, dann kann er im Grunde einfach, wenn du möchtest, diese drei Dateien lesen, wobei jede Datei eine bestimmte Funktion hat. Und es ist sehr wichtig, dass standardmäßig, wenn man eine Sitzung startet und mit Hermes sprechen möchte, standardmäßig diese Datei geladen wird, um ein wenig die Persönlichkeit zu verstehen oder einfach die Art und Weise, wie er dir antworten wird, zu erfassen.

Ich bin nicht nur Betreiber des größten deutschsprachigen YouTube-Kanals Everlast AI rund ums Thema KI im Geschäftskontext, sondern betreibe mit Everlast AI selbst die marktführende KI Beratungs und Implementierungsagentur mit über 1 Million Euro Monatsumsatz über 3500 Kunden aus dem Mittelstand, die wir die letzten Jahre betreut haben, über 6500 KI Projekten, die wir in diesem Kontext umgesetzt haben und hunderten Unternehmen aus dem Mittelstand, mit denen wir jede Woche sprechen und daher genau wissen, was der Markt gerade möchte, was die Unternehmen wirklich beschäftigt und mehr Einblicke als jede andere Agentur in das haben, was gerade in Deutschland wirklich funktioniert und was gefragt ist. sind heute strategischer Partner 100erter KI Agenturen im Hintergrund, mit denen wir das alles schon umgesetzt haben und die gerade zum ganz großen Teil schon mit diesem Use Case im Hintergrund arbeiten und dadurch natürlich den massiven Wettbewerbsvorteil haben gegenüber allen, die irgendwelchen Businessgurus nachrennen, die selbst ja immer z d Jahre den Trends hinterher rennen, weil sie selbst KI nicht an den Mittelstand verkaufen. 99,2% aller Unternehmen, 53% aller Beschäftigten, 69% aller Ausbildungsplätze kommt vom deutschen Mittelstand und der Erfolg dieses Mittelstands, der basiert auf tiefem über Generationen verfeinerten Spezialwissen, also diese sogenannten Hidden Champions, also jedes Unternehmen aus deiner Kundschaft ist davon betroffen. Wieso verschwenden Unternehmen heute immer noch Jahre teilweise mit der Einarbeitung neuer Mitarbeiter, wenn man doch das gesamte Wissen, das Führungskräfte haben aus der Einarbeitung für KI optimiert aufbereiten könnte, sodass neue Mitarbeiter mit dem Wissen chatten können, sich natürlich Videos anschauen können, Fragen stellen können mit KI und sich von KI einarbeiten lassen können. Ihr beschleunigt euer Onboarding, ihr entlastet eure Führungskräfte, ihr sichert euer wertvolles Wissen im Unternehmen, das in einer eigenen Infrastruktur, in eigenen Wissenspeichern, die nicht irgendwo in den USA liegen, die ihr sogar selbst auf euren eigenen Servern betreiben könnt, auf die ihr mit Desktop Anwendung lokal, offline mit lokalen Modellen zugreifen könnt, weil das ist ja auch das geniale.

Hier haben Sie tatsächlich ein wenig das Problem der Langsamkeit mit WhatsApp behoben, denn Hermes kann man tatsächlich mit verschiedenen Tools oder Kommunikationssystemen wie WhatsApp, Telegram, Slack bedienen und man kann eine Nachricht senden, damit Hermes diese Nachrichten ausführt. Also Vorsicht, wenn man Hermes installiert und nur die Verbindung mit Telegram oder WhatsApp nutzt, verliert man ehrlich gesagt eine Menge anderer Funktionen. Das ist eine Version, die das klassische Hermes enthält, aber darüber hinaus gibt sie Ihnen tatsächlich Zugang zu weiteren Funktionen und sie können Hermes auf eine viel flexiblere Weise nutzen und wirklich Szenarien konfigurieren und sehr einfach arbeiten. Ich kann an den Profiten arbeiten, an den Projekten arbeiten, Notizen machen und das ist eigentlich ein System, das mir sogenannte Festplattenspeicherbereiche gibt und ich kann damit viel viel einfacher und nützlicher arbeiten. Man sollte jetzt die allerneueste Version finden, also für den Agenten die Version 0.12.8 8 und für die Weboberfläche die Version 0.52.

Heute verwenden wir nur die offizielle Dokumentation von N8N, denn dank dieser, dank des MCP, dank der Kompetenzen versteht man natürlich den Unterschied und ich muss das EN8N Tool nicht mehr öffnen und das macht Hermes sehr mächtig. Gut, also was ich jetzt machen werde, ich gehe hier zu Hermes und werde ihn einfach bitten, mir die Verbindung zu N8 NMCB Server herzustellen. Ich werde ihm einfach einen Befehl geben, damit er n8 NMCB installiert und dann sage ich ihm, dass er einfach diese Konfiguration in genau diese Konfigurationsdatei einfügen soll, was sehr wichtig ist, oder? Es handelt sich also um ein Verzeichnis, das verschiedene Informationen enthält, die uns einfach die notwendigen Skills liefern, um den Agenten auszuführen und ihm die maximal notwendigen Informationen zu geben. Mit also also Chat JPDV, dem ich tatsächlich gesagt habe, er soll mit Open Irray arbeiten und da schickt er einfach die Information an Telegram.

Hermes WebUI (code : GOHERMES) : Serveur n8n illimité (code : GON8N) : Documentation : Accès à mes formations (code : LANCEMENT) : J ai connecté HERMES à n8n et il a créé mon workflow tout seul Dans cette vidéo, je vous montre étape par étape comment utiliser HERMES avec MCP n8n pour créer automatiquement un workflow n8n grâce à l intelligence artificielle. Vous allez découvrir comment installer HERMES WebUI sur un VPS, installer n8n gratuitement, connecter le MCP n8n à HERMES, ajouter les Skills n8n, configurer les instructions dans le fichier agents.md, puis lancer la création automatique d un workflow complet directement depuis HERMES. L objectif est simple : utiliser un agent IA capable de comprendre votre demande, générer un workflow n8n, l organiser visuellement, ajouter des stickers, puis le tester directement dans votre instance n8n. Si vous voulez apprendre à créer des automatisations IA, construire des workflows n8n plus rapidement, utiliser MCP avec n8n, ou découvrir comment HERMES peut vous aider à automatiser vos tâches sans coder, cette vidéo est faite pour vous. SOMMAIRE : 00:00 - Introduction : Créer des workflows n8n avec HERMES et MCP 04:18 - Télécharger ma documentation complète HERMES pour n8n 04:57 - Plan complet de la formation HERMES, MCP et n8n 08:01 - Installer HERMES WebUI sur un VPS étape par étape 10:34 - Installer n8n gratuitement sur votre serveur 11:45 - Connecter le MCP n8n à HERMES facilement 18:12 - Ajouter les Skills n8n dans HERMES 20:23 - Configurer mes instructions dans le fichier agents.md 23:05 - Lancer la création automatique de mon workflow n8n 24:55 - Accéder à mon workflow depuis mon instance n8n 27:44 - Ajouter des stickers et organiser visuellement mon workflow 30:23 - Tester mon workflow directement dans n8n 33:20 - Mon cadeau pour accéder à mes formations complètes HermesAgent n8n HERMES MCP AutomatisationIA WorkflowN8N NoCode

If you want to take this a step further, I'd even recommend using like an Open Claw as your backup account, too, cuz then you can have them not only watch over each other to make sure they're up, but then you can also take advantage of any features that one or the other comes out with that the other is not using. You can easily go in and just say, "Hey Hermes, go over to this other device and do something." So, if I want to say, "Let me know which local models I have running on the DGX Spark." I can hit enter on that and now my Hermes agent will go from the Mac Studio it's on right now, go over to my DGX Spark, which is also on Tailscale, and be able to tell me what's running on there. It then takes all those things and then figures out what it can take off my plate for me, what tasks it can do for me, what it can automate, what it needs my permission to do, and starts doing those things for me. Every day I learn more and more about what Hermes agent can do for me, and and what it can take off my plate. All you need to do to see this Hermes Kanban board is go in, type Hermes-dashboard into your terminal, whether using Ghosty or just your built-in terminal on your computer, type Hermes-dashboard.

Das heißt, die öffnen hier das Terminal und installieren dort Hermes mit einem Command, den findet man auch auf der Hermes Seite und damit läuft Hermes dann auf Rootebene und kann den gesamten Server steuern, kann sich selbst konfigurieren, Paket installieren und so weiter. Aber ich persönlich denke einfach bei einem Agenten, der 247 läuft, der auf sehr viele verschiedene Systeme zugreifen kann, dem möchte ich eigentlich nicht vollkommene Adminrechte geben, denn im schlimmsten Fall, wenn Hermes jetzt irgendwie über eine Prompt Injection vielleicht angegriffen wird, könnte er theoretisch auch einfach anfangen Dinge zu löschen, aber in so einem Container ist der Schaden eben sehr begrenzt. Das heißt, falls es mal Probleme geben sollte mit Hermes z.B., Spel, der jetzt in diesem Container läuft, falls er irgendwas nicht installieren kann, dann kann ich eben Cloud Code nutzen oder gegebenenfalls auch Codex und der läuft dann auf Root Ebene und kann den Server entsprechend verwalten und Fehler beheben. Und ich habe dir tatsächlich einen Prompt vorbereitet, den findest du in der Videobeschreibung in meinem Download Hub, den du hier einfach nur an Codex geben musst oder Cloud Code oder auch Hermes, wenn du den jetzt hier auf Rootbene installiert hast und der wird das gesamte Setup mit dir durchgehen, sodass du danach Tayscale laufen hast und all deine Geräte miteinander verbunden sind. Und ich kann jetzt einfach den Link hier zu diesem X Post kopieren und dann hier zu Hermes gehen und einfach sllearn eingeben und hier steht dann learn a reasonable skill from anything you describe und dann kann ich einfach nur diese URL eingeben und jetzt wird Hermes selbständig lernen, was die best practices sind, um mit Fable 5 zu arbeiten.

Ja, wir haben bis zu 120.000 € Ausfallkosten pro Stunde. Das Nachbarwerk in Tschechien hat schon dicht gemacht. Glaubt ihr wirklich, dass wir hier mit diesen enormen Ausfallzeiten wirklich den Laden hier noch halten können? Das heißt, entweder wir machen jetzt was oder es wird uns halt so gehen wie den Kollegen in Tschechen und der Laden ist hier in dem Jahr dicht. Das natürlich Pistole auf die Brust Last Resort, aber ich glaube, wenn wir uns die wirtschaftliche Lage mal anschauen, oft ist, wenn ich sag, okay, ich habe ein KI Assistent, der mir diese 120.000, wenn es nur 20.000 1000 € ist, den ich den ich durch kürzere Stillstandszeiten ermögliche, dann muss ich das auch der Belegschaft gegenüber machen, um die Arbeitsplätze zu sichern.

FREE RESOURCES (prompt pack brand voice template) Higgsfield MCP Claude Code Masterclass JARVIS AI Assistant Blotato MCP I automated my ENTIRE social media with Claude Fable 5 it researches my niche, writes in MY brand voice, designs the visuals with the Higgsfield MCP, and posts to every platform through the Blotato MCP. This is a full Claude Fable 5 tutorial for social media automation no code, no n8n spaghetti. You ll build a brand voice file (the secret that makes AI posts sound like YOU, not a robot), connect the two MCP tools, and watch Claude write, design, and schedule a real carousel that posts LIVE on camera. WHAT YOU LL LEARN How to automate social media posting with Claude Fable 5 MCP tools (Blotato Higgsfield) The brand voice file make AI write in your voice with your stories AI-generated visuals, carousels, and platform-tailored captions (IG, X, LinkedIn, TikTok, FB, YouTube) The weekly 15-minute routine that keeps it running on autopilot Free prompt pack brand voice template to copy my exact system CHAPTERS 00:00 Meet the Social Media Autopilot 00:27 Install Claude Desktop (2 Minutes) 01:17 Get the Free Prompt Pack Templates 02:19 Build Your Brand Voice File (the Secret Sauce) 05:38 JARVIS Demo: My AI Assistant Joins In 08:33 Finalize the Brand Voice File 09:46 Connect the Higgsfield MCP (AI Visuals) 14:03 Connect the Blotato MCP (Auto-Posting) 17:33 First Test: AI Creates a Full Carousel 19:48 Proof: It Posted LIVE on a Real Account 23:11 How It Tailors Posts for Each Platform 25:09 Schedule the Whole Week (15-Min Routine) 26:24 Troubleshooting What to Automate Next I m Zubair I build AI systems with Claude Code and show you how to turn them into income: real builds, real demos, and the exact systems inside the AI Workshop. ClaudeFable5 SocialMediaAutomation ClaudeCode Chapters: 00:00 Meet the Social Media Autopilot 00:27 Install Claude Desktop 01:17 Download The Free Resources 02:19 Build Your Brand Voice File 05:38 Jarvis Second Brain Demo 08:33 Finalize The Brand Voice File 09:46 Connect Higgsfield MCP 14:03 Connect Blotato MCP 17:33 Run A Carousel Test 19:48 Proof The Post Went Live 23:11 Tailor Content For Each Platform 25:09 Schedule Weekly Posting Routines 26:24 Troubleshooting And Final Setup Tips

Now, before we ask Fable to build an even better Fable, there's something that you have to understand before any of this is going to make sense to you, and you're going to want to make sure you follow all these five levels in the video. So, people will get like They're going to say, "Hey, why is this person so freaking productive?" Here's the thing about Fable 5 again, is it every dollar you spend or every token that you use gets you more compared to Opus 4.8, GPT 5.5, and Gemini, all the models in fact, it is the world's best model on the planet available right now, subject only to models that have not been released. And then all you're going to do is literally come down here, grab this prompt like so, I'm going to straight over to Fable 5, and Fable 5 will interview you and give you the most comprehensive breakdown that you've seen about your life. Now, if you want to actually build some websites, I'll put a link here that shows you the full breakdown of how you use Fable 5 to do that so you can get flying. Now, leveraging this incredible model is one thing, but there's a big problem, and that's the fact that if we don't have the right systems around the correct operating systems around it, we're not going to be able to get the best value out of it, which is why the next thing that we need to do is learn how to build one with agentic operating systems and understand how they work, which we're going to cover in this video right here.

In today's video, I'll show you what is Claude Fable 5, how to prompt Fable 5 effectively, all the crazy things you can build with it, and we'll actually test it out live and build something cool. Now the real trick when using claude fable 5 is to use fable for the planner and then using opus or sonnet for writing all the code that you want to. So what people do is fable 5 max reasoning as the orchestrator opus for deep reasoning sub aent task and sonnet for mechanical work and then you just set it up you create multiple agents and you use that in claw code to create any app that you want to and this will save you as many tokens as you want. You just say forward/fable prompt and then it'll ask you questions for what you want to build and it will ask you more and more questions about how do you want to build it. So I can see in the last 28 days, how many views I've had, how many reach do I have for YouTube, what are my metrics, what videos have been working really well for me, and I can see what reels are performing well for me as well.

In this breakdown we cover exactly how Cognee works, from its ECL pipeline and four core memory operations, to its architecture that collapses a four part infrastructure stack into one Postgres instance. We also go deep on how it benchmarks against other memory systems, how it integrates with Claude Code and OpenClaw through official plugins, and how it stacks up against Mem0 and Zep in terms of what each tool is actually built for. Cognee GitHub Cognee Website Host Cognee on Your Own VPS ---------------------------------------------------------------------------------------------------------- Timestamps: 00:00:00 - The Problem: AI Agents Forget Context 00:01:20 - Introducing Cognee: Obsidian-Style AI Memory 00:02:54 - Why AI Agents Need Better Memory 00:04:16 - The Failure of RAG (Retrieval Augmented Generation) 00:05:05 - The Obsidian Analogy: Why Linking Matters 00:06:30 - What is Cognee? 00:07:13 - Academic Backing and Traction 00:09:24 - The Mechanism: Extract, Cognify, Load (ECL) 00:11:25 - Core Developer Actions: Remember, Recall, Forget, Improve 00:13:06 - Infrastructure Advantages: Single Postgres Instance 00:15:45 - Integrations: Python, Rust, TypeScript, and Claude Code 00:18:22 - Benchmarking Cognee: Beam and Hotpot QA 00:21:14 - Side-by-Side Comparison: Obsidian, Mem0, and Zep 00:24:54 - Cost and Open-Source Value 00:26:08 - Potential Limitations 00:27:05 - Final Verdict: Is Cognee Right for Your Agent? ---------------------------------------------------------------------------------------------------------- USEFUL TOOLS SERVICES: FREE 50 Pinterest Canva Templates - PromoPDF AI - Systeme.io - Shopify - ---------------------------------------------------------------------------------------------------------- Follow us: Youtube - Pinterest - Buy me a Coffee - ai aiagents obsidian Affiliate Disclosure: Please note that some of the links in this video description may be affiliate links.

So, some of what we're seeing with the different colors, we can see which node is actually producing what specific frame, and right there we get a speed up of 3.7 times with 93% efficiency, and this is just a replay of the versus for one worker versus four workers, and we also have a chart right here just showing the speed. So, we can see some more, and I purposely asked this like, "Okay, can you give me a simple YouTube intro in exactly what's going on here and why this is cool?" Because it's not not visually exciting, so I think it's better to just have some real like meat to what's actually going on. Again, I understand this is not as visually exciting as like playing a cool game, which we will definitely be doing in this video, but I wanted to start out with this because this video was supposed to be out yesterday, but I got so enamored and sucked into this that I spent the entire night playing with it, and I wanted to just showcase it cuz I think it's really cool, and I believe that this may actually be an interesting benchmark for models in general, where you give them the basic setup, okay, you have this Arduino boss, and then you have four worker nodes, and you need to essentially build a distributed system for them where the work can be created to do this ray tracing task across four distributed workers, and then see if models can optimize it further and stuff like that. So, all right, I'm going to stop this now because for the purpose of this demo is probably done, but I just wanted to showcase a bit more about this cuz I think it's kind of cool and just like a different way of using AI for something. It's the entirety of the package that was put together with the sound design, the characters, the speech, some like cleverness, the intro scene, the changing music depending on year, like just like like mine the 63rd theorem, like stuff like that.

I got roasted in the last video for not understanding procedural generation, but like I was so used to AIs at that point just making like a very simple arena. I can see like a really cool like video game world being made. And you know, let me know if there's actually like any games like this, but as far as I know, this is an original game and it's playing pretty well. I I I don't know if that's too quiet for you guys to hear, but it did like a little fake talking sound like that. I'm not used to Fable 5 talking about so many bugs, but it does seem like it just fixes things like really quickly and without issue.

My playbook for growing a 1M AI agency: My FREE resources: My Tools FREE MONTH voice to text: Code NATEHERK for 10 off VPS (annual plan): Karpathy LLM Wiki Gist: I ingested all my YouTube videos into an LLM wiki and turned them into a connected second brain that my AI OS can actually reason over. In this one I show you how to build the same thing in about five minutes using Claude Code and Obsidian, based on Andrej Karpathy s LLM knowledge base idea. You drop in sources, the AI reads them, splits them into cross-linked wiki pages, and keeps the whole thing organized with routing rules so it can find anything fast. By the end you ll know how to set up the vault, write the schema, ingest a PDF and a URL, and decide when to keep your wiki flat versus structured. Sponsorship Inquiries: nate smoothmedia.co Connect with me: TIMESTAMPS 0:00 The LLM Wiki Demo 1:13 What Fable Does With the Data 2:58 Multiple Wikis in My AI OS 5:09 Where This Started Obsidian Setup 7:05 The Setup Prompt 7:59 Flat vs Structured Wikis 9:49 Ingesting Two Sources 12:38 Why It Works: Routing 14:18 Final Thoughts

Claude Fable 5 just came back today, so you know we have to make a video comparing Claude Fable 5 to the new Chinese GLM 5.2 model that is said to be better than Claude Fable 5. So, first let's go to Claude Fable 5 and let's prompt it to make Terraria from complete zero. I mean like, in my opinion, I think Claude might be winning this challenge cuz GLM 5.2 has a very high context window, but Claude Fable 5, bro, [music] it just does tasks so well. [music] And let's put this to full access, and let's paste in the same exact prompt to make Terraria from complete zero. So, I just put here again, "Make Red Dead Redemption 2 from complete zero and have most of features from RDR2 and make [music] this prompt better and plan it out for coding." So, let's send it and I'll come back when it has a complete version of RDR2.

But if all that's too complicated and having it call out all these agents, you can do something as simple as simply throwing it in plan mode, you know, having it come up with the plan, having it create some sort of markdown file, set the stage for your code base, and then simply spinning up another session with Opus and having it execute the plan that Fable laid out. But, the real point here is I want to use a lower-level model like Opus for deep research because research isn't something that requires like super high intellect reasoning level like Fable. Now, Anthropic hasn't put out any official numbers of what this looks like with fable as the advisor and having Opus be the executor or Sonnet, but we can make a few assumptions. Now, to actually use advisor in this way, you can't have your model set to fable five because whatever model you have set, that is the model that is the executor. So, if I want fable five as the advisor, and I want Opus actually doing everything, then I need to make sure my model is set to Opus.

But slowly, while playing around and then working with it, I started thinking, "Hey, these splits are actually nicer than what I currently have in tmux." And then the mouse control and the side panel and notification, it just kept telling me, "Hey, I'm a tmux rewriting rust that's a little bit more user-friendly, and you what? And there we have it, my familiar shell in a very familiar setting on top that resembles the famous catpuccin plugin for tmux as you can see from my Tmux line on top, which by the way is my setting and seems like Herder has shifted my way with the bar on top rather than the usual Tmux default on the bottom. I can run top in the first and switch to the other one, prefix and W for workspaces, shift the focus on the TUI to the sidebar, where you can use the arrows to switch and enter to attach. There's tabs and splits as any other Tmux or Zellij user knows, and there are sessions, none of which are workspaces, which you can think of as an in-between layer, letting you containerize a bunch of windows and splits without having to detach and attach from running to and from running sessions. Herder runs and now I have Herder inside Herder inside Tmux, which is a vortex of issues you'd probably want to avoid.

Dieses System gibt mir also sehr leistungsstarke Modelle, die ich vor allem unbegrenzt nutzen kann. Dass OLA Modelle im Cloudmodus bereitstellen wird, bedeutet, dass sie auf dem Olama Server ausgeführt werden, was sehr schnell sein wird, sodass Hermes laufen kann, ohne dass ich für den Verbrauch Tokens bezahlen muss, wie es bei Cloud oder GPT der Fall ist, die sehr teuer sind. Also gehst dich nach oben und klicke auf deploy, um also einfach auf Deploy klicken, also das Projekt bereitstellen. Das bedeutet, dass hier schauen Sie, das ist normal, also habe ich nur das 1 und selbst hier, wenn ich zu Provider gehe, also ich aktualisiere tatsächlich, um zu sehen und Kim wird noch nicht angezeigt. Genau jetzt starte ich also Hermes neu und danach kann ich überprüfen, ob Hermes tatsächlich läuft, also ob Kimy verfügbar ist und wir werden anschließend zur Benutzeroberfläche gehen, um zu sehen, ob das angezeigt wird oder nicht.

If something failed, say [music] so." Secondly, state your boundaries because sometimes it can take actions you never really asked for. Like if you ask it to look something over, it might rewrite the whole thing, even draft an email you never asked for. So, Fable 5 now has new safety classifiers that block more cybersecurity tasks, which also means near-term some routine stuff like coding and debugging will still be handled by Opus 4.8. And Anthropic is also scaling up its collaboration with the US government with pre-release access to models, info sharing on jailbreaks, along with joint research on safeguards. So, a lot of future models might go through the government before they reach you, and I think Fable 5 is just the first of many.

So, in this video I'm going to show you step-by-step how to do this, but on top of that I'm going to give you this one prompt website pack that you can basically just copy-paste depending on whatever niche or whatever you want to build there. We're going to be using this all inside our Claude code and we're going to connect this to a model like C Dance 2.0 and Nano Banana or GP2 image, so that way we can create everything within Claude code. So, the way to add this, you're going to click on the plus button and instead of if you click on browse connectors, it's going to show you all of the native ones that are here, but we're going to add a custom one. So, same thing like if you're creating an e-commerce website, you can just copy this and then you can take it to Claude code and say, "Hey, create something similar to this website." So, that's kind of one way to kind of clone a website, right? So, if you're uh creating a portfolio website for instance like this one, you can certainly put this and have Claude Fable say, "Hey, go ahead and now post this or uh publish this into a particular custom domain." So, this is something that you can either do it for yourself, and like I mentioned, if you are an AI agency owner or somebody who's offering this as a service, this is an incredible opportunity for you to take advantage of this and sell these as services, right?

Now, every time you create a new character, location, or prop, you need to save it as an element, so Claude can automatically reference those images in the prompt. For that, I'm uploading all the assets I've created into Claude and I'm going to say, "Write me a prompt for a video." I'm describing all the assets that I have and also the first scene. Okay, now let's copy the prompt and try making the character sheet in Soul Cinema, because it's the best one right now at generating characters from scratch. Now, let's take the new prompt and run it in C Dance, and let's see what we get. All right, now let's load all the assets into Claude, the location, the elf, the monsters, and ask for a prompt.

There's a word that's starting to get used for the people who get cut off from what everyone else can access, [music] an underclass, the ones left behind because they either never learn tools like this or never [music] got access to them. So, I'll show you why it got banned and why it's suddenly back and hand you the [music] one official method Anthropic recommended for prompting it to get the best output. You'll see it start listing out steps for itself, almost like a to-do list, then working through them one by one. Engine sounds, animations, the works, a real playable game from one prompt we didn't even wrote by hand. Now, the third thing, the one that isn't a game and the one I was most curious about.

Dieses Video von "Build an AI Website Chatbot with Memory in n8n | Google Sheets + AI Agent" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

You can now hook your up to live market charts, describe your trading strategy in plain English and Fable five writes the [music] actual bot for you. You can just screenshot a website like GitHub, paste it into Fable five and it builds you a working replica from the image alone. You can take every scattered note you've ever written, connect it to Fable five and it links everything together, fills the gaps and keeps itself updated on its own. Instead of asking Fable five one question and closing the tab, you give it a goal and a way to check its own work. I put the full step-by-step process for every use case and the exact prompts for all five into a free [music] guide.

So und da Fable 5 jetzt vor allem bei den Cyburity Fragen jetzt noch etwas durchlässig war, haben sie die Safeguards noch mal angeschraubt und im einpost sagen die jetzt vor allem auch some routine tasks like coding and debugging will fall back to Opus 4.8. So und ich würde sagen, wir testen das Ganze jetzt einfach mal und ich stresst teste jetzt Fable 5. Wir gehen in den nächsten Test und zwar versuchen wir jetzt mal Fable dazu zu bringen, diese Seite auch abzusichern mit einer Login Funktion, mit Passwort Hashing, Schutz gegen Brotforce und so weiter, um einfach zu schauen, ob da vielleicht schon der Cyburity Filter anspringt. Wenn ich jetzt noch mal in meine Nutzungslimits schaue, sehe ich, dass wir gar nicht mal so viel verbraucht haben dafür, dass wir eine komplette Landing Page generiert haben vorhin mit Videos und Bildern und so weiter mit Login Funktionen, Brot Force Schutz, Passwort Hashing und so weiter. Ein Gedanke will ich dir zum Schluss noch mitgeben, der ein bisschen über Fable 5 hinausgeht, aber seitdem Fable 5 jetzt einmal gesperrt wurde, nutzt Anthropic den Vorfall, um mit Amazon, Microsoft und Google einen gemeinsamen Standard zu entwickeln, mit dem sich die Schwere solcher Sicherheitslücken einheitlich bewerten lässt und baut die Zusammenarbeit auch mit der US-Regierung aus.

In fact, what we can probably do is we could use something like The Research, aka dynamic workflows, inside of Cloud Code, using a model like Opus 4.8, have it figure out a plan that actually makes sense. I think you sort of just use the template I gave you, which is do some deep research, figure out how that particular app actually works, figure out how you want to customize it, get your prompt in order, and then bring it to Fable 5. Take a look at how you use Claude in terms of your skills, your automations, your tasks, and then figure out what you're doing right, what you're doing wrong, and more importantly, what we can do to improve this. Now, if you're wondering how I'm coming up with these prompts in this prompt structure, this is coming from Anthropic's official documentation when it comes to prompting Claude Fable 5 because there are some nuances between how you want to use Fable and Mythos versus something like Opus. Or you can clone it and give this to teammates who aren't going to use the CLI or aren't going to use the claude app because they can add whatever they want to this and you pretty much just wire up different skills and automations that you would use because it's just doing clawed headlessp under the hood which luckily for us isn't pulling from API prices anymore since Anthropic walked that back a few weeks ago.

Google just quietly shipped the Open Knowledge Format (OKF): an open standard that formalizes Andrej Karpathy s LLM wiki pattern into plain markdown any AI can read with zero integration. You point your agent at a folder and ask it anything as long as it knows OKF! The answer is we never agreed on a format - and that s exactly what Google has now solved. In this video I break down OKF, talk about why it s so important, and even give you a bundle so you agent can immediately search through my YouTube content. - Check out Posthog, a single platform to make your products self-driving with product analytics, sessions replay, feature flags, and more: - The Dynamous Agentic Coding Course is now FULLY released - learn how to build reliable and repeatable systems for AI coding: - My open-source OKF bundle (clone this and point your AI at it): - Open Knowledge Format (Google): - OKF spec (the SPEC.md file): OKF launch blog: Karpathy s LLM Wiki: 0:00 The LLM Wiki (Karpathy s Idea) 2:04 Why We Need a Standard: Google s OKF 3:20 What OKF Standardizes 6:47 Building With the OKF Spec 8:15 Sponsor: PostHog 9:35 Why OKF Matters (Even If You Never Share) 11:00 The Gift: My AI Coding Bundle 16:25 Watching My Second Brain Query It 17:33 Is OKF Too Simple?

But, what they will do is send you a a fake email that says that your invoice is due, say through PayPal um or any other payment provider, and then they ll put in their phone number uh for you to call and interact with them, which if you were to do that they use tactics that stress a level and sense of urgency for you to then say, Well, don t worry. Like as you re as you re looking this up right now, um you know, what are some stories that you know of personally uh of people that have have had um you could be phishing attacks or other types of cybersecurity things where they they slipped up and they did a thing that they weren t supposed to and because that, they they ended up exposing themselves and and it could be financial issues or uh ransoms or whatever might have popped up. And it s also one of the things to be very aware of, especially with this whole like mm open claw craze and you know, people that are you know, downloading open claw onto their computer and then there s all these marketplaces that would have open claw, you know, add-ons and everything else and then all of a sudden you re running all these terminal commands, you re having AI run the terminal commands, there s something stuffed inside of a markdown file. And then, in terms of the notification, like So, like if I want to get notified if something like this is a red alert, and I should pay attention, and you know, you know, take bolt cutters to my Wi-Fi, uh you know, what should what like how do I get notified? What I think is great about this is what I think would be really valuable is we talked in the very beginning about this, but I do want to I want to collaborate with you to put together a little kit for somebody on like, you know, what they could buy in terms of the hardware, these workflows that you ve shown so far, um, and then, you know, any kind of like really light information kind of how to get started with this.

They re looking for somebody who has under 2 years of full-time work experience, and the only requirement they have is taking these two courses by Anthropic, which is AI Fluency Course as well as Claude 101 Course, which is very simple. You can even do the whole vlog just to see what Cloud Corps is all about and how you can use AI for good because if you decide to become a content creator, which I hope you do because it s one of the fastest way to grow your personal brand and I think that s one of the ways for us to survive, especially with AI. Now again, before you apply for this job application, you have to complete these two courses and the other thing I would do if I were you, I ll also put this link in the description below CodePath which is one of the entity who decide to put this program together. Now, to prep for interview, you can definitely ask Claude to give you some examples interview questions, because I m pretty sure the people who are working for Anthropic is going to use Claude to prep for interviews. And even if you don t get the job, that s totally fine, because that can be used towards your future career, because you re practicing how to talk to other people through camera at this point, which is going to be very helpful for you when it comes to creating content, applying for another job, or getting sale or getting a deal from your client, anything like that.

Heute wollte ich euch zeigen, wie ich hier ganz einfach ein System eingerichtet habe, bei dem ich das Cloudsystem über die Weboberfläche von Hermes verbinde, aber nicht über die API, sondern einfach mit meinem Cloud Abonnement. Also habe ich einfach eine Schulung erstellt, in der ich euch Schritt für Schritt zeige, wie ich entweder die Cloud API nutzen kann oder vor allem mein Abonnement verwende. Also hier auf meiner Hostinger Installation wurde gerade Hermes installiert und ich werde es einfach hier öffnen, um mich mit dem Passwort anzumelden, dass ich festgelegt habe beim Ausführen der Bestellung und dem Starten der Installation von Hermes. Also kann ich hier einfach einen Schlüssel erstellen und diesen Schlüssel werde ich einfach hier unter Anwendenfügen. Und dieses hier gibt mir tatsächlich einige Guthaben für Cloud Code und deshalb kann ich sie nutzen, also diese Cloudverbindung innerhalb von Hermes verwenden und das funktioniert.

In this video, we ll go through every single change with Sonnet 5, but more importantly tell you how you should be using it now, should you be using it in Hermes, should you be using Open Claw, should you be using it in Claude Code, and where you shouldn t even be touching it at all. So, one thing to note here as well from a pricing perspective, even though I like these results better from Sonnet 5 than from Chad GPT and the UI is still way better with Claude Sonnet than Chad GBT, I don t know why Chad GBT can t figure out the UI side. You want to use Sonnet for all your basic coding tasks and you want to use Opus 48 in the ultra mode for a lot of your planning and really complex tasks. So, we still use Opus 48 for the planning, but what we re going to do in a second is use Sonnet 5 for the execution of that detailed plan. A bunch of strings have been found in the clawed code around Fable 5, including looking like it is going to require API usage, as well as looking like it is going to require verification.

Dieses Video von "Dieser KI-Agent beantwortet deine E-Mails wie du – Postfach komplett automatisieren" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

Ich habe den Output hier schon mal gepinnt, damit wir den gleich wieder verwenden können und damit beim Zerschneiden an den Grenzen dieser Chunks jetzt keine Information verloren gehen, weil das natürlich zufällig passiert, also einfach an einer festen Grenze wird quasi geschnitten und deswegen wird zwischen aufeinander folgenden Chuns oft ein Überlap, also Overlap eingebaut, so dass z.B. Und das was der Prompt jetzt macht ist er schießt jetzt noch mal alle Dokumente durch den NN Workflow durch so dass ich das jetzt nicht selbst machen braucht, damit wir dann tatsächlich auch gleich alle unsere Dokumente in der Tabelle vektorisiert gespeichert haben. Also damit erstmal herzlichen Glückwunsch und damit bist du jetzt tatsächlich schon tiefer drin als 99% aller ja wahrscheinlich sogar ITiler da draußen und ich will mit dir jetzt ein bisschen tiefer gehen an der Stelle, denn das was wir jetzt gemacht haben, das ist die semantische Suche und dieser hat in der Praxis eine entscheidende Schwäche. Also, wenn du jetzt hier in die Wissensablage gehst und einfach eine PDF-Datei oder ein ganzen Ordner mit hochlädst, lass uns doch einfach mal das kmh Handbuch wieder nehmen, dass wir jetzt eh schon die ganze Zeit hatten. Jetzt gebe ich noch mal das Dokument ein, jetzt suche ich hier noch mal nach dem Dokument, das wird hier eh auch schon vorgeschlagen und stell mal so eine Frage, wie was sind die Vertragsparteien hier in diesem Vertrag?

Let me show you what an AI agent actually is and how to build one inside Claude without writing a single line of code. People spend hours designing elaborate workflows, then launch them without giving Claude any context about who they are, what they re trying to accomplish, or how they want things done. Something like a workflows folder for your instruction files, an output folder for completed work, and a resources folder for reference material you want Claude to draw from. The simplest way to build this into every workflow is to add one line in your section, Always present a written plan and wait for approval before beginning any multi-step task. A research agent produces a findings report and then a content planning agent reads that report and generates video ideas based on what the research identified.

Hier findet jetzt unser AI Day mit einigen unserer Kunden statt und genau darüber werde ich hier auch sprechen und zwar wie baut man sich denn wirklich ein recht sicheres und funktionierendes Company Brain abseits von diesen ganzen ja Second Brain Hypes und anschließend geht für mich direkt nach München, dann nach Hong Kong und China, denn ich hole mir jetzt in der nächsten Woche zahlreiche Eindrücke im Bereich der Humanuin Robotik und über all diese Dinge erfährst du natürlich als erstes hier auf diesem Kanal. Ja, das was mit Entität gemeint ist, ist ja auch dieser Multiplayer Ansatz, also dass man Cloud wirklich wie ein Teammitglied behandelt, welchem man pro Channel eben schreiben kann und alle anderen Teammitglieder in diesem Channel dann sehen, wie ist der Fortschritt der einzelnen Anfragen und Cloud vom Gesamt ja Kontext profitiert. Also du kannst kaum einen heftigeren Vendorlock haben, weil sich mir ja die Frage stellt, ja, wo liegen denn jetzt letztlich diese ganzen Daten, weil das ist ja der große Mehrwert, also dieser unternehmensweite Kontextteam und channelweite Kontext, aber wo liegt die Cloud ND? Also es zeigt vor allem, dass dies jetzt der nächste große Trend ist und dann lohnt es sich auch zu überlegen, also wenn du KI Agentur bist, dann ist das jetzt ein elementarer Use Casase. Mrell veröffentlicht mit Mrel OC A4 ein neues führendes OC Modell und Mril ist sicherlich eines der schlechtesten LMs, wenn es ums Coding geht, wenn es um normale Arbeitsabläufe geht, aber im OC nutzen wir sehr gut 6 bis 12 Monaten fast nur Mistell, also Mal war schon super, deswegen haben wir das direkt mal gebenchmarkt, also Mistal OC A3 und AC A4 im Vergleich.

Das eliminiert diese ständige Angst vor falschen Infos fast komplett, weil das Tool sich strickt und nur auf eure eigenen Quellen bezieht. Anstatt passiv hunderte Seiten durchzulesen, generiert euch das Tool mit nur einem einzigen Klick simpel schöne Illustrationen und Mindmaps direkt aus euren Texten. Jede Antwort, die die KI euch gibt, hat kleine anklickbare Fußnoten und die verlinken euch direkt und ich meine exakt, auf den entsprechenden Absatz in eurem ursprünglich hochgeladenen Dokument. Und jetzt haltet euch fest, in einer brandneuen Beta Funktion könnt ihr sogar live in dem Podcast anrufen und den Moderatoren mitten in der Unterhaltung eigene Fragen zu euren Texten stellen. Sichert eure wichtigsten Erkenntnisse immer gut ab, nur für den Fall, dass Google das Tool irgendwann doch mal einstellen sollte.

The prompt simply says study this data, clean it and build a complete interactive dashboard with KPIs, charts and slicers for a CEO level report. It identifies the columns, understands the sales numbers, and starts generating the full VBA code needed to build your dashboard automatically. If you had to do this manually, you would need to build pivot tables yourself, design charts one by one, format every single element, and align everything properly. But here, everything happens automatically with just one click. Just one prompt, one click, and a fully working dashboard ready in front of you.

Ihr werdet sehen, dass ich in diesem Dokument das mache, was man das Anbringen einiger Annationen nennt, denn man kann es bearbeiten, man kann die Informationen auswählen, die uns am meisten interessieren. Also, ich habe tatsächlich einen Text vorbereitet, den ich einfach hier einfügen werde, um tatsächlich eine spezifische Frage zu stellen, damit mir die Informationen angezeigt werden, die mich interessieren, können Sie hier eine finden. Also jetzt gibt er das Ergebnis zurück und voila, jetzt habe ich einfach das Ergebnis mit der Zusammenfassung, also wirklich schnell, wie Sie sehen, selbst wenn Sie ein großes Dokument haben, kann das System in das Dokument hineingehen. Man kann auswählen, man kann es ihm sagen, man kann ihm sagen, wo er suchen soll, man kann das Verständnis aller Fade und aller Bilder nutzen, um die Daten daraus zu extrahieren. Also, ich nehme ein komplexes, kompliziertes PDF mit vielen Daten und Informationen und zack mit diesem Workflow, mit nur einem Klick, indem ich einfach die Zusammenfassung des PDFs über Telegram als Netzwerk sende, wird es mir tatsächlich geben und in Echtzeit auf all meinen sozialen Netzwerken veröffentlichen.

Or, for example, for open web UI Hermes, we could actually create a new page and use that topic, and then we can quickly generate content with the keyword and the case study ready to go, and just generate five articles, and then we can deploy them to our website, as you can see right here. And what this does is Hermes Oracle will log into my WordPress websites, it will write the content based on the latest topic, and it will actually embed the sources inside the content, so that it references the latest news headline with proof as well. So, if you re taking a lot of time creating your content, even with chat GPT, like you ve got to quality control it, you ve got to create the outputs, you ve got to remember the right prompt, personalize it all to you, and then go to your WordPress website, plug it in, and then format it, right? And then also the cool thing about the skill is basically the agent has to follow the steps step by step, and there s 13 different steps inside the process, which means when you re creating the content, it s all quality controlled. And if you want to get the full zip file, the Oracle, the keyword engine, the auto deploy, the video agents, my SEO setup, a 30-day roadmap for weekly coaching calls, and a room with 3,700 founders doing this and building with it right now, you can get that inside the AI Profit Worm.

Today, music how to build a WhatsApp AI agent that replies to leads instantly, qualifies them, and books property viewings automatically. Meanwhile, WhatsApp has a 98 open rate music and average response time of 90 seconds. Step one, a lead messages your WhatsApp number, N8N music catches it instantly. Step three, music OpenAI sends the reply back to N8N, and N8N pushes it to WhatsApp. Every lead, music every message, every qualification answer.

Notice that I specified to Cloud which Google Drive folder to use and I also explicitly ask to take into account different structures of all the files and ignore the names and pull the information only from the data inside. That means if somebody named their file differently next time or if you add a new country, then Claude will not make the dashboard work. And just like that, in a couple of minutes, Claude built this beautiful minimalistic dashboard, which would have taken us hours or days to create on our own from raw Excel data. While we wait music for Claude to work on this live artifact, I want to invite you for my free masterclass this Tuesday music on how to become an AICFO. Let s add new files to our Google Drive and try refreshing the dashboard.

Quasi die älteren Personen kennen Hood, aber quasi Leute, die 40 oder jünger sind, ähm kennen Hood vielleicht jetzt nicht so stark, aber wenn man mit anderen Leuten spricht, ja klar, Hood, früher war ich da ganz oft und sind so dann irgendwann mal unterm Radar geflogen und seitdem wir zum Konzern gehören, ist quasi unsere eigene Transformation ganz ganz wichtig, viel selbst entwickelt und wenig bei innen gehabt und äh deswegen haben wir auch sehr, sehr wenig im AI Bereich bei Contenterstellung ähm aktuell, aber AI nutzen wir auf jeden Fall im Retail Media Business und das ist dann gut. Und man sieht das ja auch im Online Marketing, wenn wir jetzt mal vom großen G sprechen, ähm die Kampagnenstrukturen, die jetzt man da anlegt, alleine PMAX, da muss man zwar ein bisschen mehr machen als bei den Retail Media Kampagnen, aber ist ja auch weitesgehen automatisiert und quasi steckt ja eine KI dahinter. Lennard, was würdest du sagen, woran scheitert s denn aktuell noch, dass äh doch viele das entweder aus der Händlerseite zum einen das nicht in Anspruch nehmen, aber auch das Plattform es nicht nutzen, die Organisation selber oder ist es irgendwelche was meinst jetzt genau die die überhaupt, dass man Retail Media überhaupt startet, also dass man als Händler sagt, okay, ich nutze das und dass man als Plattform sagt, ich biete das Modell an für meine Händler. Nächster Step wird sein, das auch in den Kategorien wirklich verfügbar zu machen, um hier dem Händler noch mehr Möglichkeiten zu geben, Visibility mit dem Produkten zu erreichen und quasi, wenn man es jetzt jetzt starten würde, ähm nicht mit WKZ Listen oder so anfangen, sondern direkt äh quasi dynamisch anfangen, quasi äh auf die Daten, die man dann auch wirklich hat, vertrauen und äh es direkt dann vernünftig aufbauen. Die Ambition sollte nicht aufhören bei einem kleinen Retail Media Projekt, sondern man sollte wirklich schauen, wie ich das Ganze skalieren kann, weil nur wenn man es wirklich skaliert, kriegt man auch den meisten Mehrwert am Ende des Tages raus.

Und ich will an dieser Stelle auch schon eine Sache klar machen, denn wenn du jetzt planst, wirklich ein längeres Video zu drehen, also über 15 Sekunden, denn stand jetzt können diese Modelle nur bis 15 Sekunden Videomaterial generieren, dann wird es ganz sicher dazu kommen, dass du mehrmals Videos generieren musst und dann sozusagen immer nur die besten Teile aus dem Videos nimmst und diese dann nochmal zusammenschneidest in einem Bearbeitungsprogramm. Denn das Problem ist, wenn du jetzt ein KI-generiertes Bild nutzt als Referenz, wie zum Beispiel jetzt hier dieses Charactersheet, um ein neues KI-generiertes Bild zu generieren, wie zum Beispiel dieses hier mit einem anderen Outfit, dann wird die Qualität darunter leiden und das siehst du hier links auch an dem Bild. Wenn ich dem Modell aber einfach nur sage, ich möchte eine High-Budget Commercial Add für das Getränk Fresco und ihnen nur einfach sagen, in welcher Umgebung das ist und welche Personen von dem Getränk trinkt und so weiter, dann überlegt sich das Modell schon selber, was es da am besten generiert und ich habe damit wirklich bessere Ergebnisse bekommen, als wenn ich jetzt hier einen riesen langen Prompt angebe. Aber wie du jetzt siehst, ich habe hier lediglich zwei Szenen jetzt rausgenommen aus meinen Generierungen, habe die hier unten eingefügt und einfach nur ein paar Stellen gekattet und eben auch noch Audio eingefügt und am Ende ist jetzt folgendes Ergebnis entstanden. Denn das einzige, was ich jetzt hier gemacht habe, ist zum Beispiel die Stelle hier weg zu schneiden, wo ich jetzt, nachdem ich das Wasser hier eingefüllt habe, die Dose irgendwie noch mal in der Hand habe und das hier noch mal so reinfülle, obwohl das eigentlich gar keinen Sinn ergibt, weil das hier schon voll ist.

