Project Genie is live for Ultra subscribers

You can now walk inside Google’s world model

This is Google letting people step inside the model.

Google is rolling out Project Genie, a web prototype built on Genie 3, to Google AI Ultra subscribers in the U.S. starting Jan 29. It lets users generate interactive worlds from text or images, move through them in real time, and remix existing worlds, with sessions capped at 60 seconds.

What matters is not the visuals, which still wobble, but the shift toward models that generate the next step as you move, rather than a fixed scene. Watching the world extend under a moving camera, with the cursor lagging slightly, feels closer to simulation than content generation.

This points at tools for robotics, training, and game-like prototyping, not just creative demos. Limiting access to paid Ultra users signals Google is still measuring cost, control, and behavior.

If this is rough and gated now, what happens when it is cheap, fast, and everywhere?

Cloudflare wants to host your personal AI agent

People bought Mac minis just to run an AI agent.

Moltbot is an open-source, self-hosted personal assistant that runs in the background and connects to chat apps, browsers, and AI models. Cloudflare built Moltworker, a proof-of-concept that runs Moltbot on Cloudflare Workers using Sandboxes, R2 storage, Browser Rendering, and AI Gateway, with a $5 paid plan as the floor.

This is Cloudflare saying you do not need dedicated hardware to run a capable personal agent. What stands out is how much of the stack they now cover, from isolated code execution to headless browsers, all stitched together with existing platform tools rather than a new product.

If this works reliably, personal agents start to look more like hosted infrastructure than hobbyist setups humming on a desk at night. It also quietly reframes Workers as a place to run long-lived, stateful AI systems, not just request handlers.

Once agents move off personal machines, who actually controls them day to day?

Perplexity hedges its cloud bets

Perplexity just signed a very expensive insurance policy.

The AI search startup agreed to a three-year, $750 million deal with Microsoft to use Azure and its Foundry service, giving it access to models from OpenAI, Anthropic, and xAI. AWS remains Perplexity’s main cloud provider, even as the two companies are in an active legal fight over shopping features tied to Amazon’s marketplace.

This looks less like a breakup with Amazon and more like Perplexity buying leverage. After committing hundreds of millions to AWS and then getting sued, adding Microsoft gives the company options and breathing room, even if it complicates the stack.

Multi-cloud is becoming table stakes for serious AI companies, especially when model access and platform politics are involved. Microsoft benefits by pulling a high-profile AI startup deeper into Azure without demanding exclusivity.

When your cloud provider is also your courtroom opponent, how long can loyalty really last?

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.