Claude robot wins don't kill world models: why Fable 5.2 and World Labs' Atlas solve different problems
dotey · x · 2026-11-12
Pushing back on takes that Claude Fable 5.2 (64.9% on RoboDojo dual-arm sim tests) proves World Labs' Atlas is a dead end, the author argues they solve different problems: Claude-style models are action policies deciding next moves, while Atlas builds spatially consistent generative world representations. Atlas's real weaknesses are physics realism, contact dynamics, and transparency — not the world-model idea. The two routes will likely converge: general models for reasoning and decisions, world models for spatial memory, prediction, and counterfactual simulation.
More from Models
- Claude app widget bugs: edits invisible to Claude, mobile remote control broken — petergyang · 2026-11-12
- Allen-Zhu's new paper: Canon layers boost reasoning 2x, even random convolution weights help — teortaxesTex · 2026-11-12
- Ricursive releases open-weight Ripple-65B, claiming US open-weight SOTA on TerminalBench 2.1 — StanfordAILab · 2026-11-12
- A criminally underrated resource comparing tool-calling reliability across models and providers — ricklamers · 2026-11-12
- Datalab fast-mode OCR beats Opus 5.5 and GPT-6.1 Sol at 1/4 to 1/68 the price — VikParuchuri · 2026-11-12
- Claude's world knowledge impresses: model guesses user identity from one prompt — adonis_singh · 2026-11-12