You open it up, you type what you want, you get an answer, and you just kind of assume that that is the best possible answer that you could have gotten because, you know, Claude code is one of the best AI tools out there, and the models underneath it, like Opus, are super super smart. But imagine what it would have looked like if it was legitimately building a bunch of dark code, meaning you know, code that you didn t write and it s shipping features or building out automations, that s a pretty legit like big deal, which if it lies about it or does it poorly, that really could result in your business losing a ton of money. Now, how you actually do that like stress testing or the verification is a little bit different depending on what you re actually building because if you re trying to verify a landing page, that s totally different than verifying like an edited video or a data pipeline or something like that. And the whole idea of verification and checking its work on the way is where you can have it be a little bit less lazy and it doesn t actually stop until it gives you something that you can basically quickly review and shoot off because it s a complete waste of time if it gives you something and then you have to make all these changes, right? So that is what s really cool about this because we get the creativity of a model like Opus and then we get the ability for Claude code to actually do stuff like this and now we understand what all the edge cases are and what users might do.

Es geht immer relativ kurz, dieser Schorz, der hat auch langformate Videos, aber er stellt ihn immer Fragen und die Millionäre antworten dann mit ihrer Meinung und das ist sehr, sehr interessant, denn man kann sehr viel davon lernen über Lebensphilosophie, über Business und ich habe mir gedacht, dass wir einfach diesen Kanal hier nehmen und all diese Videos mit einem Programm, das wir jetzt gleich in Claude Code On transkribieren lassen. Ich habe mir überlegt, dass wir das vielleicht als allererstes in einer SQL-Lite-Datenbank machen können, also dass wir jetzt gar nicht großartig eine Rieseninfrastruktur aufbauen müssen, sondern direkt starten können und ich möchte jetzt, dass du einmal diesen Plan dokumentierst, also ein LLM-Wiki für uns aufbaust, damit wir in diesem Projekt hier mit richtig viel Kontextwissen starten können und dann würde ich im nächsten Schritt anfangen, dass wir ja die Engine bauen, die Videos transkribieren kann, bevor wir alle Videos auf einmal transkribieren, wollen wir natürlich gucken, dass wir erst mal ein einziges Video als Beispiel nehmen und gucken, was dabei rauskommt. Okay, Claude ist fertig und er hat uns hier eine drei Schichten-Prinzip, Struktur aufgebaut, nachwiesen LLM-Wiki-Prinzip, steht hier alles noch mal beschrieben, darum geht es in dem Video, aber gar nicht habe ich ein eigenes Video zugemacht und dann sehen wir hier das Startmodell, Videos, QA-Pairs, Fragen, Antworten, Einheiten, Startsekunden, alles Mögliche, ist mir erst mal relativ egal, der sagt nächster Schritt, wie ich von ihr geplant, die Engine bauen und an einem Beispiel Short testen, sag Bescheid, wer nur loslegen will, deswegen gehe ich jetzt einmal auf diesen Kanal, besuch mir einfach irgendein Video raus, kopiere mir die Adresse, schließe das ganze wieder, packe das hier rein und sage ihm, großartig dann starte jetzt mit der Implementierung, ich habe dir hier eine Beispiel URL gegeben für ein erstes Video, ich habe es mir nicht angeguckt, also es ist irgendeines seiner Videos und ich möchte gucken, dass wir das hinkriegen, ordentlich zu transkribieren, die Fragen zu extrahieren, vielleicht eben aus der Beschreibung, aus dem Metadaten auch zu extrahieren, was das für eine Person ist, falls es dort drin steht und du bist jetzt an der Reihe, setze das eigenverantwortlich um und melde dich erst bei mir, wenn es zu 100 funktioniert. Alles klar, dann lassen wir uns den nächsten Schritt gehen, ich glaube im ersten Schritt sollten wir erstmal identifizieren, welche Videos es alle gibt auf dem Kanal, es gibt ja unglaublich viele von diesen Short Format Videos, uns interessieren ja nur die Shorts und dann möchte ich das eben auch in der Datenbank irgendwie abspeichern, dass wir eben die gesamte Liste haben und auch wissen, welche schon transkribiert sind und welche noch nicht transkribiert sind. Wir könnten jetzt alles direkt durchlaufen lassen in Blocken machen und ich würde sagen, wir möchten aber einmal darüber sprechen, also wir möchten hier was Eigenes nochmal eintippen, denn ich bin mir nicht sicher, ob ich jetzt wirklich alles durchlaufen lassen will, sondern ich würde jetzt eher sowas sagen, großartig, damit funktioniert unser Test und bevor ich die Sachen jetzt alle durchlaufen lasse, wäre meine Frage, wie kann ich jetzt diese Daten sinnvoll interpretieren und sinnvoll nutzen?

I head back to the clod chat, tell it to move to the next phase, and it gives us a master build prompt we can paste into clod code, or any other tool, lovable, whatever you re using, and it ll give out a really decent looking website right off the bat. So I copy that over, and because I m using clod, I can just click over to the code icon on the left, switch into clod code, make sure this chat is connected to the correct folder I created for this project, paste the prompt and ensure plan mode is selected and click generate. Now one of the best ways to get great results fast with these tools is to turn on the audio recorder in the bottom left here, and as you go just talk through anything that s not working or you change, and in the end we get something that looks like this, which already looks great, but it s still lacking something, that something is a brain, because even though our tool technically already works, right now it doesn t actually know a single real thing. This category is one of the most slept on, not because people don t know it exists, but because they think AI image tools are just for messing around, making weird memes, and sticking your dog in a Renaissance painting. But the biggest takeaway here isn t the tool or even that you could recreate any other tool the exact same way, because this was just one example, one tiny combination of those five categories, the categories themselves are the point because once you actually understand how to use them, you can build pretty much anything.

There are thousands of so-called free AI video generators out there, but if you ve tried one, you know it s just a free trial behind a paywall. And to make that reference image, I m going to use the best image generator out there, Nano Banana Pro, which you can now get completely free inside Google Flow. Now this isn t a free one, but I want to test it and see for ourselves whether the difference between free and paid tools is really as big as people claim it to be. Once I m inside, I go to the video generation and select sea dance 2.0 as the video model, which is considered one of the best out there right now. You can spend three hours to generate a five second video that might not even look good on the first try with Juan, or go with watermarked videos like we ve seen with meta AI, or you can get access to sea dance 2.0 and all the other paid AI video generators with one single subscription inside open art.

Deshalb ja, da steckt jetzt schon drin Sport, mache ich sehr, sehr gerne, ich muss aber auch immer irgendwas zu tun haben, also mich etwas selten erwischen, dass ich mal auf der Couch lege und Sport schaue, Sport schaue generell gar nicht, sondern ich muss aktiv sein, Sport treiben, Fahrrad fahren, ab und zu mal ein bisschen Krafttraining, das wird momentan etwas weniger als mir lieb ist, ja oder tatsächlich viel, viel draußen verbringen, wo ich dann mal das komplette Gegenteil habe, also nicht digital, nicht am Handy, gar nicht erreichbar, mal ein, zwei Tage gar nicht aufs Handy schauen, dass es mir momentan sehr viel wert. Wenn man mir zum Beispiel bei Instagram schreibt, ganz egal ob privat oder über den Business Account, das wird schon relativ zügig beantwortet, nicht unbedingt von mir persönlich, sondern mit KI, aber das fließt dann alles bei mir ins Postfach mit rein, dass ich da schon täglich benachrichtigt werde, was kam so rein und da habe ich es schon gesagt mit Postfach, also ich bin tatsächlich gerade was die aktive Kommunikation angeht, wie vor ein Riesenfan von E-Mail-Verkehr, wo ich das ein bisschen sortiert haben kann, das läuft dann auch von da mit KI-Flows in mein Tool, was ich so habe, wo alle Aufträge und alles drinsteht und ansonsten Website, Instagram oder wer unbedingt möchte, kann versuchen mich zu kontaktieren, die Nummer ist ja nicht geheim, muss leider auch auf der Website stehen, nur nicht wundern, wenn das da sehr schwer sein wird. Trotzdem und das versuche ich natürlich auch jetzt schon etwas aufzubauen, ist es schön, wenn man durch verschiedene Maßnahmen, die man so ein Unternehmen betreibt, vielleicht einen gewissen Betrag hat, der, ich bin durch Abos, was ein anderer Teil ist, den wir machen, ja schon einen Betrag reinkommt, der dir ein bisschen Sicherheit und Ruhe gibt, dass du dich wirklich schwerpunkt, also dass ich mich schwerpunktmäßig auf die Aufgaben konzentrieren kann, die mir am meisten Spaß machen, sprich ja auch weiterhin das Beratungsthema möchte ich noch mehr ausbauen, um einfach noch mehr Menschen zu helfen, weil das ist natürlich immer super, wenn ich eins zu eins mit Unternehmen zusammen arbeite, das macht mir auch super viel Spaß, aber jetzt ist es natürlich so, du hast verschiedene Branchen und im Grunde genommen muss ich mich so da reinarbeiten, als wäre ich Teil des Unternehmens. Die wichtigsten Werte, das ist natürlich vor allem die Offenheit, also ich bin vielleicht auch euch aus dieser Sportbranche komme, jetzt keiner, der sich da irgendwo versucht besonders schön darzustellen und sich dem Gegenüber versucht, alles gerecht zu machen, sondern wenn ich eine Idee doh finde oder sage hey, dein Auftritt da, das ist absolute Katastrophe, dann sage ich das so, wie ich das denke, das ist ganz wichtig, das habe ich auch früher in verschiedenen Unternehmen als Angestellter gelernt, das der größte Fehler war, dass sich einfach keiner getraut hat, Nein zu sagen oder seine Meinung zu sagen, das gehört aber mit dazu, ich meine auch ich sage ja immer, wenn irgendjemand sagt oh du deine Website, hat mir vorher besser gefallen, das höre ich gerne, also ich bin offen für Kritik, nur dann nimmst du das wahr und kannst das verbessern, ja und dann natürlich auch gerne die Transparenz, also ich werde jetzt nicht versuchen da irgendwo was jemandem anzudrehen, der das nicht braucht, sondern wir schauen immer was ist jetzt wirklich die Maßnahme, die dir am meisten hilft und wenn es da vielleicht eine Anfrage gibt, wenn man sagt ich möchte mein Social Media Profil aufpeppen, aber das ist gerade gar nicht der Schritt, der nötig ist, dann sage ich auch gerne mal du entweder können wir eine andere Maßnahme machen oder ich habe natürlich auch Bekannte mit denen ich zusammen arbeite, die andere Bereich abdecken und ich selbst habe da erst mal nichts von, das ist auch okay. Zum einen habe ich lange immer geglaubt, ich leiste einfach immer mehr und mehr und irgendwann kommt jemand auf mich zu und bedankt sich dafür oder gibt dir etwas mehr dafür, das ist natürlich im unternehmerischen Bereich eigentlich nie der Fall, musste ich feststellen, ist normal und natürlich aber auch, dass ich in allen Jahren nicht nur ich, auch viele Kollegen, immer sehr, sehr gute Ideen hatte, oft frühzeitig, aber nicht umgesetzt wurden und wir dann immer sehen mussten wie Mitbewerber das dann nicht immer direkt, sondern oftmals viel viel später umgesetzt haben und das war irgendwann so viel, dass ich gesagt habe, weißt du was, lass es das doch einfach mal selbst probieren, ich meine ich habe die Erfahrung, ich habe ja schon nebenbei, also ich bin ja relativ sicher in die Selbstständigkeit gestartet, war vom Anfang an profitabel, musste mir da keine Sorgen und Gehalt und sowas machen.

Every time your agent runs, you re burning money on GPT and cloth, but what if a free model could match open? And through hugging face, you can run it completely free. Open your existing Initon agent, swap the model node, paste one link, and your free hugging face token. Same agent, same power, zero cost. The full step by step setup is in the video below.

Ну и, наконец, нажимаем здесь на плюсик, и нас интересует такая вещь, как Google Sheets или же Google таблицы. Нам нужно перейти на такой сайт, который называется как console.cloud.google.com. И чтобы ошибки всё же не было, вам необходимо зайти обратно в Google Cloud. После добавления тестового пользователя вы нажимаете тут Sign in with Google, после чего выбираете вашего пользователя, ну и, соответственно, выполняете вот этот вот connect. Таким образом, мы создали полностью автоматизированное приложение, которое теперь получает заявки от клиента, умно их обрабатывает и добавляет такие заявки в Google таблицу.

Remember, Peter, he shipped photo AI rough and started charging in week one, then made it better while people were already paying a boring tool, making a little money beats your perfect app that lives in your head forever and never gets out into the world. You can sell templates that do someone s boring setup for them, planning system that runs their whole week, swipe file of messages that actually get replies, a checklist that turns a three hour job into 20 minutes, anything you had to figure out you the hard way you can hand to someone else for a small price. Okay, so here s the last piece and I think it s the one that separates the people who make a little money from the people who completely change their whole life and their whole family s life. So find your ideal model, study them relentlessly, build your own blueprint for your special flavor, and then when you ve got the right model, you need the right people, and it s a specific way to get the most powerful CEOs to take your call. In real life, too many smart people act like the hair, so confident in their brain power that they get complacent or over-complicate everything while the less-assuming tortoise quietly grinds out the wind.

Or you guys can make the script yourself if you are confident at making your own script, or you can fill this whole thing out, and then the bottom here, leave the script blank, and then just make Claude make the script based off this information. Now, I already implemented it here, but you want to click add skill, and then create skill, upload a skill here, and I ll leave the skill file in the description as well, so you guys can download it and drag it into here. So, once you have your Claude skill installed, it s literally as simple as clicking slash and then going up to your skills here, and then using this skill right here. But if you guys have just one background, like it s a UGC, someone talking in the mic in like their bedroom just holding the product talking about it, you guys can have just a nice clean bedroom background. Now I am recording this video after I made the AI UGC video, so I m going to actually transition into live making and cut it up and just make it easier for you guys.

We put a little bit of a different angle on it just because of the fact that I think people in this space in general are just a little too positive or at least at the time I made this video, they were a little bit too positive about remote jobs. This is going to be for people who want to start a channel, but they don t know what niche to pick, what videos to make, or how to actually get started the right way. And the thing that I want them to think is, wo, this is not a generic video that s just like every other video I ve ever watched on the internet that s going to recommend that you get a data entry job or something like that. I might give it some references from other videos about things that I liked in those other videos and I just tell it what I want the video to look like and what I want the script to look like as well. Now the thing about player for education is that particular setting is really good for academic content that is meant to be watched in a classroom right that is what player for education is for and his content was a little bit educational a little bit academic but it wasn t really meant to be watched in a classroom and so when he did that really messed the algorithm up and it screwed his channel up and we had to fix that for him which we did but I have seen people mess with the settings before and screw it up and either not having the right button pushed or having the wrong button pushed can actually stop your channel from having success.

Your Way ist ein Lern-Tool von Google, mit dem KI-basiert Inhalt ist so angepasst werden, Your Way ist ein Lern-Tool von Google, mit dem KI-basiert Inhalt ist so angepasst werden, dass du sie mit deinem individuellen Lerntyp perfekt aufnehmen kannst. Literature Insights ist ein Literaturrecherchtool von Google, mit dem du wissenschaftliche Literature Insights ist ein Literaturrecherchtool von Google, mit dem du wissenschaftliche Arbeiten, strukturierte Daten und alles, was beim wissenschaftlichen Arbeiten hilft, Arbeiten, strukturierte Daten und alles, was beim wissenschaftlichen Arbeiten hilft, KI-basiert schnell finden kannst. Hypothesis Generation ist ein Tool von Google, das dabei hilft, mit einem multiagenten Ansatz, Hypothesis Generation ist ein Tool von Google, das dabei hilft, mit einem multiagenten Ansatz, basierend auf wissenschaftlichen Methoden fehlendes Wissen zu identifizieren und dadurch basierend auf wissenschaftlichen Methoden fehlendes Wissen zu identifizieren und dadurch komplett neue wissenschaftliche Hypothesen aufzustellen. Google Stitch ist das Prototyping-Tool von Google, mit dem du KI-basiert User-Interfaces Google Stitch ist das Prototyping-Tool von Google, mit dem du KI-basiert User-Interfaces für Apps oder Websites bauen kannst. Mit Google Flow Music kannst du KI-basiert Musik generieren und als letztes noch Project Mit Google Flow Music kannst du KI-basiert Musik generieren und als letztes noch Project Genie, ein KI-basiertes Tool von Google, mit dem du komplette Welten bauen kannst.

Dieses Video von "Kundendaten aus ERP und CRM direkt im KI Chat, in Sekunden. #ki #crm #automatisierung" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

Wenn du das eben nicht dazu schreibst und nur schreibst worum es gehen soll, Wenn du das eben nicht dazu schreibst und nur schreibst worum es gehen soll, dann klick einmal auf Tool, einmal auf Dokument erstellen dann klick einmal auf Tool, einmal auf Dokument erstellen und dann unter Dokument dann einfach PDF auswählen. Ich mache das Ganze hier einmal groß und hier siehst du das Ganze, Ich mache das Ganze hier einmal groß und hier siehst du das Ganze, jetzt haben wir halt eben ein PDF Dokument. Wenn wir das einmal vergleichen, mit dem was wir davor hatten, mit dem Dokument, Da war das wirklich einfach nur so ein Dokument, wenn ich das jetzt exportiere, da war das wirklich einfach nur so ein Dokument, wenn ich das jetzt exportiere, auch als PDF wird das trotzdem genauso aussehen. Und dann kannst du die natürlich auch darauf dann anpassen lassen, Und dann kannst du die natürlich auch darauf dann anpassen lassen, dass das jetzt ja wirklich erst mal nur ganz allgemein, dass das jetzt ja wirklich erst mal nur ganz allgemein, das heißt die PDF und das Word-Dokument kann auch gleich aussehen, das heißt die PDF und das Words-Dokument kann auch gleich aussehen, wenn du es eben richtig prompt ist. Und dann eben nochmal sagen, dass ihr mir eben genau aus diesen Inhalten hier Und dann eben nochmal sagen, dass ihr mir eben genau aus diesen Inhalten hier eben eine PDF erstellen soll, wenn ich die dann eben exportieren möchte und so weiter.

So I want Claude to reference my personal experience, the file I already uploaded and help me come up with ways to make my ebook unique to what s already on the market. So to promote your ebook with high quality content and better storytelling using the tools I use, head over to Storyblocks.com forward slash Sandra D and right now you can get 15 off any annual plan for a limited time, which is only available through my link. All right, now we re going to ask Claude to take all of that market research and combine it with our unique selling proposition and experience to outline and write the ebook from scratch. Can you help me create an outline for my ebook using the market research, my unique selling proposition, and I copied and pasted the USP that Claude gave me plus the supporting file in this project, which is my personal experience after you re done, please place it into a file that I can open up in Google Drive. Now you ll notice I ve only asked it to outline my ebook and send it to a Google document that I can then customize, but you can get Claude to do even more of the heavy lifting for you by writing out each section of your outline word for word and then transferring it to a Google document that you can edit.

So first, let s go to the N8n, here I type chat trigger, simply for chatting, we will take this event, this node and after that, we will use HTTP node and here we will take HTTP node and simply you have to go, copy it, import it into N8n and paste it here and import and that s it. So here, first of all, we find the API, how you can make the API, you have to go into your account and here you have to go into Access tokens, you will need to confirm the password, I have put the password here and after that, our demo API is made, we will have to copy it or make a new one. Now what we will do is, we will test it here, open chat and here we will first break the connection so that our tokens will know and here I type hello and here no information is coming, so first we will do mapping here, I will connect it here, after that we will take the chat input here, so what is this place here, what will we do? So here I have dropped the chat input, with the content and that s it, now our radio is ready, let s test it and simply if I type hello now, our response will come directly to GLM 5.2, which is one of the powerful models in the market, open source and here it has come, and here it has told us about thinking, so how we will remove it, simply we will replace it with the edit field, so here we will type edit and edit field and we will put the response in it, the model that has given hello, can I help you today, we will drop it and here we will say message. So this has been sorted, now you can replace this model in any agent, I have already done a lot of YouTube automation on this channel, I have made its agents and shared the JSON files with you, which you can directly download from the school community, you can easily replace it with existing agents, just put these two notes where we are using AI agent, and here you can make sure that you put system instruction in the content, so that your context can understand it well, so let me show you some of its benchmarks, if I go back here, and here if we simply search GLM 5.2, then we will be able to see its benchmarks, if I scroll down, here we have the benchmarks, Opus 4.8 and GLM 5.2, so if you look here, in SWE bench pro, it is almost equal to the anthropic cloud, and here it is equal to the terminal bench, and it is fully open source, its weights are openly available, you can download it and modify it, or you can run it locally, similarly in programming, the level of 4.8 in the whole competitive, and in mcp atlas, it is equal, in tool decathlon, it is almost equal, and in humanity last exam, with tools, it is equal to the anthropic, so this is the level of Opus 4.8 of the anthropic, and specifically, where you design the front end development, it is even beating the feeble 5, which recently has been shut down by the US, because of some security issues, cool, so here we have set up, let s move on to the existing working, complicated, which we modify, and integrate GLM 5.2 there, so first of all, I copy the nodes, let s see if possible, we can modify it in the node, so here I will scroll down, and here you can see daily AI update, this is our slack agent, and here we use the Gemini model, which I will replace with GLM 5.2, so here we will break its connection, and here we will break all the bottom, and we will remove it, so I will directly connect it with it, and I will connect this field directly with it, so our short-term work, so our short-term work is sorted already, and it was so simple, you can integrate it in any existing model, it is very simple, so now what you have to do, first you have to go to the AI agent, and here you have to copy the system instruction, and after that you have to come into the HTTP request, and you have to go into the expression, and here you just have to make sure, that you put the system instruction nicely, and here I have put the system instruction, and after that, here is the content, and here I have put the content, and after that I have put the chat input, and after that, what we have to do is, we have to do mapping, so instead of the content, we have to map the actual data, so how will we do that, so first of all we will just drop it here, so the way we will replace it, with our text, so the text was coming from the back, the whole text will go ahead, and that s it, after that what we have to do is, the message of this edit field, we will map it here, so in this case, we will not have it, and the data that we will give, will basically go ahead, so here we don t know how to map, so no worries, we will execute the previous steps, so it will do the mapping, so it is running, I hope so, it will run here, and then it will execute the Slack, so here we have made a dedicated video on Slack agent, how you can make a news article agent, which will summary you, daily AI news updates, and it runs in free, it is very simple, simply this news API, if it does daily news fetch, it summarizes it, and then AI model, it sends it to Slack in a free summary, that s it, so here we were using a small model, if I show you the model here, we were using Flash Light, which is not so powerful, in a good summary generation, now we will use a Cloud Opus 4.8 level, which is at GLM 5.2, and let s see what results come, let me show you the results, until this build, so if I open Slack, here we had daily AI update, today s update is like this, news would come, and the summary would come, let s see how this model gives output, here our output pattern is exactly the same, because we mentioned how output should come, so there won t be much difference in output, but here the analysis, summarization will be more powerful, here it fetches 100 articles, iterates and then moves, so here it has 100, and let s see, so here the token usage can be much bigger, and here it has failed, let s see what s the issue, it s not valid JSON, because it is breaking here, not an issue, so finally guys, this node has been executed, and how was the error here, because we are calling HTTP, but first we were using AI agent, because of which the quotation, or the line break, which was alone, but you can t do this in HTTP body, so here I just put it here, and it was modified, and it copied the error, and simply called cloud code, and cloud code fixed it, and it used code, json.stringify, basically it will make sure, it will properly string the line breaks, so that it becomes valid JSON, so we have run here, now we will execute it, and let s execute this, here the data is here, and the output is here, so we message it, and it s message is here, let s execute this, and here it s gone, so let me open Slack, and here guys, we have again 2 days, we have AI update, and 4 updates, where we have the exact pattern, but here it has done more writing, summarization, and pattern exactly, it has followed the same, and it has a great model, and open source, so here are some use cases, first of all, let me show you the Lm arena, so if you go to the code, and web dev, then overall, it is ranked on the second number, and which is really quite impressive, because here people have rate it by themselves, meaning the model has not put it by itself, it has put it in the benchmark, and people have used it, and its ranking is updated here, and if you look at the price, it is at the level of fable 5, fable 5 is not available, so our opus 4.8 is better, and if you look at the pricing, in the coding, 1.4 per million input, and 4.4 per million output, similarly here, compare, which is 5x more, which is very very expensive, and this is completely free, and here I will show you some use cases, how people are using it, and here is a user who has compared the opus 4.8 to glm 5.2, and look at the results, both are quite impressive, and here glm 5.2 is also very good, and the UI is generating, and here, how much is this cheap, if you look at the average output, here glm 5.1 was green, and glm 5.2 is here, so it performs less than, but here, if you talk about accuracy, here low accuracy, it has used 10,000 tokens, and similarly, if you want to do it then low high max is their model, where the thinking reason increases, and in max, if it solves the problems in 50,000, and how accurate it is, it is around 70 , and here, if we talk about glm 5.2, it is also almost 75 higher, so here, this line is looking at the opus 4.8, and here, it is looking at glm 5.2, so here, high, it is 75 accuracy, almost 40,000 tokens are used, but if you use max, then you will get around cloud code opus 4.8 accuracy, which is 75 , and it uses less tokens compared to 10K tokens, which means if you deploy it, or use it through open router, then you will get cheap, and here, someone posted that it is running locally, completely free, because they have good system, and codex, and they are running it free, which you can try, or deploy, but again, your cost, or your specification should be much better.

Das bedeutet, dass man mit einem Dollar praktisch hunderte von Videos erstellen kann und es ist ein System, das eine Idee direkt von Telegram aufnimmt und es wird sie einfach als Video umsetzen und anschließend wird es das Video auf YouTube und TikTok teilen. Ihr müsst euch einfach von Hosting abmelden und eine andere E-Mailadresse verwenden, so als wert ihr ein neuer Kunde. Wir geben ihm also einfach ein paar Augenblicke, damit er damit vorankommen kann und schauen uns dann gemeinsam das Ergebnis an. Also, wenn ich jetzt klicke auf TikTok und ich sollte es direkt sehen können, tatsächlich ist der Beitrag veröffentlicht worden auf hier ist die TikTok Seite, die ich eingerichtet habe und da ist es. Also klicke ich hier und das ist dann mein Video, also auf YouTube, also hier der Workflow.

The mix is two YouTube videos on the topic, one industry PDF, and one long form article. That one toggle is what lets Claude physically click around inside our notebook LM tab. If you re a hobbyist shipping one video a month, this is genuinely overkill and you should stop at workflow one. The process is the same one click as audio overview, but instead of a podcast, notebook LM generates a short cinematic video. We started with one notebook and one source set.

So you start reaching out to founders of B2B companies, founders of B of D2C brands, and you start pitching them the idea of building your personal brand and start creating revenue through that founder led sales is probably the best way today to create sales in any company. It could be, let s say you want to teach people about how to grow a bonsai tree, how to build a garden at home, how to cook a particular meal, how to do a particular task every day, how to be more productive, how to start investing in stocks, and then you simply have to create videos around them. And so if you can start building these voice agents for companies, they will end up saving a lot of time in general, and they will pay you for the value that you create for them. There are so many businesses today who want to start deploying AI agents, but they don t know how to get started. So if you are the person who is dealing and doing the sales, you should find someone who s a developer and who can create these N10 agents or can use land chain to create agents and start deploying them for any function.

क ल इ ट क हम स पल PayPal द त ह अपन ईम ल और व हम प म ट भ ज द त ह अब व प म ट आई य क क PayPal क अ दर व jस क श स कन क ट नह ह ग ऐस ह ह व प म ट गई स ट र इप क य क व ल अक उ ट क अ दर Man ब क म ट र सफर नह ह ग व प स उधर स टक ह आपक अब उन प स क व थड र करन क ल ए ल ग अलग-अलग क म करत ह क स क वह ड लर स ल कर द त ह क छ हव ल ह ड अलग-अलग तर क ह उसक करन क ज सक वजह स न कस न क य -क य ह त ह क एक त एक सच ज र ट बह त खर ब म लत ह ज ह र ह अब आपक मजब र ह र ट 280 चल रह ह व आपक कह रह ह 250 द ग य इसस भ कम फ र उसक ब द 1000 प हम फ क स फ भ ल ग जब आप ज क स ल करत ह इस तरह क आपक ऑफर स म ल रह ह ल क न आपक मजब र ह त आपक व ज हक हल ल क प स आपक म लन च ह ए थ व आपक नह म लत ट प ट प स ड क अ दर आपक ज फ यद ह ज त ह आपक य क ल म ट ड क पन ह आप यह प य क अक उ ट अपन न म क ऊपर क र एट कर सकत ह अब व ज प स आपन ड यर क ट उधर ल न थ व आप ड यर क ट इस अक उ ट क अ दर म गव सकत ह अपन क ल इ ट स य फ र अपन स ट र इप क अ दर अपन PayPal क अ दर य अक उ ट ड यर क टल कन क ट कर सकत ह और व प स ड यर क टल आपक ट प ट प स ड फ र ब ज़न स क अक उ ट क अ दर आ गए अब यह स प स आप Jaz क श इज प स ब क ट र सफर क थ र इज ल व थड र कर सकत ह क ई एक स ट र च र ज स नह और ब स ट एक सच ज र ट आपक यह प म लत ह स पल स क लक ल शन कर न 10,000 प अगर आपक स र फ 10 क र ट कम म ल त 1 ल ख क ड फर स च ह ए ज सक 50 क ड फर स म ल 5 ल ख स ध उसक गए ज क उसक म लन च ह ए थ ल क न नह म ल आप यह स प स म गव भ सकत ह अपन क ल इ ट स क ड यर क टल इनव इस भ स ड कर सकत ह अगर आप म स क ई फ र ल स ग करत ह य क स न क ई व डर स वग रह क प म ट करन ह त ह ब हर इधर-उधर त व भ आप यह स ड यर क टल कर सकत ह मल ट पल ट म म बर स यह प ऐड कर सकत ह क य क बह त स र स लप न र ह त ह बट म ज र ट ब ज़न स स भ ह त ह ज नक प स प र ट म ह त ह ज फ इन सस स द ख रह ह त ह त व यह प आप उन ह ऐड कर सकत ह ज य स र च ज़ द खत ह अगर आप अभ द ख GBP अक उ ट ह य त अभ 374 क आपक म ल रह ह Google प भ आई ग स थ ड़ ल इव करक द खत ह GBP ट PKR स 371.51 और ब र ड ट ब र ड य व र कर ग त आपक ज यह प र ट म ल रह ह 374 स एक व कर द म र प स हल ल ह ज ए ग प रश स स थ क य स मच ट ब ट ब स ड फ र ब ज़न स फ र स प सर ग द स श और ज क आप स इन अप वग रह भ कर ल न म र क पन क ड ह एचब 50 अगर आप व य ज़ कर ल त ह त ज स ह आप 500 प उ ड क ट र ज क शन कर ग उस प आपक 50 प उ ड ट प ट प स ड व ल क तरफ स फ र ऑफ क स ट म ल ज एग म ज कर उसक अच छ अभ हम एक एक ट व ट कर ग आप ल ग ज ए YouTube प YouTube ओपन कर ल य सबन अब क ल उड ए आई सर च कर YouTube प अच छ यह प न अब आप द ख ए ज म यह प ज म र प स सर च र जल ट आ रह ह य पहल ऐड ह इसक हम ऑर ग न क ल स ट ग म नह ड लत म स य एक द य फ र ऐड ह य श स क ल स ट ग ह त न य ऐड ह च र क ल उड एआई एक ब र ड क वर ड ह इसक स थ म न अभ क ई और टर म नह लग ई क क र श क र स, फ ल क र स, फल न क र स, चमक न क र स क छ भ ऐस नह ड ल ड यर क ट एक ब र ड क वर ड सर च क य ह व ड य न बर फ र प ज शन पर र क कर रह ह व य श ज़ द ख मस त 1,27,000 अब तक ह च क ह अच छ इन एज क शनल क ट ट य बह त अच छ व य ज ह और स प शल प क स त न आव म अगर इतन द ख रह ह ट प फ र म र जल ट ह अच छ इसम ज सबस मज़ क ब त ह व य ह इस व ड य क आईड य इस व ड य क ट इटल इस व ड य क थ बन ल क स क र एट ह न ह उसक इ स ट रक श स इस व ड य क स क र प ट इसक ड स क र प शन इसक ट ग स ईच ए ड एवर थ ग इज़ र टन ब य एआई व ह च इज़ क ल ड अब म र आपस एक सव ल ह अगर य स म स क ल आप ल ग क आ ज ए मस तमस त ह ज एग न ज म ज दग मज न ह ज एग आज मज त ह फ र त आज इस प र व ड य क म ब र कड उन आपक करक द ख ऊ ग ल इव म न क य प रम प य ज़ क ए ह ? य त थ ग ल बल एन ल स स अब प क स त न और इ ड य इसम कर टल अगर हम इस ट प क क सर च कर उर द ह द क अ दर न प क स त न इ ड य ब चम र क फ उ ड द स क ट ट ग प अप रर च न ट फ र ब स ज स व ड य प ज स ट प क प इस वक त 3 ल ख व य ज ह म न त अभ र डम उठ क ड ल प क स त न म उसक ऊपर क स न क ट ट अभ तक बन य ह नह त ज व र क ग ह क सक च स स आर ह ई क व र क कर ग ओनल जयक त श क र क य क क ई और ह ह नह न क स ट द ख व म झ ब ल रह ह क अभ क छ नह ह बट आप च ह त य व ल टर म ज क YouTube YouTube प आप सर च कर ल क य क म झ इस ट प क प एज क ट भ त ह न ह न त आप यह स एज क ट ह सकत ह क य क क ई ह क य ज़ नह ह आ, क ई म न ए गल नह ह अब व म झ बत रह ह क इसम म स ग क य ह ? ज आपक क र ऑड य स ह उनक प स क ई क ड ग ड ग र नह ह ल क न व ह ई ट कट व बस इट स बन न च हत ह क स-क स क प स अभ ज तन भ आप ल ग म ब ठ ह ज कहत ह भ ई म मस त क डर ह अभ एक द त न और क तन ऐस ह क यस हम क ई ऐस तर क पत चल एक एआई क य ज़ करत ह ए क हम कस टम व बस इट अपन ह स ब स ऐस बन सक अब ह थ खड कर स र उसक ब द इस व ड य क फ र म ट अब मज क च ज य र यह भ त ह सकत ह न क ड फ न ट स ब त ह म झ स र च ज त नह आत त अगर म झ क ई च ज स ख न ह त पहल म झ ख द भ त र सर च करन ह न म झ भ त कह स न ल ज च ह ए व क स ह ग उसक ल ए र फर स व ड य तक व न क ल क द त ह अब य एक आईड य ह एक आईड य ह ज सक ट इटल य ऑड य स म ग प क य ह आपक ऑड य स क ह स ब स क स ए गल प आपक ज न ह और उस स प स फ क ए गल प क न स YouTube व ड य ह य ग ल बल ह सकत क ई भ ह सकत ह ज ब स ट व ड य ह ग व व म झ र कम ड कर रह ह आप इसक ज क द ख य एग जज क टल उस फ र मवर क प ह आईड य ट ह उ ट स ल एआई एज ट स आईड य थ र ऑनल इन अर न ग ए ड फ र ल स ग एव ल य शन प ह Fiverr इज च ज ग व ह ई ट र ड शनल व ब ड वलपर स आर ल ज ग ज ब स ट एi एज ट स अभ आपक पत ह क हम र यह प व स भ य फ इबर अपवर क ड ड और य व स र च ज बह त ज य द चल रह ह व ए गल ह न क स ट ट प क क ल उड एi वर स स Google ए ट ग र व ट 2.0 ब स ट फ र ट ल स फ र फ र ल स स आईड य फ इव म इ डस ट ए ड प र डक ट व ट ज थ ड बह त एक रखत ह ज स आप कह ल क तरब यत क ल ए क ट ट क म इ डस ट क स ह न च ह ए आपक स ट प लर न ग ट क ड इन 2026 ड द स इन स ट ट और यह प अगर आप द ख त इसन स ल ख ह आ ह ईस ज ट इप क ट र वर श यल ज भ ऐप म स क डर ह अगर उसक य ब त ब ल ज ए छ ड द क ड ग आपक जर रत नह ह आप बग र क ड ग क य कर सकत ह द ल प तकल फ ह त ह क स-क स क ह रह ह य द ख ह रह ह त य क ट र वर श यल ह जब य च ज आत ह Facebook प 50 कम ट न च आत ह फ र व आग श यर ह त ह 100 उनक व ड य प आत व आग स ब लत ह अच छ य आ गय नय अभ आएग इतन स फ र ल सर अब य आक बत एग प र ग र म ग क जर रत नह म बत त ह क य जर रत ह त व आग स आग य र न ग ट व ह सह प ज ट व ह सह म र क ट ग त ह रह ह न म र व ड य प रम ट त ह रह ह न ब त समझ आई क क ट र द ल ग र द फ फड थ ड बड करन पड त ह इस तरह क जब च ज ह त ह ऑलद म इस तरह क प ग स द र ह रहत ह अपन पढ न ल ख न क क म ह उसम अल ल ह क श क र ह मस त ह ल क न कभ -कभ क स क स जगह प स ट र ग प इ ट ऑफ व य रखन च ह ए अब हमन ग ल बल व ड य क एन ल इज क य उस ट प क क हमन प क स त न ऑड य स क ह स ब स एन ल इज कर ल य इसक ब द हमन थ र ल वल एन ल स स क य ज सक अ दर हमन ग ल बल प क स त न ऑड य स ए ड द न हम र ज एक च अल ऑड य स ह उस ह स ब स य स र एन ल स स क ब द उसन म झ प च ट प क आइड य ज न क ल क द ए ज सक अ दर ज म न इस क फ र मवर क उस प बत य उसक ह स ब स न क स ट ह क आइड य ज म क छ ह क स य ज़ करत ह अपन व ड य स क अ दर ट टल व प च स छह ह क ट इप स ह व भ म न इसक बत ए ह ए ह द ख श भश प ह क य फ य चर बत त ह आपक मतलब इसक क म क य ह ? य क ल ड प भ बह त अच छ क म करत ह बल क और भ ज य द अच छ करत ज म न ई म न क य च ज़ क य यह स आप ल ग क ज च ज स खन क म ल ग व य ह क क न स ट ल क स क म क ल ए स ल क ट करन ह य बह त इ प र ट ट ह प रश स भ ई दस ल त द ज य द ह नश म चल गए म न ज म न ई क य स ल क ट क य ओवर क ल उड YouTube क ट ल उस Google क क पन ह और YouTube भ Google क क पन ह जब आप क ल उड म ज ओग न त क ल उड म स ट ऑफ द ट इम क य ह ग क व उस य आरएल क फ च नह कर प एग च ख ख ल क द ख नह प एग क य ह उसक अ दर क य क र स ट र कश स लग ई ह ई ह Google न क भ ई हम र प ल टफ र म ह न त म उस फ च क य कर ग ? ब स कल फ इल स क अ दर ज तन भ प र व यस म न स क र प ट स ल ख ह ज क YouTube क ऊपर गई ह व यह प ऐड क य ह आ ह और अभ म न बह त ज य द ड ट ऐड नह क य स र फ 1 प र ज क ट क प स ट य ज़ ह ई ह अभ इसक स थ-स थ ज म र व ड य स YouTube प ल इव ज च क ह उनक ट र सक र प ट म न यह प इस फ इल म ऐड क ह ई ह अब क ल उड क प स इ स ट रश स ह उसक प स म म र ह उस म म र म उसक म र ब र म हर च ज पत ह म र क ट ट क स ह त ह ?

Es steckt nur in den Köpfen einzelner Menschen und in tausenden Tickets, Mails und Telefonaten, die niemand systematisch durchsucht. KAPITEL 00:00 Das Problem: Wissen steckt in Köpfen, nicht im System 01:07 Drei Stellen, an denen es wehtut 02:09 Die gute Nachricht: Du hast das Wissen längst 03:14 Das Prinzip in 3 Schritten erklärt 05:03 Was du konkret davon hast 06:01 Drei ehrliche Hinweise bevor du anfängst 07:02 Zusammenfassung nächste Schritte DIE 3 SCHRITTE AUF EINEN BLICK 1. Fragen KI durchsucht euren eigenen Wissensschatz und antwortet mit Verweis auf echte gelöste Fälle SCHREIB IN DIE KOMMENTARE Wie sichert ihr euer Wissen aktuell? LASS UNS REDEN Wenn du wissen willst, wie du das in deinem Unternehmen konkret aufbauen kannst buch dir gerne einen Termin. Termin buchen: Dann schauen wir uns kurz an, ob Odoo für dich das richtige System ist oder ob eine andere Lösung besser passt.

Malva, right now, what are the best music free AI tools for creating horizontal videos, vertical videos, videos with sound, and videos without watermarks? So my recommendation would be to use Meta AI to create the rough music versions of the video, and once you know exactly how you want it to look, generate it here to get a stronger final result. Now, if what you want is to use the best video model I found right now, music we need seedings. You can animate the image without adding a prompt, but what usually gives the best result is writing a detailed prompt, music asking for a 15-second video, and generating it that way. Of course, you can also go to the image section, use the free models to create several frames, download them, return to the video section, add one as the start frame, and turn it into a video very quickly.

और य स र च ज ह ब फ र आफ टर ह ल क यह बह त थक कड ह आप इसक बह त ब हतर बन सकत ह य ज इसन भ यह क य ह ल क न ल इक फ ट वग रह और स र च ज बह त ब स क ह और वह ज स हमन र व य ज वग रह व न च ड ल थ त व स र वह ल इक ज प र पर एक च ज़ फ ल ह त ह व स र स म ह ह सबक इसन यह प फ ट बह त एक स ट र ड ल ह त य स र च ज आप इ प र व कर सकत ह अपन स इड प त ऑट ल इक ज Facebook म ट एड स ल इब र र ह यह स आपक ज ह क म प ट टर स क आईड य भ ह ज त ह क एक त क तन ल ग चल रह ह और द सर ज ह वह ल ग क अपन स ट र स क स ह अब यह प आपक प र डक ट ड स इड ह गय उसक ब द अगर आपक ऐड क र एट व च ह ए व क स च ज म इश ह ल ट स स ज हमन र स ट अभ ज ह व इटन ग स ट प स व ल क म स ट र ट क य थ आउट एस क थ र व यह प अगर आप TikTok एन ल ट क स म आत ह न आप स पल न र मल TikTok ऐप म भ ज क प र डक ट र सर च करक स र प र डक ट सर च करक ज ह व स पल आप वह स हर ट इप क क र एट व स और एड स उठ क लग सकत ह बस उसम यह ह त ह क ज भ आप एड स वग रह य ज़ कर आपन क श श करन ह क कह प भ ब र ड क न म न आए य ब र ड प क ज ग न आए ज क आप क य क ड र पश प ग कर रह ह त व च ज़ आपन ट र म कर द न ह य आपन व एड स उठ न ह नह ह ज नम ल इक क स भ ऐस क ई च ज क न म ह ग ल ट स स हम य व ल ज हम र थ यह ड क टर ड ट थ हमन ट आर ड ट बन द य और हमन प क ज ग भ च ज क ठ क ह ? अब यह स आप स ओर ज नल प स ट प क ल क करत ह त आप स ध TikTok प आ ज त ह TikTok प आन क ब द आपन क य करन ह ? अब स पल आप इस व ड य क ल क यह स क प कर Google प आए सर च कर TikTok व ड य ड उनल डर ठ क ह ? और स पल आप इन व ड य स क यह स उठ क ड यर क ट TikTok स भ य ऑट DDS क अ दर स भ आप उनक उठ क अपन म ट एड स म य ज़ कर सकत ह त अब आपक पत चल गय प र डक ट कह स ढ ढन ह ठ क ह ? और एक य न च व ल ह मज द आपक प स प र प स ह आप उसक अपन आईड य भ द सकत ह व उस प क म करक भ आपक ज ह व च ज ल क द द ग स द स व ज़ द ए ट यर ह ल प र स स ज क आई ह प क आपक समझ आ गय ह ग क प र प क थ र क स आपन क ल उड और ऑट ड एस क य ज़ करक म न अल और एi स प र डक ट स न क लन ह प र डक ट न क लन क ब द व ड य एड स भ आपक ज ऑट ड एस ह द द त ह न म ल त आपक TikTok क पत ह ग व म न आपक बत द य ह उसक ब द म आपन एआई क य ज़ करक स ट र बन न ह अपन फ ट जनर ट करन ह स र च ज ह ग ए ड स पल ए ड प आपन ज ह व अपन म ट एड स चल न ह ए ड द स व ज़ द ए ट यर ह ल प र स स ज क हमन क य थ अब इसस न क स ट ज आन व ल व ड य व ड य स ह ग उनम आपक म ट एड स भ स ख ऊ ग ए ड द यर आर मल ट पल च ज ज स क स ट र एआई स क स बन न ह ?

Milwaukee Mike here, introducing the CBD toilet wipes that s popular with electricians. Calm your ass down with this CBD toilet wipe.

Next is note taking, something most of us have to do, meeting notes, project notes, random ideas, voice memos, but they all sit there scattered across so many apps and they take time to organize or process the information. I ve been using notion AI for note taking because it s something that actually thinks with you, but here s what makes this more than just another transcription tool. So when a business email hits out in box, makes AI agent reads it, figures out what kind of inquiry it is, draws a reply in the right tone, and cues it for me to review before anything sends. Now I won t be sharing my own business finances on screen, but the point is here is you can toggle it on inside Claude co-work and connect the financial tools you use for example, QuickBooks, you know PayPal or whatever your financial stack is, then you tell it what to do. And in about 30 to 60 seconds I ve got a full presentation, properly designed, consistent styling, actual layout logic, not just you know tech s dumped into Slides, but here s what sets Gamma apart from other AI presentation tools.

