skip to content

Search
← All concepts

#world-models

Learned simulators of environments and dynamics.

6 chapters, oldest era first: one idea traced across the books.

  1. From Frames to Worlds 2018–2019 Ch. 3: Predicting the Next Frame World-model prehistory: learn a latent simulator, then act inside it.
  2. From Frames to Worlds Cross-era Ch. 18: Evaluating and Monetizing Motion FVD's blind spots, VBench, physics tests, and the economics of the most expensive pixel in AI.
  3. From Frames to Worlds 2020–2023 Ch. 17: The Third Dimension NeRF, 3D from 2D priors, and Gaussian splatting — where 3D meets video and worlds.
  4. From Frames to Worlds 2024 Ch. 9: Sora and Spacetime Patches Diffusion transformers for video, spacetime patches, and the video-models-as-world-simulators claim.
  5. From Frames to Worlds 2024 Ch. 15: Video Models Become Game Engines Action-conditioned generation: Genie, GameNGen, Oasis, and DIAMOND.
  6. From Frames to Worlds 2025–2026 Ch. 16: World Models as Foundation Models Genie 2 and 3, Cosmos, V-JEPA 2 — robotics and embodied simulation as the endgame.