Li Fei-Fei published a new article on her personal Substack clarifying the concept of "world models" in AI, which has been the hottest and most confusing term since 2025. She revisits the classic POMDP loop from reinforcement learning textbooks and identifies that what is now called a "world model" is actually three different projections of this loop: the renderer outputs pixels (observations), the simulator outputs states, and the planner outputs actions. She concludes that among the three, the renderer is the most commercially mature but has a ceiling, the planner is the most exciting but furthest from real-world deployment, and the simulator is a severely underestimated key hub. The article is also World Labs’ product manifesto; their Marble is already outputting both Gaussian splatting and collision meshes, attempting to unify the renderer and simulator into a single model. The conclusion envisions a unified foundational world model capable of freely switching between rendering, simulation, and planning based on downstream needs.