Yesterday, thousands of bright undergrads sat for the William Lowell Putnam Exam, the most prestigious university-level competition in the world. The median score is often 0. We ask: to what extent can an automated system solve the 2025 Putnam with formally verified proofs? At Axiom, we answered this question with our systems. Two minutes before the time Putnam exam concluded, AxiomProver solved 8/12 problems in Lean autonomously. After the model runs past the time limit of human exam, AxiomProver solved 9/12. In 2024, a score of 90 would've been #1 of ~4000. In 2023, #2; in 2022, #3, and in 2021, #4, all Putnam Fellow level performance. We share this result out of genuine curiosity and as a step in our mission to develop tools in service of the mathematics community. To everyone who took the Putnam yesterday: we salute you. We look forward to celebrating the human achievements when they are announced. We remember the days of camper chairs four months ago. Yet it is still day zero for Axiom. We are at the starting point of understanding what AI for mathematics and this fundamental technology of verifiable reasoning can become.
About us
Building quantitative super-intelligence. We are hiring: careers@axiommath.ai
- Website
-
https://axiommath.ai
External link for Axiom
- Industry
- Technology, Information and Internet
- Company size
- 11-50 employees
- Type
- Privately Held
Employees at Axiom
Updates
-
Thank you Ben Cohen at The Wall Street Journal for covering Axiom’s story. AI for mathematical discovery on today’s front page.
🔗: https://on.wsj.com/49RApE6 Ken Ono’s career as one of the world’s most prominent mathematicians has taken him to places that he never could have fathomed. The renowned University of Virginia professor regularly ventures far beyond campus, bringing his formulas everywhere from Hollywood to the Olympics. He’s the only number theorist who has ever been the star of a beer commercial. And for his next act, this renaissance man of math is doing something improbable even by his standards. He’s leaving his tenured job to work for a 24-year-old. Not long ago, the idea of joining an AI startup in Silicon Valley would have sounded absurd to him. In fact, before it reoriented his career and uprooted his entire life, he considered himself a skeptic of artificial intelligence. Until recently, he began talks by poking fun at the hype around the nascent technology. “My name is Ken Ono, and I am NI,” he said. “Naturally intelligent.” Now he’s the unlikeliest employee of a startup that hopes to revolutionize math with AI. At the age of 57, Ono is taking an extended leave from academia with no plans to return. He’s jumping to a company founded by one of his former students, Carina Hong, who has the sort of dazzling résumé that would make AI feel insecure.
-
-
Axiom reposted this
Exclusive: A legendary math professor is leaving academia—and joining an AI startup run by a 24-year-old.
-
Axiom welcomes Professor Ken Ono, former STEM Advisor to the Provost at University of Virginia, as Founding Mathematician. For hundreds of years, the path of a pure mathematician was clear: prove theorems, publish papers, train the next generation. That path still matters. It always will. But a new one has opened. One where mathematicians don't just discover -- they teach machines to discover alongside them. There is beauty in growing your student into your collaborator. Hardy to Ramanujan. Professor Ono to our founder Carina Hong. And now, a group of brightest mathematicians at Axiom to AI. Join Axiom to write the next chapter of mathematical discoveries. And read more from this morning's The Wall Street Journal (link in comments)
-
Incredibly thoughtful discussion where Carina Hong lays out why AI for math and AI for science are going to be the next frontier in AI that has world-changing impact — thank you Ksenia Se for bringing this episode to life!
Math is the root of all hard science. If we can build a system that survives mathematics, we get a blueprint for intelligence that scales far beyond theorem-proving. But it’s brutally hard to build a reliable AI mathematician that can actually move all sciences forward. That’s why I was so eager to talk to Carina Hong and see how she and Axiom Math are tackling this problem. Carina lays out three pillars of an AI mathematician: 1. The prover system A system that doesn’t stop at an answer. It produces full proofs, with all intermediate steps and the reasoning process exposed, so every step can be checked and verified. 2. The knowledge base A living library that tracks what’s known, what’s missing, and what just became true. As the prover generates proofs, the system adds new nodes – theorems, definitions, concepts – into a graph-structured mathematical knowledge base. 3. The conjecture system A model that proposes interesting new math questions. It challenges the prover, watches what it gets right and wrong, and turns that into a self-improving loop where both sides get sharper over time. On top of this, you add auto-formalization – the ability to turn natural language math into Lean and back. That bridge between informal intuition and formal proof is where new mathematical knowledge can actually emerge, be shared, and be reused. And this is only part of the conversation: Carina explains the difference between AGI and superintelligence in a very clean, concrete way, talks about why AI for math is the algorithmic pillar of AI for science, how to benchmark “theory building” instead of just Olympiad problems, and how she thinks about real applications – from formal verification in industry to discovering new structures in pure math. That was a very insightful episode for me, and I strongly encourage you to watch it https://lnkd.in/eXSCVswV or read the transcript (better to watch it, though – Carina’s clarity really comes through)
-
Axiom reposted this
Congratulations to Carina Hong, Founder and CEO of Axiom, on being named to the Forbes 30 Under 30 AI 2026 list. Axiom is building toward a new era of model reasoning starting with an AI mathematician — systems that can solve complex problems, generate proofs and validate their own work. Each step compounds making the next one stronger. We’re proud to support Axiom on its mission to raise the standard for what AI systems can reliably solve at scale. https://lnkd.in/eqJfPvXY
-
-
Axiom reposted this
Erdos problem 481
-
-
Axiom reposted this
Over the last weekend, AI systems developed by Harmonic, Logical Intelligence, and Axiom have made progress on formal proofs for several longstanding mathematical problems. These proofs, generated in machine-checked Lean code, demonstrate what's becoming possible at the intersection of AI and formal mathematics. The researchers and teams pushing these boundaries are doing extraordinary work - congratulations to everyone involved! 🎉 𝐄𝐱𝐩𝐥𝐨𝐫𝐞 𝐋𝐞𝐚𝐧'𝐬 𝐘𝟑 𝐫𝐨𝐚𝐝𝐦𝐚𝐩: https://lnkd.in/gzzghAHq #LeanLang #LeanProver #AI #FormalMathematics
-
-
Introducing Axiom's discovery team: Alberto Alfarano Alberto's journey runs from the math Olympiads → quant trading → using transformer to crack a 130-year-old conjecture. He found new Lyapunov functions, while no general algorithm previously existed. AI for math discovery redefines the boundary of human minds. Come join us at Axiom to solve big, open problems!
-
Introducing Axiom’s discovery team, led by François Charton: We build models that will create novel constructions, map problems into solutions and intuitions, and learn the structure of entire mathematical worlds. Built to tackle hard open problems, one at a time. Read more about this effort in our recent piece (link in comments)