Fei-Fei Li's Atlas world model rebuilds 3D spaces from 3 photos, 50-100x cheaper
Chris_Armstrong · x · 2026-09-05
World Labs co-founders Fei-Fei Li, Justin Johnson, and Ben Mildenhall join a16z's Martin Casado to discuss Atlas, their newly released world model for spatial intelligence.
- Core idea: LLMs use next token prediction and video models use next frame prediction; Atlas is built on next view prediction, the first model to unify pixel generation and reconstruction — two problems computer vision has kept separate for half a century.
- Practical result: digitally capturing a 3D representation of a space becomes 50-100x cheaper — previously 100-300 photos of a room were needed; Atlas works from just three.
- The conversation also covers technical details like the slow-motion shot.
Related event: Fei-Fei Li's World Labs unveils Atlas world model(3 posts)→
More from Embodied
- xArm 6 real-robot validation rig live; sim champion demo coming Sept 12 — markjeffrey · 2026-09-05
- Tesla Cybercab rides open to public early at 2pm CT due to demand — yunta_tsai · 2026-09-05
- Physical Intelligence solved most of a PB sandwich task in ~3 months, but success rate is 53% — binarybits · 2026-09-05
- Humanoid robots lose balance without power, so companies avoid homes with small children — binarybits · 2026-09-05
- Open-Source eInk Bike Computer Ships, AI Reverse-Engineers ANT Protocol on ESP32 — stingrae · 2026-09-05
- Tesla Cybercab picks up rider at hotel for breakfast in first-hand test — EricETesla · 2026-09-05