A lot of times you bring down your usage by being very clear in your prompt and by the way if you press the plus sign here you can add files and folders directly from here and you still use the project folder to create things inside of that folder and there s more things like plugins which is access to different apps that we re going to get into in a second. And if you go back to that folder here where we had our receipts you ll see a new document was created called output and then within that you could see that Excel sheet that is created here and you could see this dashboard was also saved as a PNG something I didn t ask for but obviously really useful to have things like this. Now actually let me get into the memory explanation here on how this works because with Codex you actually don t have to explain things over and over again so they have some things called skills that we re going to talk about in a bit but it also has long-term and short-term memory that I want to explain here. So there is a file inside of Codex called agents.md so think of this file as your onboarding document if you were hiding a real assistant right you want to give them as much context as possible so every time you start a new chat or create a new project Codex could actually read that file first so he already knows things about you about who you are and things that you want to get done. Now in this case I m going to create an agents.md file specifically for a project that you could create a new project from here remember that I m a YouTube creator I m going to have all my spreadsheets color coded by priority if I write code I want it done in Python here so anything you specifically want or want it to know about you just say that or you could actually have the interview format where you say interview me and then ask me 10 questions to learn all about this project or you could have a general agents.md file and if you send this out here it creates this file typically in that project folder if you chose one or if you didn t choose one you will just remember that long term so when it does create it if I click on it it basically is system instructions or backend instructions for how this agent should behave and you could see that right here so it says I m a YouTube creator here and you give it more information than I actually had in my prompt here the coding preference is laid out here the spreadsheet that I had and the working style and it s pretty short you could have these be much much longer and by default it will actually know how to keep its instructions like this one like an agents.md now this also has an automatic memory where you quietly learn from your past chats over time but you generally don t actually need to touch that or even see where it s at that will do that in the background now for this next demo I actually want to show you one of my favorite ways to use these AI type of agents like codecs.clot has another one that does this type of work called clot co-work so if you have messy folders on your computer like a download folder you almost never have the time to figure out what s going on in your downloads folder right so I m going to have this take care of that for me with this prompt look at every file in the download slash mess folder this is just a folder I just made but figure out which client each one belongs to what it is name it or rename it clearly sorted by different folders for different clients and give me a nice folder structure right so now you re going to give it access to go do things on your computer and move things around in different folders and in this one I actually already have that folder so I m going to say choose an existing folder and that s the folder and I ll open that folder here and this is what that folder looks like right now before it s done and I ll show you what happens after it s done this is the file structure that it created for us so let s go over here and this was the messy download folder let me go back into it and now it s nicely organized and things are renamed with the dates but it kept every single file I had in that folder is just we structured it in the way I wanted organized and if you look right here this is the folder structure and it looks exactly that way on my own computer now for the next use case I ll show you how I actually works with PDFs because a lot of times we need to extract information like we did with the receipts but we need to do it with PDFs I have a lot of PDFs invoices for example here and I ask it to pull the invoice number the dates and so on okay so we got ourselves another spreadsheet here and this time it extracted from PDFs turning invoices into a spreadsheet something a lot of us still have to do to this day manually again this did it in just a couple minutes okay let s go to the next stage of this which goes well beyond what I m showing you so far because this becomes a super app when you actually start using these things called plugins and then something called skills plugins connect your chat gpt codecs into all the different apps you already use you just got to give it permission typically one app at a time if any of them show up right here as added that s because your chat gpt account already had access not codecs get access by default here and you ll see some things like computer use this can control your computer s apps with codecs you ll see chrome this could control chrome with codecs and browse the web you ll have to give it individual access to some of these now gmail obviously calendar your entire google drive those are some of my favorite ways to give it access but again remember you are giving it access to go read write and sometimes delete files on the things you re giving it access to so definitely be careful be very elaborate with your prompt to give a very specific details so you don t burn through credits very quickly but the gmail app for example in order to add an app I uninstall this one but you just have to press add right here this will install the plugin or the connector I m on a business plan that s why it says approved by your admin here so if you re in a business plan you may need to get permission but on a plus plan it should be done by default this is something I ve been using AI for for quite a while so all you have to do is type in the adventure sign and you could bring individual plugins like this or individual apps like this in this case I m going to bring in that plugin right here and I just ask you to look through my inbox I give it access to my google account that is associated with this business here and find any sponsorships here that send me emails and research them and put them in table format and I m going to reduce this to high typically that should be fine for this one and by the way I forgot to switch projects so if you don t switch projects it creates it within the other project in this case it doesn t matter all that much but you could always go to the chat option and start a chat from here if you don t need it to be inside of a project and in this case you give it to me right inside of chat because I asked it for a table but I didn t say give me a table in an excel document so you could actually use this like you would use chat gpt and not always create things that also reduces your credit usage in this case I just wanted to quickly glance through that and this is what I ended up getting which is perfectly fine I don t actually need to create a document from it now let me show you something called skills one of my favorite ways to actually use any AI tool if you also use clod you might have come across skills before but if I go to plugins and let s say I want to create powerpoint presentations right one of the things you ll see is the plugins but the other thing you ll see is called skills let me just click on this one to show you what it does what skills are are basically recipes you are saving the recipe so chat gpt or clod or any other AI tool that uses skills could refer back to it this is enabled by default right here I could always disable a skill and I could always open it to read it but the skill is this set of instructions for how a presentation should be created and as you could see they are usually super elaborate well most likely no one sat down and wrote this skill word by word best way to create it is by actually having a back and forth conversation with chat gpt and then when you get to a point where you really love the output say save that as a skill and it will save that as a skill so inside of a new chat if I type in slash you ll see I could bring in ton of different things inside of my chat here inside of codex skill will be on the bottom and there s ton of different skills already available in my account a lot of them I ve actually created myself a lot of them come from my chat gpt account but you could bring any of them into chat now a lot of them are enabled so if it needs to use it most likely you will but if I specifically wanted to bring in a presentation and use that skill especially if I made one myself that s how you bring a skill into chat and then your prompt will be to create a presentation and then it will follow every set of instruction that skill document has I m working on a very detailed skills video because inside of clod is very relevant inside of codex is very relevant inside of pretty much every AI tool is becoming more and more relevant AI agents use skills that you could custom create so I ll get into that a little bit later but just wanted to show you that s how easy it is to create skills and to bring them inside of a chat conversation now one of the best parts of codex is remember you could do knowledge work you could also do coding you could also generate images right chat gpt has right now the very best model when it comes to generating images I could type in a prompt like create three product photos someone wearing a custom knit a sweater in a professional studio I wanted in png format and it created those three images for us and it saved that to our download folder here so that s the three that I could use and then I could then with a follow-up prompt ask it to maybe put that in a thumbnail for me or create some kind of design or a sale for me here for an ad on running for example so remember you are inside of a chat bot so you can actually have a follow-up to keep building on this which I haven t really shown you I ve shown you one-off tasks where you send a prompt you get an output but you could just continue this conversation to you get exactly what you want and reiterate which is especially useful with an AI image model that s also an AI agent inside of codex next let me show you one of the best parts of using codex which is something called computer use that s when codex could actually take over your screen move your mouse around click on things and even type for you and use the apps on your computer and you could also do this with the browser so if you didn t turn it on already you just have to go to plugins and you ll have to turn on computer use and I usually turn on the chrome use with that too so this controls chrome so it could open up browser on your computer and do things for you let me go back to this previous chat here and I ll type out this prompt and you could type in the ad mention sign and bring in computer use like this you could choose this plugin here and my prompt is going to be to open canva start a new presentation and put each of the images we just made into their own slide as a follow-up conversation here and I ll send this out so you ll see something like this in the case where canva is not installed on my computer and needs the browser it will go use the browser but he did first check if I have that app installed so if I had photoshop for example it will try to do something in photoshop which I haven t had a lot of luck with but in this case it gave me something elevated risk allow codex to use google chrome so I will allow it access now at this point I m hands off he actually opened a new tab he went to the canva website here and from this point he ll do everything on his own now computer use right now and the browser use right now is pretty slow so I don t really find it as useful it s an option that just keeps getting better and better it s been around for sometimes with different parts of chat gpt but this is still a little bit gimmicky I ll let this finish up here I ll show you the results in a bit okay he finally finished the three slide presentation he opened canva he imported the images he created this slideshow here and obviously I could take it from there but for the amount of credits you use here and for how slow it is I don t find it very practical but it s really impressive the first time you see it at work and I think this is going to be the future of AI it s just not quite there yet next I want to show you automations which is probably one of the best parts of codex where you could do things on a schedule so you could let it go to work while you re sleeping some of the really popular ones are this daily brief a weekly review a project monitor so I ll just show you this one as an example so this sets up an automation that gives a morning brief each day and it will do things based on what s on my calendar what emails I haven t read yet and anything that needs my attention now this one you really want to give it access to a lot of different connectors if you re comfortable with that because these automations don t really work well if they don t have context from your actual life right you want to give it information that is relevant to you now automations I m working on an entirely separate video so I ll make that as a part two with some other things that I m saying for a part two this one I wanted to still keep a relatively short but definitely check out the automations tab and if you go to view templates you ll see a whole lot more of them over here and then in my follow-up video I ll show you these in action next I want to briefly show you the coding ability this is after all called codex and it s one of the best coding AI tools available right now so there s a couple of different options you could build sites with something they have sites you could create new site and it will build a site with basically that site skill or you could just type in a prompt here to build you real working apps here and these could run actually on your computer too and you could publish them with more advanced settings so if I type in a prompt like this to create a modern professional website for a video production company and give it a little bit of more context I could go ahead and send this out and this will go ahead and create that website for me now while our website is getting built I want to show you this slash command if you just type in slash on your keyboard these different options that I showed you a few of are really useful one of my favorite ones is this personality option where you could change the personality on how this responds to you you could also build skills and bring in those skills to change the tone and the personality of codex the other one that s really useful is this option right here called pets so if you bring in the pets it will just create this little pet in the corner that shows you everything that you got going on to notify you when a task is done so while I m building this website for example you can see this is going to work and then the pet will notify me so this is something I would turn on if you are going to create things especially mini apps or entire websites here just have the pet so then you could do other things use codex with a new chat and then the pet will remind you here with this little check mark when something is ready okay after a bit of time he created a website and he also gave me the link let me show you this link right here and this is the website that he created for us now I didn t give it a lot of context I didn t give it a business name but just kind of show you the style of it he also created this background image it looks very very professional I obviously want to give a youtube links or vimeo links here to replay some of these videos here but overall a really fantastic looking webpage with multiple different links to different sections of it that I asked for by the way if you go to the settings section right here you could look at your usage remaining so you usually get a usage for every five hour block and you also have a weekly but you also typically have a reset option depending on what plan you have that will reset both your five hour limit and your weekly limit but you only get a couple of these typically but I did use that reset to actually make this website with our business plan and some of the other prompts that I showed you throughout this video that require one reset now for part two I m going to get more advanced where we cover more things related to plugins and automations and some more coding example to show you the coding possibility that you have with codex again make sure you grab that free guide that I link below in the description that I created with my team on exactly how to get the most out of codex with prompts included and if you haven t checked out the video I made on clod code I ll put that video over here is similar to codex for coding

Like if Alfredo had to go out there into the real world and get a real job, would anybody hire him? We had him write his own resume and now we re going to send him out into the job market and see what s out there for an AI agent just like Alfredo, which for anybody wondering is an open claw running in a combination of models right here in my studio. I m just going to pick one here and I ll ask chat, chip, it to Taylor, Alfredo s resume specifically to that job posting. or maybe you re interested in a different role, make sure to start a new chat where you just say, Hey, here s context on a job that I m interested in. Then copy paste the job listing from a different place into here and then go into customizing your resume in that same chat.

And we saw this like, you know, this cookie cutter sort of idea in so many different situations where it s like, you know, you see a person who s sleeping, deciding, okay, the car s going to drive itself and then something bad happens. But the point is, I know what I would need because if I want to make sure that s always, you know, available, I would choose something in the cloud or on a virtual private server, but would cloud do that. Everything is going to be set up about 80 of the way there and now it s just simple things such as maybe you ve got the wrong model here or maybe there s like you re off by like a variable. Many times you want to show you know some sort of stuff such as you know here s a website you know like for example this is some of the news that just came out from unthropic with fable five and my toes and of course it got canceled and so it s like I want to kind of show this and it s like if I m talking to you I ll be like hey check this out you know fable five and my toes and I mean it looks cool like it actually kind of like go like this and kind of highlight and say okay you know this is what happened over here they think that it s jailbreak so whatever right I could talk through this entire thing and kind of highlight it but it would be so cool if I could make it look really nice because it s really hard to see this and not just this I d like to be able to do it maybe with an image as I explain something because many times it s like I m explaining but then I have to have my screen on and sometimes I just don t have that capability where I could have the screen on and also you know be able to record a video at times right most of the time I m in front of this screen but sometimes I m you know somewhere else but I still want to create this video so what a b roll is is it is something which allows me to create a video that I could throw into my entire main you know roll or main video and so I said okay clawed, fable or mitos let s see what you could do with like one or two prompts right and I tested it out and I was like all right make me a b roll system I m sure I could have done this with other you know other tools but I wanted to test how good fable and mitos were at building some of this so it built this out here and so basically what you do is you choose any website that you want or you can upload an image of a website or any documentation or whatever and then when I put it inside of here I just paste the link just like that and I click on capture page and so now it s going to literally you know capture the page here and then when it does that I can draw boxes around what I want to talk about so I ll be like okay I want to talk about fable five and mitos five and then I ll say let s see here the net effect of this order is that we must abruptly disable fable five and mitos for all customers right so I m like wow that s an another important point and then what are some of these other ones let s see here we expect that perfect jail break resistance is not currently possible no one like that one out at big points to kind of grab it that s right right so let s say it s this one defense and depth strategy I m just making it so now I said okay I want to basically make it look cool so I want to be able to fly from word to word and maybe I ll come really close on the second one all right and now there s no there s no code here sorry no AI here so I m not losing tokens it automatically built this for me in like two prompts all right here s the video it auto generated so the next question here is how to re automate this right let s say another yeah breaking news heads how do we trigger it to go into this work and just automatically generate a video like it s doing here is this something you could sync up with NADN well that s the idea that s like if I took you know fable and mitos a few steps further or I could even go into opus now and continue building this out but just the fact that I was able to do this and build these b-rolls and I just click a button and now the up before downloads it s like this is time saving yeah this is like real world use case like where I m gonna use it and you know this would have taken me even just to create something like this would have taken me forever I could change away this does this too by the way I could change this for like agent avatar and then with your head talking about it too yeah yeah yeah exactly I could actually like pan through this here I could say dive right into this you know I could choose the way it does this you I could cut to that I could cut to the last one see so there we go it s going into this and boom cut right to it right so I could choose the different models of you know effects but it s like to do this before I d have to pay a SaaS company right or I d have to you know pay for the software and mitos a fable just built it like two prompts yeah no this is awesome yeah there s so many other use cases that you could do for something like this but why don t we go ahead and bring it back here and we ll conclude this episode with kind of some rapid fire AI questions the first one that I have for you do you think NADN will ever quote unquote die not exactly I would say even probably no because of what has just happened recently right you have SAP and you know NADN coming together on uncertain things and became a big big piece of news and so now NADN is evaluated at over five billion dollars you re seeing that in enterprise use day in day out and enterprises are not fast in the sense like oh yeah they re going to shift in in one shot right this is going to be like for the next 10 20 years they re going to keep a system because you know they want to make sure that it s good and so that s the whole entire idea you imagine SAP and how long it s been there so these are the biggest in the industry they re slow movers and so this is going to be like it s going to be there for many many many years look at how big short it is still on an enterprise level yeah and and you may get like for example the consumer base might move right you might see consumers going a little bit more to the clouds or whatever you know is out there but still you will have the giants who are spending big money still using these powerful tools and you know that s what I think is it s not going to die out anytime soon it s going to be there for quite some time for sure what s the most overhyped clawed code replace my entire stack claim that you ve seen yet oh yeah I ve seen those the challenge here is that a lot of people think that if it works the first time it s going to work the second or third or fourth of it and you know I would say if you could show me it running like 10 15 times in a row great you know I think that I ll have more confidence in that and for those people who are actually interested like how can I actually do that in clawed use skills okay skills are that little I like it s code there on piles yeah skill markdowns are are so good and if you want to really get to that next level mix the two mix clawed with NADN and have your clawed agents call out NADN workflows and so that s going to help you to really get the power of the two worlds but when people are saying oh yeah I m just I m going to replace NADN try running this thing for 15 20 times try running run it 100 times I am more confident that I will get maybe 99 times of a run on a workflow with NADN out of 100 then I would with clawed just being honest it s not just I think I m like it s running and times it s why don t we try this for decisions that are worth 100,000 if not million dollar decisions that we need to make sure that it runs through correctly that s right that s right and you know you want to have consistency especially where you re talking about deal use like like real world use like I m building an invoice here that invoice used to be done by hand or or a quote right maybe it s a quotation document maybe it s a proposal document those things used to be done by hand we would get like word templates or whatever that you d refill the templates or you d use some like macro that was built in and you needed to keep it consistent right well now you get these more you know probabilistic tools and you could make some really nice looking stuff but sure work the first time will it repeat that can can I build you know a thousand of these proposals with near perfect results I know I could do that with the deterministic system like any then I don t know if I m going to get that at all with a clawed type of system because every single time is a different probability yeah that makes good sense here I m going to ask you get your crystal ball out here for this next question let s say 12 months out from now what is the biggest way AI automation is going to change that you think no one is talking about a lot of people are going to probably get back to the basics I think that s the biggest key everyone is talking magic tricks all right I m just going to be honest with you they re talking about the fancy stuff that Claude can do or might those come on one time I got to cool it s all all their problems in life yep that s right that s right but you notice that what I showed you just now I didn t show you a prompt running on Claude I showed you a piece of software that I built that s part of my SOP pipeline that saves me time that I no longer need tokens for ever again that s what I built I could literally now run that without Claude ever again it s working and that s exactly it we re going to see more and more people doing that getting away from tokens and the challenge here is as we start getting into our own private systems where we re using you know offline models such as TNWN Tong UTNWN which is QN for most people who s what they ll call it all right it means a thousand questions right you re going to start seeing people using these offline these types of or came in or whatever you re using a gem is a big one from Google and we want that control you want to use a lot more you want to be able to do these things having some level of AI and the AI now is so darn good that it s like you don t really need to front your models for like I d say the majority of your stuff you just don t need to front your models you can do it with a really good you know offline model deep sea gem if you want to keep to like North America once you use Gemma very very good you could use it on a decent system all right and you can get some you know really good results keep your system on I think we re going to start getting into the time with the ejection harnesses which is this year where we ve got virtual employees virtual businesses everything is running automated I think that s what we re getting into and that s what people who are in this space are going to start doing first then as people start seeing you know people like myself and others who are just running complete companies automated and it s like I can run let s say you know 10 and 20 30 businesses they re all generating me so much money that it s like why isn t everyone doing this that s the idea I want to get to you know that s made about not using the front tier models going more towards the local side of things I can completely agree with that right you know I can t stand when there s you know fable comes out or opus 4.8 and what do we see on the hype grind here look at this morning summary look at this email look at this content that is complete overkill for writing a morning summary or drafting an email emails were good about what two years ago the models were good enough draft good enough emails in my mind so now that moving in more into the technical realm where these new models that come out like the fables and the mythos and whatever open AI has up their sleeve it s more of that technical side that s where we re seeing the developments all this knowledge work like the models are already good enough to accomplish 99 of knowledge work and I think that s a super important point that a lot of people have not grasped yeah absolutely and so you know coming back to what I was saying when it comes to making money I just you know not financial advice to your own due diligence we re going to start seeing that right even I m building out businesses I m trying to get to that level where it s like the money is automated the success is automated and you can now determine just like you can with an any network flow you could determine these businesses like that and then coming back to you know what you re saying with these frontier models and you know it s it s like you could do so much now with these offline systems that you don t need these bigger models right and you re eating up you know token after token after token and they re just charging more and it s getting frustrating where it s like I could just run this without you know any of those issues and the key here now is coming back to the basics what is your SLPs what are you trying to build how s it going to help you and a good example of what s going on right now is think about race cars you know you see these car companies coming out with these super cars which you would never drive on a regular road right so it s like sure they came out with the latest super car that s got rockets on it can go super fast I m going to show it to you and you would never drive that not even on a highway right and that s what we re in right now most of everything in fact the majority of of the large language models are like very well you know polished cars they can drive on the regular roads on the highways that s all you need how many of you are actually going out to a super car to do something yeah and that s exactly the challenge I was doing I think fabel is overkill for 99 of real world work in my opinion right now but it ll be super interesting maybe this for it s a whole other podcast episode the more my mind is spinning about it but it ll be super interesting how this plays out as with the fabel news and people are going to get fed up with these new advanced models that are super expensive and then at some point they re going to get so good where people are going to create vibe-coded versions of fabel that are free and open source or maybe not free but very inexpensive and so the frontier models know that that might eat their lunch right the big tech companies so they re going to have to do something and it s just super interesting now where we go from here because we re almost at a stalemate with the fabel five news and all that stuff and it s interesting because they re like trying to bring out these regulations say oh we want everyone to slow down but it s like are you guys slowing down? no we re not slowing down but we we we think everyone else it s slow down and it s like in the back of my mind I m like so you won t slow down but you want everyone else to slow down that s called bay of an ice cream no sense yeah it s it s called let me figure out how I could slow it out everyone from eating my lunch exactly what you re talking about because you know every other day that a new North American model is released the Chinese models come out within like a week two weeks and they are very powerful and so sure I might not get the fabel level but if I m coming darn close heck if I got the opus level I m not going to be upset I could do most of my stuff and I think for the majority of people for sure yeah that s why you have to pick up the market so much but anyways we could talk about this stuff for a versalman this is the environment I appreciate your time working people follow you online to learn more about what you re doing yeah just go to my YouTube at Solman Christ AI you ll be able to join in on many different fun live events that I ve got going on we ve got our study hall which is Monday to Friday 9 a.m.

So I m just going to highlight it, right click and click on copy, open up my terminal window, right click to paste and hit Enter. Once you ve copied it to your clipboard, I m going to open back up my notepad and everywhere I see server underscore IP address, I m going to highlight it and then I m going to right click on it and click on paste. So I m going to highlight server underscore IP address, right click, click on paste, and then the one directly underneath it again, I m going to highlight it, right click and then click on paste. So click on that, right click in the box and click on paste to paste in your activation key and then click on activate. I m going to click on the plus symbol to create a new notepad text document and then I m going to right click and click on paste.

Darin gaben 66 Prozent der Befragten an, Darin gaben 66 Prozent der Befragten an, dass sie im KI-Zeitalter das Handwerk als sicherer empfinden dass sie im KI-Zeitalter das Handwerk als sicherer empfinden als den klassischen Bürojob. Im Büro viel kann das natürlich helfen, Im Büro viel kann das natürlich helfen, aber dass die KI uns mal komplett ersetzt, aber dass die KI uns mal komplett ersetzt, kann ich mir in den nächsten 10, 15 Jahren niemals vorstellen. Doch trotz der vergleichsweise KI-Krisenfestigkeit im Handwerk bleibt offen, Doch trotz der vergleichsweise KI-Krisenfestigkeit im Handwerk bleibt offen, inwieweit sich der Fachkräftemangel dadurch abfedern lässt. Aber diese Veränderung am Arbeitsmarkt Aber diese Veränderung am Arbeitsmarkt und der Druck auf die anderen Bereiche wird, glaube ich, für uns eher und der Druck auf die anderen Bereiche wird, glaube ich, für uns eher unter dem Strich positiv sein. Kein Roboter, keine KI geht aufs Dach bei Winterbarregen Kein Roboter, keine KI geht aufs Dach bei Winterbarregen und dichtet die Wäsche ab, wird die Kernbogen durch den Altbau machen und dichtet die Wäsche ab, wird die Kernbogen durch den Altbau machen und werden die Altbau und Sanierung.

I spent the last two years obsessing over every single AI tool that came out, but one thing that I learned from this is that the whole year behind idea is complete nonsense. Because once I stopped chasing every new release and looked at how some of the best people were actually getting good with AI, it came down to three very clear phases that nobody really talks about. Now by the way, I actually made a free resource breaking down every single one of the phases and exactly how you break out of it. And so to get past this step, the one thing that ll make the biggest difference is giving it the right context, because your output is only ever as good as what you actually put inside that box. But the people who never fall behind live on the foundation, and from there, every new tool is just a fresh take on something they already understand.

music They sit in a coffee shop staring at a blank notebook, open chat GPT hoping for inspiration, or spend hours watching top business idea videos music waiting for one concept to magically stand out. Right now, there music are thousands of people online explaining exactly what annoys them, what wastes their music time, what tools they hate using, and what they wish someone would finally build. music People go there to vent about frustrating software, repetitive tasks, expensive music services, and problems nobody has solved properly music yet. music It may even point out that a single tool focused on solving one specific issue could realistically charge a monthly music subscription because the you know, the time savings already justify the costs. Claude may identify that the ADHD productivity market already contains around music 4 to 6 million actively spending users with a potential serviceable revenue music opportunity between roughly 700 million and 1 billion.

Ja, also im beruflichen Umfeld als auch im privaten Umfeld, also ich weiß gar wie scharf man das trainen kann, beschäftige ich mich eigentlich nur mit Zukunftszhen, also einfach aus persönlichen Interesse raus und ähm war auch schon früher in Blockchain, früher in KI, früh in Quanten. Ja, ich sag mal, wir haben ja jetzt durch den Börsengang durch Space X, wo ja im Prinzip auch Riesenthema ist und dann hast du ja die Frontal Labs, die jetzt auch noch ein IPO dieses Jahr machen wie Open AI oder Anthropic und das ist natürlich das ist ja das erste Layer, also die erste Ebene von KI sind ja diese Frontier Labs, also SpaceX, Opi, Anthropic und noch ein paar andere und dann hast du ja eine Ebere drunter, die Infrastruktur, was natürlich auch ähm große Chancen für Anleger bieten kann. Also, wenn man sich mit der Thematik auseinandersetzt, glaube ich, ist im Moment gerade der Infrastrukturlayer am interessantesten, weil es ist natürlich ein bisschen strittig diese Valuations, die da aufgerufen werden für die Frontals, also Anthropic, Opi, SpaceX, wie lang sich das halten kann und ob die sustainable sind, aber die Infrastruktur, die da runterläuft, brauchst du, also du brauchst die Energie, du brauchst die Chips, du brauchst die Datenzentren, um diese Technologien in der Zukunft laufen zu können. Ich das sag mal die Implementierung und natürlich auch dieser Beigeschmack, den Blockchain hat aus der Vergangenheit, weil Blockchain also ist für den Endverbrauch oder für den ich sag mal für den für den diejenigen, der nett tief in der Technologie drin ist, ein Spekulationsobjekt und der Shift passiert ist aber schon die letzten Jahre, dass es mehr reindrückt in die Okay, es ist eine Technologie und die könnte mein Leben verbessern. Also die Technologie ist da, die wird auch verwendet, aber ich glaube natürlich, dass es am Anfang hauptsächlich da wichtig ist bei den Firmen, die die Technologie in der Hand haben und natürlich neue Technologien daraus erschaffen können, weil viel mehr Rechenleistung innerhalb kürzester Zeit ermöglicht natürlich mir auch neue Technologien innerhalb kürzester Zeit sehr schnell zu erforn umzusetzen.

Mittels maßgeschneiderte KI-Lösungen kannst du dein Unternehmen automatisieren, Mittels maßgeschneiderte KI-Lösungen kannst du dein Unternehmen automatisieren, Prozesse verbessern und intelligent unterstützen. Dies ist erklärt, wie KI funktioniert, damit der Application Layer, Dies ist erklärt, wie KI funktioniert, damit der Application Layer, das heißt alle generative KI-Modelle, die ihr aktuell benutzt, das heißt alle generative KI-Modelle, die ihr aktuell benutzt, gut funktioniert und sich kontinuierlich weiterentwickelt. Layer 4 ist der Application Layer, das heißt die KI-Modelle selber. Layer 4 ist der Application Layer, das heißt die KI-Modelle selber. Wir helfen dir, deine Use-Cases, Workflows und Prozesse mittels KI zu automatisieren Wir helfen dir, deine Use-Cases, Workflows und Prozesse mittels KI zu automatisieren und KI nachhaltig, effizient und gut skaliert in deinem Unternehmen einzuführen.

Open AI-Chef Sam Altman hält die Skalierung großer Sprachmodelle weiter für den richtigen Weg und weist Kritik zurück, LLMs seien eine Sackgasse. Er geht dabei auch Forscher an, die sich zu sicher gewesen seien, was Skalierung nicht leisten könne und das Feld damit gebremst hätten. AWS hat auf dem Summit in New York neue Dienste vorgestellt, die KI-Agenten im Arbeitsalltag verlässlicher machen sollen. AWS-Context baut aus Unternehmensdaten automatisch einen Wissensgrafen, damit Agenten Informationen einordnen und weniger sicherklingende Fehlantworten produzieren. Dazu kommen neue Prüfungen beim AWS DevOps-Agent, die Änderungen vor dem Deployment gegen Produktionsanforderungen checken und passende Tests in produktionsnahen Umgebungen ausführen.

This is going to be for people who want to start a channel, but they don t know what niche to pick, what videos to make, or how to actually get started the right way. So while you re still flipping through file number one, Claude has already read all 10,000 files, highlighted the patterns, connected the dots, and handed you a one page summary of who did what and why. With that being said, there s that 1 of you out there that wants to work directly with myself or my team, one-on-one, where we literally tell you exactly how to do it in a coaching capacity, one-on-one, done with you coaching, or we just literally do it for you. With that being said, the types of people we typically work with are gonna be business owners, both online as well as some types of physical business owners, YouTubers who are getting views but they re struggling with monetization, YouTubers who are crushing it and they wanna crush it even harder, or people who are professionals that have a lot of value to give to the world, they re typically experts at something and they want to treat YouTube like a business from the beginning and they wanna skip three years of banging their head against the wall, trying to get it to work for them. So if that sounds like you, again, book that call, click the link in the description and the pinned comment below and also check out this video right here where we go over exactly how we help my client, Sean, get to 500,000 in a single month.

Dieses Video von "Die stärksten MCP-Konnektoren für Claude ⚡️" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

Eine Automatisierung ist ein fortdefinierter Prozess, eine starre Abfolge von Schritten, die genauso programmiert wurde, eventuell vielleicht noch eine, wenn dann Regel beinhaltet. Also ein Roboter, der die Teile einlegt, eventuell noch mal prüft, ob das Teil richtig gespannt ist, ist effizient, aber er denkt halt nicht mit. KI hingegen wird es erst dann, wenn das System nicht nur einen festen Plan abarbeitet, sondern eigenständig Entscheidungen trifft und aus Daten lernt. während dem Prozess, dass der Guss heute etwas härter ist als gestern und passt dann die Schnittdaten komplett selbständig an. speicher dir das Video gerne ab, damit du beim nächsten Gespräch genau weißt, wovon du redest.

For research, this is terrible, this is decent, and this is mind-blowing. For building apps and websites, don t use this, this one s pretty good, but this one is amazing. For content creation, this is terrible, this is decent, this will make you go viral. For coding, this is not good, this is pretty good, and this is the best. For making money, don t use this, this is okay, but you definitely want to use this.

If you shift from Claude to any other tool in the future, then you just have to click on export data, and all your chat workflows and everything s data will provide you with a structured way, which you can use in other tools. But Claude s feature is the most different, if I explain the artifacts very easily, then this is such a feature, which means you can make an app in Claude. So let s say I want to make a productivity tool, so I will simply click here, after which Claude will take some inputs from me, like what I want to make, if you do not have an idea, then you can select it from the given option, but I already have an idea, so I will write it here, after which again he will ask some questions, and finally he will give me a final concept, what is going to be, I think this concept is right, so I will simply say build it, after which Claude will be split in two canvases, where on the left hand side you will have a conversation, and on the right hand side, according to your requirement, he will start writing the code himself, can you imagine this, I am not a software engineer, but I can make my own app, I can see the code in front of me, I do not understand anything, but it is very interesting to see, so in a few days, the code will be completed, but here there is an error, so you can tell Claude to correct it, and the error will be corrected, after which finally my app is ready, before showing the app, I will tell you a basic, on the top left hand corner, you will get an icon, where if you click on preview, then you will see the app, and if you click on the code, then you will see the code, if you are a software engineer, then the code can be your job, but we are going to see our app from preview, now what is there in this app, in the left canvas, Claude has explained it to you, there are nine functions in this app, and in the nine functions, there are specific things, which will help in optimizing life, in the right canvas, you can see that, there are nine functions here, and in functions, there are sub functions, like by clicking on the task, I can add on the task, and by clicking on the check box, I can complete it, after which in the daily briefing, you will see that my task is pending zero, because I have completed all the tasks here, active goals are still my team, and what they are, you can see here, these goals are made on the basis of my previous data, this is basically my second free account, I am Claude, in which there was a previous conversation, on the basis of the money, he wrote my goals, even these goals are there, you will see them here, in the same way, my task is already scheduled in the calendar, in the relationship, I have to talk to the people, they are mentioned here, and right, there is something in every segment, even the AI advisor has given a feature, in which you can directly ask, which segment of my life is missing, where I should follow up instantly, and he will also give me an answer, which will be shown in my daily briefing, when you like your app, then you can click on the publish, and click on the publish and copy link, after which you will get a link, and then you can share this link with others, or open a new web, and you can see, how your app is overall, this is a very simple and easy artifact, which I made and showed you, just to show, how to make artifacts, you can literally make your game, in the artifacts, and you can play it, as soon as I recorded this game, I made it, which is not played by me, I suck at video games, but that s okay, the point is, if you learn to use skills and connectors, then you will never have to work again, your cloud will do all the work, and how will it do, you will have to understand the skill, before using skills, I will explain in an easy language, what skills are, let s say I am a content creator, so my process of making content, is that I will first search for topics, then I will research on them, after which I will follow a certain pattern, and write a script, and then according to that script, there will be storyboarding, which will be given for editing, and then I will make a video, I have made more than 100 videos, and everyone s process was the same, and for all the videos, I had to repeat this process, in the cloud, if you have done all this once, then the cloud has understood, what are your expectations, at what point you have to correct it, and at what moment you liked the result, basically it is your pattern and expectation, both have understood, and also your work flow, so what can you say, that you can make a whole workflow, which is written in this format, now next time, whenever you sit to make content, then you have to do this work manually again, you have to write the name of the skill in the prompt area, and you have to say, that make content, and prompt run, that skill will do everything for you, by following the same workflow, which you followed last time, which will be the same, as last time you liked it, and this is exactly what skills do, skills basically master, you have to say one art or one workflow, so that you don t have to do the same work again and again, and the good news is, you can make everything, first of all you have to go to the customise, where you will get two options, skills and connectors, you simply have to click on the skill, after which, if you have made any skill before, then it will be visible here, you simply have to click on the plus icon, after which you will have two options, if you click on the browse skill, then the skills installed in the cloud, will be visible to you, as you can see the skill of Canva design, this skill is basically trained with designing psychology, for PPT, and this is the reason, why the PPT I made in the beginning of the video, was made so well, because cloud automatically used design skills, whenever you give the cloud any prompt, it always search first, whether it has any related skills, and then it starts the work, if you get the skill for your work, then it is fine, otherwise, you can come to the skill option, where you will have three options, I will explain to you in detail, the deal with cloud, means that the cloud will help you in making a skill, if you click on it, then you will get a new chat, where it will ask you, what skills you want to make, here you do not have to be very detailed, you can just explain, that what skills you want related to, so let s say, I want to make a skill related to content creation, so I will write the whole process of my content creation, which I follow for every video, I have even made a format for my script writing, so I will share that format here, now if I run it, then you can see that cloud is reading the skill creation, then it will start making my skill, now there are some questions of cloud, which I have not told, so I will give the answer of it, which we had a result a few minutes later, after understanding my words, and reading my document, how the skill will work on the right hand side, that thing is shared, and on the left hand side, there is a description, where you can see, that first this skill will find the idea in phase 0, in phase 1, it will do a deep research, and then in phase 2, it will prepare the script skeleton, in phase 3, it will write the script, and finally in phase 4, it will do the storyboarding for the editors, now what will the skill do, this is the workflow of this skill, which it will follow, on the right hand side, the same thing is given to you with more details, this is basically the workflow of the skill, which it will follow itself, as you can see, in phase 1, it will search the web and find the idea, then the audience will find the gap, and then it will find the related fresh data and points, after which it will follow the scripting framework, which I gave it as a Google Doc, and it will complete the whole process, until the storyboarding, but for this, first you have to download or save the skill, now since I have made this skill on the cloud, then I can click on the save skill, and this skill will automatically be saved in the cloud, now if you go back, then you can see, here a new skill has come, which is the YouTube content creation, so in this way you have made the skill, but how to use it, this thing is very easy, you just have to open a new chat, or in any chat, just put a slash first, and as soon as you do this, your whole skill will be shown here, as soon as here YouTube content creation skill is showing you, you have to select it, and then give a normal prompt, make a video for my YouTube channel, and it will read your skill, and start the work, after which it will verify the things with me, because I had said this thing in the skill, that first verify all the things with me, and then give the final result, since it is a big skill, it will take about 2-3 minutes, so I will show you the final result, first of all, in phase 0, this skill gave me 7 video ideas, after selecting one of them, in phase 1, after completing my research, I was given an overview, of what creator has made the video, what audience is demanding, after which he started the phase 2 work, after writing my yes, and a unique angle of video click, and 3 very clickable titles are also written, that too with a thumbnail idea, and then the script is also given, that what is going to happen, after which in phase 3, he has written the whole script, and in phase 4, he has provided storyboarding, that too in detail, that which line will be aero, which line will be bero, where will be the emotion graphic, where will be the data, and everything, I will share all these things in the discussion, so that you can read its work quality in detail, but the amazing thing is, the clothes that he has done, first of all, it took at least 4 days to do that, after which my editor used to go to the script, but here all this happened in 5 minutes, and that too without my hard work, if I didn t put the approval system in it, then it would have been the final result, and if I connect it with the code work, then I would have to download it and send it to the editor, that work will also be done automatically here, I know I haven t taught you code work yet, but this is possible in it, and the same reason is that, I had said in the starting of the video, that people are sitting and doing the clothes skill, and the work is happening by themselves, now this was before the 3 methods of making clothes skill, the second method is also right skill instructions, if you click on it, then you will get a pop up, where first you can name your skill, then you have to describe the skill for which, after which you have to tell in detail, what should Claude do exactly, this process is exactly the same as the process of making a project, after which you will have a new skill saved, the third type of Claude has provided for skills, is to upload the skill, many creators or digital product sellers, are selling them by making different Claude skills, or are sharing it in their communities, if you take a skill from them, then you will get an empty file, you simply have to upload the file here, and you will have Claude skill available, these are the three ways of making skills on Claude, and believe me, you can be totally free of any small work, because our 90 work is actually repeatable, and only one work flow is followed, but if you don t find this interesting, and you are thinking that Claude will do only work, he will have to share the work from here, he will have to update it himself, he will have to track it himself, then you are absolutely wrong, Claude will actually do this too, how? To understand that you have to understand Claude Kovug, and before understanding Kovug, you have to download Claude s desktop app, because Claude Kovug does not work on the browser, and if you don t know how to download the app on desktop, then I will give a quick tutorial of it, you simply have to go to Google, and search for the download Claude desktop, and then from the official website of Claude, you have to download it, after which you will have to follow the normal installation process, and then you have to sign in, with the same ID that you used on the browser, if you don t have desktop, then you have to choose this in your phone, Claude is actually on three pillars, first is Claude Chat, second is Claude Kovug, and third is Claude Code, whatever you have learnt, whatever you did, it was all part of Claude Chat, which you could access on the browser, now we will look at Claude Kovug, for this you have to open your desktop app, and then on the top left hand side, you will see three options, which I have already taught you, beside the chat, you have the option of Kovug, when you click on it, then you will be redirected to a new interface, here on the left hand side, you will get a lot of features, which I will explain in a while, in the center you have the prompt area, and there are some settings in the prompt area, which you have to optimize, now how does Kovug work, I will show you that, right now you can see my download folder, here it is so much crowded, now usually, I sit on Sunday morning, clean up this thing for 30 minutes, and arrange my folders, and if you do client based work, I am sure you will also be doing the same, but now let me show you a magic, if I go to Claude Kovug, give a task, to organize my download folders, SF2DistData, and here in the folder, I slide my download folder, if I click on allow, and click on, don t delete anything, after which I select the Opus model, and turn this prompt, then Claude will literally organize my download folder, it will take some time in the process, but you will be able to see everything on the right hand side, here the first tab is the progress, that is to complete your task, what steps you have to do, and where Claude has reached in those steps, after that is the folder tab, that is the access to the key folder to Claude, and there is context, in which those files will be visible, whose reference Claude has taken, for now you will not see anything here, whenever you run this task, then Claude can ask you for permission, so make sure you read the things, or deny it, here in the progress bar, you can see that the task is being cut out, and in a few minutes, this prompt will be closed, after which if I show you in my downloads, then you can see, everything is organized here, everything is organized in folders, except for today s data, and for this I didn t have to sit for half an hour, Claude automatically did this, that too on a command prompt, what is kept in the case folder, Claude will also tell you here, so that it is easy to navigate, or you can make a navigation sheet or PDF from Claude, if you want, for now I will give this whole process to Andrew, after a few seconds, you can give this download folder back, this exactly happened as it was before, see, it is so easy to do any task at the time of the application, now the task we did, for that you may not have done the task, but you had given the prompt, and even though you didn t have to make the folders of Manually Bad, but you will have to give Claude instructions, if you think like this, then you are again wrong, because in Claude you can literally schedule tasks, under projects, you have an option of schedule, where when you click, then you will come to a new interface, here you can schedule your tasks, how? let me show you, to run the schedule task, first of all, it is important to be aware of your computer, that s why Claude is giving you a toggle, which if you keep it on, then your desktop will never go to sleep mode, and it will not stay on, I will close it for now, now to make a schedule task, you have to click on the new task, after which you have two options, one is the create with Claude, and the other is the setup Manually, if you click on the create with Claude, then a quiz panel will appear in front of you, in which Claude will ask, what will be your task, do research or process the file, or content creation or something else, let s say I have a task related to emails, then I will write that thing here, I want Claude to see my emails, and whatever emails I have not replied to for two days, I should tell about them in an organized way, like how many emails were of brand deals, how many promotional emails, how many of my subscribers were of K and audience, and 10, I should give a complete report of them, so I will write this thing here, after which it asks me, how many times I want to run this task, daily, weekly or one time, so let s say I select weekly here, after which the third question is, what do I have to do with the result, so here I select the most, and 10 Claude will understand the requirement, he will find a skill for it, and he will take more details from me, like when do I need a report, so let s say I need 9 am on Monday, after which I use emails, so the answer is Gmail, now there is another thing, which I have not taught you guys yet, Claude says that he has a Gmail connector, but not connected, because of which he cannot see my Gmail, the connectors do exactly this, Claude has collaborated with many brands and companies, and all the connectors are available here, the connectors basically link Claude with those apps or softwares, you can manually on them, or when you come across a chat from the Easter, you can connect from here too, like for now I connect my Gmail, for which I just have to click on it, and login in my Gmail, and that s it, the work ahead will be done by Claude, now Claude has my Gmail access, but not for reading any message, for anything else, because I have not allowed that, when you connect, you have to click on the chat, after which Claude will start the process, now your schedule task is ready, which you can do without the command, every week Monday at 9 o clock, the laptop should be on, how will this schedule task work, if you want to check, then simply run now, which will run this task once in front of you, like when I run, you can see that in progress, that he is looking for my Gmail, and it has my download folder, where he will give the report, and in context you can see, that he has written the connectors, which he is using, and you can also see the skill, which is being used here, and in a while, I had a result, here you can see that in the last two days, I have not replied to the pending emails, out of which 13 brands, 10 audiences, 6 promotional, and the rest 6 random, on the right hand side, you will see a lot of details, which mail, when it came, whose, what was the subject, and everything, this is exactly how schedule task actually works, now this task will be activated every Monday at 9 o clock, just my laptop should be on, and by chance, if my laptop is closed at that time, then whenever I will turn on the laptop, this task will automatically run, now can you imagine, Claude Chat himself was so powerful, and when you integrate him with Claude, then everything can happen, you have to do all the work, make his skills, or make a project, and schedule the shift to task, right now I have made an account, which in the beginning of every month, my previous month s business revenue, is getting account reports, to run this skill, my bank account needs details, which I have kept in a specific folder, now Claude automatically goes to that folder every month, which is the latest month s report, he reads it, studies it, and then by keeping the data of the previous month, my business revenue report, P L report, and the rest of the accounts, first he used to do this work, but now Claude does this work without me, and that s where AI is actually taking your jobs, which you have just added to Claude s co-work, similarly, if you add connectors in artifacts, and add co-work too, then it is called live artifact, where it is updated itself, with your new data, because it also has the access to your device, and under the schedule task, you have been given the option, artifacts are basically apps, which I had taught you, but the app we made in the chat, we had to utilize it ourselves, but if my notion and Google crender connectors were there, then using co-work, I could make a schedule task, where every day, my data would be updated in the app, my work would have been so much, that I would have to see what my priority list is, and follow it, I don t even have to optimize my data, and read it, and this is the real life applications of Claude, there is literally no limit, that you can do anything, a clotty limit here, is your thinking, your tokens, now I hope you understood Claude s co-work, well, and you must have learned to schedule tasks with connectors, and understood the meaning of live artifacts, but there is one more thing in co-work, which can make you happy, this feature is of dispatch, what dispatch mainly does, is that it provides remote access, now this is much easier to explain, so let s say, in my desktop, there is a folder called openartugc-add, I have to shift this folder from desktop to document, but I am not at home, and I don t have my laptop, it s just my phone, in that case, you can login to your phone, and go to the dispatch feature, and then you can tell it, what you want to do on your laptop, and Claude will do that for you, like when I ran this prompt, Claude asked me for permission to move files, and he asked me where is the folder called openart, after which it was installed on the back end, you can even click here to see, where the process has reached, and what is happening on the back end, okay, so I forgot one thing, this file is from RGB, and that s why Claude is taking time, so I will change the prompt here, instead of moving the entire file, I will just say to move an image called amazon, by doing this, within a few seconds, Claude himself took control of my laptop, and put an image called amazon in a copy document, because I did not have the permission to delete it, can you imagine, I can access my laptop from any corner of the world, without doing anything, I can give it a command, and it will work, now if I am telling the truth, you already understood Claude more than 90 of the world, and have also learned to use it in a better way, now one more thing I have not covered yet, is Claude code, now what is Claude code, before that I want to give you a disclaimer, if you are not a developer, or a computer science student, or a software engineer, then Claude code is not your job, this tool is mainly made for programmer people, Claude code basically does what, that it stays in your terminal, because of which it can read your codes, understand, can debug it, can find bugs in it, and then can fix it, overall, Claude code is an AI assistant programmer, who works with you, to access Claude code, you need to have a pro plan, now first understand, that Claude code is used for things, Claude code is basically for apps, to make software, now I know you guys must be thinking, that this could be done in Artifacts, well you are totally right, but there was a problem with Artifacts, that whatever app was there, it was made only in Claude, that is you could not use it, you did not have any files, but in Claude code you have all this, your website or call-app, you can actually deploy, you can upload it to a place like app store, again I know you guys are confused, so I will show you, on the side of Claude chat and Kobuck, you will get Claude code, by clicking on it, you will come to the interface, now the top you are seeing, is your users chart, that today you use how many tokens, how many last, how many are inside this month, basically it has no work, the main thing for you is prompt area, in this area you have to describe, what you want to make an app, here you have an option of edit, if you keep it on, then Claude basically can edit in your prompts, that is he can give you happy inputs, to make things better, there is also an option of attachment, in which you can upload your branding and creatives, if you want to upload, after that you have given a lot of models, Opus 4.8 is the best model for Claude code, but I will go with Opus 4.7, because it is their legacy model, and here the bugs problem will also be reduced, after that you have an option of effort, the less you keep it, Claude will work less, the more you keep it, the more effort it will give, Ultra code is the max level, where Claude literally will give you a full life, to make your app, but this model is completely new, that is why the bugs problem can be here, and I am telling you this from experience, that it is not written, that is why here I will choose only high, now in prompt area, I will paste a prompt, I am making an app, which is helpful for the owner, trainer, for everyone, it doesn t matter if you make your prompt in a wake term, or make it in a detailed way, Claude will make that thing, and if it gets confused, he will instantly ask you questions, but make sure, that you have written one thing in your prompt, which is, here I have refused to use any kind of API, because if Claude uses an API, then it will be chargeable, and you will have to pay for it, because the API will not be Claude, it will be of some third party, so till the time the API is not very important in the app, till then you have to write this line in your prompt, now I will run this prompt, till Claude gives something, he will understand my idea, and he will think what to do, and then when he needs to read, he will ask me some questions, and he will also ask for permissions, I can give permission, because I have given only one folder access to Claude, not the terminal, so I know that he won t do the same with the rest of the things, when this process is completed, then you will have a final result, so first of all we have login page, where our app is called fitcore app, now here three people can sign in, game owners, trainers, and members, if I sign in on the owner s account, then you will see a certain interface, or in overview, you can see total members, monthly revenue, how much payment pending is there, and how many renewals are coming, that is also showing, last six months revenue is given to you as a chart, so that you can compare, and also the membership status is also shown, how much is active and how much is not, below you are joined by the later, that is all visible, and in quick stats, you will see a quick overview of the whole gym, and all this was in the overview section, in the members section, you will see the details of all the members, where you can add a member, in the trainers section, you will see all the trainers, after which the membership has data, and in the payments, you will see the status, that whose payment is ahead and whose not, you can pay the mark and update it, so basically your app is already interactive, and below in the programs, you will see how many members are involved in each program, and what exercises they are doing, what plans they are following, that is also shared here, in the same way you have a chart of nutrition, and then you have a chart of attendance, in which you will see, regularly how many people are coming and how many are not, and at last, in the report, you will see everything, and all this was available in the owner s ID, if I sign out, and go to the trainer s account, to as a trainer, in the overview, you will see, that you have how many clients, how many days they have visited, whose payment is left, how many programs are assigned, and all the details, in the client s category, all the clients have detailed data, like how many clients have visited, what is the weight, what is the body fat percentage, what is their goal, what is the program, and how many progress they have made in that program, by clicking on the open chart, you can see all the data in a good way, in the program section, again, all the programs are available, which are providing the gym, and you can see, in that program, which exercises are there, sure, in the deal, as a trainer, you will see, with which client you have a session today, and in the message, what message this client sent, it will also be visible to you, where actually, you can interact with your clients, now, all this starts in the interface of the trainer, but again, if I sign out, and login in the member s account, then in the overview, you will have your data, how much progress you are doing, how many times you have gone to the gym, how many days of membership are left, what is the weight, what workout you have to do today, and by clicking on the workout, you can see the workout, and as the workout continues, you can cross it, here, you can see daily targets, like how much calories you have to take, how much protein you have to take, how much water intake you have to take, and what is the progress in all this, in nutrition, you will see calories, protein, carbs, fats, everything will be visible, and you will have a daily meal plan, that what you have to eat every time, on today s day, in progress, you will have a more visible form, in data, in attendance, you will be visible, where it will be marked, that when you went to the gym, and when you didn t, and in payment, payment updates, and in profile, your personal profile will be visible, this is a gym application, which is made by Claude Cod, that too within 10 minutes, when doing the same, normal developers take at least a month, the files of this application, you can see that, my Claude Cod experiment, is in the folder, now if you are a developer, you will know, how to deploy this application, and file, so that it can be downloaded, in everyone s phone, but I am not a programmer, so I can t teach you this, my job was to teach you, the features of Claude Cod, and I hope, you will have understood Claude Cod, Claude Cod, can work in many other complex ways, but again, it is necessary to have the knowledge, of the developer level, before the start of this video, you didn t know anything about Claude, but now you know, how to optimize the settings of Claude, what is the difference in Claude and GPT, how do the different models of Claude, how to save the tokens of Claude, how to use Claude Chat, what is the work of the co-worker, what can you do with this money, how can you do live, and how can you run Claude Cod, now I can say, that you have become a master of Claude, if you liked the video, then don t forget to subscribe to the channel, because on our channel, you will get such courses, free of charge, and if you want to know, how to use AI, you can make an online influencer, that you can generate passive income, then don t forget to watch this video, see you in the next video,

