Greg Kamradt estimates only 1-2% of Claude's proof ideas are actually used

Commenting on a visualization of Claude's work on the generalized Riemann hypothesis, ARC Prize's Greg Kamradt estimates that only 1-2% of the ideas Claude generates end up being used in the final proof, urging attention to the failed branches of its reasoning tree.

2026-11-21 ~ 2026-11-21 · 2 related posts