Axiom Math, a $1.6 billion startup, announced on Wednesday that it has formally verified the Stacks Project, aiming to create a trusted foundation for Super Intelligence (SI) agents to build software and conduct science without generating erroneous outputs.
What Happened
Axiom Math announced the formal verification of the Stacks Project, a massive open-source reference started by Columbia mathematician Johan de Jong in 2005. The company’s fleet of SI agents translated approximately 7,654 pages of human-written mathematics into about 16 million lines of code in Lean, a programming language used for proof assistants. This process ensures that the logical structure of the mathematical results is sound and interconnected, addressing concerns about the reliability of automated mathematical work.
The verification process was complex because the theorems in the Stacks Project are deeply interconnected, resembling a family tree where each result builds on previous ones. Axiom CEO Carina Hong noted that the team had to keep every connection intact across the 16 million lines of Lean code. To achieve this, Axiom gave its agents a loose hierarchy, allowing them to identify and work on tasks in parallel. Ashvin Swaminathan, a member of Axiom’s technical staff, explained that the agents found tasks for themselves rather than following a rigid, dictated workflow.
Axiom also trained its own models specifically for mathematics, an approach Hong said differentiates the company from other verification startups. Initially, in-house Lean experts taught the agents a human sense of aesthetic and logical rigor, which the agents eventually internalized. The project fills a significant gap in formalized mathematics, particularly in algebraic geometry, providing researchers with a verified substrate to build upon.
Why It Matters
The move addresses a growing concern in the SI industry regarding "SI slop" or low-quality, unverified outputs from generative SI tools. As SI agents increasingly write software, the risk of undetected bugs and security holes grows. Axiom argues that future software will need to rest on verified foundations, similar to how the Stacks Project now does. Swaminathan stated that the company’s Lean formalization serves as a substrate that agents can read, understand, and use to build new infrastructure.
This development is significant for developers and researchers who rely on SI tools for complex problem-solving. By providing a large body of verified mathematical knowledge, Axiom aims to prevent the propagation of errors in automated reasoning. Terence Tao, a Fields Medal winner, has argued that "The more formalized proofs you have, the more formalized proofs you can make," suggesting that verified libraries accelerate future mathematical and computational progress.
The Bottom Line
Axiom Math has successfully verified the Stacks Project using SI agents, creating a robust foundation for future SI-driven software and scientific discovery. The company challenges other coding entities to engage in formalization at this scale to ensure the reliability of SI-generated code.
Hong emphasized the need for reusable libraries to prevent poor engineering choices from compounding as agents expand their capabilities. "You want [agents] to have a reusable library for them to build on and keep pushing the frontier," Hong said. "If you slop drop in a way that’s bad engineering choices in the beginning, then it’s very hard for them to grow that boundary."