Written by Admin Alex · Fact-Checked by M.Ali · Info Verified September 2026
We review and update this article regularly as new information becomes available.
TL;DR: A cluster of well-funded startups building “world models,” AI systems trained to understand and predict physical reality rather than generate text, are staying deliberately vague about what they’re actually building. Companies like Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs have raised serious money while declining to specify product plans, timelines, or target markets, and the companies that depend on knowing what they need, like data supplier Physicl, say the silence is making their own jobs harder.
Imagine raising tens of millions of dollars and, when asked what you’re building it for, answering: we’ll talk about it when we’re ready to talk about it.
That’s roughly the posture of the companies working on world models right now, the technology many researchers consider the actual next step after large language models. Where an LLM predicts the next word in a sentence, a world model is trained to predict what happens next in physical reality: how an object will move, how a scene will change, how a robot arm interacting with a surface will behave. The potential applications span robotics, self-driving systems, video game environments, CGI production, and, per at least one reported partnership with a company called Nabia, medical software.
Two of the most closely watched entrants are AMI Labs, founded by Yann LeCun after his departure from Meta, and World Labs, founded by Fei-Fei Li. Both have attracted serious capital and serious attention. Neither is saying much about what they’re actually shipping or when.
Michael Rabbat, AMI Labs’ co-founder and VP of World Models, put it plainly: “We’re still in a research and building phase, so we’re not talking publicly about any product plans or timeline.” Pressed further, his answer didn’t move much. “We’ll talk about it when we’re ready to talk about it,” he said.
That silence has a real cost, and it isn’t just abstract. Alex de Vigan runs Physicl, a company that supplies data to world model builders, and his business depends on understanding what these labs actually need. “I wish they would tell us more,” he said. “We could build more useful data if we knew what they were working on.” A supplier that can’t see the target it’s supposed to be supplying for is stuck guessing, which is an expensive way to run a data pipeline.
Why the secrecy is a strategy, not an accident
The logic here isn’t paranoia, it’s competitive math. The moment a world model company reveals a specific commercial application, whether that’s robotics control, autonomous vehicle simulation, or game world generation, every well-capitalized lab in the space, including OpenAI and Anthropic, can pivot resources toward the same target almost immediately. Staying vague buys time. It also, not coincidentally, makes fundraising easier, since investors get to project their own read of the opportunity onto a company that hasn’t narrowed its own story yet.
The comparison that’s been floating around researcher circles is a “dark forest” scenario, borrowed from the science fiction concept where revealing your position is the most dangerous thing you can do. Every world model company right now seems to believe it’s safer unseen than celebrated.
Bottom Line
The secrecy strategy makes sense from inside a boardroom and looks a lot worse from outside one. Suppliers can’t plan, competitors can’t respond intelligently, and the rest of us are left guessing whether “the next step after LLMs” means better robots, better game engines, or something nobody’s guessed yet. LeCun and Li clearly think staying quiet is worth more than staying visible right now. Given how much money is chasing this category, they’re probably not wrong, at least for another few quarters.
