'Momentum Prior': Agents Locked Onto Initial Tasks Resist New Instructions

Developer gerardsans dubbed agents' tendency to stay locked on initial tasks the 'momentum prior,' and clarified the lock-in arises in the autoregressive loop rather than a single forward pass.

2026-09-09 ~ 2026-09-09 · 2 related posts

• "Momentum prior": the named failure mode where agents lock onto the original task and can't be steered off course — gerardsans · 2026-09-09
• One forward pass is one token — the lock-in happens in the autoregressive loop — gerardsans · 2026-09-09