Axiom Math, a Palo Alto startup, has launched a free AI tool for mathematicians called Axplorer, aimed at uncovering patterns that may help solve long-standing mathematical problems.

Axplorer is a revamped version of PatternBoost, a system François Charton co-developed in 2024 while at Meta. PatternBoost required a supercomputer, but Axplorer is designed to run on a Mac Pro, making the technology much more accessible.

The goal is to bring the power of PatternBoost — which helped solve the Turán four-cycles problem — to any researcher who can install the software on their own machine. Axiom says the tool can also match or improve on the best-known results for other graph theory problems.

The company sees Axplorer as part of a wider push, including DARPA’s expMath initiative, to encourage the use of AI in mathematical research. Charton and Axiom CEO Carina Hong argue that math is not only about solving existing problems, but also about exploring new ideas and discovering patterns that can open entirely new areas of research.

Unlike some recent headlines about LLMs solving old puzzle-style problems, Charton says Axplorer is meant for deeper, more challenging questions that have been heavily studied. The company says the new tool is faster and more efficient than PatternBoost, and that it can reach the Turán result in just 2.5 hours on a single machine.

Axiom has made Axplorer open source on GitHub and hopes students and researchers will use it to generate examples, test ideas, and speed up discovery. Still, some mathematicians remain cautious, saying AI tools are promising but not a replacement for traditional methods just yet.