Greg Kamradt estimates only 1-2% of Claude's reasoning knowledge nodes get used
GregKamradt · x · 2026-11-21
Greg Kamradt found the timeline view of Claude's thought process (which ideas worked) fascinating, but asks about the ideas that didn't — the missing "knowledge tree" of failed branches. He guesses roughly 1-2% efficiency, measured as knowledge nodes created vs. knowledge nodes eventually used in the final proof, and argues we're approaching a stable "$ → knowledge" conversion ratio: a quantifiable rate of how much spend on AI reasoning actually converts into useful knowledge.
Related event: Greg Kamradt estimates only 1-2% of Claude's proof ideas are actually used(2 posts)→
More from AGI Musings
- Riemann Hypothesis by end of 2026: bold AI timeline prediction doubles down — dioscuri · 2026-11-21
- "Any math problem is solvable with the n-th model and industrial compute" — math-engineering declared the future — basedjensen · 2026-11-21
- The AI bubble playbook: from 100x VC returns to the Fed tightening, mapped step by step — Svwang1 · 2026-11-21
- Dario Amodei: Anthropic is deliberately running at less than half the speed it could — rohanpaul_ai · 2026-11-21
- X debate: copying yourself to rig simulation predictions, a mirror of 'universal prior is malign' — jessi_cata · 2026-11-21
- The AI trade has been the compute buildout — and its end will get weird fast — matt_slotnick · 2026-11-21