At the All In conference, the emerging world-model sector was presented as one of AI’s most promising—and least legible—frontiers. Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs have drawn substantial buzz and funding around systems designed to understand physical space, yet neither has offered a crisp commercial destination.¹

The opportunity is unusually broad. World models can function as navigable representations of the real world: technology useful for self-driving vehicles, humanoid robots handling boxes, interactive video and explorable digital environments. That range is an advantage, but it also makes strategic focus difficult to read. AMI has explored manufacturing, biomedicine, robotics and medical AI through its Nabia partnership, while World Labs’ Marble has demonstrated media, game-world and CGI applications rather than a single defining business.¹

When pressed on AMI’s direction, co-founder and VP of World Models Michael Rabbat offered little beyond: “We’ll talk about it when we’re ready to talk about it.” He later said the company remained “in a research and building phase” and was not publicly discussing product plans or timing.¹

That restraint has consequences beyond the labs themselves. Alex de Vigan, chief executive of data supplier Physicl, said his company knows its data is proving useful but does not know precisely what 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.¹

Still, the secrecy is not merely evasive. With capital flowing freely into AI, a disclosed breakthrough in humanoids or Hollywood-grade rendering could quickly attract rival world-model firms, newer labs and giants such as OpenAI and Anthropic. For AMI and its peers, silence may be the price of buying time before the market identifies the target—and starts firing at it.¹