· via Hacker News – Front Page (hnrss.org)
Lean and Mathlib communities certify 17-million-line formal proof of the Riemann Hypothesis
The Lean and Mathlib communities have certified a 17-million-line Lean library from Levent Alpöge as a correct proof of the Riemann Hypothesis, verified against four criteria with independent checkers.

A landmark proof, formally certified
On November 20, 2026, Levent Alpöge released a Lean library totaling 17 million lines of code that proves the Riemann Hypothesis, long among the most famous unsolved problems in mathematics. Within about a day, the Lean and Mathlib communities published a joint statement, endorsed by the Lean FRO and the Mathlib Initiative, certifying that both the statement and the proof are correct. The announcement, which surfaced on the front page of Hacker News, expresses the communities' full confidence in Alpöge's result.
How the certification works
According to the statement, the verification rests on four criteria drawn from a checklist published on the Xena Project blog earlier in the month, and all four are now satisfied.
The first concerns the statement itself. The Riemann Hypothesis is already formalized in Mathlib, the main mathematical library for Lean, under the name RiemannHypothesis, and it has been reviewed by specialists in both number theory and Lean. This matters because a formal proof is only as meaningful as the claim it actually establishes; a subtly weakened statement would hollow out the achievement.
The second is reproducibility. Dozens of people independently compiled Alpöge's codebase, which depends directly only on Mathlib, and every reported build reproduced the same exported proof artifact with no errors or build failures.
Matching theorem to statement, and checking the proof
The third criterion addresses the gap between proving something and proving the right thing. Reviewers added a dedicated file that isolates Mathlib's Riemann Hypothesis statement and imports nothing beyond Mathlib. Two independent comparison tools, one called comparator and the other con-leche run in comparison mode, confirmed that the theorem established inside Alpöge's library is exactly that statement.
The fourth covers the validity of the proof term itself. The exported proof was checked by several programs: Lean's official kernel, the external con-leche and con-ron checkers, and multiple third-party checkers implemented in different programming languages. Independent implementations reduce the chance that a bug in a single verifier could wrongly bless a broken proof.
Two details stand out. The con-leche and con-ron checkers are accompanied by a formal argument, relative to set theory, showing they are incapable of accepting an unsound proof, a defense against the possibility that a defect in Lean's own logic could be exploited. In addition, the verification ran on a static export of the proof, insulated from any malicious code that might hide in the Lean source files, so checking could not trigger unintended execution. The statement also points readers to an extensive overview of the effort maintained at MIT.
Verified is not the same as understood
The certification settles the mechanical question of correctness. It does not explain the mathematics. The statement closes by anticipating the work ahead, saying the communities look forward to the broader mathematical community's efforts to unpack the new ideas in the proof. Turning a 17-million-line machine artifact into human-comprehensible mathematics is the next, and separate, challenge.
Why it matters
If the certification holds, this is a watershed moment for formal mathematics tooling. A problem of this stature has traditionally been validated through years of human refereeing; here, trust reduces to small proof kernels and independently implemented checkers, several of which carry their own proofs of soundness.
The verification playbook is itself a deliverable: reproducible builds, an isolation file pinning down the statement, independent comparison tools, and cross-implementation checking offer a template for certifying other large formal proofs.
For the Lean ecosystem, compiling and verifying a library of this size in roughly a day shows that the tooling, and the community around it, can operate at a scale few would have predicted. It also shifts the bottleneck from whether the proof can be trusted to whether the mathematics can be understood.
- #lean
- #formal-verification
- #mathematics
- #proof-assistants
- #mathlib