Und drittens schauen wir uns an, wie du dich auch ohne starke Hardware unabhängiger von Und drittens schauen wir uns an, wie du dich auch ohne starke Hardware unabhängiger von den großen Anbietern wie Entropic oder OpenAI machst, entweder indem du die Modelle natürlich den großen Anbietern wie Entropic oder OpenAI machst, entweder indem du die Modelle natürlich lokal hostest oder zum Beispiel günstigere Modelle über die API nutzt und damit nicht lokal hostest oder zum Beispiel günstigere Modelle über die API nutzt und damit nicht in einem teuren Abo festhängst oder die Modelle auch über einen GPU-Anbieter wie Olamah in einem teuren Abo festhängst oder die Modelle auch über einen GPU-Anbieter wie Olamah Cloud laufen lässt. Ich kann hier zum Beispiel links auf Models-Search gehen und sehe hier einige Modelle, die ich Ich kann hier zum Beispiel links auf Models-Search gehen und sehe hier einige Modelle, die ich einfach direkt herunterladen kann, wie eben zum Beispiel hier Gamma 4 mit 12 Milliarden einfach direkt herunterladen kann, wie eben zum Beispiel hier Gamma 4 mit 12 Milliarden Parametern. Zum Beispiel die Deep-Research-Funktion oder auch die Möglichkeit, Dokumente zu bearbeiten, Zum Beispiel die Deep-Research-Funktion oder auch die Möglichkeit, Dokumente zu bearbeiten, also nicht nur hochzuladen und darüber zu schreiben, sondern zum Beispiel einen Dokumenten-Editor, also nicht nur hochzuladen und darüber zu schreiben, sondern zum Beispiel einen Dokumenten-Editor, der wenn jetzt hier ein Text generiert wird, hier rechts aufgeklappt wird und man dort der wenn jetzt hier ein Text generiert wird, hier rechts aufgeklappt wird und man dort sogar dann einfach weiter schreiben kann, um den Text anzupassen. Aber ich wollte jetzt einfach diese Standard-KI-Uberfläche, wie bei Chatchi-BT oder Cloud, wo man einfach Aber ich wollte jetzt einfach diese Standard-KI-Uberfläche, wie bei Chatchi-BT oder Cloud, wo man einfach nur schreiben kann, mit Dokumenten arbeiten kann und so weiter, vielleicht auf seine E-Mails, nur schreiben kann, mit Dokumenten arbeiten kann und so weiter, vielleicht auf seine E-Mails, seinen Kalender und sowas zugreifen kann, nur eben lokal. Ich habe zum Beispiel das Quen3-Modell jetzt hier und auch Gemma4 Ich habe zum Beispiel das Quen3-Modell jetzt hier und auch Gemma4 und ich kann jetzt beiden Modellen dieselbe Aufgabe stellen, Schreibe ein Gedicht über Elefanten und ich kann jetzt beiden Modellen dieselbe Aufgabe stellen, Schreibe ein Gedicht über Elefanten und ich sehe jetzt nicht, welches Modell welches ist und ich sehe jetzt nicht, welches Modell welches ist und weil meine Grafikkarte jetzt nicht beide Modelle gleichzeitig packt, und weil meine Grafikkarte jetzt nicht beide Modelle gleichzeitig packt, wird erstmal Modell A jetzt hier ausgeführt und dann Modell B.

Und die KI selbst erstellt den kompletten Workflow, entscheidet über die notwendigen Schritte, führt den Prozess aus und verbessert den Ablauf im Laufe der Zeit sogar noch weiter. In diesem Workflow, den ich vom Deep Agent erstellen ließ, also unabhängig von deinem Tätigkeitsbereich, In diesem Workflow, den ich vom Deep Agent erstellen ließ, also unabhängig von deinem Tätigkeitsbereich, kannst du einen Workflow mit allen Elementen und allen Schritten erstellen, die automatisch ausgeführt werden sollen. Also ich werde diesen Workflow erstellen und ab dann bekomme ich all diese Daten einfach, Also ich werde diesen Workflow erstellen und ab dann bekomme ich all diese Daten einfach, indem ich sie hochlade. Schau mal, hier habe ich mein Dokument, in dem ich eine Einleitung habe, die Graphicking Schau mal, hier habe ich mein Dokument, in dem ich eine Einleitung habe, die Graphicking und am Ende das Fazit, zu dem er gekommen ist, nachdem er die ganze Analyse dieses Dokuments und am Ende das Fazit, zu dem er gekommen ist, nachdem er die ganze Analyse dieses Dokuments gemacht hat, das ich hier hochgeladen habe. Also Leute, das ist echt cool, weil du einen Workflow erstellen kannst, ohne programmieren Also Leute, das ist echt cool, weil du einen Workflow erstellen kannst, ohne programmieren zu können, ohne die Prozesse manuell ändern zu müssen, nichts davon.

I can visualize each step of my workflow and to see if it s working, let s upload an invoice here. I could throw all of that into this workflow here and I would automatically get this result without having to run a single line of code, without having to do absolutely anything. I have some other instructions here that will guide the artificial intelligence so it can build my workflow as completely as possible. Within this workflow that I asked deep agent to create for me, so regardless of your field of work, you can create a workflow with all the elements, all the steps to be executed automatically. You just need to talk to AI, ask for what you want and it will automatically build your workflow the way you want.

Das zu generieren hat 20 Minuten gedauert, hat nicht mal irgendetwas gekostet, weil es im Abo inbegriffen war und ich konnte das dann einfach als HTML Code rausspielen und in Cloud Code dann sagen: Hey, benutzt bitte immer diese 15 Templates und füll das eben mit unseren Bildern, die wir gemacht haben, unseren Texten, die wir immer schreiben, musik irgendwelchen Inspos aus meinen YouTube Scripts und mach so unser E-Mail Marketing. musik Das heißt, ich brauche nicht ein Abo bei Nano Banana, ich brauche nicht ein Abo bei Seance, ich brauche nicht ein Abo bei was auch immer, sondern ich habe einziges HixF Abo und kann aus diesem Abo auf alle guten und gängigen Modelle zugreifen. Man kann hier einfach auf Supercomputer gehen und ich kann dem dann einfach beispielsweise sagen, generiere mir bitte einfach nur einen 30 sekundigen Werbespot im Premium Look für mein Produkt. Also wirklich eine Zeile in paar Wörter ohne Anweisungen, was auch immer mit ein paar Produktbildern wirklich nur drei musik Frames von bisschen Inhalt, bisschen Cover, dann fragt er mich noch ein paar Sachen, ich gebe ihm noch musik bisschen Kontext und dann passieren wirklich wilde Dinge. Heißt natürlich nicht, dass ich nie wieder Fotoshootings machen werde und heißt natürlich nicht, dass ich nie wieder einen Grafikdesigner brauche, aber ich kann so so viel mehr einfach selber machen und ich kann so viel viel schneller und viel günstiger erstellen und viel besser testen.

Und auf einmal merkt ihr, moment mal, für meine Code-Probleme liefert die Gratis-Version von Clote exzellente Ergebnisse Und auf einmal merkt ihr, moment mal, für meine Code-Probleme liefert die Gratis-Version von Clote exzellente Ergebnisse und für Sprainstorming oder Bilder generieren, da reicht Google s komplett kostenlose Version von Gemini völlig aus, und für Sprainstorming oder Bilder generieren, da reicht Google s komplett kostenlose Version von Gemini völlig aus, oft ohne irgendwelche spürbaren Limits. Denn Investoren zahlen diese Summe ja nicht für den Istestand von heute, Denn Investoren zahlen diese Summe ja nicht für den Istestand von heute, sondern für die absolute Marktdominanz, die sie in der Zukunft erwarten. Erinnert ihr euch an die frühen 2000er, als Browser wie Mozilla Firefox den Markt aufgerollt haben, Erinnert ihr euch an die frühen 2000er, als Browser wie Mozilla Firefox den Markt aufgerollt haben, lagen die Bewertungen von solchen Unternehmen bei ein paar Zig oder vielleicht 100 Millionen Dollar. Noch letztes Jahr hieß es überall, oh, Apple hat den KI-Zug verpasst, Google ist viel zu langsam, Noch letztes Jahr hieß es überall, oh, Apple hat den KI-Zug verpasst, Google ist viel zu langsam, dann kam der Februar und Chatchi-PT glänzte mit unglaublichen 900 Millionen Nutzern pro Woche. Kann OpenAI angesichts dieser extremen Marktdynamik und einer Konkurrenz, die KI direkt in unsere Smartphones baut, Kann OpenAI angesichts dieser extremen Marktdynamik und einer Konkurrenz, die KI direkt in unsere Smartphones baut, eine Bewertung von einer Billion Dollar rechtfertigen?

When a trigger fires, N8N passes the data through each connected node from left to right, exactly like an assembly line. The platform ships over 400 official core integrations, and the catalog recently crossed 500, as the team added over 35 new nodes between January and June 2026 alone, including dedicated nodes for Anthropic Clawd, Google Gemini, and Groke. You also get a code node where you drop raw JavaScript or Python directly into a workflow, no separate service needed. SAP needed a flexible orchestration layer inside Jule Studio, so enterprise teams could wire AI agents to SAP systems, external APIs, and custom data sources without writing bespoke integration code. Compared to simpler tools like Zapier, N8N gives you self-hosting for full data control, per execution billing instead of per step billing which gets expensive fast, and genuine code access in JavaScript or Python when a node alone is not enough.

So if we are over here inside of ChatGbt, we come into the model selector, we can see it s still GPT 5.5, but there have been so many articles that have come out exactly like this with ChatGbt users reporting that GPT 5.6 got leaked on Kodak because it is so much stronger and we could see there are tons and tons of these and I want to show you some examples of what s been coming out because it s pretty insane. So what you could take away from this right now is one, go to Kodak and see if it feels smarter, if it feels better, if it feels like it s using a brand new upgraded ChatGbt method and then the other thing to take away is ChatGbt 5.6 and ChatGbt 5.6 Pro should be launching next week officially and what you re going to notice with these models is one, they re going to be way better at coding. If you remember what pulse was, it was basically where they would give you news updates every day, but now they ve completely gotten rid of that and what you need to do now is if you come over here into schedule, they ve given us a brand new experience here. In addition to that, if we click on the left hand side right here and we click on settings right here, we re going to see that we can now come over here into apps and we re going to be able to actually come over here and download a bunch of different apps. And now what this is going to go off and do here is this is going to create an interactive HTML calculator right here that we are going to be able to interact with directly from inside of here.

Hey Rocket Kannst du mir sagen wie deutschland gestern bei der weltmeisterschaft gespielt hat und damit hallo und herzlich willkommen zu einem neuen video im jahr 2026 sagt uns nämlich unsere brille wie die ergebnisse der weltmeisterschaft gewesen sind kannst du mir auch sagen wie das andere gruppen spiel ausgegangen ist und deshalb gucken wir uns die rocket glasses an die kibrile die euch helfen kann euren alltag deutlich einfacher zugesteilten 1 gegen 0 gegen equador gewonnen Ja und das ist das gute stück die rocket glasses eine smarte brille die in euer sicht fällt quasi Informationen einblendet und gleichzeitig noch als ki assistent fungiert und das ist wirklich mal ein produkt wo ich sagen muss ich habe mich Echt drauf gefreut dass ich mir das anschauen durfte weil es einfach ein geiles stück technik ist was wird mitgeliefert also bekommt hier so eine kleine bedienungsanleitung Wobei das eigentlich gar nicht notwendig ist da kriegt es ein bisschen die knüpfe erklärt aber das ist eigentlich alles wirklich wirklich selbst erklärt Und ihr seht hier auch noch mal dass das auch auf deutsch verfügbar ist also kleine bedienungsanleitung check Wir haben einen kleinen lade mir adapter sage ich jetzt mal usbc auf das proprietäre format welches die rocket glasses haben wir haben Hier noch zwei zusätzliche nasenflügel aufsätze falls man die so nennt ein kleines Brillenputztuch und dann hier die verpackung wo die rocket glasses drin sind und ich nehme es schon mal vorweg die verpackung ist auch ziemlich Genial weil die ist zwar hier so dreieckig und schützt damit die briller hervorragend vor Stößen und ähnlichem Aber das ist tatsächlich noch nicht alles weil das ding könnte nämlich auch einfach ganz einfach zusammenklappen zack zack Und schon ist es ganz schmal zusammengepackt und nimmt gar keinen platz mehr weg finde ich eine ziemlich coole idee dass man das hier halt eben so ein bisschen Ich sage meine anfühlung stürchen origami mäßig Zusammenklappen kann und dann ist hier dieses stück technik welches mich an diesem test wirklich begeistert hat und damit ist das farzisch und gesagt Wir steigen ein ins video Ja so sieht sie also aus sieht erst mal aus wie eine völlig normale brillen und jetzt erkennt ihr aber hier zwei so kleine displays oder Beziehungsweise ja ich sage mal einen display wenn man so will Das wirklich wirklich gut in die brille eingearbeitet ist und das ding sieht halt optisch aus meiner sicht sehr sehr schick aus guckt euch das mal an Wir haben hier so ein bisschen sensorik da seht ihr die kamera auf der einen seite wir haben dann ansonsten hier unten natürlich die Zwei nasenflügel aufsätze falls die dinger so heißen Und dann hier hinten auch eine ich sage man an für die normale brille da oben ist ein kleiner knopf drauf an der seite haben wir noch Ne touch ein touch element wir haben die laut sprecher hier und natürlich die mikrofone Was dafür sorgt hier an der seite seht ihr mit diesem mit diesem waagerechten Strich quasi ich hoffe man erkennt den halbwegs genau der hier das ist so die touchoberfläche zweimal drückten bringt euch direkt in die ii funktionen oder in die k ii funktionen hier auf der rückseite beziehungsweise ganz hinten dann der lade Anschluss 49 gramm ist das ding leicht und dafür ist so viel geile technik drin verbaut die ich euch gleich zeigen werde und da muss man Einfach sagen hat roket wirklich etwas geschaffen was aus meiner sicht technisch im jahr 2026 ganz ganz weit vorne mit dabei ist und auch von der Verarbeitung und von der optik hat das einen wirklich sehr sehr guten eindruck hinterlassen bei mir So und wir steigen direkt ein und es ist hier quasi ein screen recording direkt aus der brille damit ihr genau seht was passiert Und vor allen Dingen auch was gesagt wird ihr hört alles mit ihr seht alles Das ganze ist natürlich nicht in 16 zu 9 format sondern eben im hochkant format aber das sollte uns gar nicht davon abhalten dass wir uns das Ganzes mal angucken ihr seht also wir können hier einmal durch das menü durchflippen haben verschiedene möglichkeiten uns dinge anzugucken Können auch ganz einfach den notiz anschauen und aktivieren beziehungsweise an legen und wir haben in der mitte den die k ii Funktionalität also alles ganz einfach wir haben jetzt hier als erstes mal die sprachübersetzung Und da könnt ihr jetzt einfach was einsprechen und bekommt die übersetzung direkt in der brille produzieren gerade ein video für den youtube kanal we love tech Und das funktioniert halt eben auch mit umstehenden personen ihr seid also im gespräch und die brille übersetzt euch direkt was euer gegenüber euch sagt 89 sprachen sind verfügbar und lassen sich dafür eben dann auch nutzen also sehr spannend teleprompter ist die nächste Möglichkeit die wir haben wenn ihr Vortrag halten müsst bekommt ihr einfach euren text eingeblendet und er läuft dann mit euch bzw. mit eurer erzählung quasi durch und ihr seht auch dass das ganze oberhalb eures sichtfeldes eingeblendet wird in dem fall der fernseh so ein bisschen der zentrale punkt Damit ihr eben auch euer zuschauer publikum oder wen auch immer euer publikum euren zu eurer zuschauer eben noch angucken könnt Beziehungsweise sehen könnt habt gleichzeitig aber oben dann euren lauftext der euch eben anzeigt dass man oder beziehungsweise was ihr entsprechend zu sagen Also auch das schlau gelöst dass eben der text nicht direkt in der Mitte eingeblendet wird damit ihr nicht irritiert seid Falls ihr eben auch mal ins publikum gucken möchtet So gucken wir noch mal weiter die nächste funktionalität ist musik ihr könnt also über bluetooth quasi Von eurem smartphone die musik aktivieren und deaktivieren natürlich auch über die brille höher anders hat ja sie hat ja auch laut sprecher so dass ihr direkt quasi das ganze nutzen könnt die kamera funktionalität zeichel später noch mal die kann man nämlich nicht mit dieser aufnahme Diese aufnahmemöglichkeit hier aufnehmen und als nächstes haben wir die funktion untertitel gucken wir uns auch einmal an jetzt seht ihr quasi alles was gesprochen wird und gerade wenn ihr vielleicht nicht gut hören könnt oder ähnlich ist dann ist das natürlich eine große Hilfe wenn man genau das eben sieht Was andere um einen herum und die mikrofone sind ja so eine art 360 grad Lösung dass man quasi also dann 60 grad nicht ganz korrekt aber dass man quasi Auch die umstehenden hört dann bekommt man das direkt hier eingeblendet und das ist natürlich eine wirklich sehr sehr schöne geschichte Weil damit das visualisiert wird was quasi gesprochen wird Ja aus meiner sicht auch extrem spannend weil dieser text wird jetzt zusätzlich auch in der app abgelegt das heißt ihr habt wie ein Gesprächsprotokoll sozusagen und gleichzeitig ist es halt eben eine super funktion wenn ihr vielleicht nicht so gut hören könnt oder ähnlich ist Und euch dann eben die brille direkt visualisiert was gesagt wurde aber das habe ich ja gerade eben selber auch schon gesagt also von daher das Definitiv auch eine coole funktion als nächstes haben wir noch Hi Rocket Kannst du dir bitte die telefonnummer 0 1 5 3 4 8 11 11 11 merken Also die notizfunktion ihr könnt ernst einfach in die brille reinsprechen quasi über die aktivierung Sowohl über doppelklick Beziehungsweise mit zwei fingern tippen auf die touchoberfläche als auch einfach über das aktivierungswort oder die aktivierungsworte Und habe dann eine notiz die angel angelegt und abgelegt wird auch das könnt ihr nachher in der app euch wieder angucken gleichzeitig habe Dies hier aber auch unter den notizen in der brille direkt dann jetzt gespeichert Hi Rocket Kannst du mir sagen wie die bundesrepublik deutschland ihr erstes spiel In der wem gespielt hat So und das haben wir jetzt noch mal gemacht weil ich euch zeigen wollte dass selbst wenn ihr so ein bisschen ich sage mal nicht durchgängig redet Ja also das war jetzt zwischen den einzelnen punkten der frage waren ja so ein bisschen zwischen Zwischenzeitliche pausen sage ich jetzt einfach mal und das system hat ja auch erkannt dass es mehrere absätze sind aber und das ist das schlaue daran Es fügt es natürlich trotzdem zu einer frage zusammen und ihr seht hier auch noch eine sehr ausführliche antwort also zum einen natürlich die antwort auf die frage Und gleichzeitig werden euch tatsächlich noch zusätzliche informationen gegeben also die kifu Funktionalität war wirklich richtig begeistern Gemini ist da ja der partner bzw. könnt auch chat gpt auswählen und da muss ich sagen Das hat bei mir wirklich wirklich richtig gut funktioniert die antworten waren knackig und auf insbesondere auf die fragen die ich da entsprechend gestellt habe So ihr seht jetzt seht die oberfläche ein bisschen anders aus ich habe jetzt ein widget aktiviert das kann man machen wenn man möchte mit uhrzeit Text der der songs wetter und so weiter und sofort aber es geht mir um was ganz anderes denn Ich stelle einen termin für 13 uhr 5 am heutigen tag zur aufnahme des rocket videos Es ist halt ein wirklicher assistent er legt jetzt quasi einen termin an weil ich ihm das gesagt habe also die brille ist wirklich für den tag Für einigen auch für den arbeitstag eine wirklich richtig richtig coole funktionalität Weil sie einem wirklich praktisch nutzen bringt das ist das entscheidende ja also ihr bekommt Infos eingeblendet wir kommen ja nachher noch benachrichtigungen können videos aufnehmen und so weiter und sofort aber ihr könnt halt eben auch Termine notizen und so weiter und sofort anlegen rückfragen stellen euch was übersetzen lassen es ist ein alltags wertzeug muss man sagen Und das vor allen Dingen auch wenn ihr euch mal verirrt habt also hier seht ihr jetzt die navigation auch die funktioniert direkt in der brille Wir haben jetzt eine navigation gestartet gehen jetzt hier von unserem büro zum kölner dom das würde ewig dauern ihr seht das Aber wir versuchen es einfach mal und wir gehen natürlich nur los also ich laufe jetzt nicht die ganze strecke das dürfte Denklich klar sein aber hier zeigt euch das system jetzt zum einen eine größere karte an die wird natürlich insbesondere interessant wenn es nachher bisschen kleinteiliger wird jetzt sind wir gerade im grausgesumpen bereich Aber ihr könnt eben auch die große karte in dem in eurem direkten sicht fällt so ein bisschen ausblenden und habt dann eben nur die untere Information mit den anweisungen was denn als nächstes zu tun ist und da muss ich sagen Das ist auch wirklich wieder sehr praktisch für den alltag jetzt vielleicht weniger fürs autofahren aber theoretisch auch da praktisch vielleicht eher fürs fahrrad fahren oder vor allen Dingen auch für also wenn ihr Fußgänger seid könnt ihr einfach einen blenden lassen wo ihr lang gehen müsst sozusagen und bekommt immer die hinweise Wo ihr gerade abbiegen müsst und so weiter und sofort richtig richtig coole funktionalität Und natürlich ihr seht es hier bekommt ihr auch eure benachrichtigungen vom smartphone angezeigt und könnt auch anrufe über die brille führen bekommt auch eine info wenn jemand anruft also auch dieser pragmatische nutzen ist absolut dabei und das hat auch in meinem test wirklich ganz gut funktioniert sowohl telefonieren als auch benachrichtigungen erhalten So konfiguriert und gesteuert wird das ganze dann entsprechend über eine companion app die auf eurem smartphone läuft hier seht ihr wir haben verschiedene Möglichkeiten wir sehen hier auch das alles was wir mit dem ki assistenten besprechen hier nochmal dokumentiert ist wir können also alles noch mal nachlesen Was wir in der vergangenheit mit ihm besprochen haben Und wir haben noch ganz viele andere möglichkeiten das ist ein bisschen das zentrale steuerungselement und die brille ist dann das Vehikel was quasi euch zu diesen möglichkeiten dann bringt wir haben gesagt notizen können wir anlegen wir können termin planen termine Anlegen lassen für uns wenn wir das mal haben möchten wir können die übersetzungsfunktionalität uns im nachgang nochmal angucken wir können den Teleprompter den text quasi hier einstellen und auch die scrollgeschwindigkeit und so weiter aktivieren navigation haben wir uns gerade angeschaut Bei den untertiteln ich habe es vorhin schon gesagt wird natürlich euch zum einen der text angezeigt zum anderen habt ihr den aber hier auch noch mal drin Das heißt ihr könnt jetzt hier den text euch nochmal angucken der vorher gesprochen wurde wenn ihr im meeting seid wenn ihr irgendwo anders seid Wenn ihr nicht richtig verstanden habt was gesagt wurde und im nachgang das nochmal nachgucken wollt überhaupt kein thema Wir haben grundsätzlich auch die möglichkeit das hatte ich vorhin schon gesagt und euch auch gezeigt mit den widgets wetterdetails Zeitplan also euer kalender und die uhrzeit könnt ihr euch einblenden lassen das dann aber ein bisschen im sichtfeld das muss man halt schauen ob einem das ganz gut passt oder eben nicht wir können auch noch ganz viele sonstige dinge einstellen sound effecte Können wir einstellen wie lang soll das display eingeschaltet sein und so weiter und sofort also wirklich richtig richtig cool Ich habe es vorhin schon gesagt bei den k i modellen steht zum einen gemini zur verfügung aber auch chat gbt das könnt ihr euch hier Aus suchen welchen ihr verwenden wollt ich habe jetzt die ganze Zeit mit gemini gearbeitet weil das halt ganz ganz gut funktioniert hat ihr könnt auch die helligkeit des displays lautstärke und so weiter und sofort einstellen und hier seht ihr auch noch mal alle bilder beziehungsweise videos die Er aufgezeichnet hat denn es ist auch eine kamera mit verbaut 12 megapixel Die dafür sorgt dass ihr zum einen natürlich videos aufzeichnen könnt wie wir das hier gerade tun oder aber auch natürlich fotos aufzeichnen Könnt das ist dann immer so ein bisschen eine schöne geschichte weil man halt die kamera quasi dabei hat das funktioniert auch ganz easy ist oben der eine knopf die wir vorhin schon gesehen haben drücken video aufzeichnen foto machen geht super easy und super schnell und super einfach Nur noch mal ganz kurz zum themafoto als ergänzung hier seht ihr dann die einstellungen die wir zum themafoto machen können wir haben natürlich die Möglichkeit uns da entsprechend auch ein Ein format auszuwählen beziehungsweise eine auflösung auszuwählen wir können wasserzeichen hinterlassen und wir können die dauer einstellen die Das video falls wir denn auch ausnehmen möchten haben soll So gucken wir uns einmal an was jetzt zusammenfassend meine meinung zu den rocket glasses ist also ich muss sagen das jahr 2026 wir schreiben es 49 gramm technik können diese leistung erbringen die wir gerade eben gesehen haben ein ki assistent mit drin eine kamera mit drin laut sprecher mit drin mikrofon mit drin alles mit drin in einer brille die ich ganz easy bei mir auf die nase setze und damit durch die jengen laufe und das system hilft mir eben alles zu bewältigen laut sprecher sind ausreichend laut ihr versteht gut was passiert eure umgebung weniger die displays sind hell genug ich musste sie tatsächlich aber relativ hell einstellen aber das ist Denke ich klar bedienung über die knöpfe und über das touch element super einfach ihr habt natürlich auch die sprachsteuerungsmöglichkeit Und vor allen Dingen was mir besonders gut gefallen hat ist die displays sind nicht zwingend direkt in eurem blickfeld ihr könnt also quasi durch die displays durchgucken Das war wirklich richtig stark und die funktion ich habe es vorhin schon gesagt echt echt geil gerade navigation hat mich echt begeistert und auch die direkte Übersetzung von den 89 sprachen wirklich wirklich cool in der summe muss man sagen das teil ist nicht mal teuer link findet ihr in der video beschreibung Ihr bekommt dafür ein echt geiles stück technik was euch vor allen Dingen im alltag hilft der akku hätte ein kleines bisschen länger laufen können aber ich glaube ihr kommt ohne probleme über Ein tag wenn nicht so viel rum spielt wie ich das hier gemacht habe wie gesagt link findet ihr in der video beschreibung wenn auch fragen hat schreibt mir gerne in die kommentare Danke euch fürs zusehen und wenn das sage ich wie immer Alexa kamera aus okay

Hermes uses the .hermes folder, which holds all the configs and info that run the agent, and all of that lives in one single file called config.yaml. For that, you can either set max bytes directly in the config.yaml or change it to the number you need using the Hermes config command. So, if you want to increase that, you can change it with the Hermes config command and set the character count you need. When you first set up Hermes, you give it the models for different purposes, but you can set up auxiliary models as well. There's no direct way to set this up, so you actually have to do it in config.yaml, or you can just ask Claude Code or Hermes to do it for you, and it'll make the changes itself.

Probably the easiest way to do that is just to copy this prompt into a new chat, and then give it context on what you re trying to achieve with the daily task, and what stories you didn t like, and how you could modify that prompt now. But then even if you re on GPT 5.5, which is the newest one and probably the default that everybody should just be using all the time unless you have a really, really, really particular reason that s based on your experience. If it s on pro extended, it s going to generate those 10 ideas, but then it s going to ask itself like 50 or 100 different follow-up questions to stress test those ideas or evaluate if it actually read your intent correctly and how it could make these better. So, what they re doing is they re trying to interactively insert it when you write something like an essay or some other long-form text, see it appears here automatically. So, don t think of this as your complete email operating system, more like it s a really simple way to get a lot of them done, but then still go into the inbox and look at if everything actually got pulled in and handled.

So I will go to the outbound call where Tbilio and here you will see the CURL command for initiating an outbound call and if you are not sending any kind of dynamic information like name, city or address then you will simply copy this CURL command and paste it here in import CURL. So I am giving it a name, Xi API key and then for adding the API key go to the 11 labs and then search for developer section and scroll down and go to the developer section and here you have to go to API keys and click on create key. So my new API key is created and I will just copy this and paste it here in Anant and so after this you have to provide the agent ID and agent phone number ID. So I will copy this agent ID and paste it here and paste it here and after that you have to provide the agent phone number ID. So now I have pasted my agent ID and agent phone number ID and now my Anant knows that which agent he has to use in 11 labs and which phone number is assigned to this agent and after that you have to provide the number.

Matt und ich erkläre euch KI so einfach, ich kann, folgt mir gerne, wenn euch so was Matt und ich erkläre euch KI so einfach, ich kann, folgt mir gerne, wenn euch so was interessiert. Im Internet, der schreibt kaum jemand so was wie, ich weiß es nicht oder keine Ahnung, Im Internet, der schreibt kaum jemand so was wie, ich weiß es nicht oder keine Ahnung, da wird bei Unwissenheit entweder nichts gepostet oder extrem selbstbewusst etwas Falsches da wird bei Unwissenheit entweder nichts gepostet oder extrem selbstbewusst etwas Falsches behauptet. Für Jetschie Pithi sieht die Welt deswegen einfach aus wie ein riesiger Klausurordner, Für Jetschie Pithi sieht die Welt deswegen einfach aus wie ein riesiger Klausurordner, voll mit Musterlösungen, die teilweise extrem selbstbewusst aber falsch sind. Und Jetschie Pithi lernt ja Frage rein, Antwort raus und nicht Frage rein und Unsicherheit Und Jetschie Pithi lernt ja Frage rein, Antwort raus und nicht Frage rein und Unsicherheit raus. Das heißt, wir beschweren uns, dass Jetschie Pithi halloziniert, aber wenn es mal Unsicherheiten Das heißt, wir beschweren uns, dass Jetschie Pithi halloziniert, aber wenn es mal Unsicherheiten zugibt, dann fragen wir einfach nochmal nach, nur eben ein bisschen passiv aggressiver.

Right now, if you are running something like AI Profit Boardroom, you have got months of YouTube scripts, community post archives, welcome sequences, onboarding flows, member FAQs, weekly newsletters, lead magnet copy, and course content. With 1.5 million words of live context, you load everything in at once, the entire content library, the complete brand voice, every piece of member feedback ever collected, every question the community has ever asked. One prompt does it in minutes, and the output actually sounds like the brand because the model has seen everything the brand has ever said. You give it one goal, research the top AI topics generating the most discussion this week, write a YouTube script for the highest priority topic matching the existing brand voice, convert that script into a five email nurture sequence, a LinkedIn post series, and a short-form video outline. The page should explain what they get inside the community, why is the right next step after learning about this model, and make signing up feel like an obvious decision.

