World Model Startups Guard Their Secrets as AI Frontier Heats Up

The All‑In Conference Brings World Models Into Focus

At the recent All In conference, a panel on world models gave the moderator a chance to explore one of the most opaque corners of artificial intelligence. The discussion centered on two high‑profile labs: AMI Labs, led by Yann LeCun, and World Labs, founded by Fei‑Fei Li. Both have attracted significant buzz and funding, yet neither has yet demonstrated a clear path to revenue.

What World Models Actually Do

World models aim to automate spatial intelligence, a capability that could unlock lucrative opportunities across robotics, interactive video generation, and advanced self‑driving systems. The simplest implementation resembles the maps that power autonomous vehicles, but the same underlying technique can enable a humanoid robot to manipulate boxes or transform a few minutes of footage into an explorable virtual environment.

Key Players and Their Current Stance

AMI Labs, less than a year old, is still in a research and building phase, according to Michael Rabbat, co‑founder and VP of World Models. When asked about concrete product plans, Rabbat replied,

“We’ll talk about it when we’re ready to talk about it.”

In a follow‑up email, Rabbat explained that AMI is still in a research and building phase and therefore is not discussing any product plans or timeline publicly.

World Labs has released Marble, a platform that is the most polished product in the field to date. Its demonstrations include straightforward media creation, the construction of explorable environments for video games, and CGI effects. While robotics applications are showcased, the platform appears primarily designed to illustrate capability rather than to serve a commercial purpose.

Suppliers Left in the Dark

Even the data providers that feed these models are kept guessing. Alex de Vigan, CEO of Physicl, a data supplier for the world‑model sector, says his company’s datasets are useful to the labs but he lacks insight into the final products.

“I wish they would tell us more. We could build more useful data if we knew what they were working on,”

Why the Secrecy Persists

The versatility of world‑model technology makes it attractive for many industries, and AMI has already explored manufacturing, biomedicine, robotics, and even AI software for doctors through its partnership with Nabla. However, announcing a breakthrough—such as a humanoid robot or a next‑generation rendering system—would instantly draw attention from rivals, including other world‑model labs, neolabs, OpenAI, and Anthropic. In an environment where fundraising is easy, there is little incentive to lock down a single commercial direction.

Keeping quiet delays inevitable competition. The same capital that lets a startup develop under the radar also fuels potential adversaries once a market path becomes clear. This strategic silence mirrors the “dark forest” scenario popularized by Cixin Liu: if you do not know who else is lurking, it is safest not to broadcast your presence.

Looking Ahead

While the sector is rich with viable business opportunities, the prevailing culture of secrecy means that the first commercial breakthroughs may emerge only when the major players decide they are ready to reveal them. The article was originally published on September 18, 2026.