World model companies are keeping a lot of secrets
TechCrunch Russell Brandom ● Covered by 2 sources
World model startups are still hiding what they’re building. That secrecy makes sense now, but it also means nobody knows who’ll actually cash in.
Based on reporting by TechCrunch, Russell Brandom — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
I spent a panel at the All In conference trying to get a straight answer on world models, and mostly got fog. The field has two big names drawing the attention and the money: Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs. But when it comes to what they’ll actually sell, both are still keeping their cards close to the chest.
AMI’s Michael Rabbat was the closest thing to a guide. He’s a co-founder and the company’s VP of World Models, and even he wouldn’t go beyond saying the team would talk when it was ready. In an email, he said AMI is still in a research-and-building phase and has no public product plans or timeline to share. That fits the company’s age — it’s less than a year old — but it also fits the mood around the whole sector.
World models are a broad idea, which helps explain the secrecy. At the simplest level, they’re about spatial intelligence: systems that can understand and navigate the physical world. From there, the use cases fan out fast. Robotics. Interactive video. More advanced self-driving systems. A navigable world map is one version; a robot carrying boxes is another; turning video into an explorable environment is another still.
World Labs’ Marble is the furthest along publicly, at least from the outside. Its demos cover media creation, explorable game environments, and CGI effects, with robotics also in the mix. But even that looks more like a proof of capability than a clear product announcement. And AMI is already poking at manufacturing, biomedicine, robotics, and AI software for doctors through its Nabia partnership, without saying which direction matters most.
That silence reaches beyond the labs themselves. Alex de Vigan, CEO of Physicl, which supplies data to world model companies, said he knows his company’s data has been useful, but not for what exactly. He wants more detail so Physicl can build better data. Instead, everyone is waiting for the same thing: one lab to finally show its hand, and then the rest of the field to pile in.
My take — AI-written commentary, not fact-checked reporting
This is the oldest trick in the AI playbook: raise easy money, say almost nothing, and call it research. It works right up until one company blurts out a real product and turns the whole “mysterious” market into a stampede. Then the dark forest gets loud very quickly.
Read more about this at: TechCrunch