So now every single day Gemini research is the latest in AI and hands you a ready made infographic enough to feed an entire content channel without you lifting a finger and it connects to the tools you already use including Gmail, Calendar and Docs as well as services outside Google like GitHub, Asana, HubSpot, MailChimp and Salesforce. Now let me open a tool most people never even find it is called Google AI studio and here is the best way to picture it. Let s use this tool to build a music app together design its logo record an actual AI podcast for it and build it a proper business pitch page one honest heads up before we begin these screens look a little more technical than the normal Gemini app do not let that scare you 90 of what we do is simply prompting. This main screen is called the playground look at these six blocks think of them as a menu of everything Google AI can do for you up here you ve got coding and chatting then there s image generation video generation speech and music and real-time voice and video. Now click this drop down on the right it shows you every model Google has ever built all in one spot Gemini for chatting and coding nano banana for images Vio for video Lyria for music and plenty more you can choose to use any of these models because they all live in one window and by the time we are done you will have used four of them to build a final product there s also these little settings you need to know first is tools in simple words these let the AI actually go and do things before preparing a response for example look at this one grounding with Google search when you turn it on the AI goes and does a Google search before it answers you so you get today s information fresh not something dusty from a week or month ago there s another that handles code and a third tool that opens Google Maps then you ve also got an environment section which is nothing but a clean space for the AI to work and build entire applications in finally let s look at the last feature you absolutely need to know which is system instructions if you remember nothing else from this tour remember this one think of it as an instruction where whatever you write the AI obeys it every single time so you never have to repeat yourself watch what we tell it you are here to think with me not agree with me treat everything I say as a claim that needs verification stay skeptical by default double check facts and don t accept anything without reasoning I am not always right and neither are you so focus on evidence clarity and accuracy if my logic has flaws point them out if something is unclear or missing tell me prioritize truth over politeness and challenge me whenever needed so we reach the most accurate answer possible now this prompt matters a lot in the kind of responses we get from this AI if you ve noticed AI generally operates like a yes man you say something and it agrees with you even when you are completely wrong this one instruction flips that now it will push back double check facts and question us let s name this instruction learn with me and save it now that we ve set it up once it ensures execution every time we use this tool okay now for the part you have been waiting for let s build the music app let s click on new app and just describe what we want build me a music app similar to Spotify it recommends playlist and suggests music for me it has a quiz that figures out my music taste and shows me how to explore new styles and it provides a personalized dashboard that tracks my listening history and my discovery progress name the app news quest make it a mobile first app focused on engagement and discovery it s a detailed prompt that s well thought out and lists a numerous set of features let s see if google can execute it let s hit build and straight away it gets to work first it asks how we would like it to look and offers a few design styles we like the first one so let s click select this design now it s writing the code for us and finally here is the app it built rather lovely is it not it s got a good design a tidy layout but watch closely let s click a track to play it but as you can see there s no audio coming out just silence it looks every bit like a music app but it does not actually play music yet but don t worry we do not go in and fix it ourselves we simply tell it what is wrong if I click on any music track make use of the lyria 3 model to add music in the background of whichever track I click on for those who don t know lyria 3 is google s music making AI it does not pull songs from a library it composes the music itself from scratch so we are telling our app reach out to lyria and the moment someone taps a track create real music on the spot we send that off and it starts wiring it in okay now watch closely let s go to synthwave and hit play and just listen that is real music created right now this second by lyria inside an app that did not exist 20 minutes ago we never uploaded a single song the app is composing the music all by itself that is genuinely wonderfully wild and it is not just a player let s step into this curator section we pick a vibe say coffee focus and hit start discovery quiz give it a second and there it is a custom set of tracks for that exact mood ready to save as your own playlist then there is this lovely personality quiz it asks you playful questions your ideal weekend how you listen to a song you adore and at the end it hands you your music personality with tips on what to explore next and remember this is the very quiz we asked for in that one paragraph it genuinely built it there is even a stats page tracking how many genres you have explored and how much you have discovered that personalized dashboard we asked for sitting right here exactly as promised and remember we asked for mobile first let s click this mobile button and there it is the same app perfectly reshaped to fit a phone nothing broken everything tidy ready to use in your hand one honest thing to know if you ever want to put this live for real users the publish button asks you to set up billing first so building and playing with it like this is completely free going fully live is the one paid step and you can connect it to your other google tools drive docs forms whenever your app needs them so let s take a breath here we typed one paragraph and we got a working music app with real AI music a quiz a dashboard and a phone version no code anywhere now let s give this little thing a proper brand every brand needs a logo and normally that means a designer but let s just make one ourselves right here let s click image generation we get two options nano banana two and plain nano banana we will use the plain one because it is free the pro version costs money but for a logo free is more than enough and as usual we describe what we want build a logo design for my music brand the brand shortlist playlist quizzes you on your taste and helps you discover different genres of music my brand colors are black and purple it should feel a bit like Spotify and in a few seconds it hands us a logo clean purple a little music waveform there is one tiny problem though it named the brand melody lab that is not our name so once again we simply talk to it the name of my brand is muse quest not melody lab it politely apologizes and fixes the name in an instant this is our new logo here s something to know about these ai generations you will almost never get it perfect on the first try and that is completely normal you just keep refining it and giving it suggestions you can say put a guitar inside a circle make the design white on a purple background make the guitar bigger tuck the name inside the guitar take off the extra text outside the circle just keep talking to it until it looks exactly right and the moment you are happy download it and now you have your own custom brand logo now let s go back to our muse quest app and we literally pick up the logo we just made and drop it right into the chat telling it this is my brand logo add this brand logo to my app it thinks for a moment updating the header and there it is our logo sitting at the top of our app on the desktop version and the phone version too now let s say you want to promote muse quest with a little podcast clip you do not need a microphone or two people or any recording gear you can direct ai voices like little actors let me show you let s go to speech and music and look it offers us ready-made templates and everyday assistant a storyteller an ad voiceover a podcast host we will click the energetic co-host because we want that podcast feel here is how it works you set the scene two friends chatting into mics in a studio and then you build each voice one at a time this next bit surprises everyone so stay with me for our first speaker let s write out a whole personality we want an authoritative main news anchor make the person sound like a dictator dominating the conversation somebody who puts his points straight up then we get to shape everything about how it sounds the style let s set to newscaster because it s professional and authoritative the pace rapid fire because it s fast and energetic and for the accent let s choose american south and then we sample a few voices clicking through them until we find the one that fits let s go with this one akkad now let s work out the same thing for our second speaker the co-host let s give her a personality too we want a professional field correspondent with over 25 years of expertise in AI and marketing who puts a point forward now let s pick a warm female voice for her let s go with call your hoe so now we have two AI characters each with their own personality accent and pace ready to talk you can also add speech blocks of what you want the podcast to look like right now let s just hit run and see what comes out welcome back to the show today we re diving into the intersection of AI and creative expression exactly i ve got so many thoughts on what happened this week it really is shifting daily i mean did you see the demo they dropped on tuesday oh absolutely it blew my mind but it also raised some pretty huge questions about how we define art in the first place right so let s get right into it because we have a lot to unpack today did you hear that that is a proper back and forth podcast around AI and creativity one of them even laughs the pacing the energy the personalities we dreamed up all of it is there and remember we did not record a single second of audio we simply described our characters and pressed a button you can do this and a lot more using speech and music feature the only limit is your imagination one more lovely thing you should know because it saves you so much time you do not always have to start from a blank screen this is the gallery these are apps and games other people have already built and you can open anyone use it and remix it into something of your own just look at the range a hide and seek game where you play against the AI on google maps multiplayer games music apps built on lyria design tools built on nano banana video tools on veo and whole 3d worlds so if you ever feel stuck on where to begin then do not start from scratch simply pick one of these see how it works and make it your own i have to show you something else quickly because it honestly feels like the future let s go back to the grid of six and click on real time this is basically a live translation and it handles over 70 languages speech to speech almost instantly so you could speak in english and hear it come out in another language as the words are still leaving your mouth and the second part is even more useful for learning you can share your screen point at whatever has you stuck and just talk to it like a friend while it watches your screen alongside you picture this if you are stuck on something on your laptop instead of googling for an hour you just show it the screen and ask your question isn t that amazing well we ve discussed a lot of features and built a lot of things like the app logo podcast but this last one ties the whole story together let s make a pitch for this like a real business for this we need to go to code and chat and ask it to build a partnership pitch page here is the dream muse quest our little music app wants to team up with t-series one of the biggest music labels in the world create a premium partnership landing page for muse quest proposing a strategic collaboration with t-series position muse quest as an emerging music discovery platform and invite t-series to distribute their music catalog to reach new audiences use t-series inspired brand colors a modern high end design and messaging focused on innovation audience growth and long-term partnership let s send it off and watch it code by itself about a minute later it gives us a result which we just need to download and open look at this it s a professional pitch website it s got a headline about bridging india s biggest music catalog with audiences across the globe to be honest that s a little longer than i like but it can be changed later then it has a four-phase partnership plan then it s got the vision section and a part explaining why a label should partner with muse quest even the technical details of how the two would connect and a contact form at the bottom to request a proposal deck and consultation so that was google ai studio for you if you use this tool to build anything upload it on social media and tag us we d love to see the results you got now i have only just scratched the surface with these first two tools let s carry this momentum and move on to the next one this next tool comes under google s research tools it s one of a kind and you ve definitely used it before however almost nobody i know uses it to its full potential and that s a real shame because it might just be the quietest powerhouse on this entire list the tool we re talking about is notebook lm it can not only read the numerous things you feed it from youtube video transcripts to entire books to websites and even audio files it can also turn that pile of sources into almost anything you need a grounded research report a debate style podcast a narrative animated video mind maps study guides flashcards quizzes infographics and even clean data tables so let s show you some of these use cases by building something exciting together and i hope you leverage it to its full potential the next time you use the tool let s dive in when you sign in with google this is the dashboard you land on if you ve used it before all your notebooks or basically research projects live here and if you re new there s just one button that matters now which is create a new notebook so let s click it it spins up an empty notebook and immediately asked for a source you can pull sources from your documents websites youtube videos or your own notes you can even search the web or pull projects straight from your google drive folder before you start the research you ll see a little globe drop down with two modes fast research and deep research fast is great for quick results deep is for when you want a proper in-depth report great now let s upload our files we ve got a document on tesla so let s drop that in the moment it lands it reflects on the left hand panel tesla dot doc x ticked and ready and notebook lm even reads it and titles the whole notebook for us tesla 2026 the triad of autonomy robotics and ai and now you can see the whole workspace which is really just three columns on the left your sources everything you feed it in the middle the chat where you talk to those sources like a person who s read them and on the right the studio where you can turn your document into an audio overview a slide deck a video overview and all these other options there s even an add note button for anything you want to jot down yourself that s the whole map now let s start filling it in the first thing notebook lm does extremely well is deep research so let s make it go out and research the open web for us in the chat we ask one simple question what is tesla looking into in the next three years there s a little button that switches the search from reading only your document to reading the whole web so we flip that to web then we open the drop down and choose deep research and we hit submit it makes itself a five step research plan and works through it one step at a time and here s what came back a complete deep research report called the physical ai pivot with 40 sources discovered underneath it we click import and every one of those 40 sources cascades straight into our panel on the left indexed and ready and each source has a checkbox so you can stay in control of the data you want you can tick exactly which ones you want notebook lm to draw from we ll select a handful of the most relevant the 2026 update the fsd and robotaxi roadmap the optimus deployment and a few more now automatically what happens is that notebook lm writes up a clean overview based on your sources it lays out tesla s strategic shift from a traditional car maker into an autonomy and robotics company you can hit save to note to keep that answer and you can even hit convert to source which turns the answer itself into a new source the tool can cite later and then notebook lm does your thinking for you suggesting the next questions to ask it throws up three how is tesla shifting focus from cars to ai and robotics what are the biggest risks facing tesla s new technology bets and how does tesla s robotaxi progress compared to competitors like waymo that third one is interesting let s run it a full four-part competitive breakdown of tesla versus waymo covering fleet size technology vehicle design and cost comes back now here s one more thing i want to show you see those little numbers dotted through the answer hover over one and it shows you the exact source it came from click view source and you re taken straight to the original quote in your material and one more small button worth knowing auto label sources by topic hit it and notebook lm sorts all 40 odd sources into clean folders on its own autonomous driving battery technology energy products financial performance market competition new vehicle models robotics and ai and a miscellaneous catchall expand any folder and there are the exact sources that belong to it basically in one click your whole research library got organized now let s open the studio on the right where this document can turn into 10 different things we ll start with audio when you click the settings on the audio overview card it asks what format you want the audio to be in whether it s a deep dive or a brief or a critique or a debate where two ai hosts actually argue your topic out let s go with debate and now for my favorite feature if your first language isn t english language you can choose between hindi marathi Bengali gujarati english spanish and plenty more options to really push notebook lm will generate the first one in german and then do a second round in english to compare then it asks what the host should focus on so here s the actual prompt we give it debate if tesla s new cybercap production methods represent true innovation or just desperate survival discuss if existing competitors like byd have already surpassed tesla s core technology and analyze if the manufacturing process is truly scalable for global mass production what we ve done here is we ve handed the host a real argument with two sides and three specific questions to fight over it starts cooking in the background side by side let s run the same prompt again this time in english okay so it took some time but now the results have come back let s check out the german one first that was two ai hosts debating tesla s manufacturing in fluent german with a completely natural back and forth entirely from our document then we ve got the english version about five and a half minutes of two hosts arguing the same question do tesla s radical new production methods represent a revolutionary leap in manufacturing innovation or uh just a desperate survival tactic the stakes are definitely high exactly i argue this unboxed process is a fundamental technological breakthrough that completely reinvents how we build hardware and i contend it s a forced high-risk gamble one driven by a shrinking market share and honestly overwhelming global competition listen to this it sounds so natural there s a portion where one of them even laughs and the content is landing too the potentials of this are insane educational content can be made by schools using this tool to cater to students across different geographies and languages if you have any more ideas of the applications of such a tool drop them in the comments okay let s move on to the next use case slide decks a large portion of the time of educated professionals goes into researching for and designing powerpoint presentations let s cut that time down today when you open slide decks on notebook lm it offers two formats a detailed deck which is a presentation built for reading and sharing via email and then presenter slides which are clean slides with just the key talking points to speak alongside let s go with presenter slides it s going to use the same resources we uploaded for tesla but we still have to describe what exactly we want in the presentation so here s our prompt create a deck for beginners use a bold and playful style with a focus on step by step instructions now hit generate here s the slide deck we get titled the second s-curve teslas pivot from high volume automaker to hyperscaled ai and robotics platform they ve also given a proper financial dashboard the headwinds the war chest the diversification across energy and services and a strategic takeaway all pulled straight from the sources and if you want to go deeper on any part you just hit revise and tell it what to expand the deck reshapes itself around whatever you need it s that easy to make a pitch deck now similarly notebook lm can also make videos when you open the video overview card you ll get three format options cinematic explainer and brief let s choose brief then it gives you some visual styles as options we ve got classic whiteboard kawai anime watercolor retro print heritage and papercraft will go with papercraft and here s the prompt made using the focus box and suggested modifiers compare tesla s market position against competitors like byd lucid and jayomi in 2026 contrast the production speed and manufacturing efficiency between these major global automotive players analyze how the new unboxed manufacturing method impacts tesla s long-term competitive cost advantage once you hit generate the video starts rendering this is the final result it s called 2026 ev market shift it s about two minutes long in the papercraft style we asked for let s play it in the 2026 ev market the ground is already look at byd they actually dethrone tesla with 2.26 million battery electric deliveries tesla s big issue right now an aging lineup meanwhile chinese rivals are pumping out rapid smartphone style updates what was just built was a fully narrated animated explainer done entirely by notebook lm you don t need expensive editing apps and skilled workforce to create such a draft anymore the rest of the studio features work the same way pick the format give it a prompt and hit generate let s blow through five more of them super quick first let s explore mind map we will follow the same steps and give it a one line prompt tesla 2026 overview and future scope we hit generate and it starts building a clickable tree of the whole topic while that s happening let s ask it to build the rest when you click on reports it gives you a menu of formats like briefing doc blog post and study guide for revision we ll pick the study guide and let it auto generate for the quiz we choose a standard number of questions at easy difficulty and prompt this generate a 30 question quiz telling what exactly tesla is looking into next for the infographic we need to pick a language we choose english for orientation let s take portrait and for style we get options run from kawai and clay to sketch note anime and editorial let s take clay we also set the level of detail to detailed then prompt it use teslas brand colors and put three stars under the infographic and generate an infographic based on that and finally for the data table we prompt highlight the key findings from the resources with the title author and key results make sure the data is in bullet points and a tabular format and tell me the key insights with the metric each thing is being calculated in every one of these studio features is being generated at the same time in the studio panel let s walk through each of the results now let s start with the mind map we expand it to full screen and there s the whole topic as a living tree with tesla 2026 overview and future scope at the center click any branch and it unfolds autonomous mobility opens into cybercap production and one more click drills into the specifics every other branch goes just as deep from optimus to teslas energy storage the vehicle lineup and the full financial and ai strategy when you zoom out you can see the entire thing at once end to end down to the smallest detail it s a brilliant way to take everything you ve gathered and actually hold it in your head now let s look at the quiz it gives you 30 questions like what is the primary design purpose of the tesla cybercap pick the wrong one and it flags it red with a not quite and lights the right answer green while giving an explanation for each you can go through all the questions to understand what you know and don t know it s a brilliant study tool next look at the data table it turns everything into a clean grid the title author key results key insights the metric each figure is measured in and a source link for every row so you can always trace a number back to where it came from next the infographic came back the title reads tesla 2026 the great pivot to ai and robotics in the clay style we picked with a consistent color palette and small animations it s even added the three stars at the bottom exactly as we asked tesla s colors that we asked for have not been highlighted enough here maybe a different style language would suit what we re looking for better now finally let s look at the report we asked for a study guide and got a full revision pack a short answer quiz with questions and model answers a complete answer key essay questions on the bigger themes and a glossary of every key term for anyone prepping for an exam that s hours of work completely done for you so let s step back and look at what we just did from one document on tesla we pulled a 40 source research report a full competitor breakdown debate podcast in two languages a slide deck an animated video a mind map a study guide a quiz an infographic and a sourced data table anybody can use it for free so for all you students watching just drop your lecture audios and textbooks and walk out with free study material and quizzes since everyone is based on your own sources you do not have to be worried about ai hallucination either that was three free tools we covered let s carry this momentum and move on to the next one since we are learning about free tools around education here s one called learn your way it basically takes a topic from a textbook and rebuilds it in different ways shaped around how you personally like to learn the premise is that everyone learns differently some of us are auditory learners some are visual learners some of us learn with mind maps instead of handing a normal textbook to everyone this tool allows you to learn in whatever way suits you it s been tested and proven to increase knowledge retention by 11 percent in students compared to normal textbooks that s a decent result for a free tool so let s start and experiment with it this is the homepage it shows you the four ways that they reimagine the content in immersive text slides and narration audio lesson and mind maps now scroll down and you ll see the subjects google already has content around astronomy biology chemistry computer science economics and a handful more note that these are sample lessons google built so you can try the tool out without any setup you can also drop in your own textbook pdf your own chapter all your own notes and it runs that exact same transformation on your own material for now let s use one of the ready-made lessons so you can see how it works let s go with this psychology lesson here what is learning it asks you what personalization do you want to explore basically which example would you prefer a high schooler who likes skateboarding or an undergrad who likes music let s go with the high schooler google is now transforming and personalizing your study material improving the readability then splitting the material into different sections the first format we get is immersive text it gives you the learning objectives then some visuals to help you understand the concept these visuals help you remember the topic a different way so it s easier to recall than a normal definition or paragraph now if this was not enough it even quizzes you on what you ve just read look at this it s asking for a key characteristic of a reflex let s go with option d it marks it as correct while also giving a line of explanation now let s deliberately get one wrong and see what happens the tool gives you multiple options from a hint to an option to show you the answer or even skip it if you ask to see the answer it also provides an explanation of why it was correct so that was the first method immersive text let s look at the other three the second format is slides and narration let s hit play hello everyone and welcome to our lesson introduction to learning today we re going to explore what learning is how it works and the different ways we acquire new knowledge and behavior just like that it s built you a full 25 slide narration of the same chapter with clean slides and a voice guiding you through it you can explore it a little more yourselves but for now let s move on to the third format the audio lesson let s click to start it oh welcome to our lesson today before we dive into the topic let me ask you a question can you think of something you do automatically without ever having been taught how to do it basically it turned the entire chapter into a conversational audio lesson you can simply sit back and listen to while traveling walking or doing any other chore it s faster and easier than reading a textbook you just got a free audiobook basically the last format is the mind map to help you organize the information from the entire chapter easily it breaks down your topic into segments which also branch out themselves for example when we click on innate unlearned behaviors it splits into reflexes and instincts which branch into their own explanations so you can start right at the big picture and drill straight down into the single part you re stuck on without having to reread everything surrounding it then zoom all the way out and there s your entire chapter sitting on one single screen this is exactly what you want the night before an exam so overall what this tool learn your way does is it takes your topic and hands you back the reading and visual material a quiz a narrated video a podcast and a mind map and you get to pick whichever one you prefer it s basically the new form of personalized education now we re done with all the research tools let s go ahead and build something this next tool we re looking at allows you to describe any app you want in simple language and it goes ahead and builds it for you it s called opal and we re going to use it to not only make an app but also to reach out to companies to spot business opportunities for the tool let s start slow this is what the opal dashboard looks like a gallery google has already built a whole pile of ready-made apps which you can just open copy or improve based on your needs as you scroll you ll pass dozens of them a Diwali planner bookrex business profiler claymation explainer fashion stylist game concept builder interior designer marketing maven product marketing and so much more each of these is a working mini app so for a head start just grab one and tweak it let s open bookrex a little app that recommends books this is the whole app three boxes connected by arrows you ve got gather book description then find books then output basically you tell it what you want it acts as a literary assistant and finds five books with the description for each then it lays them out on a clean easy to read page let s run it it asks the kind of book i m looking for so let s give it a genetic ai explainers first it finds the book recommendations then gathers the explanations and purchase links for each one then it generates the html and designs the interface so it s building you a small web page to hold the answer here s the result five real books each with a proper description and a buy on amazon link let s look at the first one agentic artificial intelligence it s described as a strategic jargon free roadmap for enterprise leaders moving from generative ai to autonomous agentic workflows the second is agentic ai for dummies is an accessible introduction to agentic systems that breaks complex tech down into simple concepts similarly it provides three more now you can either download the result as an html file straight to your computer or click the app tab on top which turns the same result into a clean full screen grid of all five cards you can share with anyone that was a pre-built app right now let s build one of our own let s ask it to make something an e-commerce brand in india would pay money for for example an automation that creates product catalog images for a fashion store on autopilot every week let s give it a detailed prompt read this week s new product arrivals from mintra and add them to a google sheet run gemini with search to pull current seasonal and festival trend queues for each product category in the sheet apparel gets a model and lifestyle treatment accessories get a flatlay and detail set and festival wear gets a festive editorial treatment use the nano banana model to generate two catalog images per product one model shot and one flatlay with legible price and product names write the image urls back to the sheet as a new column drop a message saying today s mintra catalog imagery is ready set the chain to run every monday at 6 a.m and there it is it took that whole paragraph and built a complete workflow and it even named it catalog genie for us it s asked for your mintra link and google sheet then it works through the steps one by one first it fetches the mintra content next it pulls out the catalog data then it researches the current trends for those products after that it saves everything into your sheet finally nano banana creates the images and the last step drops those image links back into the sheet that was how to create an automation but now let s build something you can actually make money with say you run an agency or you re freelancing and you need to send cold emails to land clients the hard part is never actually writing the email it s researching each company well enough that the email doesn t sound like spam so let s build a tool that does both for you let s give this prompt to opal act as an elite b2b tech strategist use google search to research the business models of the three distinct companies listed in target brands then draft a highly tailored hyper personalized cold outreach email for each company pitching our ai automation agency services based on your offer slash purpose signing off professionally as an ai ops partner now look at these two bits at target brands and at your offer the at the rate sign makes them blanks you fill in each time you run the app so you build this once and then you can reuse it for any company and any offer you want basically the same tool but with new clients every time okay so now it builds the app calling it strategy outreach ai with two inputs target brands which are the three companies to research and your offer which is your agency s value proposition the middle step is research and draft emails where it acts as a b2b strategist googles each company s business model and writes the emails then in the final step generate html outreach lays out the results on a clean page let s head to the app view and use it now it asked for my three companies so let s use three real indian brands mokobara blue toki coffee roasters and the sold store then it asks what i m actually selling so let s describe our offer an audit and customized multi agent ai pipeline to automate customer operations cart recovery and conversational support to scale efficiency by 40 percent and now it goes to work analyzing the objective researching each of those three brands on google then drafting a different email for each one and finally designing the dashboard to hold them all let s give it a minute it s given us three different well researched and written emails let s read mokobara s first this is a premium luggage brand so the strategy it s chosen is high ticket conversion it pitches cart recovery over whatsapp something for warranty claims and a travel concierge it says this will increase efficiency by 40 percent now let s look at blue toki the coffee brand this one gets a completely different angle focusing on subscription value and retention since coffee is something people reorder it pitches a whatsapp layer to manage subscriptions profiling that learns each customer s taste and recommends coffees for them and an agent that nudges customers and then finally let s look at sold store the fan merch and apparel brand this one gets a different angle again it focuses on return optimization the email notes they ve mastered fan merchandise but are facing a sizing and returns bottleneck so it pitches solutions for the same including an online size and fit agent to cut returns and smoother exchanges to keep money inside the brand did you notice how these were all three genuinely different fully researched ready to send emails that were customized to each brand each one had its own strategy subject line analysis of the business and tailored data backed proposal so that was opal it lets you package ai into a little machine that does a whole job for you use it to build different solutions for your business and let me know what you build in the comments meanwhile let s move to some more exciting tools this next one might be the most underrated tool in google labs not too many people know about it either using this tool you can build web apps without knowing both coding and design it s called google stitch and it basically designs user interfaces and hands you the code so you can build and host a real website using it you only need a google account to run it so today we ll do three things with it first we ll go over some quick hacks to get the best results from stitch second we will clone zomato and design a new food delivery app using it as reference and finally we will build a web app from scratch and export it into google ai studio to make it a working product this tool is for those of you who have had an app idea sitting in your head that went absolutely nowhere now s the time to go back and build it so open the tool alongside me and follow along let s get started before we touch the tool let s check out the walkthrough made by the team behind stitch themselves whenever you write a prompt for stitch it helps to think in four boxes the first box is idea which is simply what the thing is a landing page a food app a dashboard the second is theme which is the vibe you re going for the mood and the colors and the overall feeling whether that s modern and edgy or soft and minimal the third is content which is the real words and sections that live inside the app things like the headline the buttons and the menu items and the fourth box is image which can be used if you have a reference picture or a website you can just drop it in and stitch matches that look take this example let s say the idea is a landing page for a running podcast the theme is modern edgy high contrast black and white with hard angles and the content is a hero section with a specific headline and links to the podcast instead of just saying make me a podcast page you fill in all four boxes it s the difference between a generic result and exactly what you want don t stress about getting it perfect on the first try the goal is to get something on the screen first and then fix it from there the single most important rule is to make one change at a time pick one screen select a specific element and tell stitch exactly what to do if you try to change everything at once it gets confused now stitch has three models you can use the first is fast for getting screens on the canvas quickly then there s thinking which is slower but smarter and is used for complex layouts finally you have a redesign which handles visual makeovers and polish use the modes wisely depending on what it is that you wish to create okay the next thing you need to know is this word bank bookmark this to come back to it later it s basically a list of the exact style words that understands and responds to for layout you ve got words like bento grid and editorial for texture there s glass morphism and clay morphism which is that frosted glass look you see everywhere now this word bank is incredibly effective dropping one of these into your prompt instantly changes the entire look so whenever a design feels flat you come back to this list grab a style and drop it into your prompt the last thing you need to know before we dive in is this you can also utilize a design dot md file on stitch which is a single brand rulebook containing your colors fonts and spacing if you hand that file to stitch every result it makes follows your brand guidelines automatically okay now that you ve understood all this let s go build something this is what the homepage of stitch looks like it s a simple interface with just one prombox let s click the plus icon and paste in the link to zomato stitch uses that link as a visual reference pulling its layout feel and colors to work from we also give it this prompt build a foodie app the user sees the nearby restaurant searches by cuisine adds items to the cart and lands on a live order tracking screen with a delivery rider on a map zomato vibrant red palette clean mobile first layout let s pick the redesign model and hit generate and look at what happens here because it doesn t just spit out a single picture there s an agent working away on the left narrating what it s building as it goes and on the canvas three separate screens start forming all at once the first screen is the home screen with the line hungry let s get food delivered you ve got food categories running across the top and restaurant cards sitting below them the second screen is the menu and the cart it s filled with real food items neatly organized into starters mains rice and noodles and desserts and there s a working cart on the side that fills up as you add things to it and the third screen is the live order tracking complete with the map and the delivery rider exactly like the prompt asked for let s zoom out for a moment so you can see all three screens sitting side by side what we have here is the clone of zomato now we can move on to step two and refine it let s say you want this menu screen to feel a bit more personal to do that let s click on just that one screen to select it you ll see a blue outline appear around it and that outline is what tells stitch to touch only the screen and leave all the others exactly as they are here s the prompt we give it create the add to cart with specific user preferences showing a personalized recommendation like the best seller and something they ve repeatedly ordered mark that in the restaurant menu and cart let s run it and watch what it does stitch builds a fresh version next to the old one so you can compare the two directly zoom in and there it is a brand new personalized recommendation block the rest of the app stays exactly as it was it just executed a cleanly made change now let s go ahead and use that design dot md file i told you about which holds all of a brand s design information here s the easy way to do it stitch already comes with ready made design systems basically presets you can apply in a click when you open this drop down you ll see names like alexandria bow house and glass here each one is a complete look with the colors and the overall feel already decided for you let s pick alexandria and then give it this prompt use the design dot md and redesign the ui of the app being built with the specific colors from the design file then click generate and look what happens the whole app flips from that bright zomato red to these deep green tones the same screens now have a completely different mood and from here you just keep going you can select a screen ask for a tweak pick a color and repeat it s basically that simple changing one setting in the design file can update the look everywhere instantly now let s level up let s build a brand new web app from a custom design dot md file and for this build we are making something i really need a personal newsroom the idea behind it is an app that watches every new ai tool that drops works out what each one can do and then tells me where i could actually use it in my own work we give it a quick prompt select the web toggle which basically allows it to make a web app instead of a phone app and finally we keep it on the fast model then we click generate watch how it works before drawing a single screen stitch builds the design system setting primary colors fonts and headline styles because it establishes the brand first every subsequent screen matches perfectly and now the screen start to come through here s the dashboard there s a section where you pick which ai model you re working with a live feed of raw updates coming in and a grid for spotting the important shifts as they happen now let s say you also want a reports area in this app you don t have to build it manually you simply ask for it in a single line create a separate ui for the report section for me and watch the agent get straight to work it builds a brand new report screen complete with a report vault full of real example entries like a weekly tech audit market sentiment and compliance checks and this section of the app matches the design of everything else around it but so far it s still a simple design we need it to become an app in order to do this we need to click on the triangle on top what that does is it turns it into a prototype which is basically a clickable version of your design and there you go this isn t a static picture anymore let s click the reports tab and watch how it actually moves and takes us through to the reports page the buttons work the navigation works and you could hand this to someone and let them click through your app before it s been hosted on the final website you can also view a mobile version of the screen by clicking the little computer screen icon on the right side you can even redesign what it looks like until it feels right to you okay so far we ve designed the app and we ve made it clickable now for the final step turning it into a real working product this is the point where stitch connects to that other google tool we covered earlier ai studio let s hit export and choose build with ai studio and just like that it carries every one of our screens straight over into google ai studio in a brand new tab ai studio s job from here is to stitch all of those screens together into one working app you can actually use so we simply tell it to build and it gets to work turning what was a pile of separate designs into a single unified web app that works across your phone tablet and desktop and look at the result this is our newsroom running for real now let s click into reports where the filter works across risk strategy audit and legal the other sections work too and the settings page is right there everything we designed back in stitch is now a living functioning app inside ai studio i m going to share one last thing with you on this app it s a shortcut you can use you don t have to start with a blank box every time if you scroll down on the home page there s a whole gallery of designs that other people have built and shared everything from dashboards to landing pages to full apps let s open one of them do you see this button here remix this project it lets you take someone else s finished polished design and make it your own so their layout instantly becomes your starting point let s try it out look at this the full landing page the dashboard and all the variations including the light theme that bright lime green theme the hero sections the call to action blocks and the footer are now in your workspace ready for you to change and make your own so to summarize whether you describe an idea from scratch or remix an existing layout stitch can take your app from an idea to a designed exportable product now the next tool brings everything we ve covered so far together we ve seen google s tools for learning research app building and productivity but none of those tools answer one very important question how do you actually market what you create that s where pomelly comes in building a brand usually means hiring designers copywriters photographers advertisers and web developers that can get expensive and slow but now you can just give pomelly your website and it learns your style and creates assets that actually match your brand from ad campaigns to product photo shoots to complete websites by the end of this section you ll know how to create an entire marketing stack inside a single tool and i ll also show you a hidden feature that can push these assets directly into google ads let s start head over to pomelly website and you will land on the homepage that says easily generate on-brand content for your business let s get started the onboarding screen shows you the four core capabilities that pomelly unlocks which are build your business dna get campaign ideas and assets generate a product photo shoot and create your brand book and website click let s go and you reach the input screen that asks for your website from here you have two complete paths you can take the first path is for when you already have a website the second path is for people who do not have one yet this is where things get really interesting but let s start slow and go with the first one so you understand how the core engine works for this walkthrough let s use h and m as the test brand because it is a global retailer with a deep product catalog a clear brand identity and a heavy website that puts real stress on the tool let s copy the h and m homepage url from the browser head back to the pomelly tab paste the link into the input box and click continue pomelly now starts generating what it calls your business dna you will see a live starter screen that shows each step as it happens including analyzing your website gathering your colors studying your brand values determining your visual aesthetic picking your brand fonts pulling images from your website summarizing your business finding your logo learning your tone of voice and writing your tag line this typically takes a few minutes depending on how heavy the website is and how busy the servers are once the analysis finishes you land on the business dna dashboard which becomes your home base for everything you build from this point forward on this dashboard you can see the generated brand overview which includes the logo fonts color palette tagline brand values and aesthetic tone of voice and a full business overview for h and m the tagline pomelly generated reads fashion and quality at the best price in a sustainable way which actually matches the real h and m positioning surprisingly well on the right side of the screen you will notice the pomelly agent panel which is a chat interface that can directly trigger campaigns photo shoots websites and more let s leave the agent panel closed for now because we will come back to it with the second brand later in this video we click let s go at the bottom of the dashboard and a pop-up appears that says your business dna is ready with four next step options which are campaigns photo shoot website and brand book let s walk through every single one of them in order let s click on campaigns first the moment you land on the screen pomelly has already pre-generated three campaign concepts based on your business dna for h and m you will see suggestions like summer starts here golden hour in every stitch and your summer escape style click on the one you like and within a few seconds pomelly generates four full vertical poster creatives for that campaign each poster uses real product images from the h and m catalog with professional layouts and auto-generate headlines like from beach to city nights and matching descriptions click on any creative and the right sidebar opens up the full editor where you can change the header text the font the colors and the description you can also toggle the call to action on or off and when you turn the cta on pomelly adds a button like shop the edit to the poster from the same sidebar you can edit the cta link the button text and the button font right below the creative there is a button called fix layout and this feature is incredibly useful when you click fix layout pomelly automatically adjust the spacing the font sizes and the overall composition so nothing on the poster feels cramped or oversized so basically we don t need to waste any amount of hours on those tiny little design changes to make everything look professional once this finishes you can use the version history toggle at the top to compare your edited version against the original and you can download any version as a high resolution png by clicking the download arrow in the top right corner the file will then land straight in your downloads folder ready to post on any platform now if you don t want to use one of the pre-generated campaign concepts you can build your own custom campaign but before that you need to know two of these underrated features look at the left sidebar you ll see two tabs the catalog tab and the assets tab the catalog tab shows you every single product pomelly scraped from your website and each product comes with an auto generated name and description click on any product and you can directly create a campaign or a photo shoot for that specific item without going back to the main dashboard then there s the assets tab which shows you every visual element pomelly pulled from your live website including your logo your lifestyle imagery and your product photography you can download any asset individually or you can delete the ones you do not want pomelly to use in future generations that last part is a much bigger deal than it sounds because it lets you quietly train the tool to ignore older or off-brand visuals over time now let s create a custom campaign in the prom bar at the top of the screen let s type out a clear brief create a social campaign to promote our new spring linen collection targeting young working professionals age 25 to 35 time for the start of summer with a fresh and breezy tone below the prom box you can change the aspect ratio depending on where you plan to post and you can also manually select up to six products from your catalog to feature inside the campaign now click generate brief and pomelly first builds a structured campaign brief that includes the title the description the goal and any optional offers or promotions you can edit anything in this brief before locking it in and once it looks right click confirm pomelly then generates the full campaign with four ready to post creatives i m really impressed with pomelly these look genuinely studio made with on brand styling and clean typography now here we can also add more creatives to the same campaign by clicking add creative and the new ones automatically inherit the same brief and visual identity okay so pomelly can generate people well but how good is it with products what if i want to generate a professional product photo shoot for this let s explore the photo shoot tab on the left sidebar click create a product photo shoot and pomelly asks you to pick a product image from your catalog or your uploaded assets pick any product you want and then as usual choose your aspect ratio depending on whether you are posting to instagram a story or a website hero section the most powerful part of this section is the template gallery pomelly gives you photo shoot templates broken down by category such as fashion beauty home and consumables the most popular template is model try on which places your product directly on a photorealistic ai generated model in a setting that matches your brand aesthetic click generate wait about a minute and pomelly returns four complete variations of your product worn or used by a model when you expand them the lighting the styling the setting and the outfit choices look genuinely like a real studio shoot you ve just saved a lot of money and time that would have gone in getting a photo shoot done by professionals now you can save all four images or you can even click create campaign directly from the photo shoot screen to instantly turn it into a marketing campaign now the final core feature in pomelly is the one most people outsource and pay large amounts of money for website building but what if you could launch a high converting website in 60 seconds with a tool that already knows your brand s colors fonts and vibe let s try this out do you see the websites tab click on it and without doing anything pomelly starts generating a full website directly from your business DNA after about a minute the website editor opens up and you can click preview in new tab to view the full thing the way a real visitor would see it for h m pomelly built a complete landing page that includes a hero section a product grid a text block a feature row and a large image gallery at the bottom the fonts the colors the imagery and the copy all match the original h m aesthetic back inside the editor you can give pomelly a prompt to change any section you can update the live link you can delete the site or you can toggle the publish button to push it live on a hosted pomelly url the published version works on mobile and desktop and you can share the link anywhere you want now up until this point everything we built assumed that you already have a live website that pomelly can scrape but what happens if you do not have a website yet what if you are starting a brand from scratch or you are freelancing for a client whose site is not built yet or you are just experimenting with a brand idea before you commit this is where the second path comes in and it is the path most tutorials online skip completely head back to the home screen click let s go again and on the input screen click the small button that says no website yet you will get two new options you can either build your business dna from a pomelly template or you can build it completely from scratch using the pomelly agent let s go with the from scratch option for this walkthrough let s build a fictional handcrafted julie brand called gulabi based in jpore the pomelly agent panel opens on the right side of the screen and the first thing it asks you to do is upload any files that help the agent understand your business you can drag in product photos lifestyle images brand documents or even pdfs if you do not have product photos yet you can generate them in seconds using nano banana or chat gpt s image generator and then drop them straight into pomelly for gulabi let s upload around six to eight handcrafted kundan and minakari julie photos directly into the chat pomelly analyzes each image extracts visual details from the products and once the analysis is complete a button appears in the chat that says generate business dna click that button and pomelly starts populating your dashboard the same way it did for h m except this time it is working purely off your uploaded images instead of a website scrape to make the dna much sharper type a clear brand description directly into the chat for gulabi you can paste something like i run a small artistry julie brand based in jaipur called gulabi v handcraft traditional kundan and minakari julie including necklaces earrings and mantikas using designs rooted in classic rajasthani artistry our brand feels regal handcrafted and timeless our colors are deep emerald green royal gold and jaipur pink our tone is elegant warm and proud of its heritage we sell to women who want statement julie for weddings festivals and special occasions the moment you hit send such a detail prompt the dashboard updates the brand aesthetic the tone of voice the color palette and the brand values to match your description exactly the agent then gives you three tagline options to choose from for gulabi the third option timeless rajasthani craftsmanship worn with pride feels the most aligned with the brand description so let s lock that one in as the tagline now that we have a complete business dna for this new brand we made we click let s go at the bottom of the dashboard and you ll see the four options pop up let s click on campaigns and the moment we do that pomele suggests festive campaign ideas like the summer wedding guest edit based on your julie products and your description but we want to create a custom one so what we do is we go with a prompt like create a festive campaign to promote our handcrafted kundan necklace sets targeting women shopping for weddings and diwali with a regal and elegant tone that celebrates traditional jai poor craftsmanship let s change the aspect ratio to 9 to 16 for an instagram story and click on the arrow in a few minutes pomele builds a campaign brief which consists of a title description and a goal once you confirm it within about a minute you receive four high-end festive creatives one even features an ai model styled in coordinated outfits wearing our uploaded jewelry with elegant typography that reads timeless jay poor artistry the other three are regal stunning and instantly usable on social media too let s go on to the photo shoot tab next click on create photo shoot select some product images let s go with the earrings click generate photo shoot and soon enough we get four results pomele placed those exact earrings onto a few photorealistic indian models one is even wearing a matching magenta and red traditional outfit that complements the jewelry that was a success now let s even try the brand book tab after that out of the cover options it gives me let s pick the rich textured cover it looks stunning after we click confirm we need to choose supporting images let s go with the logo some jewelry images and one model image now pomele instantly builds a full multi-page brand book for gulabi with sections like brand overview logo variants minimum size typography color palette imagery and brand voice now that we have that the final thing we can do is create a website for this so let s click on the website tab and like last time all it took was a click and it has already generated the first draft of the website look at this there s a landing page that includes a legacy section then a grid with timeless pieces a full product gallery and a core values block complete with placeholder links you can swap with your actual store url whenever you go live the colors and typography used in website are very regal the whole thing is stunning for a first draft this reminds me one thing we missed out on was business details when you go to the taskbar and see overview you ll find a section on business details here you can add your location business hours phone numbers call to action links and your social media handles this will help fill in the website too and now let s come to an even more useful part of the whole thing this feature is going to be a blessing for marketers do you see the three dots in the top right corner of your dashboard click on it and you ll see connected apps from the drop down menu what you can basically do on this is connect your google ads and integrate it here on pomele once your google ads account is connected every campaign you generate inside pomele every creative variation and every aspect ratio can be pushed straight into google ads without leaving the tool if you manage multiple ad accounts pomele lets you pick the correct one at the time of sharing that means your entire workflow starting from brand identity moving through product photo shoots then campaign creatives and ending at paid distribution on youtube and google search all leaves inside one free google product your marketing team doesn t need to use multiple different tools to manage campaigns anymore google quietly collapse them into a single interface this is a free gold mine by google if you run a small business and you have been waiting for the right moment to take your marketing seriously this is it if you are a freelancer you can now hand clients a full brand kit campaigns and a website without hiring a single person the people who start using tools like this now are going to look like they have a whole team behind them while everyone else is still stuck waiting on three different people to get back to them this brings us to the final tool in our google free ai tool masterclass and it s easily one of the most impressive ones it s called google flow and it is essentially an ai powered filmmaking studio built by creators for creators inside flow you can generate scenes create characters organize projects arrange shots on a timeline and stitch everything together into a finished film from one place so let s put the tool to the test let s use flow to create an entire cinematic sequence from scratch and see just how much of the filmmaking process ai can handle today to use google flow you don t need to install anything all you need to do is head over to labs dot google slash sign in with your google account and you are inside the studio before we open a project i want to show you something on the home page that almost nobody talks about and it s called google flow tv you ll see the icon for it on the top of the home screen and when you click into it what opens up is essentially a library of ai generated videos that other creators have made inside flow organized into channels and short films now this can be a cool workflow trick for you you see every single video on flow tv comes with the full prompt that was used to generate it sitting right underneath the clip so if you want luxury aesthetics you can search the word luxury inside flow tv find a clip that matches the visual direction you re going for copy the exact prompt that produced it and use it as your starting point you can also hit shuffle all at the top which plays a continuous stream of clips across every category and it s the fastest way to get a feel for what the model can actually do before you spend your own credits once you ve explored flow tv you head back to the home page and hit the new project button to open a fresh canvas and this is your main workspace for everything we build today quick tour of the interface so you don t waste time hunting for buttons the left sidebar is where all your assets live the all media tab is the master folder that holds every image and every video you ve ever generated inside this project the characters tab is where you build reusable people that you drop into any future scene and we ll spend real time on this in the second demo because character consistency is the biggest unlock in flow the scenes tab is where you actually assemble your final video by dragging clips into a timeline and the tools tab is the most underrated section of the entire product which will come back to at the end the bottom of the screen is your prompt bar and the small settings icon next to it is where you choose between image or video set your aspect ratio pick how many variations you want and select your model one thing worth knowing before we start image generation inside flow costs you zero credits which means you can generate as many images as you want completely free video generation cost credits depending on length and you get 1000 credits on the pro account which is enough to make a full 60 second mini ad without burning through your balance but today let s try building something from scratch quick note if you like such tutorials and helpful tips around AI you should follow our channel so that you don t miss the next video we drop okay let s click on new project and try and turn this blank canvas into a finished one minute film the first thing you do is click the plus button at the bottom of the screen that says agent the agent is essentially an AI creative director that sits inside flow and helps you plan structure and generate your entire project through a conversation instead of you having to write technical prompts one at a time so if you ve never written a video prompt in your life this is what removes that learning curve completely now let s give it a prompt here is the brief I am giving inspired by the prompts we saw inside flow tv act as my creative director for a 75 second luxury jewelry film inspired by Cartier s panther emblem break it into a six-shot storyboard where a panther cub explores an empty library and discovers a glowing jewelry box and for each shot give me the setting the camera move the lighting the mood keep the whole thing warm and intimate so essentially you gave the agent a role to play you gave it a duration to work with you gave it the brand reference for tone you gave it the narrative arc and you told it exactly what you want back before you kick off open agent settings you said confirm before generating to always so the agent checks in with you before burning credits on something you did not want you set your default aspect ratio to 16 by 9 for youtube friendly output you pick nano banana 2 for the images and omni flash for the actual video generation and hit save now you hit enter on the prompt and watch what comes back the agent thinks and sends back is a complete six-shot storyboard titled the midnight cub with every shot specified down to the lighting setup and the camera move shot one is the silent mason shot two is curiosity shot three is the library exploration shot four is the discovery of the box shot five is the invitation and shot six is the final reveal of the jewelry if you re happy with the direction go ahead now the agent doesn t blindly start generating all six shots the moment you approve the storyboard instead it gives you three options you can generate the panther cub first or the library setting first or you can go straight to the full six shot storyboard grid the reason for this is consistency if the panther cub in shot one looks even slightly different from the panther cub in shot four the entire film looks uneven the same way if the library in shot two looks like a different room from the library in shot five the illusion breaks so the rule you want to remember forever is this you lock the character first you lock the location next and only after both of those are locked do you generate the actual shots that use them as references so click the first option which is generate the panther cub the agent asks you for confirmation before it spends any credits if you click approve it generates character concept images both renders look good the fur looks realistic the eyes have that deep amber color and the lighting on the cheekbones gives it a mysterious quality the agent asks which look we prefer let s type both into the chat because we want the agent to blend the aesthetic of the two going forward you also have the option to upscale and download either of these renders right from the chat because there s a button on each image that lets you bump it up to 2k resolution before saving that s useful if you ever want to use the character in a thumbnail or poster now let s move to locking the location let s choose to generate this library setting these two renders of a grand Parisian library come back both are lit by warm lamp light and gentle moonlight from the windows both have antique furniture and deep shadows let s upscale one of them to 2k and download it just to have the asset on hand now that you ve locked the character and locked the location the agent asked if you re ready to see the full storyboard grid yes let s generate it see the agent has produced a single image that contains all six shots laid out side with the locked panther cub appearing consistently in every panel inside the locked library setting executing each of the six shots that the storyboard originally specified the character looks identical in every frame and the lighting stays consistent across the panels note that this was generated well because of the order in which we did the steps first we log the character then location then generated the shots that use both if you skip the locking step and try to generate everything in one go you might get shots that look like different short films stitched together but this is good so let s tell the agent to continue and move on to the next phase which is turning the storyboard into the actual moving video now this is where our credits finally start getting used so be cautious first let s tell the agent to generate all the shots one by one and it automatically queues them up the first shot is a slow push in through the empty library with realistic lighting real depth of field and ambient sound of the room you also notice something underneath the clip in the editor there s a small text input field that says describe your edits so if you want to change the color temperature or end the shot on a different angle you can just type that instruction into the field and the model will reframe the clip for you we ll test out this feature later okay so now the remaining shots come in one by one the panther cub walking across the floor of the library the cub approaching a side table the red julie box opening the cub looking into the box and finally a macro shot of a gold panther ring with emerald eyes for the final reveal once all six clips are generated you go to the scenes tab on the left sidebar which opens a timeline at the bottom of the screen here you can drag and drop your generated clips in the order you want them to play we click the add clip button and select each shot from the media library to add them the timeline now holds all six clips lined up in the correct order now let s hit play what we have on screen is a finished 60 second kathie panther film you can also save the project at any point all you need to do is click the project title at the top and rename it kathie then you click the three dot menu in the top right and select download project which gives you a zip file of all your assets and the final cut ready to share or repurpose anywhere i have a special surprise for you i am about to hand you something that is going to simplify your life google publish the official prompting guide for gemini omni the model that google flow is based on the people who build the model are handing out an instruction manual for it they are telling you exactly what kind of language omni responds to what kind of edits it can pull off on the first try and how to phrase your asks to get something cinematic out of it instead of mediocre let me walk you through the whole guide quickly and then i am going to show you a trick that lets you carry this guide with you into every project you open here on there s five core building blocks in omni the first block is short framing and motion you tell omni whether you want wide angle medium or close up and whether the camera should glide gently or rush suddenly the result becomes cinematic if you use this feature the second block is style this is where you decide whether the whole thing should feel realistic or cinematic grounded or majestic and google s own advice on this is interesting they tell you not to over engineer this part you tell omni the name of the effect you want and let the model work out the details on its own based on its training the third block is lighting and this is the unsung hero of every great frame in cinema you tell omni where the light is coming from whether it is the sun a street lamp or something off screen and then you tell it the feeling you want the light to create whether it is crisp warm or ethereal one line of lighting language does more for your output than almost any other line you could write the fourth block is location and the guide makes one useful point here you do not have to describe every leaf every brick and every detail of the setting you can write something as simple as an alien landscape with clear azure water and omni s reasoning will fill in the rest using what it already knows about that kind of environment the fifth block is the action itself which covers what your subjects are doing who the characters are and how they are moving and interacting inside the frame if you have any changes for these blocks edit through natural conversation you do not need to re-prompt your entire scene to change one part the guide shows you an input video of a boy looking at a butterfly and with one sentence change the butterfly to a b the model easily executes it then if you want to change the b into a small swarm of fireflies just type that and there you go then they take a generated video of a violinist playing and with one line change the camera angle to be over the violinist s shoulder the entire shot reframes perfectly then they take a dark apartment building shot at night and with one line the lights of the apartments start turning on in sync with the music every window starts lighting up to the beat of the audio they fed it the next section is application of real world knowledge with the older google vio model you need it to be precise and descriptive with omni you do not you can write something as broad as explain the difference between regular computing and quantum computing visualize the sentence with a flat infographic style and omni s reasoning will build the entire animated sequence for you because it already knows what quantum computing is and how to visualize it then the guide moves into text rendering which is honestly one of the cleanest things this model does you pick the type of text the placement on screen the animation style and the exposure timing and omni will render the text in perfect sync with your visuals word by word beat by beat after that the guide covers reference complex actions where you ask omni to apply a specific kind of motion across the entire video without describing it frame by frame they show a skateboarder rolling through a street and they prompt it one line saying edit this keeping everything the same and add animated motion effects coming out of the wheels the skateboarder is now leaving a trail of animated fire and smoke behind him across every frame the final sections cover directing the camera with specific videography language like push in punch in dolly zoom locked off and static plus camera types like film camera natural smartphone zoom and webcam style the guide also covers referencing anything which means combining images videos text and audio together as multiple inputs into one prompt applying new styles like anime claymation or watercolor while keeping the original motion intact and keeping characters and objects consistent across a scene by anchoring them to a reference image whether that reference is from real life or generated through nano banana now here is the move i want you to steal from me you copy the url of the entire guide head over to claude and paste the link into a new chat the prompt you give claude is hey claude this is the prompting guide from google deep mind for gemini omni and from now on whenever i ask you for an omni optimized prompt you have to refer back to this document and give me a prompt that follows their structure you hit enter claude reads the entire guide comes back with a clean summary of the principles and confirms it will follow them but you are not stopping there because if you leave this in one chat it disappears the moment you close the tab you are going to turn this into a skill which is claude s way of saving instructions permanently so it applies to every future conversation let s click the create skill button title the skill gemini omni prompt optimization guide and claude walks you through a quick setup it asks how you want it to behave let s let claude ask us a couple of quick questions first then write the prompt then we choose the format the and scope the skill should cover click save and that is it the skill is now permanently saved into your claude account and it fires automatically every time you ask for an omni prompt from this day forward to prove it works let s open a new chat and ask give me a gemini omni prompt i want to visualize the poem twinkle twinkle little star claude pulls up the skill in the background applies everything it learned from the guide and comes back with a fully structured prompt that has visual style camera movement lighting direction and text rendering instructions all written in the exact format omni response best two let s copy the prompt switch tabs over to gemini itself because omni does not only live inside flow it is also baked into the regular gemini app so if you do not feel like opening flow for a quick one-off video you can generate the whole thing from inside gemini in seconds in the gemini sidebar let s click create videos and a video generation panel opens up on the right we set the model to flash change the thinking level to extended because extended thinking gives you the cleanest output for anything cinematic paste the optimized prompt into the input box and hit enter omni goes to work and the loop is complete and here is the final result hi and that s a wrap on the complete google ai tools masterclass at this point the tools are free the workflows are free and now you know exactly how to use them so here s my challenge to you go build something a website an app a marketing campaign a study assistant a short film anything at all and then come back and tell me in the comments what you built i read every single one and i d love to see what you guys create and if you want more masterclasses like this one subscribe so the next one shows up in your feed automatically check out my previous masterclass on cloud code it s a complete beginner friendly guide to building software without needing any coding background i ve put the link in the comment section thanks for watching and i ll see you in the next one

Also, ich werde hier einfach mal einen kleinen Sticker platzieren, also den Sticker. Also habe ich jetzt den ersten Schritt gemacht, nämlich tatsächlich zu klicken, um meinen Auslöser zu erstellen, also der gesamte Workflow. Also hier werden wir einstellen, dass ich möchte, dass es also in einer freundlichen Sprache ist, also etwas, das klar und so weiter ist, aber du kannst auch etwas professionelles einstellen, wenn du möchtest. Und hier das Modell, also das Modell, weil ich das Modell tatsächlich hier einfüge, aber das liegt daran, dass es manchmal vorkommt, dass man das Modell ändern kann. Diesen Code werde ich einfach hier einfügen und ihm sagen: "Voila, ich möchte, dass du dich an die letzten fünf Unterhaltungen erinnerst." Das ist tatsächlich interessant, weil ich ihm so ein paar Informationen schicken kann.

Also heute habe ich gerade eine Schulung für Anfänger mit dem Tool N8N gemacht und wir haben genau diesen Workflow erstellt und da habe ich mir gedacht, hey, das ist wirklich interessant, das mit meiner Community auf YouTube zu teilen, um Ihnen diesen Workflow zu zeigen, den du jeden Tag nutzen kannst. Stellen Sie sich vor, dass dieser Workflow, wenn ich hierherkomme und ihn tatsächlich veröffentliche, bedeutet, dass ich ihm sage: "Los, du gehst jetzt in den Ausführungsmodus." Was passieren wird ist, dass jedes Mal, wenn ich hier eine kleine Nachricht über Telegram sende, eine Veröffentlichung auf allen Netzwerken erfolgt. Das heißt, die Sprache, ja, du kannst ihm also sagen: "Schau, ich möchte, dass du nur auf Französisch, Italienisch oder Englisch arbeitest." Und was das Modell angeht, ich habe Grock gewählt, aber du kannst tatsächlich auch andere existierende Modelle auswählen und auch der Stil oder die Art und Weise, wie das System schreiben soll, sowie das Bildformat, die Auflösung, all das wird einmalig konfiguriert. Also, was ich jetzt machen werde, ich werde den Workflow ausführen und jetzt wartet er darauf, dass ich ihm die Information über Telegram schicke. Grock ist wirklich ein extrem leistungsstarkes System und jetzt werden Sie sehen, dass tatsächlich alles auf den sozialen Netzwerken veröffentlicht wurde.

