Keep pulling the thread on Carina Hong.
In December, Axiom Prover achieved a perfect score of 120 out of 120 on the Putnam exam, becoming only the sixth perfect score in the exam's 98-year history.
Axiom Prover solved four open mathematical conjectures that had never been solved by any human mathematician, submitted by professors from universities including Technion, Boston College, and Williams College.
The output of Axiom Prover is a computer program that can be executed to verify the correctness of a mathematical proof, allowing non-experts to confirm its validity.
Professor Ken Ono stepped down from his position as Vice Provost of the University of Virginia to join Axiom.
AWS spent three to five years with human experts writing 260,000 lines of formal language code to verify a component of its hypervisor for CPU optimization.
Axiom's core thesis is that an AI mathematician can be generalized to serve as a verifier for other logical domains, such as software code.
Axiom's mission is to build AI reasoners that can perform at a superhuman level.
Carina Hong, CEO of Axiom, dropped out of a combined PhD and law degree program at Stanford University to found the company.
In 2019, a paper co-authored by Francois Charton and Mistral co-founder Guillaume Lempel demonstrated that a Transformer-based model could outperform traditional computer algebra systems like Mathematica in symbolic integration.
Francois Charton, who co-authored a 2019 paper on using Transformers for symbolic math, is a full-time member of the technical staff at Axiom.
The best-performing large language model on the Putnam exam, DeepSeek by Mass Arena, scored 103 out of a total 120 points.
The top human score on the most recent Putnam exam was 110 out of 120, achieved by an MIT undergraduate.