Free AI Tool Puts Math Discovery Power on Your Laptop

Solving the world's hardest math problems just got a lot easier for researchers everywhere.

Axiom Math, a startup in Palo Alto, California, released Axplorer, a free AI tool that helps mathematicians discover new patterns and solve puzzles that have stumped experts for decades. The breakthrough? It runs on a regular Mac Pro instead of requiring access to massive supercomputers that only big tech companies can afford.

The tool's predecessor, PatternBoost, made headlines last year when it cracked the Turán four-cycles problem, a famous puzzle in graph theory that helps us understand complex networks like social media connections and supply chains. But that success came at a cost: it ran for three weeks on tens of thousands of machines at Meta.

Axplorer achieved the same result in just 2.5 hours on a single computer. François Charton, a research scientist at Axiom who co-developed the original tool, calls the old approach "embarrassing brute force."

The timing couldn't be better. The US Defense Advanced Research Projects Agency recently launched expMath, an initiative encouraging mathematicians to develop and use AI tools for discovery.

Other AI tools like Google DeepMind's AlphaEvolve have shown promise, but they remain locked behind corporate walls. "You have to go and ask the DeepMind guy to type in your problem for you," says Charton.

The Ripple Effect

Breakthroughs in mathematics create waves across all of technology. New mathematical discoveries are crucial for advances in computer science, from building better AI systems to improving internet security.

Unlike AI chatbots that solve existing problems by rehashing past solutions, Axplorer helps mathematicians explore uncharted territory. You give it an example, it generates similar patterns, and you select the most interesting ones. The tool learns from your choices and keeps generating better suggestions.

Axiom Math founder and CEO Carina Hong, herself a mathematician, designed the tool to be approachable. While some AI tools require users to train their own neural networks, Axplorer walks researchers through the process step by step.

The code is open source and available on GitHub for students and researchers worldwide. Hong hopes democratizing access to powerful discovery tools will spark new insights that nobody has ever had before, opening up entirely new branches of mathematics.

Sometimes the most important breakthroughs come from patterns that were always there, just waiting for someone to spot them.