Handelt es sich um ein Thema, das mit der Rechnungsstellung zu tun hat oder ist es ein Thema, das mit dem technischen Bereich oder dem Account, also einfach mit der Verwaltung des Kundenkontos zu tun hat? Und hier, wenn ich also in mein E-Mail-Postfach schaue, sehe ich tatsächlich, dass das System die E-Mail gesendet hat, also sie empfangen und einen Entwurf vorbereitet hat. Wenn ich hier auch ein bisschen weiter nach unten gehe, sehe ich, dass ich ihm hier, also GPT 4.1 gegeben habe, aber ich kann ihm auch ein anderes GPT aus der 5er Reihe geben, das etwas weiterentwickelt ist, weil es sich hier lohnt, das System ein wenig zu beschleunigen und ihm 5.1 zu geben. Stellen Sie sich vor, das ist nur ein Beispiel, aber stellen Sie sich vor, Sie haben mehrere Abteilungen, mehrere mögliche Fälle, mehrere Systeme und Sie würden alles in einem einzigen Agenten zusammenfassen, dann wird er tatsächlich niemals ein wirklich gutes Ergebnis liefern. Natürlich bleibt das immer noch ein Beispiel, aber selbstverständlich musst du die Informationen weiter vertiefen, damit tatsächlich jeder Agent in deinem Unternehmen spezialisiert ist, indem du ihm einfach die nötigen Details und sehr präzise Anweisungen gibst.

You also have DeepSeek V4 Flash and Pro, GLM 5.1, Kimiko 2.6, Mistral Medium 3.5, Minimax M2.7, Qwen models, Gemma models, Cosmos models, safety models, OCR models, retrieval models, speech models, and a lot more. You're now using Nvidia NIM APIs inside open code, and this is why I think this is so good for people who want to use AI coding tools but don't want to keep paying for 10 different API subscriptions. You just choose the OpenAI compatible provider, set the base URL to https://integrate.api.nvidia.com/v1, paste your NVIDIA API key, and then enter the model ID exactly as shown on the NVIDIA build page. You basically change the base URL to NVIDIA's endpoint, use the NVIDIA API key, and use the NVIDIA model ID. You can literally go to Nvidia build, get an API key, connect it in OpenCode, pick Nvidia NIM, select one of these models, and start coding.

So, basically what happens here is that fusion takes panel of top models, usually like three to five, ask them all the same question at the same time, and then you've got like Claude and Gemini coming in with answers. And so, the dashboard where you can plug all your agents into one place, Claude, Open Claude, Hermes, all of it with one screen, one shared memory. And the whole reason I built this is so that, you know, if something changes, like let's say for example, you can't use a model like Feable 5 anymore, you don't just scramble, you just lean on the others. So, from my own tests, and maybe it's just the models that we chose on the panel, but you can see here that it was a lot more expensive to use Fusion, and also it was a lot longer to get the answer. Then run it through fusion, watch the panel work, see for yourself that you don't need like fable five to get the best results, and then stop relying on one model, right?

僕が気になったAIニューストピック9個 ①米国政府の指令の結果 当社はすべてのユーザーに対するClaude Fable 5へのアクセスを一時停止します ②Claude Codeのヘッドレスモード SDKは 6月15日以降サブスクの通常レートリミットで使えなくなるので注意が必要です ③GLM-5.2 は現在 すべての GLM Coding Plan ユーザー Lite Pro Max Team プランを含む で利用可能になりました ④Coding Plan のサブスクライバーは ZCode 内で 150 の使用量クォータが付与されます ⑤OpenRouterは 複数のAIモデルが並行して動作し より優れた結果に 融合 する新しいモードであるFusionを発表しました ⑥新しいプロジェクト Omnigent をオープンソース化できるのが本当に楽しみです これはAIエージェントのためのメタハーネスです ⑦Gemma 4 12B Coder が登場し ローカルコード生成のゲームチェンジャーです ⑧先月 The InformationはDeepSeek V4.1が6月にリリース予定だと報じました ⑨噂では OpenAIが6月23日にGPT-5.6をリリースする可能性があるそうです 動画タイムライン 00:00 オープニング 00:20 ①米国政府の指令の結果 当社はすべてのユーザーに対するClaude Fable 5へのアクセスを一時停止します 01:21 ②Claude Codeのヘッドレスモード SDKは 6月15日以降サブスクの通常レートリミットで使えなくなるので注意が必要です 03:04 ③GLM-5.2 は現在 すべての GLM Coding Plan ユーザー Lite Pro Max Team プランを含む で利用可能になりました 05:32 ④Coding Plan のサブスクライバーは ZCode 内で 150 の使用量クォータが付与されます 07:43 ⑤OpenRouterは 複数のAIモデルが並行して動作し より優れた結果に 融合 する新しいモードであるFusionを発表しました 08:55 ⑥新しいプロジェクト Omnigent をオープンソース化できるのが本当に楽しみです これはAIエージェントのためのメタハーネスです 10:04 ⑦Gemma 4 12B Coder が登場し ローカルコード生成のゲームチェンジャーです 11:55 ⑧先月 The InformationはDeepSeek V4.1が6月にリリース予定だと報じました 13:28 ⑨噂では OpenAIが6月23日にGPT-5.6をリリースする可能性があるそうです 15:52 雑談 サッカーW杯 オランダ戦について 17:14 エンディング 生成AIの研修 セミナー AI顧問の依頼はこちらから 整体院向けGASで開発した 予約管理アプリ を配布してます Google Keepにワンクリックでメモを保存できるChrome拡張機能 GitHubリポジトリ管理を楽にするWEBアプリ DevBoard デヴボード 個人開発したブックマークアプリ ClipTuck 自分だけの本棚 体験版 noteでもAIニュースやTipsをまとめています noteメンバーシップでは ぶっちゃけ日記を書いてます Xアカウントはこちら 僕の整体院のHPはこちら 整体院の公式Instagram

But if you look at the benchmarks fable 5 has there are some parts for example this knowledge work has a massive jump and of course like cyber security which has the biggest jump which is supposedly also the reason why fable is banned in the first place by US government but generally speaking it's the best model out there in a lot of scenarios in a lot of use cases so let's take a look at how this thing which opener is saying this fusion API apparently this is not a model this is just an API works better than fable so they say we benchmark fusion on 100 hard research tasks and found that these panels of model consistently outperform individual models beyond frontier performance can be achieved with frontier panels and panels of budget models can surpass frontier models at a much lower cost. So if you look at this what this basically means what they're saying is that we take a bunch of models let's say you know opus here and GPD here and let's say miniax let's say here and synthesis means that each one of them actually answers their own response differently right so this is answer one this is answer two and this is answer three now when you actually go ahead and combine them not exactly combine but compare them so you'll have like another opus 4.8 date let's say which is sort of a judge here right so this judge opus it's going to look at all these three answers over here and it's going to output a final answer based on whatever they have said so this is what they say by saying that we are doing like this increase in the benchmark score is from synthesis which is what I explained you and diversity is different models right so if you look at these graphs you're going to see opus and opus I'm assuming opus is the base model and opus is also So the judge here for you know evaluating the right response and over here we have diversity which even results in better response because obviously the training data is different the way the model approaches the problem is different in diversity case notably the budget panel was also comparable with claude fable 5 in performance a panel of Gemini 3 flash kim 2.6 and deepseek v4 pro fused together beat solo GPT 5.5 and solo opus 4.8 outright. Next we have Opus 4.8 date GPD 5.5 and Gemini 3.1 Pro at 68.3 which is you know still I mean statistically it's just as good as Fable 5 then we have Opus 4.8 and GPD 5.5 then just two Opus 4.8s I don't know like which one is judge and which one is uh just the models here so we have that information here so the judge here is opus 4.8 trade in case of Fable and GPD. Draco also does not include long horizon tasks which is where Fable shines and which is basically the whole programming right if you're building any sort of feature debugging anything you know just figuring out what task is there then Fable still is exceptional at that where Draco at least the test suite that they have does not include those tasks so I would I would have actually it would be super interesting if somebody could run this on actual programming benchmarks as well in order to figure out that is this difference like similar to what we are seeing over here for programming tasks comparable to fable 5 and I'll be really surprised if that is the case because at some point the intelligence of the model also would start to carry a lot of weight right the ideas that fable is able to generate would not be possible with just raw models and even if they come again and you know just combine their approaches I'm sure like it'll result better than any single solo model as we have seen in this. So in a way if you are able to combine these models together and sort of like you know I don't know like if it'll actually be better than opus or fable but I I have seen like those models sort of tend to get slow if you're using like maximum or extra high thinking but yeah these would be slow it it's no doubt like you would not get responses as fast as the raw model itself but I think it's it's an okay trade-off if you are willing to you know just let a good answer come out instead of just whatever garbage that you know a simple a smaller model is giving you out.

So now one thing that we have to talk about is of course the pricing which is the Fable 5 solo pricing versus the fusion models pricing because there's no point trying to get Fable level intelligence if it's just going to cost the same amount. So we click this and then of course with this one, what you can either do is run it from the API if you want to, you know, have it for an application or what you can do is just run it in the playground or you can just go to open router.ai AI/fusion and then this is what will pop up. So here you can see you have the user interface for the model fusion where you can run multiple models side by side run an analysis and then fuse it into the best result. And this is really interesting because this is where we get to look at where the models disagree and what differences the models do have because some of these models have been trained on different pieces of data, trained in different ways. And so when you ask a model a question and it gives you a response, often your first response is to just take that response at face value, but often times there are many different blind spots that you don't even know how to ask the question about that because the model is just giving you what you answered.

And then you can actually use this, for example, inside a chat, like you can see here, and you can choose which models you have working together. So we've actually set up the fusion boardroom inside our agent operating system and so we can ask a question here and then we can see what we can get back. Now if you're wondering, "Okay, how does that work?" Well, you you can see an example of how it created this actual tool where you can paste in a OpenRouter API key, ask it anything, and then it will come back and and come back to us. So, you might say, "Right, act as a you know, act as an SEO content council for the keyword best AI community, search the web for what currently ranks, then synthesize search intent, the angle competitors are missing, a recommended H2 outline, three questions every article forgets to answer." And then we can click convene and the judge and the panel will start working together to do that via the API with Open Router, which is pretty mind-blowing when you think about it because now you've got something that is potentially Fusion level, but at the same time, you don't have the problem with all the you know, the if you were using Fable 5 on an API like well, that would be insane for using up tokens, right? And so, if you look at, for example, how goal mode is working and also how fusion mode is working, it's like something that's really good as best practice is to have a judge in place who checks the work, quality controls it, and then fuses all the answers or tells the agents to iterate from there.

Das heißt, er wird die Aufgaben nach Ausführungsreihenfolge anordnen und ihr werdet sehen, dass er sie dann abarbeitet, in den Ausführungsmodus wechselt und sie nacheinander, also sequentiell bearbeitet, außer wenn du die Aufgabe anderen Agenten zuweist. Also, ich gebe ihm tatsächlich die Priorität vor und basierend auf meiner Antwort schlägt er selbst eine Liste konkreter Aufgaben vor, die er mit mir gemeinsam erledigen kann oder die er selbst übernehmen kann, um mir zu helfen bei dieser Priorität voranzukommen. Ich sage ihm also jeden Morgen um 9 Uhr möchte ich, dass du eine Routine startest und dass du mich fragst, was die Priorität des Tages ist. Und ich, anstatt ihm jedes Mal eine Aufgabe zu geben, manchmal vergesse ich es, manchmal habe ich Aufgaben zu erledigen, gebe ich ihm einfach eine Unterfuggabe, oder ich bitte Herr Mess mir zu helfen, nur um tatsächlich eine Idee weiterzuentwickeln. Also, wenn ich jetzt hierherkomme und nachschaue, sehe ich z.B., dass er angegeben hat, dass er darauf wartet, dass ich ihm die Antwort gebe.

So it is interesting to see here that they are running the model basically on just a deep research task benchmark and claiming as if the model is just better in all places which seems a bit misleading because using multiple models for text is always better in my opinion. It might be good for some tasks, but for the majority of tasks, this is just a bad model that costs more and may even give you worse results than just using one model. I don't know why a company of open router stature is making such misleading marketing claims and talking as if a fusion model can solve and surpass Fable, which is not the case in reality at all. It's like those GPT3.5 projects that used to make the models perform well by asking the model in multiple turns to make the response better which worked then but now it is mostly just diminishing returns. I think open router should stick with what they do best which is model routing and work to make it easier faster and stuff like that rather than focusing on trying to become an AI model lab or research company at this point.

Get the Agent OS Want to make money and save time with AI? Join here: Video notes links to the tools Get a FREE AI Course Community 1,000 AI Agents Get a FREE AI SEO Strategy Session Get 200 Free AI SEO Prompts Get our SEO link building book here: Still dragging cards around a to-do list and doing all the work yourself? In this video I break down the Hermes multi-agent Kanban drop in one sentence, approve the plan, then watch named agents research, design, and code your project while you go do something else. Idea to finished, shippable product, mostly hands-off. 00:00 Intro one idea to a finished app, hands-off 00:54 The Problem ideas pile up, nothing ever ships 01:22 What s New multi-agent Kanban that runs itself 02:08 Step In Once approve the plan, then walk away 02:16 The Workflow capture, classify, approve, build, ship 03:12 Live Build full SEO blog from a single sentence 04:00 Memory Layer agents remember, no rebriefing 05:25 Self-Checker agents review their own work 06:09 The Model Claude built for long, parallel builds 07:29 Pro Tip sharper input, better finished product

Deshalb stellt man fest, dass wenn man eine Website, eine Anwendung, ein System oder ein CRM erstellen möchte, man auf ein KI Tool zurückgreift, aber man denkt, dass dieses Tool einem tatsächlich das gewünschte Ergebnis liefern wird. Wenn man versteht, dass jedes dieser Tools eine bestimmte Funktion oder ein klar definiertes Ziel hat, werden Sie sehen, dass ich wirklich das richtige Tool auswählen und es schaffen kann, das Beispiel umzusetzen oder die Software zu erstellen oder zu entwickeln, die ich aufbauen möchte. Das ist also etwas, dass ich persönlich entwickelt habe, weil ich viele Anwendungen nutze und möchte, dass diese Anwendung, wenn ich eine Anwendung für meine Kunden erstelle, denn ich verkaufe die Anwendungen weiter. Aber dieses Tool ist wirklich ein Trumpf, wenn ich sehr sehr schnell ein Werkzeug erstellen möchte und einfach die Idee in eine Anwendung übersetzen, um sie zu sehen und zu testen. Du sagst ihm: "Schau, ich möchte, dass du das gleiche wie bei diesem Projekt machst." Und du änderst das Logo, das Bild, den Kontext, den Text und du schickst die Informationen ab und das System gibt dir wirklich etwas Visuelles, dass man navigieren kann mit Menüs.

Ich bin aber ein großer Fan vom Openmi Abo, falls du schon ein Abo hast bei Opmi, gerne auch nur das kleine für 20 € kannst du damit relativ viele Kredite bekommen und dein Agent kann auch sofort Bilder mit stellen. Du siehst, ich habe das Ding Hormes Root genannt und das hat natürlich wieder Zugriff auf diesen Server und genau deshalb kommen wir jetzt zu Chron Shops, denn Chron Shops sind tatsächlich vor allem praktisch, falls du Hormes auf einem VPS laufen hast. Natürlich kann man auch ein nachden auf einem Server installieren und Hermes kann ein kontrollieren, kann Workflows direkt erstellen, kann sich über MCB verbinden mit einer und du kannst ziemlich coole Dinge machen. Falls du ein noch nicht installiert hast auf deinem Server, dann kannst du sogar zu HMES sagen: "Hey, installiere ein Übrigens das kannst du theoretisch auch sogar lokal machen, falls du magst. Du kannst das Ganze aber auch gerne so machen, dass du ganz einfach Hermes das Gitter Breot zu diesem ein nach deinem CB Server gibst und eventuell noch zu diesen Skill.

Das bedeutet, dass es wirklich ein Projekt bearbeiten, die ein gutes Ergebnis liefern und vor allem ein Ergebnis liefern wird, dass man ausführen kann. Also Slashgoal bedeutet ganz einfach, dass Hermes Mess ein dauerhaftes Ziel gegeben wird. Das ist sehr wichtig, denn wenn du das Ziel nicht richtig entwickelst und ein Ziel setzt, das nicht klar ist, das nicht gut ausgearbeitet ist, bei dem alle Details fehlen, dann kann es natürlich sein, dass er enorm viel Energie verbraucht, das Ziel nie erreicht oder manchmal denkt, er hätte das Ziel erreicht, weil es zu Waage und nicht gut erklärt ist. Ich kann also Enhropic verwenden und sie wissen, dass man hier besonders wenn man ein Ziel vorgibt aufpassen muss, denn es wird nicht aufhören. Also heute ist es sehr wichtig, ein bisschen zu verstehen, wie das Gedächtnis funktioniert und wie ich es verbessern kann, um wirklich einen leistungsstarken Speicher zu haben, der es Hermess ermöglicht, einfach ein extrem mächtiges Werkzeug zu sein und vor allem der mir Zeit und Geld sparen wird.

Also, die Idee ist einfach, ich habe die künstliche Intelligenz mit nur einem einzigen Prompt gebeten, mir diese Website zu erstellen. Ja, es ist also einfach ein kleines Gehirn innerhalb von Chat GPT, das ich erstellt habe und dass es jedem ermöglicht, wirklich jeder Person, die einfach ein sogenanntes Lastenheft erstellen möchte, für jede beliebige Website dieses kostenlos zu nutzen. Also habe ich natürlich im Hintergrund so viele Anweisungen darin programmiert, dass dieses System mir tatsächlich ein vollständiges Lastenheft für jede beliebige Tätigkeit erstellt. Ich kann ein Konto für die Nutzer erstellen, ich kann Formulare machen, ich kann ein Nachrichtensystem einrichten, ich kann sogar ein System zur Terminereinbarung erstellen. Deshalb denke ich, es lohnt sich wirklich zu sagen, hier, ich habe eine Plattform oder eine Website oder sogar einen E-Commerce Shop, wenn ich die KI tatsächlich darum bitte, mir etwas schöneres, stärkeres und äh rentableres zu erstellen, weil wir ihr sagen, wir wollen, dass die Leute, wenn sie auf die Seite kommen, dass es konvertiert.

We didn't broadly release our previous model with this level of capability because when we finished training and testing it, [music] we saw that the model, Claude Mythos preview, was finding thousands of cybersecurity vulnerabilities. A model that can find flaws like that can also [music] be used to exploit them. So instead of releasing it, we handed it to the people who protect [music] the world's critical software and put it to work fixing the holes before someone could break through them. Our safety systems for Fable 5 automatically [music] review requests that touch on high-risk areas like cybersecurity or biology. [music] We do that intentionally so people can continue to benefit from the capabilities of a powerful model like Fable without the cyber and biology risks that come with it.

Master Claude Code: FREE community with Skills Need custom work? Book a consult Obsidian Graphify is a game changer. By combining the repo deconstructing power of Graphify with the organizational layer of Obsidian, we have the ability to merge our knowledge graph insights into whatever larger Vault we are working with, improving Claude Code s answers in the process. TIMESTAMPS: 0:00 - Graphify Obsidian 2:45 - How it Works Demo 15:06 - Final Thoughts RESOURCES FROM THIS VIDEO: Master Claude Code: My Website: claudecode

Er wird also das Bild analysieren, den genauen Zweck der Veröffentlichung verstehen und das Bild beispielsweise in ein 3Dbild, ein animiertes Bild oder ein Werbebild umwandeln. Und die gute Nachricht ist, dass man heute tatsächlich Nacht 8 Nacht, also auf Hostinger, mit einem Rabatt bekommen kann, wenn man zum ersten Mal einen Server bei Hostinger nimmt. Ich habe ihm dieses Bild geschickt, also eine Flasche, von der ich einfach schnell ein Foto gemacht habe. Sobald ich also mein Bild habe, gehe ich zum nächsten Schritt über, um dieses Bild tatsächlich in ein animiertes Bild zu verwandeln, das sich bewegt. Wenn ich also hier auf die History klicke, werdet ihr sehen, dass wir ein Bild finden, das gerade erstellt wird.

Get the Agent OS Memory Galaxy Want to make money and save time with AI? Join here: Video notes links to the tools Get a FREE AI Course Community 1,000 AI Agents Get a FREE AI SEO Strategy Session Get 200 Free AI SEO Prompts Get out SEO link building book here Obsidian Memory Galaxy: Turn Your Notes Into a 3D AI Memory System (Goldie Setup) The script introduces the free Goldie Memory Galaxy setup that connects an Obsidian vault to an agent operating system so AI agents like Hermes can share and retain context through a visual 3D star map where each note is a star, links form constellations, and recently updated notes glow brightest. It explains how this solves scattered, forgotten notes and blank session AI by letting users fly through their second brain, open notes by clicking stars, and feed agents real business context such as clients and writing voice. The five layers are outlined: stars, constellations, glow, fly-through navigation, and continuous improvements via a feedback loop where agents create and update markdown notes (including exports from Omi to Obsidian). 00:00 Memory Galaxy Overview 01:18 Why Notes Stay Invisible 02:30 From Scattered Memories to Stars 04:00 Get the Free Setup 05:01 Five Layers Explained 05:41 Agents Plugged Into Obsidian 08:03 Old Way vs New Way 09:17 How the 3D Map Works 10:11 What You Can Do With It 11:40 Objections and Recap 12:31 Agent OS Demo and Features 14:36 Community Support and Coaching 15:15 Final Encouragement

Man musste jedes Mal WSL2 starten, damit man unter Ubunto ist und erst danach konnte man Hormes launchen und dann hatte man Hormes im Terminal und das mögen auch wie gesagt einige Leute nicht und mittlerweile ist alles was du machen musst nur genau auf diesen Link zu gehen. Etwas Weiteres, dass ich probiert habe, ist, dass ich zu Hormes einfach nur gesagt habe: "Hey, ich habe gesehen, du hast einen Conf Skill, kannst du den sofort verwenden, um Bilder zu erstellen?" Und ich konnte nur noch sagen: "Hey, mach ein Bild von einer Katze z.B. Falls du nie irgend ein Abo hast, also weder bei Chat GPT noch bei Untropic oder sonst wo, kannst du gerne auf Open Router gehen. Als nächstes sollte ich dir das Messaging zeigen, denn falls du mit einem Agent gerne über Telegram, Discord, Slack oder sonst was schreibst, kannst du dich im 0, nichts damit verbinden. Falls du da oben drauf klickst, kannst du die Sidebar auch komplett verschwinden lassen und links unten kannst du auf Home klicken, falls du gewisse Dinge verstecken oder zeigen möchtest.

Ich habe dazu auch schon mal ein Video auf meinem Kanal gemacht, aber wir sehen hier ein lokales Modell Quen Code, was ich mir mal heruntergeladen habe. Wir können hier ganz normal auf New Session klicken, haben hier eine Übersicht der ganzen Skills und Tools, also der ganzen Fähigkeiten, die Hermes hat und die können wir dementsprechend hier aktivieren oder deaktivieren. Das werde ich gleich zur Verfügung stellen und dann hätten wir hier eine Liste der ganzen Artefakte, also der ganzen Produkte und Objekte, die uns Hermes erzeugt. Dann heißt es Token kopieren, das in Hermes zur Verfügung stellen und dann brauchst du ein Telegram User ID und dann gehen wir wieder zurück zu Telegram. Ich finde das richtig geil, weil das wäre ja eine Aufgabe, die ich eigentlich selber hätte machen müssen zu überprüfen, ob dieser Skill, den ich herunterlade und generell alles, was man vom Internet herunterlebt, sollte man gegenchecken und prüfen und dass ich hier eine Prüfung habe, ich werde das jetzt nicht auflösen, aber finde ich sehr, sehr geil.

Ich würde dir in dem Fall Cloud Code benutzen, weil du dann eben ein besseres System aufsetzen kannst und da kannst du dir da wie gesagt dann einen Topic Cluster mit aufbauen lassen, wo du ganz genau siehst, okay, diese Videos haben zu dem Thema gut performt, das sind welche, die zu dem Video gut performt haben und dann kannst du eben auch die Automatisierung, die du dahinter einsetzt, eben auch besser ja mit den richtigen Daten füttern. Wichtig ist, dass du das Ganze auch visuell darstellen kannst und visuell sehen kannst, dann ist es auch für dich leichter und du kannst eben auch dann leichter mit dem Auge ja die Top Outlier Videos, sage ich mal, identifizieren und dann ist es für dich auch leichter, das Ganze zu bewerten, weil ganz wichtig ist dieses System, was wir hier auch setzen, ein Content System ist an sich nur so gut wie du es dann auch bewerten kannst und das gehen wir dann eben auch weiter zum nächsten Punkt. mit der Automatisierung gemacht hast, dann hast du einmal die Top 100 Hooks aus den letzten beispielsweise 30 Tagen aus Shortform Content einmal identifiziert und dann kannst du dann natürlich auch wieder ein Cloud reingehen und natürlich auch Formatkluster wieder da identifizieren und auch aufsetzen, dass du einfach wieder mehr Informationen hast, nicht nur basierend auf Bauchgefühl, sondern eben auch auf Daten, dass du das Ganze nicht manuell machen musst und es einfach ja leichter bewerten kannst. Das ist eben auch der Part, der meistens ja komplett weg bleibt, weil wenn du jetzt halt das Ganze manuell machst, wenn du jetzt hal auch vielleicht schon so ein, ich sag mal, kleinen Chat GPT oder Cloud Workflow gemacht hast, wo du Content erstellst, dann kannst du das Ganze nicht wirklich messen und skalieren, weil es eben kein in sich stimmiges System ist. Dann natürlich auch jede Hook pro Plattform, da einfach auch jeweils ähm würde ich dir empfehlen, immer Kontext Files mitzu erstellen, dass du ganz genau dem System gibst, okay, so funktioniert ein LinkedIn Post, so funktioniert ein Instagram Real, so funktioniert ein Karussell, so funktioniert ein YouTube Video und dann hat er immer genauen Kontext und kann dir dann eben Stück für Stück auch die perfekten Hooks für die Plattform mitgeben.

Was dürfen Brands in Social Media 2026 auf gar keinen Fall machen? Ich bin mal sehr sehr gespannt auf deine Meinung dazu. Immer mehr Unternehmen nutzen ja KI für ihren Content und das Problem ist nicht, dass sie KI nutzen. Das ist der größte Fehler, den Brands im Umgang mit KI machen können. User nehmen Brands übel, wenn sie KI generierten Content nicht entsprechend markieren.

Mit Hermes hingegen haben wir tatsächlich ein wirklich stabiles System und genau deshalb möchte ich euch heute einfach zeigen, wie man Hermes richtig installiert, um diese Power wirklich zu beherrschen. Ihr nehmt also den Server, ich klicke hier aufwählen und ihr werdet sehen, dass er jetzt tatsächlich tatsächlich vorschlagen wird, die Anzahl der Monate auszuwählen. Wenn ich also OLAMA anschaue, wir gehen einfach hier zu den Modellen, seht ihr, dass OLAMA immer lokale Modelle anbietet, die man herunterladen kann und er bietet auch Cloudmodelle an. Also jetzt kommt er also das Ganze mit Kimi laufen zu lassen und natürlich werdet ihr sehen, dass er gerade antwortet. Der nächste Schritt nach der Installation von Hermes besteht einfach darin, nach Fähigkeiten zu suchen, um dein Hermes zu verbessern, damit Hermes einfach ein leistungsstarkes System wird, das in der Lage ist spezifische Aufgaben auszuführen.

Beim anderen mit katischen Korrosionsschutz Beispiel Offanlagen bei Imersion 2 Beispiele Hafenbrech Stahlbur Schleusentoren Schleusen können wir auch noch mal prüfen in den Quellen jeweils. Jetzt sehen wir, es spielt keine Rolle, ob es Tabellen sind, ob es wie gesagt Excel Dokumente ist, ist alles nie davon möglich. Jetzt springen wir noch mal zurück und diese Fragen, die ich gestellt habe, können natürlich auch auf Englisch gestellt werden. Das heißt, ich gebe dieselbe Frage jetzt auf Englisch an und der spielt keine Rolle, welche Sprache die Dokumente selbst haben, denn die Dokumente, wie Sie gerade gesehen haben, waren nur auf Deutsch. Was jetzt auch noch Hauptpunkt ist, den ich darstellen möchte, ist, wenn wir Fragen stellen, jetzt nicht abgedeckt werden in unseren Systemen, in den Informationen unserer Systeme, in den Dokumentarten, wie z.B.

So the way this is going to work uh if you think about an auto loop we have a worker a judge and a loop that's effectively what's happening with the system now this works and has been released across different platforms codeex released at the end of April then claude and then our beautiful Hermes agent has also come to the table now before we go on to the long-term goals the real Jewish we have to understand the foundations of these skills and the kind of capabilities it unlocks in Claude and Hermes first want to do is get a short-term goal skill like so I'm going to come over here and grab it this will be available for down below in the description. And essentially what we're going to be able to do guys is build out longer term goals using our agentic operating system which we can trigger with Claude or we can trigger with Hermes. Now on my Hemy's agent, let me go ahead and drop in that prompt lexone and this will cover everything and I'll explain what goes into this, but essentially we've been really specific on exactly what it needs to do in terms of creating this long-term goal. So what's really cool here is that it's really broken down and this probably took a lot of time when I was building this out to really get this specific but breaking it down into okay like if this is the midterm goal like specifically what mini goals would we need to achieve? So, when we want to hit those mid-term goals, like actual meaningful things, not just like running around in circles, not really accomplishing much, we can really leverage both things together, the human and the agent.

I find myself waking up, going straight to Hermes, check what it is doing, but I brainstorm with it, I ask it for content ideas, I ask it to do research, I ask it to help me fix problems, and the funniest part is that I have now used Hermes help me fix open claw when open claw breaks. Open Claw felt like an agent, and at times it felt like a team member, but with every single Hermes updates, I'm starting to see what a team member can actually look like. Hermes is starting to feel like an operating system, an entire operating system for agents, a whole lot more than Open Claw is. If your open claw works, your workflows are stable, and you do not feel pain, I would not tell you to rip everything out just because Hermes is hot right now, or Hermes is a better alternative today. But if you're starting today, or if your open claw setup keeps breaking, or if you're spending more time maintaining the agent than using it, I would consider installing Hermes first.

And then the other one that I was going to say and this is really good especially if you're trying to find if you're trying to find like trending topics is Grock and Hermes agent because it can do Twitter search and with those three setups you can basically use Hermes agent to do web research for you. So you can create a second brain with OMI where it records what you do all day and then what it can do is export those memories over to Obsidian and then Obsidian has some stored inside your agent operating system like you can see over here, right? So, if you want to get access to the new video model for Gemini with Gemini Omni or however you want to pronounce it, you can just go inside the chat here, click on create video, and then you'll see create with Omni and that's how you can get access to it. And Damon says, "Where should I market my new AI SAS?" What I would recommend here is you want to do like videos daily showing how to use a tool or using for example relevant tutorials where you show how to use your tool and what it can be used for and then people will see that and they'll become interested in your tool and you'll generate more leads that way. And we have one final question here from Jason who says, "How do you find the best setup for all this stuff?" So if you're using claude code or gemini or whatever if you don't have a lot of resources so if you want to reduce the resources required I'd recommend that you go with a free setup for example like you can plug in an API like our alpha with open router and that's a free API you can plug into for example claude or for example n and you can see our alpha right here.

So what you can actually do is browse all this stuff here and you can see for example you can like search and it can search like eBay for you and stuff like that. And so what we can actually do is as an example of this, we can say, okay, give me a list of what you can do, right? So you can browse pages, you can take snapshots, you can for example, you can click around, it can navigate, you can also use your mouse, you can browse sessions, you can actually manage your sessions with these. So you can see it running right here inside Hermes agent and then you can see it's navigating here. You can get the full Hermes and agent operating system setup guide or browser use agents right here.

Also, es ist offiziell, Grock kann ab heute mit Hermes verbunden werden, aber ich spreche nicht von der Grock AP, ich spreche von meinem Abonnement bei Grock. Mit meinem Grock Abonnement kann ich jetzt tatsächlich Hermes laufen lassen, wobei Hermes dann das offizielle LM verwendet, nämlich Grock. Er sagt mir: "Ja, ich kann ein hochausgestattetes Bild machen, aber für mich ist das Video interessanter und deshalb musste man unbedingt diese Option aktivieren und Enter drücken und hier die neuen eingeben und Enter, um diese Funktion nutzen zu können. Das bedeutet, wenn ich die Anfrage starte, wird sich hier nicht erscheinen, aber sie wird einfach im Hintergrund bearbeitet und das System gibt mir die Kontrolle zurück, wenn sobald es einfach die Antwort gefunden hat, kann ich also weitermachen. Codex, dank eures Abonnements bei Chat GPT, dank der Verbindung mit Hermes und jetzt wird Hermes noch leistungsfähiger und sehr fortschrittlich, um einfach alles zu entwickeln, was programmiert wird, alles was automatisiert wird, jede Website dank Codex, das offiziell verbunden ist,

Das heißt, ich wäre gezwungen, jedes einzelne aufzurufen und jetzt diese Oberfläche, die ihr gerade seht, zeigt mir einfach die laufenden Projekte und die, die bereits abgeschlossen sind und ich kann schnell in dieses Projekt hineingehen, um einfach auf eine Frage zu antworten oder den Prompt anzupassen. Ihr werdet sehen, wenn ich jetzt hier tippe, also wir starten Cloud und dann gebe ich hier einfach Agent ein, weil das Wort Agent bedeutet, dass ich das Agentensystem starten möchte, dass es ermöglicht Agenten zu starten. Also jetzt starte ich den zweiten Print und ihr werdet sehen, dass wenn es hier gelb wird, das bedeutet, dass er gerade wartet, dass ich ihm eine Information schicke. Also überall in meinem System kann ich den Ordner auswählen, der mich interessiert und dann kann ich für diesen Ordner dem System mit der BG Funktion sagen, dass ich möchte, dass genau dieses Projekt an ein bestimmtes Projekt gesendet wird. Und das Goal ist also auch so etwas wie ein Prompt, der es mir ermöglicht zu tatsächlich etwas viel ausgefallteres zu erstellen, dass es mir einfach ermöglicht, meinem Agenten zu sagen, hör nicht auf, bis du diese Aufgabe erledigt hast.

in dem Fall Cloud Code und unser Dashboardsystem eben immer Kontext, immer konsistent Kontext, weil er weiß, okay, ich habe vor 500 Tagen das und das geschrieben, das ist die Regel und hier habe ich eben das Wissen dazu und dann kannst du eben dann darauf eben auch aufbauen und dann klingt eben der Content auch nach dir und es ist ein in sich stimmiges System, weil du nicht immer wieder dem System erklären musst, wie es eigentlich zu funktionieren hat. Und das ist eben das Mächtige, wenn du das System richtig aufgebaut hast und eben auch ein YouTube Video dazu produziert hast, was eben dem System, sage ich mal, genug Input gibt, genug Informationen gibt, dann hast du eben auch ein in sichstimmiges System und du hast sozusagen eine Content Maschine, was dann im nächsten Schritt auch der Hebel ist, weil du kannst es nicht nur auf eine Plattform erstellen, weil ich habe das immer schon wieder gehört auch von Leuten, mit denen ich mir erst Gespräch bin. Referenzvideos, die zu dem Thema schon gut ranken und dann kannst du natürlich hier auch noch mal Pingpong spielen, kannst sagen, mach mir das noch mal mehr, kannst mir so da noch mal mehr Recherchepunkte geben und basierend auf dem Punkt hat man hier auch schon mit dem Script Grade Script Generator auch schon ein fertiges Script, wo du dann auch die Hooks noch mal individuell anpassen kannst und dann hast du hier auch schon ja mit wenigen Klicks ein perfektes Script, was du dann auch sofort direkt aufnehmen kannst, wenn weiter runter scrollen, kannst hier auch noch mal mit den Titeln etwas rum optimieren und das Besondere ist, Das ist jetzt hal nicht einfach nur irgendwie einen ChatGBT Prompt, der dahinter ist, der dir irgendein Output gibt, sondern es ist eben ganz genau das Cloud Code System, was dahinter ist, weiß eben ganz genau, okay, ich habe dieses Video hier recherchiert, ich möchte zu dieser Idee ein Video machen. Das ist eben das mächtigste und auch wie ich finde so das besonderste in dem Ökosystem, weil du jetzt halt basierend auf einem YouTube Video, was du jetzt hal mit der Recherche davor erstellt hast, was an sich schon sehr sehr gut ist, weil du eben eine richtige Recherche gemacht hast, du hast eine genaue Zielgruppenverständnis, du hast eben genaue Datenpunkte bekommen und alleine deswegen wirst du schon ein hochkonvertierendes YouTube Video erstellen können. Du hast hier eben auch die Hero Assets, die du mit einsetzen kannst und natürlich dadurch, dass wir das ganze YouTube Video eben mit dem richtigen Konzept aufgesetzt haben, mit einer guten Content Outline, mit einer guten Recherche und natürlich auch mit dem guten Inhalt, was an sich auf YouTube schon sehr sehr gut funktioniert, können wir natürlich dann auch repurposen auf den anderen Social Media Plattformen und dann hast du eben ein in sich stimmiges Ökosystem, was natürlich dann auch Stück für Stück für Stück immer besser wird und je mehr du da postest, desto besser wirst du natürlich auch mit dem System und du wirst sozusagen nicht nur einfach jemand, der Tools bedient, sondern jemand, der wirklich so die Vogelperspektive hat und eben ein gesamtes System bedienen kann und dann fängt's auch wirklich an Spaß zu machen.

Und natürlich, genau, ich habe für diesen Agenten alles komplett übernommen und füge jetzt tatsächlich den allerletzten Agenten hinzu, also den Agenten hier, der dafür verantwortlich sein wird. Tatsächlich ist diese Anfrage, also das ist ein Prompt, bei dem ich Ihnen darum bitte, dass ich vier Agenten habe und möchte, dass er Missionen für diese Agenten erstellt, die er ihnen dann zuweist. Also werde ich ihm wirklich alle Missionen geben und hier bitte ich Ihnen zunächst, die Reihenfolge zu bestätigen, einen Vorschlag zu machen und tatsächlich das Briefing für die Mission zu verfassen in Form von einsatzbereiten Prompts, die ich einfach bei Escot kopieren und einfügen kann. Sagt mir also in den Kommentaren Bescheid, wenn ihr möchtet, dass ich euch zeige, wie ich direkt über das Terminal die drei Agenten aktiviere, also diesem Workflow die Kontrolle gebe, damit dieser Agent die Möglichkeit hat, Befehle auszuführen, dann lasst es mich wissen. Also, wenn ihr möchtet, dass ich euch tatsächlich zeige, wie ich den Zugriff gewähren und das Problem lösen kann, denn es bleibt letztlich immer ein Zugriffsproblem zwischen den verschiedenen Containern, dann schreibt es mir in die Kommentare und ich werde euch in einer separaten Umgebung Schritt für Schritt zeigen, wie ich das aktiviere.

Hermes Desktop is a separate community project not built by Noose Research themselves, and it's an Electron-based native app that wraps around the actual Hermes agent install. If you're on Windows and you want to run Hermes, the most reliable path is to install the actual Hermes agent inside WSL2 and use the CLI directly because native Windows support for the agent itself is still labeled early beta. The desktop app gives you a panel where you can browse all the bundled skills and any custom ones you've installed, edit them, or disable the ones you don't want. You can toggle each one on or off per profile, which matters a lot for safety because giving an agent shell access on your main machine is not something you want to do casually. If you want a quick recommendation, if you're already comfortable in the terminal, just install Hermas agent directly with the official install script and skip the desktop app entirely.

HERMES (coupon : GOHERMESAI) : Documentation : Accès à mes 32 formations (coupon : AI2026) : Dans cette vidéo, je vous montre comment j ai créé une véritable équipe d agents IA avec Hermes Workspace, un outil puissant pour automatiser des missions, organiser des tâches et faire collaborer plusieurs agents intelligents comme une vraie équipe. Nous allons voir étape par étape comment choisir un serveur VPS, installer Hermes Workspace, connecter l API Claude, créer un premier agent IA, puis construire des sous-agents spécialisés capables de travailler ensemble sur des missions précises. Je vous montre aussi comment distribuer les tâches, automatiser des actions quotidiennes, créer un Agent Swarm, et configurer les conteneurs sur mon VPS. Si vous voulez comprendre comment utiliser les agents IA, les multi-agents, l automatisation IA, Claude, Hermes Workspace et les workflows intelligents pour gagner du temps et créer un système autonome, cette vidéo est faite pour vous. SOMMAIRE : 00:00 - Introduction : créer un workspace Hermes avec 4 agents IA 01:42 - Comment choisir le bon serveur VPS pour Hermes 04:17 - Installer Hermes Workspace et connecter l API Claude 05:48 - Comprendre le rôle et la mission d un agent Hermes 08:01 - Lancer mon premier prompt avec un chef d équipe YouTube IA 11:37 - Pourquoi les sous-agents IA sont essentiels dans Hermes 16:36 - Créer mes premiers sous-agents dans Hermes Workspace 19:33 - Corriger le bug Kanban dans Hermes Workspace 21:38 - Lancer des missions automatisées avec mes sous-agents IA 25:14 - Comment distribuer les tâches entre plusieurs agents IA 28:11 - Programmer une tâche automatique tous les jours 33:26 - Créer un Agent Swarm avec Hermes Workspace 40:12 - Configurer les conteneurs de mon VPS pour Hermes 41:05 - Mon cadeau : 32 formations IA offertes gratuitement HermesWorkspace AgentsIA

Hermes Agent just went from being one of those interesting open-source projects people were quietly testing to suddenly sitting at the center of one of the biggest agent stories right now. And the crazy part is that in the space of a few days, Hermes overtook Open Claw on Open Router's global daily app and agent rankings, started processing around 224 billion tokens per day, pushed past Open Claw's daily number, and then dropped a massive update that makes it feel like a personal AI operating system that can keep learning from your work. Open Claw still has the bigger all-time footprint with more than [music] 370,000 GitHub stars and more than 9 trillion cumulative tokens compared to Hermes at roughly 6 trillion. You drop the Higgsfield MCP endpoint into your Hermes config, reload the agent, and suddenly Hermes can start generating videos, images, [music] ads, landing pages, and other creative assets directly inside the pipeline. One article described Hermes AIOS as combining Hermes agent with the Ion UI, creating an open-source platform that can [music] manage tasks directly on the computer.

The other thing that I would say here is like some people say well this stuff is only for developers and programmers but Hermes agent it works in plain chat right and if you use something like desktop you can see how easy it is to just kind of use like chat GBT is it's integrated really nicely as you can see and people say well I I'll set up desktop and open web UI when I have more time but you know the people who set this up today are building systems that save them hours every single day so every day you wait is another day you competitors are pulling ahead. So we said for example you know what what is what skills do we have inside Hermes actually automates it by category gives us the examples the category how we can use it etc and that sort of thing right and that actually gives us some ideas of like what we could do with an example right here right and and some ways that we could actually use it which is pretty powerful so this is a delegation machine this is what separates people who use AI as a toy from people who use AI as a weapon right as as a powerful way to to to grow with AI automation. I I will be doing a video later on openclaw AI SEO but in general I don't tend to use it as much just because Hermes is so good right Hermes is so smooth it's like why would I use two agents when I can just use one I do switch between them you know week to week but if you look at like okay what am I using the most dayto day it's Hermes and it's Claude everyone good to see you here Michael thanks for joining all right so some people going to say you Not I'm not technical enough for doing this. Step 3.5 flash is back on this research portal and it's free for the next 15 days which means that you can use Hermes agent for free over the next 15 days which is pretty awesome right so you can use one of the most powerful agents in the world for free using step as you can see right now how do you set this up so what you want to do is you go over to terminal like this and then from here you're going to run Hermes setup Right. You can use something like uh Google search console or a mega indexer and then that's basically how it works right now and some people say well you know AI content is low quality and Google doesn't like it right I actually think that Google ranks helpful content and open claw when set up correctly produces highly helpful content right so Google's helpful helpful content update you know that targets thinn useless bad content but it doesn't target well ststructured genuinely useful AI assisted content and that's what we're creating with open claw that's a pipeline that we're using now people say well I need to be a developer to use something like openclaw or hermes agent and you can do all of this inside hermes as well honestly like openclaw runs um pretty cleanly once you set it up and it's like one command to install it so it's pretty quick and easy to use obviously people say well setting up openclaw will take me months honestly if you have the right system you can have openclaw writing and publishing content within one prompt, which is what we've done here.

They also added more generic platform plugin hooks, so new adapters can be built without changing the core. There are also new platform allow lists across Slack, Telegram, Mattermost, Matrix, and DingTalk, so you can restrict which channels, rooms, or chats are allowed to interact with Hermes. No OAuth also persists across profiles now, so you can sign in once and have profiles inherit the session. There are also new optional skills, including Shopify, Here Now, a personal shopping assistant, Anthropic financial services skills, a Kanban video orchestrator, and SearXNG search. SearXNG also connects to the web tooling changes, where Hermes can split search, extraction, and browsing across different backends.

So, I've got Hermes Desktop set up over here, and I've got Hermes open WebUI over here. Now, one thing I would note here is like some people say, "Why wouldn't you just use Hermes directly inside the chat here? Why don't you just use it on the TUI, the terminal user interface?" The thing is, compare this, which is like great for technical people, great for developers, but 99% of people are not coders, not developers, and don't really know how to use a terminal, right? And if you use something like Desktop, you can see how easy it is to just use like Chat GPT. We also have over 156 pages of wins and testimonials from people winning inside the AI Profit Room like you can see right here.

