Key takeaways

  • On November 21, 2026 the Lean and Mathlib communities said they have full confidence that Levent Alpöge's Lean proof of the Riemann hypothesis is correct.
  • Independent checkers confirmed the proof targets Mathlib's vetted statement and uses only Lean's three standard axioms.
  • MIT's Andrew Sutherland says the code also proves the Generalized Riemann Hypothesis for Dirichlet L-functions; a human-readable paper has not been published.

The Lean and Mathlib communities said on November 21, 2026 that Levent Alpöge's proof of the Riemann hypothesis is correct. In a joint statement on the Lean community blog, endorsed by the Lean FRO and the Mathlib Initiative, they wrote: "We have full confidence that Alpöge's proof of the Riemann Hypothesis is correct." The proof, which Alpöge says he found with Claude, was released as Lean code one day earlier.

That is a much stronger statement than anyone could make on Friday. A computer-checked proof doesn't need a referee to read 250,000 pages. It needs people to confirm that the code proves the right statement, that it assumes nothing extra, and that the program doing the checking can be trusted. The Lean community says all three checks have passed.

What was checked

Lean is a programming language in which every step of a proof is checked by a small core program called the kernel. Mathlib is the shared library of mathematics written in Lean, and it already contains a definition of the Riemann hypothesis that experts have vetted.

According to the statement, the checks covered four things:

  • The statement. The proof targets Mathlib's own RiemannHypothesis definition. The community added a file that states the theorem and imports only Mathlib, and two tools confirmed that what Alpöge's code proves matches it.
  • The build. Dozens of people compiled the library, about 17 million lines that depend only on Mathlib, without errors, and got the same exported proof file.
  • The proof. Lean's kernel and several independent checkers, written in different programming languages, accepted the exported proof. The statement says two of them, con-leche and con-ron, come with a formal proof that they cannot accept a false proof.
  • Tampering. The checks ran on an "inert export" of the proof, so any malicious code hidden in the sources could not affect them.

The MIT check covers GRH too

MIT mathematician Andrew Sutherland keeps a verification page with the details, and says he ran the checks with Claude Code. As of its update at 12:55 UTC on November 21, it reports that all 175,187 Lean modules compile with zero errors, and that the main theorems use only Lean's three standard axioms (propext, Classical.choice and Quot.sound), so nothing extra was assumed.

The page lists five theorems. Besides the Riemann hypothesis itself, they include the Generalized Riemann Hypothesis for Dirichlet L-functions, a family of functions that do for primes in arithmetic progressions what the zeta function does for all primes. More general L-functions are out of scope. The compressed proof file alone is 7.7 GB.

Sutherland lists what was still running: two of Lean's own kernel modes and part of the lake comparator tool. He also notes the build was not sandboxed and that a pass depends on the kernels and Mathlib's definitions being right.

What this does and doesn't settle

A formal proof answers whether the theorem is true. It doesn't explain why. The Lean statement says an account of the proof's ideas is future work, and Alpöge has said he will write a paper. Until mathematicians can read the argument, nobody can say which new ideas it contains or whether they apply elsewhere.

The Clay Mathematics Institute's rules for the $1 million prize also require publication in a qualifying journal and a two-year wait, so the prize question is years away.

The result also puts the competing work in a new light. Mathematician Ravi Vakil announced a separate, all-human draft proof of the Generalized Riemann Hypothesis by Alessandro Masullo on November 20, which Brian Conrad has been reviewing for months. That one is still being checked by hand.

For the background, see our reports on Alpöge's original claim, what the Riemann hypothesis is and what formal verification is.

Next to watch: Sutherland's remaining kernel checks, and the human-readable paper Alpöge has promised.

Sources

  1. Lean verification of the Riemann HypothesisLean community blog
  2. Verification files for the Lean proof of the Riemann HypothesisAndrew Sutherland, MIT
  3. Remark (statement by Levent Alpöge)Levent Alpöge
  4. Hurried announcement: Alessandro Masullo, GRHRavi Vakil
  5. Millennium Prize rulesClay Mathematics Institute