Nick Cammarata: CoT monitoring is starting to fail, but interpretability has never looked brighter
nickcammarata · x · 2026-09-03
Former Anthropic researcher Nick Cammarata says the decline of chain-of-thought monitoring is sad and worth fighting, but argues interpretability is the real way to know what neural networks are actually doing — and he's never been more optimistic about it.
In the follow-up thread, mayfer adds that if interpretability works, massive RL runs could presumably self-correct away failure modes detected by mechinterp, and speculates CoT degradation may just be drift toward AI-flavored lingo.
More from AGI Musings
- Ben Goertzel outlines a general recipe for superintelligence 'basic drives': what it improves vs. protects — bengoertzel · 2026-10-21
- Torch vs Flame: researchers debate whether AI's endgame serves humanity or life itself — danfaggella · 2026-10-21
- 2022 interview resurfaced: AI safety researcher says competence, not consciousness, is the risk — davidmanheim · 2026-10-21
- Silicon Valley is a stock market for people: narrative eventually catches up with substance — karinanguyen · 2026-10-21
- Half of humanity is fed by one century-old reaction — time for AI machines to hunt for more — shyamalanadkat · 2026-10-21
- Agent identity verification will be a mess: restaurants ignore AI emails — alliekmiller · 2026-10-21