Yann LeCun’s World Models Raise $1 Billion - DTNS 5222

March 10, 2026·29 min·1 clip

AMI Labs raised $1.03 billion to build Yann LeCun's world-model approach to AI.

World models get the marquee slot. The hosts start with LeCun's new company, AMI Labs, which raised $1.03 billion after his Meta exit last November. They do not declare LLMs dead. The sharper point is that LeCun thinks they may be a dead end, while world models become the AI term everyone suddenly has to pretend they knew already. Money makes that argument harder to ignore. A billion dollars and change turns a technical disagreement into a market signal. Then the conversation moves to Anthropic's code review beta, which lands right in the messiest part of AI-assisted development: how much trust should a team put in work checked by agents before code ships? The tool sounds useful. Multiple agents can review new code blocks for bugs and other problems, which fits the idea of models checking each other. But trust gets weird fast. One host asks whether the review agent needs its own review, and the joke has teeth. Someone still owns the result. The grounded answer is that a human in the loop has to take responsibility when code goes live. That is not hand-waving. The hosts admit how code review already works, including the familiar habit of approving small or trusted pull requests without reading every line. So the bar is practical. AI review does not need to be magic. It needs to become another signal engineers understand, doubt, and calibrate. The mood is cautiously optimistic. These tools may earn trust over time, especially since developers already ask agents to find bugs in their own code. The tension stays useful: automation may catch more problems, but accountability still belongs to the person who has to explain and fix the breakage.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Daily Tech News Show