At the All In conference in September 2026 I moderated a panel about world models, which highlighted a relatively opaque corner of the AI industry. Prominent players include Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs. Both organizations have attracted attention and funding, but so far they appear to place less emphasis on immediate monetization.
What is a world model?
At a basic level, world models seek to automate spatial intelligence. That approach can be applied in many directions: robotics, interactive video, more advanced self-driving systems, media creation, or building explorable environments for video games. The concept’s flexibility helps explain why companies might be cautious about revealing specific commercial plans.
Few direct answers about product plans
Michael Rabbat, co-founder of AMI Labs and the company’s VP of World Models, joined the panel. When pressed about what AMI is working on, he declined to provide specifics: “We’ll talk about it when we’re ready to talk about it.” In an email, Rabbat added that AMI is “still in a research and building phase, so we’re not talking publicly about any product plans or timeline.”
Given that AMI is less than a year old, some reticence is understandable. Still, this reluctance is common across the world-modeling field. World Labs’ Marble is likely the most developed product in the space, with demos that include media generation, explorable game environments, and CGI effects. Robotics use cases appear as well, but the platform’s demos mainly showcase capabilities rather than announce concrete commercial deployments.
Secrecy reaches suppliers too
On the conference sidelines I spoke with Alex de Vigan, CEO of Physicl, a data supplier to world-model projects. De Vigan said he knows Physicl’s data has been useful to their customers’ work, but he does not always know exactly what the customers are building. “I wish they would tell us more. We could build more useful data if we knew what they were working on,” he said.
Funding, competition and a "dark forest" dynamic
Part of the calculus is strategic. Many viable businesses can be built with world-model technology, and as long as fundraising remains easy there’s little pressure to focus narrowly. But that same funding also enables competitors: if one lab publicly reveals a promising product — say a humanoid system or an advanced film-rendering pipeline — other labs, new entrants, and large players like OpenAI or Anthropic could rush in.
That creates an incentive to delay revealing market-facing details. The situation resembles the “dark forest” analogy from Cixin Liu’s fiction: if you don’t know who else is in the woods, you avoid drawing attention.
Conclusion
Research into world models attracts capital and interest, yet companies are intentionally vague about product timelines and use cases. The technology’s versatility and the risk of triggering rapid competitive response make secrecy a rational strategy for labs that remain in research and development.
(First published: September 18, 2026.)
