Robotics is an intelligence problem: GPT-6.2 jump reportedly pushed OpenAI to restructure robotics team

DJiafei · x · 2026-11-22

Robotics researcher Chris Paxton argues robotics is fundamentally an intelligence problem, and current robot learning will hit the same fundamental limits as foundation models. Context from Andrew Curran: an unusual capability jump — similar to what Anthropic observed in Fable 5.2 — appeared in GPT-6.2, significant enough that OpenAI restructured its dedicated robotics team.

Related event: GPT-6.2 hinted for robotics as capability leap prompts OpenAI team reshuffle(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →