World model companies are keeping a lot of secrets

At a panel discussion on world models at the All In conference, participants highlighted how much of the emerging field remains shrouded in secrecy despite substantial funding and market interest. The discussion identified AMI Labs, led by Yann LeCun, and World Labs, associated with Fei-Fei Li, as high-profile actors in the space, but concrete information about product roadmaps and commercialization timelines was scarce. Michael Rabbat, AMI co-founder and VP of World Models, declined to detail the company’s plans onstage and described the organization as still in a research-and-building phase.

World models aim to automate spatial intelligence, a capability that could be applied to diverse areas including robotics, interactive video, advanced self-driving systems, manufacturing and biomedicine. World Labs’ Marble product was noted as one of the more developed demonstrations, showcasing uses from media creation and explorable game environments to CGI effects and early robotics concepts. Nevertheless, observers said many of these demos appear oriented toward proving capabilities rather than signaling immediate commercial products.

The lack of transparency extends beyond the labs themselves to their suppliers. Alex de Vigan, CEO of Physicl, told the author that his company provides data used by world-model developers but does not always know how that data is being applied, limiting its ability to tailor datasets to customers’ needs. Industry participants said this opacity is partly strategic: revealing a clear path to market could invite rapid competition from other labs and larger players such as OpenAI or Anthropic.

The current funding environment makes secrecy easier to sustain. With ample venture and research capital available, labs face less pressure to pick a single commercial focus quickly, and they can use discretion to postpone revealing promising directions. Observers compared the resulting dynamic to a "dark forest" scenario: staying quiet reduces the chance of attracting competitive attention until a team is ready to disclose its work publicly.