What changed
Axiom Math announced it has formally verified the Stacks Project, one of the largest bodies of mathematics ever written by humans. The company used AI agents to translate the project's 7,654 pages into about 16 million lines of computer-checked code in the Lean programming language. This verification ensures the mathematical structure is logically sound and consistent.
What this means in practice
The verification of the Stacks Project provides a reliable foundation for future mathematical and software development. Mathematicians can trust the logical consistency of the theorems and definitions, while developers can use this verified code as a building block for new software. The project's interconnected theorems and definitions, which trace back to earlier results, were a challenge to verify, requiring the team to maintain all connections across the 16 million lines of Lean code.
Limitations and availability
The verification process involved a team of AI agents working in parallel, with some autonomy in task selection. Axiom trained its own models to be good at math, an approach not commonly used by other verification startups. The result fills a significant gap in formalized mathematics, particularly in algebraic geometry. While the verified Stacks Project is now available, the process of formalizing mathematics remains complex and resource-intensive. The company's approach may influence future AI development by providing a reusable library for agents to build upon, reducing the risk of producing unreliable results.
