Debate on LLM Cognition: Is Chain-of-Thought Just a Record?

teortaxesTex · x · 2026-09-02

This post responds to the claim that increasing non-CoT cognition makes CoT less informative. The author argues that 100% of LLM cognition happens "not in CoT," which is merely a record of that cognition. Increasing intelligence per token, whether via Engram or scaling, is essentially increasing "moar neuralese."

Original post →

More from AGI Musings

AGI Musings channel →