Riemann Hypothesis Lean Proof Receives Community Certification, Claude-Assisted 17M-Line Code Passes Multiple Validations

Mathematician Levent Alpöge released a 17-million-line Lean proof library for the Riemann Hypothesis built with Claude assistance, and the Lean and Mathlib

On November 20, 2026, mathematician Levent Alpöge released a Lean proof library for the Riemann Hypothesis, generated with Claude’s assistance, containing 17 million lines of code.

Factual Reconstruction

On November 21, the Lean and Mathlib communities completed an independent compilation check of the proof library. The check showed that the codebase depends directly only on Mathlib and meets four established criteria: the statement is correctly formulated; it compiles without error; the proof content is consistent with the RiemannHypothesis statement in Mathlib; and the Lean proof is valid. Multiple participants reported generating identical raw proof export files, with no build failures or errors.

The community additionally added an isolation file that imports only Mathlib to compare the statement of the proof in Alpöge’s library. Two independent tools confirmed that the two match exactly. The exported raw proof was checked by the official Lean kernel, con-leche, con-ron, and checkers implemented independently in other languages by third parties, and all passed.

Mechanism Breakdown

The checking process emphasizes inspection of lazily exported files to avoid the influence of potentially malicious code in the Lean source code. The con-leche and con-ron checkers come with formal proofs relative to set theory showing that they will not accept false proofs, thereby reducing concerns that Lean’s logic itself may contain flaws.

The entire process relied on resources contributed by multiple parties in the community; dozens of people participated in reproducing the compilation, and the proof library was ultimately formally certified as correct.

Industry Impact

This certification shifts the discussion of the credibility of AI-assisted mathematical research from “claimed solutions” to a “machine-checkable” level. The formal verification system becomes the gatekeeping checkpoint for AI mathematical capability, meaning that similar work in the future will need to pass the same multiple independent checks before it can gain recognition from the mathematical community.

The relevant parties, including Lean FRO and the Mathlib Initiative, both endorsed a joint statement, indicating the practical feasibility of the formalized toolchain in handling large-scale proofs.

Strategic Assessment

[Analysis] Looking at historical precedents, previous AI mathematical claims mostly remained at an informal stage. This time, 17 million lines of code passing zero-error validation by multiple checkers may encourage more researchers to incorporate formal verification into AI-assisted workflows and change the mathematical community’s acceptance threshold for AI-produced results. However, this judgment is based on a single current event; the long-term landscape still requires observation of more cases.