World model labs keep plans under wraps

World models are one of the most closely watched corners of the AI industry, but the companies building them have little to say about what they are actually making. At the All In conference, a panel on the subject made that secrecy plain: the biggest names in the space are accumulating funding and attention, yet they rank low on any scale that measures how close they are to making money.

That gap between hype and disclosure is not accidental. It is a deliberate posture, according to the people working inside the field, and it extends even to the suppliers who feed these companies their data.

What a world model actually is

At their core, world models are about automating spatial intelligence. That single capability could point toward several different commercial futures, from robotics to interactive video to more complex self-driving systems. The field's breadth is part of its appeal and part of its opacity: a modeling approach that helps a car navigate traffic might also help a humanoid robot move boxes, or turn a few minutes of footage into an environment someone can explore.

The two companies most associated with the space are Yann LeCun's AMI Labs and Fei-Fei Li's World Labs. Both have accumulated a lot of buzz and funding, but both rank pretty low on the trying-to-make-money scale.

AMI has already dipped its toe in manufacturing, biomedicine, robotics, and even AI software for doctors through its Nabia partnership. Whether it intends to pursue all of those directions, or only one or two, is not something the company has said publicly.

A founder declines to get specific

The closest thing to an authority on the panel was Michael Rabbat, a co-founder of AMI Labs and the company's VP of World Models. When pressed on exactly what the company was working on, he was cagey.

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

— Michael Rabbat, co-founder of AMI Labs and the company's VP of World Models

Over email, Rabbat clarified further: “We're still in a research and building phase, so we're not talking publicly about any product plans or timeline.”

AMI is less than a year old, which makes a quiet period reasonable on its face. But the same caginess runs through the whole world-modeling sector, not just its newest entrant.

The most developed product is still a demo

World Labs' Marble is probably the most fully developed product in the space. Its demos range from straightforward media creation to building explorable environments for video games and CGI effects. Robotics use cases exist as well, but the platform as a whole seems more designed to demonstrate capabilities than to serve a defined market.

That pattern — a polished demonstration standing in for a commercial roadmap — is what a reader runs into when trying to trace where the technology will first generate revenue. The answer, for now, is that no one involved is saying.

Even the data suppliers are in the dark

The secrecy extends past the labs themselves. On the sidelines of the same conference, Alex de Vigan, CEO of Physicl, a data supplier for the burgeoning world model business, said he knows his company's data has been useful for whatever its customers are building, but he remains in the dark about what that is.

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

— Alex de Vigan, CEO of Physicl

That is an unusual position for a supplier to be in. Data vendors typically tune their output to a customer's stated needs; here, the vendor is producing blind, guessing at what would be most valuable.

Why the silence is rational

Part of the mystery comes from how versatile world models are as an idea. The simplest version is a navigable map of the world, similar to the AI models that power self-driving cars. But the same approach scales sideways into robots, into video, into simulation.

No one doubts that there are lots of viable businesses to be built on world model tech, and as long as it is easy to fundraise, there is no particular pressure to focus on one. In fact, there is a good reason not to.

If AMI announced tomorrow that it had built a humanoid OpenClaw or a next-generation Hollywood rendering system, a lot of other labs would suddenly be very interested in the space. The lab would then face potential competition from the other world model companies, the neolabs, and even OpenAI and Anthropic.

In some ways, this is the flip side of easy fundraising. Competitors can fundraise too — and the same money that lets a company build under the radar is also funding lots of potential rivals once the path to market becomes clear. Even if competition is inevitable, delaying it for as long as possible is the better outcome, which means keeping quiet about exactly what you are building.

Cixin Liu fans will recognize this as a dark forest scenario: if you do not know who else is in the woods, it is best not to attract attention.

The disclosure problem for everyone else

For observers trying to figure out where the industry is heading, the practical effect is a sector that produces panels and demos but very little in the way of specifics. The people with the clearest view of what is being built — founders, and the suppliers selling into them — are the ones declining to describe it.

That leaves a small set of verifiable facts. AMI Labs was founded by Yann LeCun and is less than a year old. World Labs, founded by Fei-Fei Li, has produced Marble, the most developed product in the space. AMI has worked across manufacturing, biomedicine, robotics, and AI software for doctors through its Nabia partnership. Physicl supplies data to the sector.

• Two leading world model companies: AMI Labs and World Labs
• AMI Labs is less than a year old
• Marble, from World Labs, is the most fully developed product in the space
• AMI has touched manufacturing, biomedicine, robotics, and AI software for doctors via its Nabia partnership

What the quiet means for buyers

Anyone evaluating world model technology today is doing so against a backdrop of deliberate ambiguity. There is no product roadmap to compare, no timeline to plan around, and in at least one case, a data supplier who would build differently if he knew what his customers needed.

That does not make the technology less real, but it does make it harder to assess. The demonstration videos exist; the commercial commitments do not, at least not publicly.

For now, the field's most concrete output is its fundraising. The most concrete statement from one of its leaders is that there is nothing to say yet.

Why this matters beyond the panel

If world model companies continue to withhold product direction, the burden of reading the space falls on indirect signals — who is hiring, which suppliers are growing, which partnerships surface. That is a slower and less reliable way to understand an industry, and it favors observers with informal access over the public record.

It also cuts both ways for the labs. Silence may delay competition, as the dark forest logic suggests, but it can just as easily delay customers who need to know what they are buying. If the easy fundraising environment that currently sustains this approach changes, the same companies may find that the case for openness has become considerably stronger.

Reporting based on original coverage from TechCrunch.