On November 20, 2026, Levant Alpöge published a proof of the Riemann Hypothesis as a public Lean library of 17 million lines of code, and on November 21, Lean FRO and the Mathlib Initiative issued a joint statement that the proof has been verified and is correct. This is the first case where one of the great unsolved problems of mathematics has been claimed as proven and simultaneously formally verified by independent checking kernels.
What Happened
The proof is fixed in the Deicyde/riemann-hypothesis repository and compiles without errors in Lean v4.33.1 and Mathlib v4.33.1, which contains 175,187 modules. The raw export of the proof was accepted by independent checking kernels nanoda and con-leche, and the theorem statement was cross-checked against the RiemannHypothesis assertion in Mathlib using the comparator and con-leche --compare tools. According to the author, the proof establishes five theorems: the Riemann Hypothesis itself, its verbal formulation, the absence of zeros to the right of the critical line, and the placement of zeros on the line for all Dirichlet L-functions, including the form for primitive characters, i.e., GRH for Dirichlet L-functions. The key idea of the proof is iterated theta lifts between unitary groups, yielding multiplicity over the integers.
Context
The Riemann Hypothesis is a 165-year-old unsolved problem on which error estimates in analytic number theory depend. Previously, major mathematical results were published as articles and verified by journal peer review, not by machines. The methodology applied here—a public repository, raw proof export, acceptance by independent kernels, and statement cross-checking—is reproducible and can be transferred to other proofs. Verification took the community one day, reflecting the maturity of the Lean/Mathlib toolchain and community.
Why This Matters for the Industry
For the industry, this is a precedent: "machine-verified proof" has become applicable to problems of this scale, and the scheme of "raw artifact plus several independent checkers plus a public results page" can become the standard for future major proofs. Teams already working with Lean/Mathlib receive a reference for scaling, and the formal verification market receives a public precedent that lowers the trust barrier for clients who previously required journal peer review. This opens a window for proof-as-a-service with reproducible auditing and for pipelines where LLMs generate Lean code and verification serves as an automatic control stage.
Why This Matters for Users
If the community verdict is confirmed, the Riemann Hypothesis will be considered solved. The proof is fully fixed in code that anyone can compile and check, rather than relying on trust in journal peer review. Independent verification results are published on the page math.mit.edu/~drew/grh/ and in the Deicyde/riemann-hypothesis repository, where readers can follow the verification process.
What Is Still Unknown / Limitations
From a scientific perspective, the result is currently a machine-verified claim, not a peer-reviewed article: available sources do not contain the full text of an article or an independent assessment of the novelty of the key idea. The scope of the assertion is broader than the classical Riemann Hypothesis, and the extended formulations require separate verification. If remarks or errors appear during verification, this will become a case study on the limits of machine verification and the fact that compilation does not equal final scientific confirmation.
Sources
- Lean verification of the Riemann Hypothesis — joint statement by Lean FRO and Mathlib Initiative
- Claimed Proof of the Riemann Hypothesis (RH) — note by Levant Alpöge (PDF)
- RH Lean proof: verification files — Andrew Sutherland (MIT)
Author
Look at AI, editorial team