Karpathy Claims Models Learn Spatial Reasoning From Text Alone

Karpathy argued that models' spatial and map understanding comes entirely from text prediction, with no image input beyond 2D text layout. Researchers pushed back, questioning whether such models were trained on GIS data.

2026-10-02 ~ 2026-10-02 · 2 related posts

• Karpathy: it's all text prediction — no images involved, only 2D arrangement of responses — karpathy · 2026-10-02
• Karpathy says spatial skill is pure text prediction; researcher counters models train on images and GIS data — gleech · 2026-10-02