Mathematicians spar over whether a 250,000-page Lean proof is just a useless oracle
jessi_cata · x · 2026-11-21
Responding to mathematician Arturo Magidin's claim that "a 250,000-page Lean dump is basically just an Oracle saying yes/no — answer by itself is useless," jessicata calls it a bad pun and a howler: it runs together three distinct things — (1) having a proof of Y conditional on one of X, (2) having a proof of X, and (3) knowing whether X is true. Part of the ongoing math-community debate over Anthropic's Riemann hypothesis proof claim.
More from Research
- SWE-bench Co-author Ofir Press: Good Benchmarks Should Start Near 0% Accuracy — OfirPress · 2026-11-21
- Claim: Anthropic mathematician and Claude produce 250,000-page Lean 4 proof of Riemann Hypothesis, verification ongoing — Electronic-Weekend66 · 2026-11-21
- Levent Alpöge announces Claude-assisted claimed proof of Riemann Hypothesis — sudoraohacker · 2026-11-21
- NVIDIA researchers present continuous diffusion LLM with 8B benchmark results — ArashVahdat · 2026-11-21
- Two new benchmarks push AI evals past static testing: recursive self-improvement and KradleBench — typewriters · 2026-11-21
- Claude's claimed Riemann proof mocked: 250,000 pages of unreviewed Lean, author calls it 'slop' — GaryMarcus · 2026-11-21