Axiom Math Aims to Safeguard Mathematics from AI Inaccuracies

As [artificial intelligence](/topic/artificial-intelligence) continues to permeate various fields, mathematics is facing a unique challenge. Axiom Math, a startup valued at $1.6 billion, has taken a bold step to ensure that the integrity of mathematical proofs is maintained amidst a flood of AI-generated solutions. On November 5, 2026, the company announced the formal verification of the Stacks Project, a monumental library in the field of mathematics, consisting of over 7,654 pages and 16 million lines of code.

The past few months have seen an explosion of AI-generated mathematical manuscripts, with [OpenAI](/topic/openai) alone releasing 719 purported solutions to complex problems. However, the rapid pace of AI advancements raises concerns about the accuracy and reliability of these solutions. Axiom Math believes that unchecked AI outputs could lead to significant errors in mathematical proofs, potentially undermining the discipline itself.

Axiom Math's CEO, Carina Hong, emphasizes the importance of verification in mathematics. "We want to build a verified foundation that AI agents can rely on to write software and eventually do science," she stated. The company’s approach is not to slow down the integration of AI in mathematics but to ensure that it operates on a solid foundation.

The Stacks Project, initiated by mathematician Johan de Jong in 2005, serves as an open-source reference for a wide array of mathematical concepts, from basic algebra to advanced theories in geometry and number theory. Esteemed mathematicians regard it as one of the most significant contributions to the field, comparable to the Bourbaki books that reshaped mathematics in the 20th century.

Axiom Math's recent verification of the Stacks Project marks a significant milestone in the intersection of AI and mathematics. By translating the entire library into a formalized programming language called Lean, the company has created a robust framework that allows for rigorous checking of mathematical proofs. This formal verification process ensures that each theorem is logically sound and interconnected, akin to a family tree where each result builds upon its predecessors.

Despite the ambitious nature of the project, Axiom Math faced considerable challenges in maintaining the integrity of the Stacks Project during the verification process. The interconnectedness of theorems posed a risk of inconsistencies, which could lead to paradoxical results. Hong expressed initial doubts about whether the verification process could preserve the beauty and coherence of the original work.

"Things that are big and delicate in life just don’t go together," Hong remarked. The team had to ensure that every connection across the 16 million lines of Lean code was intact, a daunting task that required innovative strategies.

To tackle the complexities of the Stacks Project, Axiom Math employed a unique approach. Instead of imposing a rigid hierarchy on their AI agents, the team allowed them to operate in parallel, giving them the autonomy to identify tasks and collaborate on different aspects of the verification process. This strategy not only enhanced efficiency but also fostered creativity among the agents.

Ashvin Swaminathan, a member of Axiom’s technical staff, explained, "With Stacks, we imposed a hierarchy, but it was not a complete hierarchy dictating what every single agent was doing. The agents were finding tasks for themselves and identifying things they could work on in parallel." This flexibility proved crucial in successfully verifying the intricate web of mathematical results contained within the Stacks Project.

As Axiom Math continues to push the boundaries of what is possible with AI in mathematics, the implications of their work extend beyond the Stacks Project. The company aims to set a precedent for how AI can be integrated into mathematical research without compromising rigor and accuracy. By establishing a verified foundation, Axiom Math hopes to inspire confidence in AI-generated solutions and pave the way for future advancements in the field.

In a landscape where AI is rapidly evolving, Axiom Math stands firm in its commitment to safeguarding the integrity of mathematics. The successful verification of the Stacks Project is just the beginning of a journey that could redefine the relationship between AI and mathematics for years to come.