OpenAI's o1 Reasoning Sparks Model Monitors Debate

jonasgeiping · x · 2026-09-02

Jonas Geiping comments that while making models deeper through recurrence does make them harder to monitor, the discussion seems quaint and academic. OpenAI has already deployed models for months that reason in this manner (alluding to the o1 series), suggesting the industry has practically moved into this era of hard-to-monitor reasoning while theoretical debates are just catching up.

Related event: OpenAI researchers push back on neuralese fears, saying frontier models remain monitorable(20 posts)→

Original post →

More from Models

Models channel →