GPT-6-controlled robot arm breaks gripper during LEGO grasping; team seeks force-aware control

paigeinsf · x · 2026-09-21

paigeinsf's team ran OpenAI's GPT-6 Astra through a custom agent-control harness on YAM arms for LEGO grasping — and broke a gripper, suspectedly from excessive force. They're asking for recommendations on force-aware control for LLM-driven robots: feedback signals, controller-level limits, and better action interfaces. A concrete example of how unconstrained model outputs can damage real hardware.

Original post →

More from coding & agent

coding & agent channel →