World model companies hide far too many secrets.

This week, journalist Russell Brandom hosted a discussion about world models at the All In Summit. The session gave him a chance to dive into one of the most mysterious corners of the AI world. Following his observations, let's take a look at what these world model companies are hiding.

Russell Brandom writes that the major players in this field are Yann LeCun's AMI Labs and Li Feifei's World Labs. Both firms have accumulated massive hype and funding, but rank surprisingly low when it comes to "actually making money."

At their core, world models are about automated spatial intelligence. As a result, this field could move toward many exciting, highly profitable directions: robotics, interactive video, and more sophisticated autonomous driving systems.

But when he started pressing for clarity on where exactly this technology will be commercialized, things became very vague.

Even Executives Refuse to Spell Out the Details

The closest thing to an authoritative source Brandom could find was Michael Rabbat, co-founder of AMI Labs and the company's Vice President of World Models. But when he asked what exactly the company was working on, Rabbat was cautious: "We'll talk about it when we're ready to talk."

Over email, Rabbat further clarified: "We are still in the research and building phase, so we will not publicly discuss any product plans or timelines."

To be fair, AMI was founded less than a year ago, so its silence is understandable. But Brandom notes that this cautious attitude extends to the entire world modeling sector.

Even the Most Mature Products Look Like Capability Demonstrations

Brandom mentions that World Labs' Marble is probably the most mature product in this space. Its demos range from straightforward media creation, building explorable environments for video games, to CGI special effects. There are also robotics use cases, but the entire platform seems to be designed more to showcase capabilities rather than follow a clear commercialization path.

This culture of secrecy even extends to the suppliers of these companies.

At a side event of the same conference, Brandom spoke with Alex de Vigan, CEO of Physicl, a data provider for the emerging world model business. de Vigan said he knows Physicl's data is useful for what his clients are building, but he still has no idea what that end product actually is.

"I wish they would tell us more. If we knew what they were working on, we could build much more useful data," de Vigan told Brandom.

The Concept Is So Versatile That No One Can Clearly State What They Are Building

Brandom believes part of the mystery comes from the extreme versatility of world models as a concept.

In its simplest form, it is a navigable map of the world, similar to the AI models that power self-driving cars. But the same modeling approach that helps Waymo navigate through traffic can also help humanoid robots move boxes, or turn minutes of video footage into a fully explorable environment.

AMI has already dabbled in manufacturing, biomedicine, robotics, and even AI software for doctors through its Nabia partnership. Brandom judges that it certainly will not pursue all of these directions — but maybe one or two of them are starting to stand out?

Lack of Focus Could Be a Deliberate Strategy

Brandom points out that no one doubts many viable businesses can be built on top of world model technology. And as long as fundraising remains easy, there is no particular pressure to narrow down and focus on just one direction.

In fact, there are good reasons not to focus.

If AMI announced tomorrow that it had built a humanoid OpenClaw or a next-generation Hollywood rendering system, many other labs would suddenly become extremely interested in this space. Very soon, the lab would face potential competition from other world model companies, new startups, and even OpenAI and Anthropic.

In a way, this is the flip side of easy fundraising: your competitors can also raise capital. And the same money that lets you build under the radar will also fund dozens of potential rivals the moment the path to market becomes clear.

But even if competition is inevitable, it is best to delay it as much as possible. That means staying silent about exactly what you are building.

The Dark Forest

Brandom writes at the end of his article that fans of Liu Cixin will recognize this as the dark forest scenario: if you do not know who else is in the forest, it is best not to draw attention to yourself.

World model companies are keeping so many secrets not necessarily because they have no story to tell, but because at this stage, silence itself is a competitive strategy.

This article is from the WeChat official account "Tech Business", and is republished with authorization from 36Kr.