One option is where you can just go into your dedicated CLI and then use the Hermes computer use install command and this will then natively install it within your instance. Speaking of canban, if you haven't seen it yet, Hermes has a canban and it just got a massive update cuz it is something that allows you to create unlimited boards projects inside Hermes agent and even a capability to subscribe to project updates directly within your home channel through the gateway messenger that you have configured. And that is where you can just type in Hermes and then type in update to make sure you have the latest fetched update for the Hermes agent. But with Hermes's /go command, it is something that manages the the orchestration memory and hands off between different agents all tracked through the single Hermes canband board that we had just talked about while autonomous agents continuously run the loop with the goal command. And I truly believe that Hermes is going to be one of the go-to agent workflows that is open source and that many people will start using more and more as time goes because this is truly a remarkable open-source self-improving system that is going to be providing a lot of value.

Uruchom Claude Code, n8n czy inne aplikacje - bez kodowania - na sprawdzonym serwerze VPS: -10 z kodem NORBERT Film zawiera fragment sponsorowany platformy Hostinger. W tym odcinku serii Claude Code Guys z Robertem Szewczykiem pokazujemy, jak ogarnąć chaos w notatkach z Notion, telefonu i kartek i podłączyć to do AI. W tym filmie: - Czym jest Obsidian i dlaczego jest lepszy od Notion do pracy z AI - Jak działa graf wiedzy i co daje przy odpytywaniu Claude a - Instalacja Obsidiana synchronizacja iPhone Android przez iCloud - Wtyczka terminal; Claude Code wewnątrz Obsidiana - Pakiet 5 skilli do Obsidiana (Markdown, bazy, JSON Canvas, CLI) - Gotowy claude.md integracja WhisperFlow Obsidian od Roberta - System tagów i cotygodniowy rytuał porządkowania notatek z Claude - Komenda setup; Claude dopasuje konfigurację pod Ciebie WSPÓŁPRACA Szkolenia i wdrożenia dla firm: Społeczność Operatorów AI: Narzędzia które polecam (reflink): RIZE: FLOW: Ty otrzymujesz 30 dni za free i ja też 30 dni. Win-win :) INNE FILMY, KTÓRE POMOGĄ CI ROSNĄĆ: Playlista serii Claude Code: 2 Kompetencje AI na 2026: 3 Najlepsze funkcje Gemini: Kolejne odcinki z serii Claude Code Guys już wkrótce BĄDŹMY W KONTAKCIE: LinkedIn: Instagram: Operatorzy AI: Kanał Roberta: skutecznieefektywny SPIS TREŚCI: 00:00 - Chaos w notatkach 01:13 - Czym jest Obsidian? 06:46 - Fragment sponsorowany 08:07 - Jak zainstalować Obsidian 12:10 - Jak działa Obsidian 17:20 - Konfiguracja Claude Code Obsidian 32:40 - Podsumowanie 36:20 - Q A Plik z filmu: Plugin nr 1: Plugin nr 2: Plugin nr 3: Wtyczka Obsidian Web Clipper: ClaudeCode AIwPraktyce Poradnik

Wir hätten die Zeilenbrüche also nicht einfach so lassen sollen, also damit es sich um eine funktionierende URL handelt. Also gebe ich ihm diese Berechtigung hier und Sie werden sehen, dass er mir einen Code gibt, den ich kopieren werde und dieser Code, ich werde ihn einfach hier einfügen. Und dann sagte er zu mir, hören Sie, wir mußen tatsächlich noch einen hinzufügen, also Benutzername am Ende von Boot, also werde ich es genauso machen. Also sage ich ihm: "Ja, das ist also meine Hauptid." Ich drücke Enter und um es zu testen, werde ich ihn fragen, welches LM auf diesem System installiert ist. Sie haben also gesehen, dass er einfach einen Austausch zwischen dem, was ich hier schreibe, gemacht hat und dass das LM funktioniert und mir ein gutes Ergebnis liefert.

Now that it's set up and the project is deployed, you can click on the open token right here and you'll get this screen where you type in [music] your username and the password that we just copied and you can sign in now. For example, right now Gwen 3 coder is free on open router or you can use some other models such [music] as Deepseek V4, Minimax 2.7 or Gemma 4 or Devstrol 2 [music] which are really really cheap in comparison to the top tier models that you have right here. You can simply ask it something like what model are you using [music] and it will show you exactly right here it's using minimax 2.7 the context window as well as how [music] much of the context window is being used and how long the session has been going on. If we go back to Kimmy K2.7 and we go by apps, we can see that Hermas agent is taking up most of their usage right now [music] because it's just such a powerful model that is also affordable. If you want to see how to use paperclipip to set up multiple of these agents, let me know down in the comments because I just made a video on paperclipip right here that you might like to check out next.

Das heißt aber auch,   ihr könnt nicht einfach draufklicken  und es läuft und fertig, sondern ihr   müsst euch ein bisschen damit auseinandersetzen. Wenn die App jetzt läuft, könnt ihr  wirklich einfach dieses Briefing,   was ich jetzt quasi habe mir in einer  Markdown-Datei zusammenfassen lassen,   hier reinstellen. Aber   ihr macht dabei etwas Sinnvolles, ihr nehmt den  Rechner mit und macht Sessions mit uns zusammen,   wo es darum geht, wirklich schrittweise  zu verstehen, in den 4 Wochen, mit über   12 verschiedenen Sessions, reinzukommen in  die verschiedenen Themen, die es braucht,   absoluter Pro zu werden mit Agents, bis hin auch  zu, wie funktioniert eigentlich ein Agent Harness,   was kann ich mit Cloud Code machen, was ist  der Unterschied zu einem Chatbot, wie arbeite   ich damit? Hier könnt ihr sagen, wer sind die  Chat-Dealer, also mit wem wollt ihr   schreiben, wen wollt ihr befragen von denen,  die hier sind. Wenn ihr jetzt noch eine bessere   Schritt-für-Schritt-Anleitung  wollt, wie man es installiert,   ihr habt ja gerade gesehen, Christoph  Magnussen erklärt die Sachen manchmal   einfach eher im Überblick.

So we can say, can you build out something like a three sheet financial dashboard with income statements, revenue breakdown with charts as well as conditional formatting and different variances. So right now we have one agent working on the spreadsheet task making it work upon doing financial analysis where we can have another agent that we can create that is powered by Hermes to execute something else where it can work upon I don't know maybe doing analysis on a report like creating a white paper for a rural rural EV charging infrastructure. This is like some simple task that you can have powered by the Hermes agent and you can have multiple agents being deployed simultaneously to do different things off of your computer. It can even be a smart desktop automation tool where it can even run schedule tasks for you. Please make sure you take a look at our second channel, join the newsletter, join the Discord, follow me on Twitter, and lastly, make sure you guys subscribe, turn on notification bell, like this video, and please take a look at our previous videos so that you can stay upto date with the latest AI news.

This is where you can manage all of these different individual tools like web search, browser use, terminal, file operations, and many others. You have the ability to set up cron jobs which is where you can set up multiple scheduled tasks like you would with cloud code or codeex gateway is for you to connect the Hermes agent desktop app with many of these other platforms like telegram, discord, iMessage and many others so that you can even control it from your phone and then obviously the settings is where you can easily have it so that you can change the theme as well as the network. Also something important to note is that if you have a preconfigured open claw session or an instance that you have multiple tools and skills and you want to migrate it to Hermes, you can easily click on migrate to Hermes with this desktop app so that you can use all of your configurations, API keys and sessions and skills directly within Hermes. And with this desktop app, you have so many different use cases like using it as a self-improving large language model wiki where it is a second brain to assist you on multiple workflows or different described components that you can use for further generations. Or you can consider joining our private Discord where you can access multiple subscriptions to different AI tools for free on a monthly basis plus daily AI news and exclusive content plus a lot more.

Weil wenn jetzt hal deine Mitbewerber zehn Videos machen, wo die relativ ähnlich aussehen, wo wenig Authentizität stattfindet, wenn du jetzt hal das ganze ja einfach authentisch rüber bringst und einfach du selbst bist, wirst du damit automatisch auch wenn du den gleichen Inhalt produzierst mehr hervorstechen, weil die Leute einfach sehen, du bist ein Mensch und da kommt auch eine emotionale Connection irgendwie bei rum. Das ist mirich extrem wichtig, dass wenn du auch einen Kanal haben möchtest, der dann auch Anfragen generiert, du musst deine Messages, deine Aussagen, deine Kernaussagen immer wieder wiederholen, weil dadurch, dass du es wiederholst, merken sich die Leute das und wissen: "Ah, okay, du bist derjenige, bei dem es um Thema XY geht, der steht dafür, der hat diese Aussagen getätigt, das ist sein Standpunkt und je mehr du es wiederholst, desto mehr ist es auch in den Leuten drin und sie wissen ganz genau, wer du bist, was du machst und was dich vielleicht auch von den anderen Mitbewerbern so differenziert." Was ich auch immer wieder sehe, besonders im Zeitalter von KI, das theoretische Wissen ist überall und die Leute, die wirklich einen Kanal haben, der wächst und auch wirklich etwas Besonderes produzieren, da ist die Execution einfach King. l natürlich auch eigene Testimonials auf deinem Video, auf deinem YouTube-Kanal hoch, um einfach zu zeigen, hey, ich habe hier ein Angebot, das klingt interessant, verpacke das Ganze vielleicht auch noch mal mit Mehrwert, weil die Leute sehen einfach, was du dann machst, dass du auch wirklich das hältst, was du versprichst und du musst da jetzt hal nicht eine Verkaufssow machen und irgendwie hyper kompliziert pitchen, sondern einfach sagen: "Hey, ich habe ih ein Angebot, das habe ich mit dem gemacht, so und so kann das aussehen und alleine dadurch werden die Leute sehen, wenn das denn für sie interessant ist, ob das denn auch das richtige Produkt ist und genau da kannst du natürlich das Ganze dann einfach auch verbinden. Das nächste, was du beachten solltest, wenn du einen Kanal haben möchtest, der auch verkauft und auch wirklich ein perfektes YouTube Video zu konzipieren, baue Thumbnails, die dir nicht gefallen, sondern eben auch deiner Zielgruppe, weil wie du weißt, Thumbnails sind eben 80% des Erfolges eines YouTube Videos und stell dir dann immer die Frage, worauf springt denn deine Zielgruppe an? Das heißt, wenn ich jetzt im letzten Video von dir eine Präsentation bekommen habe, ein Talking Head Video und jetzt auf einmal ist ein Entertainment Video, wo du einen Vlog machst oder wo du ein Podcast machst, klar kann es vielleicht auch die gewissen Interessen ziehen, aber es ist einfach die falsche Erwartungshaltung, weil wenn ich mir jetzt ein letztes Video von dir angeschaut habe, was ein Talking Head Video war, dann will ich auch wieder ein Talking Head Video, weil das Format mir gefallen hat, weil mir die Art, wie du es rüber bringst gefällt etc.

Deshalb gehe ich hier zurück in mein System, also in unserer Installation sage ich ihm jetzt: "Hören Sie, wir gehen zur ersten." Also Konfiguration, ich wähle einfach eine aus, ganz einfach und starte, um zum nächsten Schritt zu gelangen. Also wollte ich wirklich, dass dieser Abschnitt hier tatsächlich aufgenommen wird, um Ihnen zu zeigen, dass man jedes Mal, wenn ein Fehler oder ein Problem auftritt, mit dem Agenten sprechen muss. Also gibt er mir hier ein paar Befehle, die ich ausführen soll, um das System zu aktivieren, aber ich kann ihn auch bitten, das für mich zu machen. Aber das Hinzufügen von Front, also diesem Workspace Teil, bleibt immer ein sehr interessanter Aspekt und dass ich natürlich jederzeit arbeiten und die Qualität der Arbeit verbessern kann, indem ich eine visuelle Übersicht habe. Es sollte also einen sehr ordentlichen Status haben, also es gibt keine Fehler, aber wir werden es kopieren.

Also sagt er, ich gehe das Risiko ein und erwähnt mir eigentlich, dass er es mit dem Modell startet, dass mir also lassen wir es noch weiterlaufen. Schauen Sie, das ist ein bisschen tatsächlich machen wir jetzt den Teil, von dem ich Ihnen erzählt habe, dass er erzwungen wird, eigentlich der Start von Gateway Open Cloud, jedes Mal, wenn der Container oder der Server neu gestartet wird und dadurch wird vermieden, dass man das in Zukunft noch einmal machen muss, also die Installation oder das Starten des Gateways manuell durchzuführen. Also bevor wir mit dem Test beginnen, ihr müsst euch einfach diesen Link hier notieren, also genau diesen, wenn du möchtest, der ist dir ermöglicht, die URL mit dem Passwort abzurufen oder du kannst, wie wir gesagt haben, einfach nur also hier genau diesen Link mit deinem Token verwenden. Also, also wir starten jetzt und er wird mir tatsächlich einige vorschlagen und ich weiß eigentlich, dass es ziemlich viele davon gibt, tatsächlich wie Jimny und andere, die ebenfalls sehr leistungsstark sind und im Cloudmodus laufen. Du siehst, als ich hier "Hi" geschrieben habe, hat er geantwortet: "Tatsächlich ist das unser System." Also hier ist es so, ich kann hier das liegt daran, dass ich einen anderen Telegram Account erstellt habe, aber das hier ist eigentlich das Haupttelegram und dadurch kann ich die Unterhaltungen tatsächlich so führen, wie es sein soll.

If you  want to take a look at all the different terms   there are, you can click over here and here  you see new issues, working on it in progress,   blocked or done. upload them all over here and then paperclip will   choose which one will be used from within  the media library because also over here,   it took one of those images. So what I would prefer over here  is that these will be titles because now I think   it could be displayed better, or I should make  them purple like the titles over here. Um,  I use chat CPT before, so you can just ask what   are some great paperclip use cases that can help  me with my, and then you can name your business,   my YouTube channel. And over here,   you can click on the plus, you can create a  new company, use ggpt to create a mission,   all that stuff.

Hermes says the background review fork is more rubric based now, prefers updating the skill the agent just used, can handle reference and template files, and properly inherits the parent runtime, including provider, model, and credentials. So the API task depends on the schema task and the test task depends on the API task. Instead of hoping three background agents are doing the right thing somewhere, you can see what is running, what is ready, what is done, and what got stuck. If a worker process dies mid-flight, the dispatcher can detect that the process is gone, release the claim, move the task back to ready, and let a fresh worker try again. For example, one run can be marked crashed because of an outofmemory problem and the next run can be completed with metadata explaining that the worker switched to a chunked strategy.

Also könnt ihr beispielsweise Gemini, Codex oder auch Deepak alles unter einer Haube von Cloud Code gesteuert in der Codingumgebung eurer Wahl vereinen und somit ein mächtiges Tool schaffen, wo ihr auch nicht von Codex zu Geminite zu Cloud Code zu was auch immer halt switchen müsst, sondern alles unter einer Haube für euch einfach erreichbar und das zeige ich euch in diesem Video mit ein bisschen Bonuscontent, aber dazu komme ich gleich. Dann gibt ihr jetzt hier ein Cloud, also Cloud müsst ihr auch installiert haben und hier müsst ihr dann jetzt wieder auf die Website gehen und diese Sachen wirklich Schritt für Schritt einfügen, also damit das auch wirklich alles drin ist. Damit ihr jetzt Deeps installieren könnt, ist es ein bisschen anders der Prozess, weil Deepsc ist ein Open Source Modell, da gibt's jetzt nicht diesen Plugin, da müsst ihr eine andere Version von Cloud Code halt nutzen und das ist alles nur möglich, weil Enhropic, also das Unternehmen, was Cloud Code halt entwickelt hat, aus Versehen die Codebase für Cloud Code released haben, ohne dass sie es eigentlich wollten. Hat jetzt hier verschiedene Bilder geholt, verschiedene Tassen halt auf jetzt auf ein anderen Sprache kann man eigentlich könnte es man auch ändern, da kann man natürlich iter tief rangehen, aber er hat das Design von Ferrari halt verwendet und man sieht hier noch ein paar Fehler. Das finde ich halt so powerful, weil man ja auch dann letztendlich mit den richtigen Verbindungen könnte man auch seine Modelle lokal laufen lassen und alles lokal halt dann mit diesem Cloud Code hier verbinden, wo man dann letztendlich dann die einzigen Kosten, die man dann hat, ist die Hardwarekosten für sein PC, weil man braucht dafür eine starke GPU und halt die Stromkosten und das sind die einzigen Kosten, die man hat, ohne halt dann API Kosten oder sich irgendein Abo halt machen muss und das finde ich halt einfach so eine starke Möglichkeit und spart einem langfristig halt auch eine Menge Geld.

Do you want to make AI images? Do you want to make content? Do you want to make a website or product? Do you want your business to show up in AI search results? If you want free AI courses, hit follow and comment muffin and I ll DM them to you.

Vorweg sei aber gesagt, ich werde euch heute nicht mit Theorie langweilen, denn hier auf YouTube gibt es schon genügend gute Tutorials zur Entstehungsgeschichte von Hermes und die die Frage beleuchten, warum gerade Hermes so viel besser sein soll als alle anderen KI-Agentensysteme. Und ich habe lange überlegt, was ich euch denn zeigen kann, denn wie gesagt, Analyse von Webseiten oder irgendwelche ja Zusammenfassungen, KI News, das sind alles Sachen, die mich relativ wenig interessieren, die euch relativ wenig interessieren, die haben keinen wirklichen Nutzen, aber etwas zu finden, was für euch alle interessant ist, ist natürlich gar nicht so einfach. So, jetzt ist das Problem, dass ich bis heute tatsächlich wirklich noch nie in der Lage war, das genau zu bestimmen, einfach weil ich es nicht ausgerechnet habe, mir nicht die Zeit dafür genommen habe. Und genau deswegen habe ich mir schon vor längerer Zeit vorgenommen, all diese ganzen Laborberichte einfach mal zu digitalisieren und dann mir ein Programm zu bauen oder etwas, womit ich tracken kann, wie denn gerade meine Blutwerte sind. So, das ganze zu machen, dafür habe ich leider keine Zeit gefunden und deswegen werden wir jetzt Hermes das Ganze geben.

Wenn du gerne mit Cloud Code anfangen würdest, aber dir das ganze Setup mit Terminal oder Entwicklungsumgebung bisher zu technisch war oder wenn du Cloud Code auch schon nutzt und überlegt hast, ob du in die Desktop App wechseln könntest, dann ist heute vielleicht dein Tag, denn Anthropic hat ganz still und leise ein großes Update für Cloud Code in der Desktop App veröffentlicht und das Update ändert wirklich, wie die meisten Leute Cloud Code von jetzt an benutzen werden. Ich persönlich nutze das auch immer noch, weil ich mich einfach dran gewöhnt habe und ich das sehr übersichtlich finde, aber mittlerweile ist die Cloud Desktop App so gut geworden, dass man Cloud Code auch theoretisch nur hier nutzen kann. Das heißt angenommen, ne, man startet jetzt hier einfach eine neue Session oder neuen Chat und ich kann jetzt in dieser Session hier Cloud Code Zugriff auf einen Ordner geben, lokal auf meinem Rechner. Hier sieht man so, welche Ordner ich zuletzt verwendet habe, aber ich kann ja auch einfach einen neuen Ordner öffnen und ich habe mir jetzt hier einen neuen Ordner erstellt namens Lernkartenapp und den Ordner kann ich dann hier Cloud Code zur Verfügung stellen. Normalerweise müsstest du die App jetzt erstmal in einem Browser öffnen und kannst sie dann selber testen und du müsstest erstmal Cloud über das Chrome Plugin, ne, also die Chrome Extension im Browser Zugang geben, damit er die App selbst einsehen kann und testen kann.

The run command tells Olama to start a model, and Llama 3 is the name of the model from the library. Use the search bar to look for models like Mistral or Gemma and click on any model to see details and the exact command you need to run it. If you want to download a model without opening a chat session, you can use Olama pull model. You can run Olama list to check what's installed and remove anything unnecessary with Olama RM model. As you explore more of these features, I am sure that you'll start to see how flexible local AI can get, especially when everything is running directly on your own system.

Image generators, video generators, without really understanding what any of them are and how they actually work. Just learning the basics of how AI works can make you so much better using every single one of these tools, whether it s chat GPT, image generators, or even more advanced systems like clang and seed ants. What AI actually is, the five types of AI tools you can use right now, how they work under the hood, and most importantly, how to prompt each one so you can get results that are actually worth your time. The models are trained on massive data sets of videos paired with descriptions, and they learn not just what things look like, but how they move: spatial relationships within each frame and temporal dynamics across frames. When you give a video model a prompt, it generates frames one by one, each building on the previous one, maintaining consistency in how objects look music and move.

This is what s actually music going on with free AI video right now, and a lot of the lists you find online are already out of date. For this video, works music means I d actually recommend it to someone who wants to make AI videos for free without spending hours music fighting limits. There s a director mode that improves your prompt for the model before generating, and unlike most prompt enhancers, music this one actually feels like it knows what its own model needs. music Images, video, image to video, four images at once, horizontal, vertical, custom animation where you tell the model exactly how the image should move, restyling. From there, you ve got vertical text-to-video, music image generation that produces four images at once, image-to-video, image editing using what looks like a Nano Banana style backend underneath, an AI avatar library bigger than what most paid tools music have.

Only music five were worth using. Watches your screen and walks music you through any software Figma, Excel, Photoshop. A local AI financial advisor that connects music to your bank and answers questions about your actual spending, debt, and savings. An AI overlay on Windows that stays completely invisible during Zoom, Meet, music and Team screen shares. An open-source second brain that connects all your notes and research music with AI generated wiki summaries and citations.

Dadurch entsteht jetzt ein System, wo man wirklich überlegen muss, wie man Skills lädt, dass sie sicher sind, dass sie auch verifiziert wurden, dass diese wirklich diese vom Hersteller sind und dann haben wir natürlich auch den KI Agenten selbst besser überwachen kann, dass er jetzt nicht einfach dann die Daten löscht, z.B. damals mit JetGBT als JetGBT Memory gekriegt hat, habe ich ein JetGBT System gebaut, wo man einen Agenten kompromitiert wurde über Prompt Injection und dann im Speicher des Agenten wurde eben abgespeichert, dass er zukünftige Instruktionen laden sollte von GitHub und jetzt mit diesen ganzen neuen Agenten, die es da gibt, da hat man ja viel mehr Möglichkeiten und da habe ich eben gezeigt, dass man diese verschiedenen Agenten, alle OpenCla, Nanoclaw und so weiter, dass man die dann eigentlich einem neuen Command and Control Center hinzufügen kann, also ist wirklich einfach ein neuer Weg, wie man so wie wir als Benutzer Computer jetzt über Agenten verwenden mit Prompts, so werden Angreifer diese Computer auch verwenden. Es war das Cloud oder an Tropic hat den Codeinterpretor gebaut, also ein Computer, den Claud verwenden kann und diesen Computer haben sie Internetzugriff gegeben und ich habe mir dann genau diese Mittel Default Einstellung angesehen mit diesen erlaubten Domänen und da ist mir eben aufgefallen, dass Antropic ihre eigene Domäne erlaubt und das bedeutet, dass man mit dem Antropic API kommunizieren kann. so wie ich es gerade erklärt habe und das hat eben den Hintergrund, dass die nerdige Persönlichkeit in den Trainingsläufen im Reinforcement Learning immer wieder Wörter wie Gablins produziert hat und diese durch Reinforcement Learning belohnt wurden durch meinetwegen Zufall und daraufhin das Modell eben immer weiter diese Worte produziert hat, weil es eben gesehen hat, dass es dadurch mehr belohnt wird und diese nördige Persönlichkeit, die ist durchgesickert auch in jede andere wie auch die Standardpersön Persönlichkeit, was dann letztendlich dazu geführt hat, dass man sogar im System prompt die Wörter Goblins, Gramlins und so weiter, wie es Ry geschrieben hat, unterdrückt hat im Nachgang, um dieses Verhalten eben zu verhindern. All das muss man als Nutzer im Hinterkopf halten, dass solche Kleinigkeiten in den Trainings Runs, wie wir es jetzt bei Goblins gesehen haben, sich auch in jedem anderen Bereich widerspiegeln und man deswegen natürlich wie immer gerade jetzt auch ein guter Gelegenheit Codex und Cloud Code in vielen Aufgaben parallel laufen zu lassen oder als Model Council laufen zu lassen, um nicht sich auf einen Modell Output tatsächlich zu verlassen, gerade eben nicht bei kritischeren Geschäftsentscheidungen.

अगर sanस 2.0 क आप ल ग क फ र म ए ड अनल म ट ड य ज करक क छ इस तरह क एआई व ड य जनर ट करन ह त आप सह जगह पर आए ह क य क आज क इस व ड य म म आप ल ग क त न स क र ट एआई ट ल द ख न व ल ह ज सक आप अपन स म र टफ न स भ य ज कर सकत ह प स स भ य ज कर सकत ह और स डस 2.0 क ड फ न टल फ र म य ज कर प ए ग त हम र ल स ट म ज सबस पहल ए ट ल ह उसक न म ह व ड य आईए व ड य आईo क ज ल क ह म न ड स क र प शन म द द य ह त वह पर ज क इसक आप एक स स कर सकत ह अपन म ब इल य फ र प स स स ट र ट पर क ल क करन ह उसक ब द आप ल ग क यह पर ट र ई स ड स 2.0 ल ख ह यह पर क ल क करन ह ल ग इन पर क ल क क ज ए इसक ब द यह पर स इन अप फ र फ र ल ख ह आ ह यह पर क ल क करन ह यह पर आप ल ग क अपन ईम ल आईड ह व ए टर कर द न ह अपन न म ए टर कर द न ह और यह पर ओट प क आप ल ग क व र फ ई कर ल न ह ल ग न ह त ह द स त आप यह पर स क र न पर द ख सकत ह हम ल ग क 500 फ र क र ड ट द खन क म ल च क ह और अगर आप ल ग क ब र ग स क व श चन क क प ल ट कर द ग त एक स ट र 200 क र ड ट स म ल ज एग य न क यह पर 500 प लस 200700 क र ड ट स हमक म ल च क ह अगर आप ड ल ल ग न कर ग इसक ड स कर ड क ज इ कर ग त आपक और एक स ट र क र ड ट म ल ज एग ट र ई स ड स ल ख ह आ ह यह पर क ल क करन ह फ र यह पर आप ल ग क व ड य म डल क अ दर आक स डस 2.0 फ स ट ह व स ल क ट करन ह आस प क ट र श य यह पर ड य र शन यह पर र ल य शन सब क छ आप स ट कर सकत ह फ र यह पर आप ल ग क अपन ज प र प ट ह व प स ट कर द न ह और यह पर जनर ट पर क ल क करन ह व द इन अ फ य म नट क अ दर द स त आपक फर स ट व ड य ह व बनकर आ ज एग कर हन क आव ज हम र ल स ट म ज स क ड एआई ट ल ह उसक न म ह न ट ज pट न ट ज pट क य ज़ करन क ल ए आप ल ग क इस व बस इट पर आ ज न ह यह पर एi व ड य जनर टर क अ दर आ ज न ह त आपक स डस 2.0 स डस 2.0 द न एi म डल ह व द खन क म ल ज त ह ब ल क ल इज ह क छ नह करन ह प र फ इल पर ट प करन ह ल ग इन व द ईम ल ल ख ह आ ह यह यह पर क ल क क ज ए स इन अप पर क ल क क ज ए इसक ब द आप ल ग क च ह ए एक ड ट edu ड म न त इसक ल ए आप ल ग क इस व बस इट पर आ ज न ह यह पर ईम ल आईड क प कर ल न ह यह पर आक इसक प स ट कर द ज ए अपन नय प सवर ड ह व स ट कर द ज ए क छ इस तरह स स ल इड करक सबम ट पर क ल क करन ह त ऑट म ट कल एक ईम ल व र फ क शन ह व आप ल ग क एक ट म म ल पर आ ज एग त क य करन ह ? यह पर र फ र श ल ख ह आ ह यह पर क ल क क ज ए थ ड स स क र ल कर ग त आप ल ग क न ट ज pट क म ल आ च क ह अब इस म ल क आप ल ग क ओपन कर ल न ह फ र यह पर ल ख ह आ ह व लकम ट न ट ज प ट इस पर ट प करन ह त व न अ फ य स क ड क अ दर द स त आपक ईम ल आईड ह व सक स सफ ल यह पर व र फ ई ह च क ह व पस स आप ल ग क एi व ड य पर ट प करक एi व ड य जनर टर क अ दर आ ज न ह अब यह पर आप द ख सकत ह sड 2.0 म आप क ई भ एi म डल क स ल क ट कर ग त यह पर 100 स भ ज य द क र ड ट य ज ह त ह अब यह पर आप प र फ इल पर ट प कर ग त अभ आपक प स स र फ 15 ब स क क र ड ट ह और ज र प र म यम क र ड ट ह त क ई ब त नह आप ल ग क क य करन ह ? अपग र ड पर ट प करन ह फ र इस ग फ ट आइक न पर ट प करन ह व पस स अपग र ड न उ ल ख ह आ ह यह पर ट प करन ह व य ऑल प ल न पर क ल क करन ह और यह पर आप ल ग क एज क शन पर ट प करन ह ड फ न टल द स त ऐस व ड य क स न भ नह बन य ह त व ड य क ल इक कर द न क य क 00 पर म थ म व द उट क र ड ट क र ड आप ल ग क 1000 ब स क क र ड ट ह व म ल ज त ह ग ट वन म थ फ र फ र ल ख ह आ ह यह पर ट प करन ह ए ड सक स सफ ल द स त आपक अक उ ट ह व व र फ ई ह च क ह आपक प ल न ह व अपग र ड ह च क ह अब यह पर आप ट प कर ग त आप द ख सकत ह ज ब स क क र ड ट ह व 1000 ह च क ह और प र म यम क र ड ट 100 ह च क ह अब इसस आप सड स 2.0 क ज व ड य ह व जनर ट कर प ए ग यह स आप अपन र फर स व ड य य फ र इम ज अपल ड कर सकत ह यह पर अपन प र म प स ट कर सकत ह यह पर र ल य शन, ड य र शन, ऑड य और प र म क एनह स करन ह त सब क छ करक जनर ट व ड य पर ट प कर द ज ए व द इन अ फ य स क ड क अ दर द स त आपक फर स ट व ड य बनकर आ ज एग क छ इस तरह स अब इस व ड य क ड उनल ड करन ह त यह पर ब ल आइक न क अ दर ड उनल ड ल ख ह आ ह यह पर ट प करक इसक ड उनल ड कर सकत ह आग ज न स पहल ब त कर ल त ह हम र आज क स प सर क ज सक ह ल प स 2.0 क आप फ ल ल क ट र ल कर सकत ह और क छ इस तरह क ह ल व ड ल वल क आप व ड य जनर ट कर प ए ग व भ आपक फ स क स थ इतन ह नह इसक अ दर आप यह पर 1080p क अल ट र ह ई र ल य शन क यह व ड य जनर ट कर प ए ग SDS 2.0 क अ दर त ज ए ट ल ह उसक न म ह वन ए ड ओनल HFL AI HFL क ल क आपक ड स क र प शन म म ल ज एग त यह पर आप द ख सकत ह sड स 2.0 अब 1080p सप र ट करन लग ह त इसक य ज़ करन क ल ए व ड य क अ दर आप स डस 2.0 क स ल क ट क ज ए ज सक अ दर आपक स डस 2.0 और स डस 2.05 क ऑप शन म ल ज त ह आप इज ल स ल क ट कर प ए ग क तन स क ड क आपक व ड य जनर ट करन ह क न स आस प क ट र श य च ह ए और यह पर 2.0 म डल स ल क ट कर ग त 1080p क आप र ल य शन जनर ट कर प ए ग व भ बह त ह कम क र ड ट स क अ दर अब यह पर इस म एक बह त ह अच छ फ चर ह व ह यह पर फ स ड ट क शन क अगर क ई भ फ स यह पर ड ट क ट नह ह , एल ज बल नह ह , त न ट एल ज बल ल खकर आ ज एग बट अगर आपक फ स यह पर एल ज बल ह च क ह , त इसक ह ल प स आप अपन ख द क फ स क स थ य व ड य जनर ट कर प ए ग त यह पर म न अपन इम ज ह उसक अपल ड कर द य एक छ ट स प र स कर द य अगर आपक ज नन ह इसक ड ट ल म प र ट क स क र एट करन ह त आई बटन म च क कर वह पर ल क म ल ज त ह अब यह पर प र प ट ए टर करन क ब द और क छ म न स ट ग च ज नह क य ए ड जनर ट पर ट प क य त क छ इस तरह क आउटप ट बनकर आ च क ह और यह पर ट प करक इसक ड उनल ड कर सकत ह र इट न उ HP य एकल त AI ट ल ह ज क म र क ट म सबस च प स ट प र इस म आपक SDS 2.0 क फ ल ल य ज़ करन क ऑप शन द त ह त ड स क र प शन च क करन आपक वह पर उसक ल क म ल ज एग हम र ल स ट म ज थर ड एआई ट ल ह उसक न म ह र ट एi इस एआई ट ल क ल क म न ड स क र प शन म द द य ह यह पर आप च क कर ग त व द उट स इन अप आपक 10 क र ड ट म ल ग व थ स इन अप पर आप ल ग क 100 क र ड ट म ल ग ज स पर sided 2.0 क य ज कर सकत ह बस यह पर ल ग इन पर ट प करन ह अपन Google अक उ ट स य फ र अपन ट म अक उ ट स ल ग न कर ल न ह फ र यह पर आपक ओट प व र फ ई करव ल न ह त आपक क छ इस तरह क ड शब र ड ह व ओपन ह ज एग अब यह पर आप ल ग क ट र ई न उ ऊपर ल ख ह आ ह यह पर क ल क करन ह त क छ इस तरह स द स त ह क सफल क ज स एक य आई ह व ओपन ह ज एग और यह पर आप ल ग क 100 क र ड ट स ब ल क ल फ र म द खन क म ल च क ह अब यह पर स डस 2.0 क य ज़ करन क ल ए आप ल ग क यह पर अपन इम ज ह व अपल ड करन ह त व कर सकत ह प रम प र स करन ह त व कर सकत ह इसक अल व यह पर आप ल ग क ज स हमन आग बत य थ स डस 2.0 स डस 2.0स ट द न क ऑप शन ह व भ म ल ज त ह यह स आप ड य र शन स ल क ट कर सकत ह आस प क ट र श य स ल क ट कर सकत ह र ल य शन स ल क ट कर सकत ह और यह पर अपन व ड य जनर ट कर सकत ह त यह पर म न प र ट ह व प र स कर द य और यह पर जनर टर पर ट प क य त क छ इस तरह क व ड य बनकर आ च क ह द स त य द रख ए य सभ फ र वर जन म आप क छ ह स क ड क व ड य ह व जनर ट कर प ए ग और अनल म ट ड य ज़ करन क ल ए आप ल ग क ट म म ल क र क व यरम ट ह ग ड उनल ड करन क ल ए द स त यह पर ट प क ज ए त य व ड य आपक ड व इस म ड उनल ड ह ज एग स द ट स इट फ र ट ड स व ड य ग इस आई ह प य त न स क र ट ड आई ट ल क य ज़ करक आप SDS 2.0 स ब ल क ल फ र म व ड य जनर ट कर प ए ग और ड स क र प शन म आप ल ग क य स र ट ल क ल क म ल ज एग ए ड HFI क भ च क आउट करन मत भ लन क य क सबस च प स ट प र इस म स डस 2.0 क अगर आपक य ज़ करन ह प र फ शनल व स और क छ इस तरह क व ड य बन न ह त ग इस उसक ल क आपक ड स क र प शन म म ल ज एग और इस व ड य पर ट प करक आप Can 2.0 क फ र ए ड अनल म ट ड व ड य क प र ट वन व ड य ह व च क आउट कर सकत ह

Wenn du Claudi AI nutzt, dann macht dieser einfache Tipp einen riesengroßen Unterschied und die meisten übersehen diese Funktion. Claud hat eine Einstellung, mit der du einmal festlegst, wer du bist und wie Claud mit dir kommunizieren soll. Dazu gehst du einfach in die Einstellung, dann auf allgemein und da findest du dieses Textfeld. Welche persönlichen Präferenzen soll Claud bei Antworten berücksichtigen? kannst du da reinschreiben: Ich bin Social Media Managerin, antworte immer auf Deutsch, schreibe kurz und direkt.

even directly in each field you can write one JavaScript method you can write like two uppercase and it just works right so like you don t even need to do a whole code node or open a separate file you can do everything visual and if you need a little bit of extra power right there you just do it right in code so right now I m sure we all can see the landscape since you re here you can go anywhere and build an agent it s so easy to build agents right One of the problems we re seeing and where the winners are going to lie is seeing what your agent can do, knowing what it s doing, seeing what went wrong and being able to tweak it and fix it. So it can be when a form is submitted which we can say uh lead and then that will be name and then the native NAN form will come up here and say my name is Liam right so that s a trigger there s also stuff like schedule and hundreds and hundreds of others there s web hook and uh API calls and everything like that so you have a trigger and then there s actions things that happen in other apps like Google create a contact, send an email, um, and like Salesforce if you want to create a contact just gluing stuff together. uh in here the in the system message that s where you would normally put all of your different stuff but I like to make it a little bit more modular by I only put my like email prompts like oh don t use m dashes don t do whatever in the actual description which maybe some people would disagree with that approach I really like it because it s modular I can copy a tool from different workflows and it will just Um I m sure people are probably ahead of me now in from a tool perspective of adding them. ever then it would run once per hour do those actions and these can be connected to Slack to just ping a channel so it s still human in the looping for everything but just running in the background you don t even have to initiate it so you can imagine all the things in the background what I use this for if there s a GitHub issue or a PR or something I have it come in I I have it go through scan the scan the code do whatever and send off messages to people asking them details that I know I ll when I read through it and then I just have the message come to me because I don t want to AI message co-workers or clients or anything without seeing it first. uh let s say you design a workflow in uh and and and you want to um to use it as an API and call it from someplace else like in my company we have we are able to deploy Microsoft stuff the rest is complicated but use it as an engine and then just call it and hide the is it that is that possible absolutely I make um full rest APIs and n all the time and I make you know the fancy uh like swagger docs and anything like I just like I mentioned with the coupon generation.

Now, it has a desktop app on Mac and Windows, and it can do work across your whole computer, not just inside one tool, across everything. Codex can now see your screen, move its own cursor, click on things, and type into apps, like a person would. Codex can now remember things about you, your preferences, the way you like things done, the tools you use, the mistakes it made before, the corrections you gave it, all of it, across sessions. Codex can take a screenshot, make a new image based on what it sees, then drop that image right back into your design tool, all in one flow. This means Codex can pull info from your tools, take action in your tools, and keep your work moving without you having to copy paste between five different apps.

Cloud Chat, Cowork und Projekte. Die meisten nutzen nur Cloud Chat und fragen sich, warum es sich wie Chat GPT anfühlt. Cloud Chat ist vor allem für schnelle Fragen. Ohne Erinnerung oder Kontext startet Cloud Chat jedes Mal leer. Ich sag dir, du wirst Chat in Claw zu schnell nicht wieder nutzen.

AVISO: este vídeo es el fragmento específico de HERMES de este otro vídeo por lo lo quieres ver completo. Este vídeo es un tutorial de instalación pero si quieres construir un sistema 24 7 robusto para tu día a día, en este club cientos de personas estamos construyendo uno paso a paso. Si quieres construirlo acompañado, de otros socios como tú y donde preguntar dudas, compartir recursos, skills, herramientas, igual esto te interesa: Más de 30.000 personas ya han pasado por nuestra newsletter con este enfoque y están construyendo sistemas personales donde la IA se apoya en su propio conocimiento, no en prompts sueltos. Nuestra nesletter aquí abajo: COMANDOS 1) curl -fsSL bash 2) openrouter.ai 3) hermes status 4) hermes CRONOLOGIA 00:00 Qué es Hermes Agent? 01:04 Diferencias entre Claude Code vs Hermes Agent 02:30 Skill para mejorarse a sí mismo o núcleo automejora 04:12 Cómo instalar Hermes Agent 05:00 Elegir modelo LLM en Hermes Agent: nanotron 05:22 Configurar OpenRouter: enrutador de modelos IA 07:22 Elegir modelo nemotron 120B a12b de NVIDIA gratis 08:36 Cómo exportar skills de Claude Code a Hermes 08:42 Hermes estado status 09:24 Ejecutar tareas en Hermes

And I also gonna reveal some industry dark secrets which actually prevents more open source models going live. It s not that complicated, but it also has quite a lot of details. It also has 3D generation, but it was quite often used together with Huion 2.1 to achieve, at that point, the best quality 3D models. And Lattice, by the way, is behind Huion 2.5 and Huion 3, which is really, really good quality, even better than Trellis. But we can see quite a lot of artifacts there and there.

Today, we how to verify Docker container health on a VPS step-by-step clearly. Use Docker logs container ID to see the container's log output. Another option is to run Docker execute container ID bin bash to access the container shell. Rebuild the container with the updated settings using Docker compose up build or Docker build. After making changes, restart the container using the command Docker restart container ID.

They operated under the assumption that we could just load up the context window of something like Claude Code or Codex with, you know, a billion lines of documentation explaining here s what you should do, here s what you shouldn t, and hoping that because we shoved all that context in there, we could get JSON that actually worked. If something s wildly complicated, well, I m just going to use normal code, quote unquote , inside of Claude Code and go down that route. you re going to have to exit Claude Code, start it back up, and then just do forward slash MCP, and you will see n8n MCP now connected. And now to use it, all you have to do is just talk in natural language inside of Claude Code because once you set up that MCP, it now has a connection to n8n itself. Just writing straight-up code inside of Claude Code or Codex is going to be better, but there are situations where n8n is great.

Now, the way that this is going to work is what you re going to be able to do is you re going to be able to connect it to your different Claude Code and Codex subscriptions. Now, the one really cool thing with NimbleList is what you re going to be able to do is as you interact with the agents, it s going to move the particular task that you re working on, the sessions that you re working within. And one thing that I do want to emphasize with this is as you begin to move tasks through their respective swim lanes on the Kanban boards, is as you have all of these different agents spawn off different tasks, this is going to allow you to have a view, which is essentially a way on how you can orchestrate all of your different agents. One of the big differences with the shift that we re going through right now is what you ll notice is while you still can access the code, like you can go in and access all of the respective code that you want to edit, all of these nice pieces where you have the abstractions that are above that. So, within here, the way that you re going to be able to provision different agent providers is you re going to be able to authenticate, like I mentioned, within Claude Code or Codex.

So and what I say that is if I as a security analyst see a CVSS 10 and that s because it s being exploited in the wild on publicly facing um web servers but uh and by the way vulnerability scan tools often will just go oh it s a 10 right not taking into account where it is and what it s used for and so forth then my point is if it s a 10 but let s say it s in a internal network you know basically in a sandbox then maybe the risk to the organization is lower so having that context and using AI to um um to feed it that context could help by saying well okay this is a 10 but based on your environment um it doesn t need to have top priority like it could be a little bit further down. But what I m thinking about is there s a lot of small operators that probably don t have those capabilities and certainly in those environments I m probably seeing a lot more of the offensive stuff kind of taking over because you know, they re not going oh, I m going to go put in this to try and you know, protect because you know, clears throat they re they re much more organizations. Um so, here, you know, there s a lot more um trust when it s, you know, face-to-face video, um or, you know, some sort of video call, um but, you know, this is where it s really important to think about some of those verification steps. We also have you know honeypots and and DOS attacks so basically you know traditional honeypots you know AI can can assist in obfuscating them again from detection we ve also got I guess they could be made well honeypots can also be made more realistic but then attackers may use AI to detect decoys so yeah this is kind of that back and forth sort of thing. But you know, those types of roles where you need you know, human oversight, context, you know, the kind of yeah, those are the types of roles I think that are you know, still going to be needed for quite a long time.

But again we were doing something very different to what we had been used to when we were building AI applications in the past where with these agents we were really leaving the decision making to the large language model itself which brings us to one of the other products we released during this time and that is the llama agent workflows. So we started to build these Llama agent workflows around not just the OSS Llama index framework that you saw before, but also the Llama Pars document processing tooling that allows us to parse which makes really the core of the platform but also extract structured information, classify, split documents per class or category of the specific path in the document and so much more. What would be the advantage of picking llama index over say small agents which me was talking about in the last um presentation or langraph like I know they all have different approaches so it s not don t want to pit them against each other but if you re you re kind of coming into agents why would I pick one framework over another um I think about like a few years ago um a lot of it was around um the specific developer experience that certages or teams may even provide you how it was maintained, how a package was maintained, what kind of tooling they had available, etc. I I had the experience of building a React agent uh like a slightly complex one um last year and and that providing context with tools is so difficult like it s it s a very different way of thinking about software going from deterministic which obviously you can do with lum index and other agentic frameworks to snorts here s some tools and I m gonna let you decide but you didn t decide it right. Um but we ve built these multi- aent systems where a single agent has access to a specific set of tools and is responsible for a specific set of tasks and then we have this orchestrator agent um that sits sort of in the middle delegating tasks etc which we actually do have this type of agent in llama index but I didn t show it in my talk today.

Software engineers can give agentic  AI systems a set of capabilities,   which they will then use to autonomously achieve   their objective such as fulfilling an  order in an online management system. The agentic AI system could look  up the best price for the jeans,   assess the pair against competing  prices, back order out-of-stock items,   and provide alternative prices for several  delivery schedules. For example, orchestration would require the  agentic AI system to validate the customer s   user account before checking prices  to fulfill the jean order request. When attempting to fulfill an order  request using choreography, for instance:  Th agentic AI system could identify that the  pair of jeans the customer wants is out of stock. The system could raise a back order and Request the notification system send an   email to the customer to alert him or  her when the jeans are back in stock.

Im KI Governance Bereich geht es darum, dass man eigentlich seine eigenen Assistenten, seine eigenen Skills oder whatever you call it eigentlich nur noch orchestriert. Das heißt, der Mensch muss mit seinem Wissen, mit seinem Verstand eigentlich nur noch die KI befähigen, auch die KI also sich selber quasi einzuhalten in diesen Regeln, die sie selber vielleicht mit uns zusammen aufgestellt hat. Das eine ist das Orchestrieren, das andere ist aber auch weiterhin das überprüfen, weil es dürfen wir nie vergessen, KI eben nicht deterministisch ist, ne? Also KI, ich sag das immer so, KI ist genauso schlecht da drin, sich an Regeln zu halten wie Menschen.

