World model companies are keeping a lot of secrets

At a recent All In conference panel on world models, participants described a field that has attracted significant attention and capital but remains opaque about concrete products and timelines. The strongest public players cited were AMI Labs, founded by Yann LeCun, and World Labs, led by Fei-Fei Li; both organizations have drawn buzz and funding while emphasizing research stages over immediate commercialization.

Panelists and follow-up interviews illustrated how circumspect leaders are about specific plans. Michael Rabbat, an AMI Labs co-founder and its VP of World Models, told the moderator the company would disclose product details only when ready and described AMI as still focused on research and building. World Labs has a more visible demonstration product—Marble—which showcases capabilities such as media creation, explorable game environments and CGI effects, with documented robotics applications but appearing primarily designed to show what the technology can do.

The secrecy extends beyond the labs themselves to their suppliers. Alex de Vigan, CEO of Physicl, a company that provides data to world-model labs, said he knows his firm’s data has been useful but does not have a clear picture of the specific projects it supports; he suggested that greater visibility would enable more targeted data products. That dynamic highlights a practical consequence of limited disclosure: partners who could tailor inputs to use cases lack the information needed to do so.

Observers say part of the reticence stems from the breadth of possible applications for world models. Techniques that produce navigable spatial representations can be applied to self-driving systems, humanoid robotics, interactive video and other domains; AMI has publicly explored workstreams spanning manufacturing, biomedicine, robotics and a partnership delivering AI software for clinicians through its Nabia collaboration. With venture funding readily available, labs face little immediate pressure to narrow focus, while revealing a compelling commercial path could quickly attract competition from other labs, large AI firms and new entrants—creating an incentive to stay quiet until a clear lead emerges.