Genie 4 landed Monday from DeepMind. Every AI video tool you've used works the same way. You write a prompt, you wait, you watch a clip. It's finished before you see it. A world model doesn't do that. You prompt an environment and then you're inside it, moving around, while it generates frame by frame in response to what you do. 🌐 Genie 4 is the first one from DeepMind that makes sound natively. Not music added afterward. The wind, the water, the room tone, all generated alongside the picture. Characters in the scene talk back and react to what you do. You can also change it while it's running. Type something mid-session and the world adjusts. Here's the structural bit I find interesting. There's a Gemini model sitting in the planning seat, breaking the scene down and deciding what happens, while Genie handles what it looks and sounds like. DeepMind calls these hybrid world models. One model directs, the other renders. That's the same shape as the video I made a few weeks back. Opus 5.5 wrote the code, a renderer turned it into frames. Reasoning model up front, generation behind it. Different tools, same architecture, and I think that's the pattern most production work ends up using. Now the part most posts about this will skip. 🛑 You can't use it. Research preview, trusted testers only. Google AI Ultra subscribers are next on a waitlist with no date attached, and the developer API is a plan rather than a product. Sessions run about two minutes. And Google didn't publish resolution or frame rate this time, which the previous version's launch post did. When a spec disappears from a launch post, that's usually not because it improved. Also worth noting that Runway shipped a playable world model with audio a couple of months ago, so "first" is carrying some weight in that sentence. So what do you actually do with this today? Nothing. Don't build a plan around a waitlist. What I'd take from it is the direction. Video production is moving from rendering a finished clip toward generating a space you make decisions inside. When that lands properly, the skill stops being prompt writing and starts being closer to directing.

Hello, I have been working in the field of software engineering for several years. I specialize in AI automation, LLM agents (such as OpenClaw), and prompt engineering, with a focus on building new AI agents and SaaS solutions. By mastering the latest trends in model training, I am able to utilize and manage multiple agents simultaneously, significantly maximizing productivity. I am particularly skilled in the use of coding agents; I possess deep expertise in prompt optimization, ensuring task continuity and accuracy, as well as implementing self-testing, self-reinforcement, and learning mechanisms for these agents. I am capable of independently fulfilling multiple roles including architect, coder, engineer, and tester all at once. Furthermore, I conduct in-depth research into the strengths and limitations of state-of-the-art models such as Codex, Claude Opus, Fable, and Gemini as well as the complex, unresolved challenges inherent in these systems. Addressing these issues is a core aspect of my work in model training and the development of AI agents. If you are considering hiring for this field, I am open to working with you starting with a part-time role and potentially transitioning to a full-time position. I have no restrictions regarding scheduling or language, and I am confident in my ability to deliver results that exceed your expectations.

Most people use Claude to answer questions. This course teaches you to build with it. Claude Code Masterclass is live in the Classroom for Premium members. It's a written, step-by-step course that takes you from zero to shipping real software, apps and automations, even if you've never written code. Here's what you actually build: ✅ Your first web page from one prompt ✅ A polished one-page site for a real business ✅ Tallyloop, a full-stack lead tracker with a database, sign-in, Slack alerts on every new lead, AI summaries of lead notes, and a live deploy on your own domain ✅ Your own skills, an MCP connection to a tool you already use, hooks that run automatically, and a code-reviewer subagent ✅ Capstone: my client demo framework, built, deployed as a live link, then turned into a reusable skill so every new prospect gets a demo in minutes It's ready to go right now. Open it and you're building in the first lesson. No live calls, no cohort, no waiting for next week's module to drop. You move at whatever pace you want. Every lesson has numbered steps, real screenshots, copy-paste commands, what you should see when it works, and a "Stuck?" section with the actual errors you'll hit and how to fix them. Every module ends with a checklist, a mini challenge and a post to share your progress. Where a module lines up with a Claude Mastery video, the start page links you to it first. 🛠️ You need a computer, a paid Claude plan, and VS Code. Module 2 walks you through the install. Open Classroom → Claude Code Masterclass → read "Start here: how this course works" → Phase 1 → Module 1. Check it Out Now! https://www.skool.com/automation-academy-7703/classroom/abac31d4

Claude Duet went live yesterday in research preview. Every AI tool you've used waits for you to open it. Duet doesn't. It watches your connected apps, sends you a briefing in the morning, and pings you during the day when something actually needs a decision from you. It runs on its own cloud computer with its own browser, so it keeps working after you close the app. Built on Opus 5.5, connects to over 6,000 apps including Gmail and Slack, and it carries over your existing memory, skills and connectors. It can also buy things through Stripe Link and Shop Pay. Here's what caught my attention though. 🛑 Look at what Anthropic says it won't do. No trading securities. No moving money between accounts. Payments need your approval. It never sees your passwords. There's a separate classifier checking every message for data leaks. Compare that to Grok Bot, which shipped with no spend cap at all and routes every customer's traffic through shared IPs. Same category, same week-ish, completely different posture. You can tell a lot about an agent product by reading the list of things it refuses to do. One detail nobody's mentioning. 💸 Duet uses your existing plan limits, and its background checks count toward them. So an agent that's monitoring your apps all day is quietly drawing from the same pool as your actual work. Watch that in week one before you let it run on everything. Anthropic's own caveats, which I appreciate them stating: it might misread a thread or get urgency wrong, and it's weaker on personal stuff like booking tickets than on work. Research preview, waitlist, US only, Pro and Max for now. The real shift here isn't that it's smarter. It's that the thing now starts the conversation. Which means the stuff worth handing it isn't the tasks you remember to delegate. It's the ones you keep forgetting.

Not what you've read. What's actually running.