Infographic cover

📅 November 22, 2026 • The Decoder

On November 20, 2026 Levent Alpöge released a proof of the Riemann Hypothesis as a Lean library of 17 million lines of code. His own note says: "I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis", and that "Claude and I have also formalized the arguments without assumptions". On November 21 the Lean and Mathlib communities, backed by the Lean FRO and the Mathlib Initiative, published a joint statement: they have full confidence that Alpöge's proof is correct and certify the statement and the proof.

Dozens of people reported correct compilation of the codebase, reproducing the same raw proof export file without errors or build failures. The statement was checked separately from the proof: the hypothesis is Mathlib's own RiemannHypothesis definition, carefully vetted by experts in Lean and number theory, and a community file isolating that statement was compared against the theorem in Alpöge's library with two independent tools, the comparator and con-leche -compare, and both confirmed an exact match with the Mathlib statement.

The proof was checked by several proof checkers, Lean's official kernel, the con-leche and con-ron checkers, and independent checkers added by third parties in different languages. The con-leche and con-ron checkers carry a formal proof, relative to set theory, that they cannot accept a false proof, addressing the concern that Lean's logic might have flaws. All checks ran on an inert export of the proof, out of reach of hypothetical malicious code in the Lean sources, and all four criteria from the Xena Project post of November 12, 2026 are satisfied.

Reproduction comes from outside the release team as well. The MIT verification page of Andrew Sutherland reports that all 175,187 modules compile with zero errors and that 175,187 of 175,187 modules were replayed through the kernel, and an independent check on the LUMI supercomputer reports the same compile result. Main.lean carries five theorems, the Riemann Hypothesis plus analogous statements for Dirichlet L-functions, and those theorems rest only on three standard axioms, propext, Classical.choice and Quot.sound.

Key figures

IMPACT: For practice this is about checkability, not model correctness. Per The Decoder, Metaculus forecasters had expected the proof by May 2035 and Terence Tao said last month he expected the problem to fall to a combination of humans and AI around 2050, while The Decoder calls Alpöge a Stanford mathematician and per MLQ.ai and ai.info he is a mathematician at Anthropic, with the result described as Claude-assisted. The note claims the method extends to the Generalized Riemann Hypothesis, and the author expects the Extended Riemann Hypothesis with more time.

Sources:
https://leanprover-community.github.io/blog/posts/riemann-hypothesis/
https://the-decoder.com/lean-and-mathlib-community-officially-confirms-levent-alpoges-proof-of-the-riemann-hypothesis-is-correct/
https://math.mit.edu/~drew/grh/
https://alpo.ge/remark.pdf
https://mlq.ai/news/lean-and-mathlib-communities-certify-levent-alpoges-claude-assisted-proof-of-the-riemann-hypothesis
https://github.com/Hertaspace/alpoge-rh-lean-verification

IMPACT checklist

Continue the conversation: https://t.me/AIAutomationArchitecture · Bluesky

AI Automation