Lean and Mathlib Communities Certify Claude-Assisted Riemann Proof

Unite.AI
Read full story· 5 min
The Lean and Mathlib communities certified that Levent Alpöge’s 17-million-line Lean formalization proves the Riemann Hypothesis, verified through extensive compilation and proof checking on November 21, 2026. Alpöge discovered and formalized the proof using Claude, with multiple independent tools confirming its correctness and logical soundness.

What happened

On November 21, 2026, the Lean and Mathlib communities certified that Levent Alpöge’s 17-million-line Lean formalization proves the Riemann Hypothesis. The certification followed a day of compilation, proof checking, and statement-matching by multiple independent tools and community members. Alpöge developed and formalized the proof with assistance from Claude, an AI system. The proof was released as a Lean codebase, with its correctness confirmed by several independent verification systems.

Key details

  • Certification date: November 21, 2026
  • Proof size: 17 million lines of Lean code
  • Lead author: Levent Alpöge, using Claude for discovery and formalization
  • Verification: Multiple independent tools, including Lean’s kernel, con-leche, con-ron, and a Rust kernel
  • Mathlib version used: 4.33.1; 175,187 modules compiled with zero errors

Why it matters

Mathematicians and formal methods researchers now have a machine-verified proof of the Riemann Hypothesis, a major unsolved problem in mathematics, formalized in Lean and checked by several independent systems. This gives the mathematical community a concrete, reproducible artifact to study and potentially build upon, though understanding and simplifying the argument remains a challenge. The use of Claude in both discovery and formalization introduces new possibilities for collaboration between mathematicians and AI systems in future research.

More on this story

  1. Hours after Anthropic claims a Claude proof of the Riemann hypothesis, Stanford releases an all-human competing draft
    · The Decoder
  2. In a year of AI solving maths problems, an Anthropic worker says he may have cracked the biggest
    · 20 sources

More in AI Research