Der Preis von dem guten Abendessen, wie wir es nennen, klingt charmant, aber 62 meiner Zielgruppe vergleichen den Preis nicht mit dem Restaurant, sondern mit Mail Gim Free, mit Brevo für 25 oder mit dem Status Quo. Wir erstellen also einen New Secret Key, nennen den einfach Mirofish Hostinger, weil dann wissen wir, das ist Mirofish auf unserem Hostinger Server. Erstellen uns hier ein neues Projekt, erstellen einen neuen API Key, kopieren uns den raus und hauen auch den einfach hier rein. Und dann drücken wir auf Deploy und Hostinger baut uns jetzt hier einfach alles auf und das dauert ungefähr eine Minute und dann hat man hier im Docker Manager den Open Button und die Miro Fish Oberfläche wird gleich laufen. Und kurzer Hinweis, wenn du schon ein Hostinger VPS hast, musst du nicht über den Katalog gehen, einfach im Age Panel den Docker Manager aufmachen, Mirofisch suchen, deployen und das ist der gleiche Prozess.

This was this was the other which was also so this was not able to generate the report because it hit the rate limit of number of tokens and now I will show the report successful report generation. So, it s a AI platform and simulate and it it predicts the future based on the context and So, in this uh like something like 23 25 agents So, uh many uh agents works and each has uh some uh its own persona. Like some uh works, some will talk like investor, some will talk like some enterprise buyers and uh like one one entity main entity in this is Palantir, which also work data data work, which also do data works. So, in this how to build a narrative like investors posted that Nexxar created a narrative for uh innovation, right? So what I do is uh I will take some All right, so it will be a very big document, all right.

Ähm mit so einem wie so eine Notizblock, die man sich als Mensch halt vorstellen kann, den man sich aufschreibt mit Sachen, die man macht und diesen Pfeil wird er halt iterativ verbessern, je öfter der an den Sachen arbeitet, wenn ihr dem das richtig halt auch sagt nach jeder Aufgabe. Also er baut sich seine eigene Datenbank sozusagen auf, wie er halt durch die verschiedenen Skills und und so halt rumswitchen kann, damit er dann auch jeden Skill so anwendet, wenn er den braucht und den dann halt auch jeweils auch immer verbessern kann. Ich habe es jetzt wirklich ganz grob nur gesagt, man kann ihm da halt natürlich genau ganz genau seine Nische sagen und was man da halt auch letztendlich eingibt, ist einem selber lassen. Äh jetzt baut er sich dann auch einen neuen Skill auf, mit den er halt dann diese PDFs halt besser raussucht, weil E-Mails aus Webseiten Fischen ist ja eine andere Fähigkeit als PDFs aus Earningsreports zu fischen und er hat jetzt hier die Internehmen Nvidia Alphabet, Apple, Microsoft und Amazon rausgesucht. Dann könnte ich sagen, jo, such mal hier die verschiedenen Daten raus und rechne mir Price to earnings Ratio oder was auch immer raus und ihr seht, wie wie gut man das automatisieren kann und wie er sich halt wieder iterativ verbessert und jetzt am Anfang nur PDFs gesucht hat, aber rausgefunden hat, dass Microsoft Worddateien halt nur publish und dann hat das halt auch angepasst, dass er auch Worddateien findet und so wird es halt immer besser.

Today we have a way to enable Nginx reverse proxy with Docker on VPS full guide step-by-step clearly. Once Docker is set up, create a new directory for your Nginx configuration files. Use the command deer Nginx reverse proxy to create the directory. Point to Nginx.conf on the host and etc/nginx/conf.d/default.conf in the container. To start the Nginx reverse proxy, run docker-compose up -d.

It takes your idea, just a sentence or two, and turns it into a fully animated video with a script, storyboards, character design, animation, voiceover, lip syncing, all of it from a single chat conversation. Now, here s where it gets serious for anyone who s tried to use AI for animation before, the number one problem with every other tool has been character consistency. You upload a reference image to lock in your character s face and design and then that character stays visually consistent across every single scene you generate, no matter what style or environment you put them in, this is massive because it means you can build a recurring character for a YouTube series, maintain a brand mascot across dozens of videos, or create a consistent cast of characters for an entire animated show, and they ll look the same every single time. You can now produce recurring character-based animated content without mastering professional animation tools and without outsourcing professional animators and studios. In one platform, from one chat conversation, you get script writing, storyboarding, image generation, consistent character design, animation with cinematic camera movements, professional voiceover, and lip-synced final video output, all available from day one when you sign up.

What I'm going to do just so it knows which one it's doing, I'm also going to screenshot it and I'm just going to put it in there just to be safe. You want like a researcher agent going, you want a marketing agent going, and you want a developer agent going. If you're like very confident in what you're building, you can just go on to the next thing and say, "Hey, let's build this." But one thing I really like to do, even if I am confident in what I'm building, is reverse prompt. I'm going hit enter on that and it is going to start going and building out the live data in here. Going to take that command for convex, paste it in here, and then we are going to take our API key and we're going to put it in right there at the end.

Now, when you re making dashboard, obviously it makes sense to upload or copy and paste actual data here, so you get a dashboard that makes sense, but you could also do that when you re developing the app inside of Claude code here, cuz right now we only care about the design portion of this right here. The comments allows you to again change a portion of it and then you have the present option so you could just present this here in full screen just straight out of Cloud Design if you want, but in most cases you probably want to share this so you could actually share it with teammates if you have a teams plan and then they could also help with this project or you could export it as a PDF or a PowerPoint. You re probably only going to get one design system if you re on a lower plan here, and you could add fonts, logos, and assets by dragging and dropping things over here, too. But once this is copied, I just need to open up the Claude app on my desktop and I ve already created my folder here, so I just have to paste what I got from Claude design and I could go ahead and send this out. But, let me show you this part right here as far as usage, so you could understand where to find it and how to see how much you re using here inside of the Claude website.

And so in this video, we're going to go through what Codeex from OpenAI is, how to build anything with Codeex, and also how you can build an application in just one shot using GBT 5.5 inside of the Codeex app. So, if you don't know about the new codeex app from OpenAI or much about GPT 5.5, I want to show you exactly how you can start using it and how you can use it to build apps in one shot inside of the Codeex desktop app. what it's built and it's done this in 38 minutes and 58 seconds is the Tori app shell the the file system settings and ACP command layer folder tree claude code sidecar bridge with streaming AI UI soft editorial design from the designd folder launch scaffolding including the readme contributing md and the license road map is updated at 56 out of 56 complete steps and it's verified all of these to actually run the application The production app bundle was created at scribe.app here. We don't have the ability to add new files here which is something that is and I do want to get rid of this markdown heading but you can see how quick and easy it is with codeex to get an app up and running from a PRD from a product road map and then start to iterate and refine and fix anything that's broken until you get to a point where you have an application that you're happy to release. All we have to do is line up all of those changes in these different chats with the right context in codecs and you're going to get to a point very very quickly where you have an application that matches the thing that you want to build and where you can even get more complex functionality to work like this claude code subprocess here that we have running in the background to edit and read and write our document that we're creating.

Das ist eher die Tatsache, dass Codex eine Anwendung öffnet, die er in diesem Kontext noch nie gesehen hat, die Oberfläche liest, anfängt damit zu arbeiten, ohne dass ich ihm erklärt habe, wie Schach aufgebaut ist, wo die Felder sind, wie man Züge eingibt. So wie wir das auch von Cloud kennen, kannst du mit Codex selber Fähigkeiten erstellen und wenn du eine Aufgabe erledigt hast, womit du sehr zufrieden bist, kannst du Codex einfach anweisen. Das Besondere daran ist, dass Codex und GPT 5.5 so stark ist, dass ich mit wenigen Information etwas bekommen habe, womit ich eigentlich nicht gerechnet habe. Ich kann mir etwas ausdenken, dass ein vernünftigen Prompts umwandeln und dann baut ihr Codex mit GPT 5.5 eine sehr sehr geile Lösung. Das geile bei Codex ist, dass es ein Problem erkennt, was du nicht vorher gesehen hast und es geht einfach einen anderen Weg und liefert trotzdem ein Ergebnis und das ist der Unterschied zu einem Tool, das einfach nur ausführt, was man ihm ha genau sagt.

If you want to stay ahead in AI, this is the video to watch because the people still using chat GPT [music] in 2026 the way they used it in 2025, they are about to get left behind. They took the version developers have been using all this time and built it into an app anyone can open powered by the latest model GPT [music] 5.5 underneath. Now the part you saw at the start not going to ask codeex to design my screens directly instead going to use a tool called paper. Don't worry about the term just know that codeex can now drive paper like a car. So, I type, "Hey, only build the first four screens for now, but I don't want to interrupt the agent and break the flow." So, I click this little drop-own option called steer, which means talk to me right now while you're working.

And here, in the side image option, the option of video option is, we don t have video on video. So, let s make a free free video for 10th credit. next wara aet U alakcoins studio ae ai jápelevalanquen studio 에isko fingerprintend cuill kis heop fae till ever Very good I think that costume headed in Kobe 嗯 today, notebook automatically We may not know about it But there are all any оно over SNS whoever has this video will be blessings When we become Simple sizeкой on his manifesto He always bring us energy to feel freedom Keep Galaxy is not so great And at the very least ALL our mission is to block the laptop Veer.com VHWRB here. And here we have to generate a video of the video. Sai Naap up karo gata bap up 70 credits me lenge I need total 70 plus 20 90 credits up kustartik me milja and daily ke bis ke red sub kumil te red sub kumil te red and kaur kao s bata invite friends ke you koi up kodes me lenge agarab you kodes up ne dos ne on go do go go go go go go go go go wo bis sain up karo lenge pop kup ke chas kredit saur s adh me milja and kaur kao s vath vai is kis kinder cdance 2 bi vali vali vali vali AND the ekmet is to kong pomp s?

I m going to show you the easiest way to make money with Claude right now, and you can actually do this quicker using a thing called Claude design that was just released recently. So, this video is not going to be very long because I m going to show you a really simple way, well, the easiest way to make money with Claude right now. We re going to do a few of these because this is important on how we actually make money with this, and we re going to now going to see what it can do. Now, what you want to do is um this I m going to let this do its thing, but what you want to do is think of different types of icons in your head. What you want to do is once you ve done this with Claude, so we re making money with Claude here, you want to go to this website here called creativemarket.com.

From here, just to make a little bit more space, I can literally just hide the file tree, maybe make the sidebar a little bit smaller, maybe make the browser a little bit smaller if I if I really want to because you can do all types of things here. And then you can since today because GPT 5.5 was just released and a new version of Codex, you can actually say something like at browser use, "Test the form in this website." And now Codex will control the browser. So yes, if you're using a lot of high models or extra high models, and you're using the fast variant, which is this little flash right here, which is about one and a half times faster, but costs twice as many like quota basically, then you will run out if you're using it a lot like me every day. To tell you a real story about this, when the conflict, because I'm in Dubai, when the conflict broke out like a month and a half ago, I used ChatGPT pro a lot to find out how can I get out of the country, how can I like be safe because it was the first day, I didn't know what was going to happen, right? And just like that, you can see that my system is actually fully prepared because obviously, I've been coding for a while, but in your in your case most likely it will just have a couple of axes here and it will offer to replace them or install them rather.

In diesem Video nehmen wir uns die Zeit und schauen uns ganz genau an, welche Tools euch den größten Mehrwert bieten, ohne dass ihr auch nur ein Cent dafür bezahlen müsst. Trotzdem das Gesamtpaket ist für den Start auf jeden Fall unschlagbar und wenn man dann merkt, dass man das wirklich auch intensiv nutzen möchte und braucht, kann man sich immer noch einen Pro Plan dementsprechend holen. Im kostenlosen Plan habt ihr Zugriff auf alle Kernfunktionen, die auch zahlende Kunden haben, von den besten KI Modellen über Stemd Downloads bis hin zur Videogenerierung. Das Besondere im kostenlosen Plan habt ihr nicht nur Zugriff auf Googles beste Modelle, sondern auch auf Konkurrenzmodelle von Anthropic und teilweise sogar Open AI. hat auch einen relativ einfachen Grund, denn im Gegensatz zu beispielsweise Open AI oder Thropic hat Google einfach durch ein langjähriges erfolgreiches Geschäft finanzielle Mittel und auch was die Rechenleistung angeht massive Ressourcen und davon können wir aktuell stark profitieren auch ohne selbst etwas zu bezahlen.

But in this case, you can see it went ahead and done four different tool calls and then it s going to go ahead and based on the instructions, give me some form of output. But, however, if we were to go ahead and replicate the same workflow on something like N8N, we can go ahead and see that it s going to output consistently every single time. So, in this case, we have our Bluehost VPS running our N A N instance, but I can go ahead and as long as I give something called an API key, so in this case I go to the settings N A N API, and you can see that I ve got a key here, but in this case I can go ahead and create another API key, and I can go ahead and say, for example, let s just say YouTube demo, right? And also, the benefit of having it hosted on a VPS is that once you have gone ahead and deployed it, the URL that we re actually on right now, which we got to by clicking manage on the Bluehost dashboard, you can simply go ahead and access that now anywhere in the world and simply log in with your NAN account. So, something like Open Claw, it has that intelligence layer where we can basically give it a fuzzy piece of text and it can go ahead and determine and reason and figure out what our goal is and then can use the tools required to go ahead and create.

I ll tell you right now that one of them does all the functions in case you don t want to install them separately, but if you want to have the most powerful one for each task, I recommend saving all of them. If we want to use these voices, or the ones we create afterward, to create long audios, we re going to download them and upload them to clone voice. The ones we re mainly interested in are these down here, where we can use these sliders to choose how the voice we want to create will sound. In this option here, we re going to paste the text we want to turn into voice, and if we click generate, we ll see that it follows exactly what we asked for. Now, like I promised at the beginning of the video, I m going to show you how to get the free PDFs with all the information from today s video.

Und wenn wir uns einfach den Stand der Technik angucken bei agentischen Systemen, bei Coding Agenten, aber auch allgemeinen Agenten, das Thema OpenCla beispielsweise hatte ich ja auch schon angesprochen, also Agenten, die jetzt eigenständig auf vielleicht sogar eigener Hardware laufen können und schalten und walten können, dann stellen wir doch relativ schnell fest, wie viele Aufgaben mittlerweile schon von diesem System übernommen werden können. Und ohne jetzt darüber ein finales Urteil abgeben zu wollen, ob wir an dem Punkt sind, dass wir sagen, wir haben AGI oder nicht, glaube ich, dass wir uns eingestehen müssen, dass diese Systeme einen sehr großen Impact haben werden auf unsere Arbeitswelt, auch auf die Gesellschaft und dass wir natürlich idealerweise lernen sollten, wie wir diese Systeme sinnvoll einsetzen können und wie wir mit diesem System bestmöglich umgehen können. Dazu gehört aus meiner Sicht aber auch etwas ein Phänomen, das jetzt zunehmend Leute beschreiben, nämlich das Phänomen, dass wir, obwohl die KI sehr viele Aufgaben übernehmen kann, anstatt dass wir weniger arbeiten, haben viele Leute das Gefühl, dass wir jetzt mehr arbeiten, dass Arbeit intensiver ist. Ich hatte ja in einem vorherigen Video auch schon die Harvard Business Review Studie dazu einmal gezeigt und das ist auch etwas, was ich persönlich tatsächlich berichten kann, nämlich dadurch, dass ich jetzt viele Dinge parallel bearbeiten kann, viele Projekte parallel bearbeiten kann, habe ich auch tatsächlich das Gefühl, dass ich letztendlich mehr arbeite als zuvor. Und das ganze ist jetzt, sag ich mal, mit relativ wenig Aufwand entstanden, wenn wir da natürlich jetzt noch mehr Anleitung hinterlegen, wenn wir der KI noch mal beibringen, wie diese Berichte strukturiert werden sollen, das alles ist technisch tatsächlich möglich und ich persönlich finde tatsächlich, müssen wir hier ehrlich sein, das sind Aufgaben, die uns wahrscheinlich sehr sehr viel Zeit ersparen werden und auch Aufgaben, die zumindest bis vor kurzem dann noch sehr, sehr viel Zeit auch gekostet haben in Unternehmen.

What's also great is that you can even push commits directly from the codeex app where you can write in where you can even commit and then create a PR as well. But in the same manner, you can essentially work with it to set up automations for code quality where it can scan recent commits and it's able to analyze any bugs, propose minimal fixes as well or you can even have it automatically fix various sorts of critical issues. Now within the main dashboard of Codeex, something important I want to mention is that you can change permissions where you can have it so that it has the default permissions where Codex automatically runs commands in the sandbox only or you can also have it on auto review where it runs commands in a sandbox and uses auto review for elevated requests where it will use your own personal guidance to essentially execute that task. This is quite important cuz you can create an imple implementation plan first and then have GPT basically create that detailed specdriven workflow so that you can then use one of these different models to essentially code out whatever you had planned. But let's say if you want to use a plugin like browser use, you can essentially use the act command, then select the browser use plugin or whatever plugin you want to use and then you can type in your prompt like open up YouTube and find the world of AI YouTube channel.

Down below, I could copy the message,  or over here, one of my favorite features,   you could also open the email directly within your  email client. Right over  here, I could click on the plus icon, and here,   I could upload a file or a photo. Here I have my PDF   with all the research and over here we also have  that artifact that we just created a moment ago,   all here in one place. Back in the main chat view,  let s click on the plus icon and right here,   you see the option to add connectors. Right here, I ll click  on the plus icon and here you can manage your   skills or you could add a new skill.

Google न य ज स न य ज क फ च कर ग ठ क ह ? त 1 स ल क इतन म पड रह ह त आप अपन प क ट क अक र ड ग च ज़ कर सकत ह उसक ब द स पल क ट न य प क ल क करन ह त ह अब आत ह स क ड र ल अगर आपक व क ड च ह ए और आपक ड स क उ ट च ह ए त आपक यह पर Google क बटन पर जब आप क ल क कर ग त ऐस ईम ल आईड स ल ग न आपक करन ह ज आपन उसस पहल कभ ह स ट गर प ल ग न कर न ह मतलब क एक नई ईम ल आईड बन ल न आपक Google प क ल क करन ह और अपन नई ईम ल आईड क स ल क ट कर ल न ह ठ क ह ? मतलब च ट ज प ट स ल खव त ह त हम उनक एप आई च ह ए ह त ह ज क फ र नह ह ल क न इतन मह ग भ नह ह क आप अफ र ड न कर सक ओपन एi क एप आई आपक 5 य 10 क क र ड ट ल न प म ल ज त ह और ब स कल इसम आप बह त स र आर ट कल स ल खव सकत ह अगर आप इम ज न बनव ओ, स र फ आर ट कल ल खव ओ त आप 10,000 स ज य द आर ट कल ल खव सकत ह 500 स 800 वर ड स क मतलब 5 10 क अ दर-अ दर त आइए म आपक बत त ह ओपन एi प अक उ ट क स बन त ह ल क न एक ब त क ध य न रखन यह प ज क र ड ट स म लत ह अगर स र फ आर ट कल ल खव ओग त त बह त स र ल खव सकत ह ल क न इम जज़ स थ म जनर ट ह ग त 10 म आप एप र क स म टल 20020 आर ट कल ल खव सकत ह ज क म र द न इनफ ह त ह भ ई त चल ए आपक बत त ह ओपन एi प क स तर क स अक उ ट स टअप क य ज त ह त द स त ओपन एi प एप आई अक उ ट बन न क ल ए आपक प ल टफ र म ai.com न म स इस व बस इट क ओपन कर ल न ह अब यह प द ऑप शन आए ग आप च ह त ईम ल एड र स भ ड ल सकत ह य फ र क ट न य व द Google प क ल क कर सकत ह त यह प हम क ट न य व द Google क बटन प क ल क कर ल त ह अपन ईम ल आईड क स ल क ट करत ह ईम ल आईड क स ल क ट करन क ब द आपक यह प व ट करन ह और क ट न य क बटन प क ल क करन ह इसक ब द स पल थ ड स प र स स ग ह ग और आपक स मन ओपन एi क प ल टफ र म ख ल ज एग और यह प आपक अक उ ट बन ज एग यह प आपक प स ऑप शन आएग क ऐड क र ड ट क ठ क ह ? त यह म न ट इम ग स ट क य ह आ ह आप इस आर स म नट स, स क ड स ज मर ज कर सकत ह व आपक च इस ह ल क न म प र फर कर ग आरस ह रहन द इसक ब द यह प आरएसएस फ ड ह म न यह प Google न य ज ल य ह आ ह आप क ई भ ल सकत ह ठ क ह ? ब ल क ल व स ह म हर ब र ब लत ह एक क ल स म 40 स ट ड ट ह 40 म स क छ ट प करत ह , क छ प स ह त ह , क छ फ ल ह त ह जबक स म ट चर स म ब क ह त ह , स म एनव यरम ट ह त ह य स र फ हम र म इ डस ट ह त ह इसक ब द स क ड च ज यह प आत ह जब भ आप य एआई ऑट म शन रन कर ग त इसक स ट र ट करन स पहल अपन ब ल ग प 20 आर ट कल ख द स ल ख ल और उसक ब द एडस स अप र वल ल न क क श श पहल कर उसक ब द ऑट म शन क रन कर ब त आत ह ट र फ क क 2026 क अ दर ट र फ क ल ग क बह त कम म ल रह ह Google क तरफ स क य क एआई आ च क ह त ट र फ क क ल ए आपक य ज़ करन ह स शल म ड य ज स Instagram, Facebook यह प अपन फ न फ ल इ ग बन न ह क य क ब ल ग प क म त आपक करन ह नह पड रह व त एआई ख द कर द ग एक ट इम प आक अप र वल ल न क ब द ल क न आपक अपन द सर स शल म ड य प ल टफ र म स पर अपन फ न फ ल इ ग बन न ह वह स ट र फ क यह पर ड यवर ट करन ह और जब आप यह कर ल त ह , न क स ट स ट प आत ह Google न य ज और Google ड स कवर य ऑट म ट कल ऑन ह ज त ह अगर आप स शल म ड य स ट र फ क ल रह ह जस ट आपक करन क य ह ?

Open AI just published a fascinating article about the new open source agent orchestrator to help overcome the biggest bottlenecks they encountered when trying to scale autonomous coding agents. Open AI encourage you to point your favorite coding agent at the spec and then have it implement its own version in whatever language you want, and also get it to orchestrate against whatever coding agents you want, not just Codex, so you don t have to have a Codex subscription to run this. So, instead of using a meta-prompting framework, where we might ask the AI agent to reset the context, the outer harness can actually deterministically terminate the session, clear the context, read the task state from disk, inject the relevant files, and then work from there. They re broad, open-ended, and agentic throughout, but they can still have plenty of deterministic scaffolding around them, such as computational checking of citations or multiple layers of LLM reviews before they re passed back for human review. If you want to get started with the Symphony open-source spec, it s available on the GitHub repo links below and you can get your own coding agent to create an orchestration system based on this spec in whatever language you want or you can also use their out-of-the-box demo example.

And if you know someone who is already using Claude code, Open Claw, Hermes agent, or any of the other tools we re going to cover today, share this video with them because this one is directly relevant to how they work. When you have Open Claw, Hermes agent, Codex, Quinn code, Goose AI, or any of the other supported tools on your machine, Ion UI finds them and surfaces them inside the same workspace. Each one addresses something that developers running multi-agent workflows genuinely run into, and understanding how they work individually is what makes the full picture of Ion UI click into place. If you re already running Claude code, Open Claw, Hermes Agent, or any combination of the other supported command line agents, and you re managing those sessions across separate terminal windows with separate configurations and no shared history between them, Ion UI is a direct solution to that specific problem. The common thread across all of these situations is that Ion UI lowers the operational cost of working with AI agents seriously, not the cost of the models themselves, and not the cost of the individual tools, but the cost of managing them, configuring them, coordinating between them, and building workflows that rely on more than one of them at a time.

Besonders bemerkenswert ist, dass Figure seine Produktionsrate von einem Figure 3 pro Tag auf einen pro Stunde gesteigert hat eine 24-fache Durchsatzsteigerung in weniger als 120 Tagen. Während ein Roboter pro Stunde physische Präsenz schafft, erfordert die Verwaltung einer wachsenden Flotte von Hunderten einen Wandel von der individuellen Wartung hin zur autonomen Steuerung. Um dies zu gewährleisten, hat Figure ein fortschrittliches Alarmsystem zur schnellen Fehleranalyse implementiert und Software-Ausweichmechanismen entwickelt, die es einem Roboter ermöglichen, seine Leistung sanft zu reduzieren oder sich von nicht kritischen Fehlern zu erholen alles ohne menschliches Eingreifen. Diese Infrastruktur erfasst in Echtzeit den Zustand, den Standort und den Betriebsstatus jeder Einheit. Doch diese Integration von KI in die physische Welt beschränkt sich nicht nur auf Roboter, sondern dringt nun auch in die Software ein, die zu deren Entwicklung verwendet wird.

And then when I was bored and hacking on a bug and I was like you know the thing that I do now is I go check Twitter when I m annoyed and I need like a my brain needs a break but what I would do then was I would I would I had like the jet brains IDEs which have like really good like code navigation and I would just go read the code of the libraries we were using and so I go read all the code of the internals of the Google collections libraries. Yeah I was talking to him last night and he was basically saying he one of the principles that he really likes is he has I forget the names he gave them but it was basically like you have semantic functions and then you have I forget the other word but there s basically two categories of like you have code that s allowed to be ugly and doesn t necessarily need tests and it s like it has an outcome or a goal it s like check out get repo in sandbox and then those sit on a layer of like library functions that require like comprehensive testing and basically you are only allowed to import from library functions the like app functions are not allowed to import from each other and so that way you have like two different levels of code standards as far as like okay this thing is designed to be used so it needs to be well tested the contract needs to be good we need to like make sure the code is really high quality and they actually relax the quality bar on the app stuff cuz it s like much more pragmatic. I mean so our founding engineer started about like a month or two ago so he s like there was a for a while it was like you re going to do like small stuff that s kind of isolated and off the hot path and like go figure out some experimental features and things like this but now he s writing that he s like we have a feature we want to ship by the end of like probably like ship to ourselves by the end of this week and then ship to customers within a week or two after we ve tested a little bit and so he s he s owning that and I m like yeah Kyle can t own the giant features anymore because he s he s also on call. So I don t know if that answers your question but our process is basically like if a feature is big enough you take your design discussion and you review it with the team or you send it to one other person you have a yap about it maybe we ll do some whiteboarding if it s small it s it s kind of on you it s like if you are worried that your your pull request might get a bunch of feedback that you re like  __  now I got to go redo this and like I worked to get the test passing and I got it looking how I now I got to go redo it the design then it s kind of everyone chooses if when and how they share their design doc or their like outline of how they ve broken it up into tasks. But there s another thing is like now that I am posting a lot more online and getting a lot more input like there s a lot of people who come in super super negative and like they just like to dunk on people and things like this and like sometimes they have good points and sometimes they don t but like my my new question is like I don t block people right away if they re being shitty.

Building multi-agent AI systems costs 39 per developer monthly on competing platforms. Microsoft Agent Framework is completely free. This powerhouse orchestrates complex multi-agent workflows across both Python and net with graph-based flows that include streaming, checkpointing, human in the loop debugging, and even time travel capabilities to replay agent decisions. You get a full visual development interface, built-in observability with open telemetry, and support for multiple LLM providers without paying enterprise licensing.

OpenAI obviously was had a nonprofit arm, and now, you know, they ve started they did like growth funds and all this stuff and they re trying to become a this is they are for-profit now and that s always been a problem with Elon. And XAI and Grok models, they re not the most popular, but you see a company like Cursor that has good branding, they have a stupid IDE that people use, and they also have clout and have raised quite a bit of money and I believe the founders are controllable by the big boss Elon, so it seems like a good move. So, you know, maybe maybe there s some like, you know, maybe they wrote that maybe they wrote it in the in the prompt to be like, Hey, if you see Claude just like destroy it. It does seem like OpenAI was quiet for a while, then they all of a sudden they re shipping all this stuff all at once, but they did that in the past, and then a lot of like 50 of that stuff never stuck around, and you don t know whatever happened to it, so we ll see. So, Anthropic s always done these kind of interesting research type projects, where they, you know, they gave Claude uh access to like control the store, to control the economics of a store, and they try to try to like learn how the model reacts in these different situations.

And really, it s an issue with a distraction that s going on within the public cloud companies, where they re focused on a genetic AI and innovating and changing their platform to adapt it to using a genetic-based principles, and they re not focusing on the core capabilities as much as they should. In other words, they re showing me where their genetic technology, their genetic AI technology, is going to start improving the core platform and how it s going to be built in the core platform. And the cloud providers, you know, every time you went to a conference, they would, you know, show you how the improvements they re making in their core infrastructure, the storage, their databases, you know, even even the AI stuff, the machine learning stuff, which has been there for a long time. Uh it s not necessarily going to be something that s going to harm everybody, and certainly, there s companies out there that are waiting for these genetic AI capabilities to show up in the cloud because they want to leverage them. But, for the most part, the majority of people out there are not going to be using the genetic AI services that these cloud providers are providing, and they re going to be hurt by this.

So, like I I I think it s worth mentioning that like we started Monday with an idea to bring larger teams and even massive companies, enterprise companies, their ability to have a product that is like fun, that people actually like using it, okay? That it s simple, that you can build super complex stuff, and some of our customers, we have like a a customer over 60,000 seat seats that manage like their entire company, and the whole process is there, and we have like uh companies of like two people, and I know um we have manufacturing plants, people do clinical trial research management on of like a wide variety of stuff that you can manage on Monday, and that s because it s simple, easy, and you can adapt it to whatever you want. I m running a team and we may talk about it of uh agent talent like a service we just launched and a developer joined that team and like initially like he was a bit terrified cuz we had an agent working on the platform and I kind of like send a lot of messages of fixing like a small UI tweaks. And I remember when the mobile phone came then like good photos like professional photographers or people who had a passion for it said like, We need to wreck the the camera off phones cuz it s like diluting the quality of like taking a photo. Not telling them exactly what to do where you need to like iterate and like like uh, working with Claude code, you need to do every step and not even more high-level, but like to be able to direct them.

So, you see, the real driver of business value isn t music just buying the fanciest AI. We re already seeing totally new roles pop up, like agent orchestrator or chief of staff for AI. The companies that are going to win in this new era are going to be the music ones who invest in training their people to manage, guide, and collaborate with these new digital teammates. The AI agent is on track to become a core music part of our professional lives. And so, the music real question I want to leave you with is, are you ready to lead your new digital team?

Uh just one little note about it is that Data API builder or SQL MCP server in this case, we call it SQL MCP server, but it s just Data API builder with MCP enabled. And anytime we re talking about these different capabilities, it s across kind of whichever of the SQL Microsoft SQL family things you re using, whether it s SQL Server or Azure SQL Database or even SQL Database in Fabric. So, for having a database in source control, one of the fundamental ways that you can do this for free with SQL Server, Azure SQL Database, SQL Data Fabric, is by using SQL Projects. But it becomes part of the extended properties on the database so that if I have a stored procedure that has not great practices like dynamic SQL that s right for SQL injection, when I interact with SSMS Copilot, it ll pick up that constitution that s within the database. They don t really impact licensing and things like that, as well as the internal capabilities around AI and everything so that it makes it so that choosing SQL just means tomorrow when your boss is like, It s time to do the AI thing.

Well that was our cloud expert talking about the module wise learnings that you should be following and before I talk about this program and talk about the certification I also want to add you that there are a couple of bonuses there are four beginner level courses and then on top of that there are also four certifications and then before I talk about this all these things this is the offer that we have which is cloud code mastery which is a 13 module 50 plus lessons 30 plus hands on lab one project and there ll be weekly live session. If you already know cloud you will still get one or two few things definitely helpful like we cover things like cloud service model which is we cover infrastructure service platform as a service software as a service. So you look Azure cloud portal, groups, resource groups, resources, how do you how subscriptions, Entra ID, which is deployment of users, groups, roles, policies, compute services like virtual machines, Kubernetes cluster, also app services on Microsoft Azure cloud. You also cover and learn data fundamentals or a data for beginners like type of data, data structures, as well as how data when you re deploying AI, how the data comes from various multiple sources into data stores, and then how ETL extract, transform, load works, how exploratory data analysis works using Python or from a multiple sources, data visualization, SQL, and so on. What is difference between AI, machine learning, deep learning, agentic AI, generative AI, agents, multi-agents, and then machine learning concepts like supervised learning, unsupervised learning, reinforcement learning, deep learning, neural networks, things like what is large language models, foundation models, prompt engineering, context engineering, long-term memory, short-term memory, vector databases, RAG.

Dieses Video von "The Future of Agentic AI: Inside Microsoft Agent Framework 1.0#microsoft #agentic #chatgpt #nextgen" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

When you toggle it, instead of Claude going and eagerly making changes to your code base based on the prompt you gave it, it s going to give you a detailed plan and instead of immediately rewriting half of your application, Claude Code is going to read your existing code, understand the patterns, maybe ask a clarifying question or two, and then propose a plan of attack. So, if your use case is, you know, for example, performing a PR review or doing an an accessibility audit, in that case it might make sense to make an accessibility skill that outlines, you know, all of the things that that you want the audit to do, all of the things you want the agent to check in terms of accessibility, and then create a sub agent and point him you know, point them to that skill and say uh you know, when I m doing a security review or a PR review, use the accessibility sub agent that is then going to use the accessibility skill. Um and and the reason you would want to do this is that, you know, if that accessibility skill is uh applicable to to multiple use cases and you know, you may want the skill to be pulled in when an agent is also writing code and as opposed to just like you know, checking for accessibility features, uh you may want to share that expertise across multiple sub agents and even your main agent, but you may want to have a dedicated sub agent uh that has access to that skill. Yeah, another really really good question and um you know, at the end of the talk I I kind of gave a couple of tips on like, you know, if if you re totally new to Cloud Code and not sure where to where to how to get started, like just download the tool, uh you know, get it set up on your machine, open up a code base and just run slash init to have Cloud kind kind of go through your code base, generate a cloud MD file and um Cloud Code is really one of those tools that that I feel like uh you learn it best by experiencing it and and using it yourself and seeing just how far you can go. So, I m just going to pick the second And it s just what what what what what what what what what what what what what what what what what what what what what what what what what what skills Let s talk about ones that you ve used that are not real that you ve seen for a skill.

music music music So, I spent four of the best years of my life here in Athens, Georgia, earning my MIS degree from the University of Georgia. And they loved this place so much that they named an album, arguably their best, Automatic for the People. It isn t automatic for cheaper food or more streamlined service, or even automatic for more profits. Now, there are a lot of key components to make all this work, things that we talk a lot about today, agent to AI, hybrid agents, complex orchestration, um to even managing the role of humans and machines. But after all these years, I m still inspired by these words, from a humble restaurant to a massive band, to a goal for us today, let s all strive to build CX that is automatic for the people.

Um so the chat GPT boom that was the first thing where okay this is what I m going to sort of by default if I want AI assisted coding it very quickly changed the workflow when strong uh CLIs came out terminal interfaces that I could chat with it at the terminal and not only uh do I have a chatbot but I have a chatbot that can use tools manipulate my environment and fit in with my workflow very well. that um CLI was kind of like the next big bang of how we use AI and coding in my opinion and I m making a pretty audacious statement here which is that something Ralph like is likely to be the next thing. But as we get further and further away from the chat part of the agent experience, you know, this is a highly scientific representation of what s happening to the user. We don t even have really like a user interface that supports that yet or the next wave of user interfaces. But even just doing that by manipulating these markdown files or other like scripts I m running, I have found this is sort of good enough to become my default.

Well, a federal court ruled that even if a   human explicitly hands an AI agent the digital  keys to their account to act on their behalf,   that user consent does not override a platform s  right to block the agent. Hearing  all of this, I mean, the gold drift,   the multi-agent contagion, the FBI data, you d  think companies would just abandon the technology,   you would think, but then I looked at the adoption  metrics in the dossas, and it suddenly makes   sense. If   an agent loses them a few million dollars, they  write it off, deploy a massive corporate legal   team to try and recover the funds and just move  on. I love this concept because a paper  policy is just an internal memo that says like,   hey, don t let the AI touch the payment gateway,  which fails 100 of the time the moment an agent   actually goes rogue. Just like banks   have to know your customer to verify a human s  identity to prevent moneyaundering, systems must   now dynamically verify the exact authorization  limits and safety architecture of an AI agent   before letting it interact with an API.

I would want an agent that s versatile across a bunch of different tasks cuz I m not going to be babying it through, you know, things it s not so good at. And I would want, because I m customizing, we re exploring the frontier of how we use this, I would want something open source so that I knew that at the end of the day I could uh I could suit it to my needs. Uh we know that it s high autonomy because we have not just tried it out and vibe checked it, but we have benchmarked it across a wide variety of software engineering tasks, not just the uh very popular coding agent benchmarks like Terminal Bench or Sweet Bench, but uh you know, we ve actually innovated in how these agents are benchmarked, published uh you know, papers on this. More recently, we ve got the Open Hands Index, which is an aggregate across a bunch of different kind of agentic tasks, whether it s um you know, programming across multiple different languages, uh whether it s optimizing um code, whether it s writing tests. Now, we uh do support very high control of uh the agent s behavior if you script using our software agent SDK, which is the same engine that we built the rest of the product around.

Mach jeweils ein Deep Research zu jeder Zielgruppe, simuliere aus diesen Deep Research Reports dann mögliche Fragen dieser Zielgruppen und pack das dann händisch zusammen. Mit dem geht das auf Knopfdruck und zwar nicht nur ein bisschen, sondern ihr könnt ganze Shit Storms oder politische Kampagnen oder noch größer tausendfach simulieren mit künstlichen quasi Ex Feeds und Reddit Feeds und Gesprächen. Ihr müsst das ein bisschen anschauen, muss durchklicken, das ist nicht auf Knopfdruck, aber ihr könnt das im Prinzip anlegen und dann fangt ihr an über euer Produkt oder über das, was ihr machen wollt, möglichst viel Kontext reinzugeben und zu erzählen, was ihr vorhabt. diese Simulation, da werden dann ganz viel von diesen Zielgruppen und Unterzielgruppen erstellt und zwar nicht nur fünf wie bei euch vor, sondern tausende und die ackern sich dann über ein paar Tage simulierte Tage durch und ihr bekommt am Ende einen fertigen Report zu eurem Produkt und könnt sogar mit diesen Zielgruppen mit bestimmten Leuten dort Gespräche führen. Aber denkt immer dran, ein Modell ist sehr, sehr wichtig, um sowas gut laufen zu lassen.

Wenn ich die Frage dagegen noch einmal kopiere, jetzt hier einen neuen Chat öffne, das Ganze einfüge, den erweiterten Modus ausschalte und diesmal das Heiku Modell benutze und das Ganze jetzt noch mal abschicke, dann kommt er hier schon nach fast 3 Sekunden zurück. Gehen dann zurück in Cloud und alles, was du jetzt machen musst ist das was Chetchbt dir ausgegeben hat einmal einzufügen und wenn du jetzt auf diesen Button gehst, dann wird jetzt sozusagen die ganzen Informationen von ChatBT in sein eigenes Gedächtnis übertragen. Hier in Cloud kannst du jetzt sicherstellen, dass diese Instanz mit Gmail verbunden ist, indem du einmal auf das Plus gehst, dann auf Connectoren und jetzt sicherstellst, dass hier dieser Gmail Eigen aktiviert ist. Und wenn ich jetzt hier auf mein Gmail Postfach gehe und einmal in die Entwürfel reinschaue, dann sehen wir hier genau den Entwurf, den Claud uns jetzt erstellt hat und den wir hier theoretisch genauso abschicken können. Dieses Artefakt ist jetzt also deine persönliche Trainingsplan App und hier kannst du jetzt mit Cloud drauf zugreifen und du kannst das Ganze auch mit Cloud ganz easy bearbeiten.

The next thing we need to do is get two plugins, one for Gemini and one for ChatGPT, that will connect it to Claude Code, which then basically means that we can access those within any of our Claude Code sessions. And then within anti-gravity, again, you re just going to open up Claude one more time by typing in Claude, and then just paste everything in. And I m going to say something like either I d like it to use Gemini to go ahead and analyze this video just to ask it to describe what happens in the video and just give me a bit of a breakdown, then save it for me on the left-hand side as video analysis, okay? And in addition to that, guys, I can also give him a prompt like this that just says, Hey, I want you to use three different models, ChatGPT 3.5, Claude, and Gemini. This drastically improves Claude s performance, but to get the most out of Claude, you need to know how super skills properly, which we re going to learn by watching this video right here.

You can create beautiful landing pages and directly inside Codex, you can generate images that you can place in your site on Codeex. So after 7 minutes it is now done and we can very easily just open this Excel project up directly inside codeex and you can see that it has created this file. I can type at learn and see I can actually at@mention any file that's created within this seven capabilities folder and I can say please turn this into a landing page a good learning resource for my viewers and so I can take any document within this folder and I can just atmention it and say please turn this to a landing page and because it can control my computer and go look at all of the documents that are organized very easily. And if we go to the notion plugin that we tested, we can see here that it created this document and it went through my notion, found all the scripts that I had made previously and it said, "Do you ever feel like AI coding tools are getting more powerful every week, but also somehow more confusing?" And this is literally in my voice because it went through all of my scripts and it created this document using the notion plugin. I love it." And so even once you've used a skill, maybe you've used it for a while and you like it and then you get surprised by something valuable like this diagram right here or these this formatting, you can always just say, "Hey, I want you to always format it like this." And so every time you use the skill is an opportunity to make the skill even better.

You made a comment that was something along the lines of like if your dev team isn t using cursor or similar tools in 2026, they re either lying to you or they re behind. Well, I ll answer the question from the perspective of of the engineers and you know, the first part of the question is what s the incentive for them to do so or admit that they are or not doing so one way or the other. If AI is generating the code, then there s some anxiety that comes with this about how s my performance going to be measured when now I m the orchestrator, I m not generating the code. Uh they may be saying that they re using it, but they re not really embracing it like your power users. Or they might be on the other end of the spectrum where they re using the heck out of it, but they re concerned about the quality because they re not taking the time to really evaluate the pull request.

Dieses Video von "TRAE AI Talk #3: Beyond the Spreadsheet - Data Analysis Reimagined with AI Agents" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

Wenn ihr es nicht habt, dann könnt ihr das aber auch an den KI Agenten sagen, dass ihr das Video halt so angepasst haben wollt, dass das dann beispiels soz sozusagen für jeden Frame, für jeden Frame ist ein Frame ist hier, das ist ein Frame, ein Bild, das ist alles ein Frame, das für jeden Frame ein Timestamp eingebaut wird, damit das KI Modell, dass das dann gleich diese Website erstellen wird, das dann halt auch nutzen kann. Wenn man Cloud Code hat, wie ihr hier seht, habe ich jetzt auch schon mein Limit für bis Montag schon ganz gut ausgebraucht, aber ich bin am Pro, als ich diese Website halt vorher dann erstellt habe, diese Beispielwebsite, auf die ich noch mal hier zurückkomme. Jetzt müssen wir einfach warten, den Kollegen arbeiten lassen und dann kann ich euch gleich auch noch mehr Funktionalitäten zeigen, wie man dann iterativ an verschiedene Elemente rangehen kann mit Hilfe von Cloud Design, ohne das alles selber schreiben zu müssen, sondern halt auch in der Website selber einzeichnen können, hin was hinschreiben können, damit das Modell auch wirklich guten Kontext hat, um so iterativ seine Webseite auf sich anzupassen und dann am Ende das bestmögliche Ergebnis zu haben. Ich könnte jetzt beispielsweise hier auch noch mal reingehen und dann den letzten Frame als Endfame machen und dann neuen Frame, wo ich dann wieder in Chat GPT das reingebe und ich sage ihm dann: Jo, erstell mir dann Video von innerhalb dieser Welt und dann mache ich das wieder so, dass er dann noch weiter reingeht und dann kann ich ihm das auch wieder geben und dann kannst du auch sagen, j, dann geht er wirklich durch die Welt und dann sobald er in dieser neuen Welt ist, macht das und das, was dann halt aufkommt. Also, da geht ihr dann jetzt einfach hier, seid jetzt hier in eurem Projekt, habt das einmal in Visual Studio Code geladen, hier einfach mit hier auf File und dann auf Open Folder und dann geht ihr auf den extrahierten Folder, den ihr halt gemacht habt, den ich hier halt auch sehe, denke groß.

Dieses Video von "Demo: Workflow-Automatisierung mit n8n & AI Agents" enthaelt keine Beschreibung und kein Transkript. Bitte das Video direkt auf YouTube aufrufen fuer mehr Informationen.

Und diese Skills findest du hier auch noch mal über Fähigkeit und das macht tatsächlich auch Sinn einige dieser Skills zu installieren, vor allem die empfohlenen, also den PDF Skill, ja, den Playr Skill, sowie diese Skills hier, die ich bereits installiert habe, sodass Codex auch wirklich befähigt ist, perfekte PDF-Date Docs und auch Präsentationen für dich zu erstellen. Also Conix erstellt jetzt die Website und die mobile App und währenddessen will ich jetzt noch mal einen realen US Case testen, den ich tatsächlich schon länger vorhabe und den lasse ich jetzt auch tatsächlich mal gegen den Cloud Code antreten. Hier haben wir jetzt wieder den Barcode und tatsächlich hat er jetzt die Vorschau generiert, Steve Jobs, die autorisierte Biografie des Apple Gründers und ich sage jetzt Open Shelf und hier habe ich jetzt auch meine gespeicherten Bücher. Also, du hast jetzt eine ganze Menge gelernt über GBT 5.5, vor allem über Codex, das aus meiner Sicht tatsächlich eine der interessantesten Alternativen gerade für alle ist, die sich mit Agentic Coding befassen möchten, ohne im Terminal arbeiten zu müssen und die vor allem immer nach Lösungen schauen, die gerade äußerst kosteneffizient sind, da Codex von Open AI gerade eben unfassbar subventioniert wird, wovon du natürlich direkt profitierst. doch Open AI sieht das genau anders herum, denn im Gentic Coding Loop ist die Bildgenerierung tatsächlich elementar, den Teil eines jeden Software Development Prozesses, aber auch nahzu jeder Abteilung, die in irgendeiner Form mit Präsentationen, mit Websites, mit Bildern zu tun hat, gibt es zu Haufe Designer, Product Manager, nicht Entwickler, die rein über einen Agenten ihre Arbeit verrichten wollen und letztlich nur noch mit einem Prompt beschreiben können, wie die Software aussehen soll, sich dies eben visuell generieren Mockup rändern und eben in gleichen Atemzuge beispielsweise über Codex dann auch direkt den Frontend Code dazu bekommen.