World Model Companies Are Keeping a Lot of Secrets
World model companies have accumulated significant buzz and funding, yet the field remains remarkably tight-lipped about what it’s actually building, a pattern that becomes obvious the moment anyone starts asking specific questions.
The space’s biggest names, Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs, have both attracted considerable investor attention despite ranking relatively low on efforts to actually generate revenue so far. At their core, world models focus on automating spatial intelligence, a foundation that could theoretically lead toward robotics, interactive video, or more sophisticated self-driving systems.
Why Getting a Straight Answer From World Model Companies Is So Difficult
Pressing these companies on where the technology will actually get commercialized tends to produce vague, noncommittal responses rather than concrete plans. Michael Rabbat, co-founder of AMI Labs and the company’s VP of World Models, offered a telling response when directly asked what the company was working on: “We’ll talk about it when we’re ready to talk about it.”
In a follow-up email, Rabbat clarified further, explaining that AMI remains in a research and building phase and isn’t currently discussing product plans or timelines publicly. To be fair, AMI Labs is less than a year old, so some caution is understandable. But this same guardedness extends across the entire world-modeling space, not just one young company.
World Labs’ Marble platform stands out as probably the most fully developed product in the category, with demos spanning straightforward media creation, explorable video game environments, and CGI effects work. Robotics use cases exist too, though the overall platform seems designed primarily to demonstrate raw capability rather than reveal a specific commercial direction.
Even Suppliers Are Left Guessing
The secrecy surrounding these companies extends surprisingly far, even reaching their own suppliers. Alex de Vigan, CEO of Physicl, a data supplier serving the growing world model industry, acknowledged knowing his company’s data has proven useful for whatever these labs are building, without actually knowing what that end product is.
“I wish they would tell us more. We could build more useful data if we knew what they were working on,” de Vigan explained, highlighting an odd dynamic where even close business partners remain in the dark about the ultimate application of their own contributions.
Why World Models Are Inherently Hard to Pin Down
Part of this mystery stems from just how versatile world models genuinely are as a concept. The simplest version functions as a navigable map of the world, similar to the underlying AI models that already power self-driving cars. But that same core modeling approach could just as easily help a humanoid robot carry boxes, or transform a few minutes of raw video footage into a fully explorable virtual environment.
AMI has already tested the waters across manufacturing, biomedicine, robotics, and AI software for doctors through its Nabia partnership. The company surely won’t pursue every one of those directions simultaneously, though it remains unclear from the outside which one or two applications might actually be gaining real traction internally.
The Strategic Logic Behind Staying Quiet
No one seriously doubts there are multiple viable businesses waiting to be built on world model technology, and as long as fundraising remains relatively easy, there’s little external pressure forcing these companies to publicly commit to just one direction. In fact, there’s genuine strategic logic in avoiding that commitment for as long as possible.
If AMI announced tomorrow that it had successfully built a humanoid robotics platform or a next-generation Hollywood rendering system, competing labs would suddenly take notice and start pursuing the exact same opportunity. The company would quickly face competition not just from other world model startups, but potentially from major established players like OpenAI and Anthropic as well.
This dynamic represents something of a flip side to easy access to funding. Competitors can raise money just as easily, meaning the same capital that allows one company to build quietly under the radar is simultaneously funding potential rivals who could pounce the moment a clear path to market becomes visible. Delaying that inevitable competition, even briefly, means staying deliberately quiet about specifics.
Science fiction fans familiar with Cixin Liu’s work will recognize the underlying logic here as a version of the dark forest scenario: when you don’t know exactly who else is operating in the same space, the safest strategic move for world model companies is avoiding unnecessary attention until they absolutely have to reveal their hand.