What happened
Computer scientist Louis Castricato founded Overworld, shifting focus from large language models to "world models" that understand and navigate physical environments. This pivot sees prominent AI scientists, including Fei-Fei Li (World Labs founder) and Yann LeCun (Advanced Machine Intelligence Labs founder), dedicating efforts to AI systems that learn the statistical structure of space, time, and physics to predict action consequences. Venture capitalists, such as Kindred Ventures' Steve Jang, are investing in these world model companies, including Overworld, Causal Labs for weather prediction, and Extropic for specialised chips.
Why it matters
AI development priorities are shifting from text-based chatbots to physical interaction, redefining future application landscapes. World models provide AI with an understanding of physical laws and environmental dynamics, enabling a faster, cheaper path to "physical AI" and robotics, as noted by Carnegie Mellon University's Martial Hebert. This contrasts with the inherent limitations of large language models, which cannot physically interact with the world. Platform engineers and product architects should prepare for infrastructure demands supporting AI agents interacting with real-world physics, moving beyond purely linguistic model deployments.
