AI Turns Math Library Into 16 Million Lines

Axiom Math's research-stage work formalizes algebraic geometry text so machines can verify each proof step.

Updated on Nov. 6, 2026 in Mathematics

AI Turns Math Library Into 16 Million Lines

Live Poll

Do you trust AI-generated work more when it has been computer-verified?

Axiom Math, an AI startup founded by Carina Hong, announced its AI agents converted about 7,654 pages of the Stacks Project into 16 million lines of code. The conversion is a research-stage formalization effort, not a shipped product.

Why it matters

AI systems can generate mathematical proofs faster than researchers can check them by hand, so formal verification tools that check each step by computer are becoming a bottleneck-clearing technology. Axiom's work on a shared, machine-checkable library of algebraic geometry suggests that verification at scale is becoming feasible.

The company reports 16 million lines of formalized code from about 7,654 pages, a figure that counts code output rather than independently audited theorems. The Putnam result is a company-reported benchmark: the system solved all 12 problems, while human competitors completed 8 within the time limit, but the extraction does not detail the evaluation conditions.

The players

Axiom Math

An AI startup building systems that generate and formally verify mathematical proofs using the Lean proof assistant.

Carina Hong

Founder of Axiom Math, who left a Stanford mathematics PhD program in March 2025 to start the company.

Johan de Jong

Mathematician who launched the Stacks Project in 2005 as a free online reference for algebraic geometry.

Menlo Ventures

Venture capital firm that led Axiom Math's $200 million funding round in March 2026.

The details

Axiom's AI agents worked on thousands of mathematical problems at once while keeping the collection consistent. The system converts Stacks Project material into a form that Lean, a proof-checking program, can examine step by step, verifying that each proof step follows logically from the previous ones. This process uncovered missing conditions and untrue claims in some original Stacks Project statements.

Timeline

  1. 2005: Johan de Jong launched the Stacks Project as a free online reference for algebraic geometry.

  2. March 2025: Carina Hong left a Stanford mathematics PhD program to found Axiom Math.

  3. 2025: Axiom raised $64 million at a $300 million valuation.

  4. January: Axiom's AI solved all 12 Putnam problems.

  5. March 2026: Axiom raised $200 million led by Menlo Ventures, lifting its valuation above $1.6 billion.

The Tech Race

Axiom's conversion of the Stacks Project pushes formal verification from individual theorems toward large shared libraries, building on the Lean proof assistant that mathematicians already use. The company's aim to apply the same approach to software development points to a wider race to verify AI-generated code.

The technology is not yet available to general users, and the extraction gives no release date for public tools. Engineers interested in verification could eventually benefit if Axiom's stated software development application reaches the market, but no timeline is given.

The takeaway

Watch for independent review of how completely the Stacks Project conversion caught errors, since the company's error-detection claims are not yet verified externally. The next concrete signal is whether Axiom releases its software verification product, which the company has said it aims to build.

Further reading

For broader context on how computers are being used to check mathematical proofs, see Mathematics coverage.

Source note: This article includes information reported by VnExpress International – Latest news, business, travel and analysis from Vietnam.

Live Poll

Do you trust AI-generated work more when it has been computer-verified?