World Model Startups Keep Product Plans Under Wraps, Even From Their Data Suppliers

The publication's author, who moderated a panel on world models at the All In conference, described pressing panelists on where the technology would actually be commercialized and getting little in return. Michael Rabbat, a co-founder of AMI Labs and the company's vice president of world models, appeared on the panel and declined to describe what the company was building. "We'll talk about it when we're ready to talk about it," he said, adding later by email that the company remains in a research and building phase and is not discussing product plans or timelines publicly.

AMI Labs is less than a year old, which makes early silence reasonable, but the same reticence runs across the field. World Labs' Marble is described as probably the most fully developed product in the sector, with demonstrations spanning straightforward media creation, explorable environments for video games and CGI effects. Robotics applications exist as well, though the platform appears designed more to demonstrate capabilities than to sell a finished service.

The secrecy extends to the companies' suppliers. Alex de Vigan, chief executive of the data supplier Physicl, told the publication that he knows his company's data has been useful to clients but remains unaware of what they are building. "I wish they would tell us more," he said. "We could build more useful data if we knew what they were working on."

World models are intended to automate spatial intelligence, and the underlying approach is broad enough to support several businesses at once. The simplest version is a navigable map of the world, similar to the models that power self-driving cars. According to the account, the same modeling used to help a Waymo vehicle weave through traffic could also help a humanoid robot carry boxes, or turn a few minutes of video footage into an explorable environment. AMI has already touched manufacturing, biomedicine, robotics and AI software for doctors through a partnership called Nabia, though it is unlikely to pursue all of those at once.

No one quoted in the account doubts that viable businesses can be built on world model technology, and easy fundraising removes pressure to choose a single direction. The article argued there is also a strategic reason to delay that choice: if AMI announced, for example, a humanoid product or a next-generation Hollywood rendering system, other labs would quickly take interest, and the company could face competition from other world model developers, newer AI labs and even OpenAI and Anthropic. Because rivals can raise money just as readily, the same capital that lets a company build quietly also funds potential competitors once a path to market becomes clear. The article compared the dynamic to the "dark forest" scenario in Liu Cixin's fiction, in which the safest course for anyone in the woods is to avoid attracting attention.

TechCrunch AI said the piece was first published on September 18, 2026.

Editor's Summary

AMI Labs and World Labs, the two best-funded developers of AI world models, are declining to describe their product plans, according to TechCrunch AI, and the silence reaches as far as their data suppliers. The technology could serve robotics, interactive video and self-driving systems, but the companies appear to be delaying any public commitment to keep larger AI rivals from entering a market they are still defining.