FULL STORY

AI Agents Formalize the Enormous Theorem in 24 Days

Axiom Math announced that a cluster of dozens of AI agents completed the first fully machine-verified Lean proof of the Classification of Finite Simple Groups in 24 days, with follow-up confirmation that every logical step passed the Lean kernel check.

2026-10-27 ~ 2026-11-03 · 2 episodes · 36 posts

Episode 1 · Axiom's AI Agent Cluster Lean-Formalizes the Classification of Finite Simple Groups in 24 Days (2026-10-27, 34 posts)

In late October, Axiom Math announced the first complete machine-verified proof of the Classification of Finite Simple Groups (CFSG), nicknamed the "Enormous Theorem": a cluster of dozens of AI agents formalized the roughly 15,000-page proof—spread over 500+ papers by 100 mathematicians across generations, and the longest proof in human mathematics—into tens of millions of lines of Lean code within 24 days, verified line by line by the Lean kernel.

Confirmed

  • Axiom announced the complete Lean auto-formalization of CFSG, the first fully computer-verified proof, checked logically statement by statement by the Lean kernel; the theorem states every finite simple group belongs to known families such as cyclic groups of prime order and alternating groups, and is called the "periodic table of symmetry."
  • Original proof scale: 15,000 pages, 500+ papers, 100 mathematicians over decades, scattered across thousands of journal pages; multiple posts note no single mathematician has been able to independently verify it for 40+ years.
  • Formalization took 24 days using a cluster of dozens of AI agents; per a post by docdavkitty, agents coordinated via GitHub issues and Slack with mutual review, and verification—not generation—was the bottleneck. Reported code volume varies across posts from 38 million to 50 million lines, commonly cited as 43.1 million or "over 43 million," among the largest Lean codebases ever.
  • rohanpaulai relayed that 34 errors were caught; burnytech relayed over 5,000 errata found, all fixable; other posts (including FuSheng0306) claimed a decades-old gap in the original proof was found and fixed.
  • OpenAI's Kevin Weil publicly announced the result; Terence Tao and others praised it. Some posters noted the project may be among the largest open-source software projects; CFSG had previously been considered among the least likely proofs to be formalized.
  • Rutgers mathematician Alex Kontorovich relayed Timothy Gowers's assessment: given the long-standing disputes over CFSG's correctness and its spread across tens of thousands of journal pages, its formalization is major good news.

Unconfirmed

  • Exact line counts vary across posts (38M–50M); figures such as "decades-old gap fixed," "5,000+ errata," and "34 errors" come from reposts without a primary technical report and may refer to different measures.

Why it matters

  • The "Enormous Theorem" received line-by-line machine verification for the first time. Gowers and Kontorovich both noted CFSG's correctness could never be fully confirmed due to its sheer size and dispersion; formal verification provides a definitive endorsement and cleaned up thousands of fixable flaws in the literature.
  • Completing verification of a decades-long human effort in 24 days demonstrates AI agent clusters' capacity for large-scale collaborative work; some posters noted verification rather than generation was the real bottleneck. Commenters see potential for agent swarm coordination and a signal that formal verification could scale to major conjectures like Fermat's Last Theorem or the ABC conjecture; some (e.g., FuSheng0306) argued AI intelligence has already surpassed human mathematics.

14 more related posts →

Episode 2 · Classification of Finite Simple Groups Fully Formalized in Lean (2026-11-03, 2 posts)

Axiom Math has completed a full Lean formalization of the Classification of Finite Simple Groups, spanning about 43.1 million lines of code and 2.05 million declarations. The effort also uncovered proof issues in the published literature.