After 25 years in venture, the founders who stay with me longest are the ones whose love of their subject turns out to be their sharpest strategic insight. Carina Hong is that founder. Her thesis: math is the starting point of reasoning, and therefore the foundation of AGI.
From that conviction she built a company that scored 120/120 on the Putnam exam, outperformed every major LLM on formal verification benchmarks, and is solving unsolved math problems autonomously. In our most recent episode of Founded & Funded, I sat down with Carina, founder of Axiom, and a few things that have stayed with me since we sat down are:
1) When she describes the relationship between math and code, it's not a product pitch. It's a mathematician describing something she finds genuinely beautiful. The commercial implication is almost a byproduct.
2) Google, DeepMind, OpenAI, and Meta all ran formal math research programs. Each deprioritized it when commercial pressure arrived. Carina's read: they started too early. She started at the right moment, when informal reasoning systems had advanced far enough that formal ones could benefit from them, and committed when everyone else was pivoting away.
3) When the field defaulted to unstructured math data, scraping textbooks and chain-of-thought labeling, she chose structured formal data. Less of it, but verified. Her argument: better sample efficiency, and scarcity puts you in contact with the hard problems faster. She was right.
4) She identified hardware verification as the first commercial market before it was obvious: cycles running 3–4x longer than design, 3–4x the headcount, and production-grade formal checking tools that can't scale without human intervention. When they tested Axiom Prover against real circuits, it verified things the standard tools could not.
Carina's read on where this field is going has been more precise than the field itself. That's what I keep learning every time I talk with her. This episode is worth a listen.
