The Starting Point for Reasoning
Joined September 2025
- Tweets588
- Following54
- Followers16.7K
- Likes556
Axiom retweeted
One thing I love about academia is the opportunity to work with the smartest folks on incredibly interesting stuff. It's the adult equivalent of working at a toy store. And the cherry on top are that these folks are also some of the nicest persons I've ever met! This week we soft launched a passion project of ours that was in stealth mode for too long. While still at a nascent stage, We hope #matheconlib will eventually lay the foundation for theoretical research in economics and adjacent disciplines in the new age of AI aided discovery. Thank you @skominers and @KenOno691 for leading this endeavour and to @chen_ruize3962, Ben Eltschig, @JujianZhang,
@bocowgill,
@PiotrDworczak,
@NikhGarg ,
@ZivHellman,
, and Daniel Lyng for the fun collaboration! See you at our next weekly.
Axiom retweeted
Every math teacher has written "show your work" in red pen. @CarinaLHong turned it into an output format.
The @axiommathai team built an AI for really hard math that writes its proofs as code a machine can check. Watch Carina's story 👇
Axiom retweeted
Thx for reading. I really like this theorem. I would have bet my life that there would be an elementary proof. But maybe not because the question becomes Eisenstein cohomology in disguise. I hope folks search and hopefully find a simple algebra proof.
I really like the AI declaration in this paper:
"This paper was written through human–AI collaboration ... The collaboration included literature search, numerical experimentation, and the proposal and rejection of candidate strategies before the proofs took shape. The authors set the direction, made the mathematical judgments, and take full responsibility for the correctness of the results. Theorems 1.1 and 1.2 and the essential argument leading to them are formalized unconditionally in Lean 4.35.0.."
The optimist in me says "This is the future".
The pessimist has a vision of AI as a God-like oracle, with the ability to answer any mathematics question we could pose, and possessing better taste and aesthetic judgement (and even the capacity to tune it for each person) than we could ever have.
The realist acknowledges that we don't yet know, and that we live in interesting times.
Axiom retweeted
First paper with the legend @KenOno691. As a fun data point, gpt-5.6-sol Ultra tried to solve this for 2 weeks without success.
arxiv.org/pdf/2610.19309
RT @CarinaLHong: Wonderful to watch this open-sourced econ library being brought to life by the Axiom team @chen_ruize3962, Ben Eltschig, @…
This quoted post is unavailable.
Axiom retweeted
A great day! So happy to share MathEconLib, bringing machine-checked proofs to economic theory. We’d love for folks to dive into the Lean 4 repo and build MathEcon! @leanprover @skominers @bocowgill @axiommathai
We are open sourcing MathEconLib and made it a lot more polished than before, in collaboration with leading economists!
Axiom Math co-founded the project and continues to support its development. We welcome other organizations’ support - please reach out to the devs about this!
Check out the brand new website: mathecon.dev
A math neolab many of you know and love 🥰
Axiom retweeted
Tomorrow I’m speaking in “AI and the Future of Mathematics” seminar, supported by Mathematics Münster.
An honor to follow Andreas Thom, who I understand drew 300 online viewers.
See link in the screenshot. @axiommathai @leanprover
🤖 Made with AI
Axiom retweeted
1/ Today I heard @CarinaLHong, founder of @axiommathai, speak at @DigEconLab on "AI That Can Prove It." Two ideas I keep thinking about 🧵
Axiom retweeted
I don't need to imagine. That's exactly what we do at @axiommathai
Axiom retweeted
Most of my followers have never met me, so let me introduce myself.
I got my math PhD from the University of Washington. I have three Masters degrees, about 20 papers, a few were published in math journals, most in ML conferences. I've worked in mathematical biology, scientific machine learning, numerical PDEs, sampling, diffusion models, formalization and lately AI. I've done four different Google internships. At the University of Washington I taught the first Lean class in the math department, back in 2024. I co-founded the UW Math AI Lab, where we have done lots of formalization, first by hand and nowadays with AI. I finished my PhD in June and joined Axiom Math to continue working on formalization. I am the creator and maintainer of Lean Pool, the largest maintained formalization library. My short term goal is to formalize ~all known math.
I live in the bay area. In my free time I like rock climbing, I'm proud to have done an outdoor V11! If you are in the bay area and would like to discuss formalization, math, or go climbing, DM me!
Axiom retweeted
That was quick !
hexagonmath.org/2610.00200v1
Axiom retweeted
Check out this preprint by Cureton and Thorner using AxiomProver: hexagonmath.org/2610.00200v1
github.com/AxiomMath/Dedekin…
Some fascinating developments following OpenAI's shock breakthrough proof of the Quasi-Riemann Hypothesis, which was released publicly on October 6:
1. Building on OpenAI's proof, a recent Hexagon preprint (released on October 12) by Cureton and Thorner establishes that ALL finite-order Hecke L-functions are zero-free to the right of 1 - 1/6000.
2. And now, today, an arXiv preprint by Frei, Somos and Widmer uses this result to prove unconditionally the 2007 Ellenberg–Venkatesh bounds on ℓ-torsion in class groups of number fields.
I was amused to see the AI declaration in the second paper: "No artificial intelligence tools were used in this work."
#1 of the Annals Bench
Axiom retweeted
Axiom Proved The Quasi Riemann Hypothesis (Generalized) For All Dedekind Zeta Functions
Important complexity consequence of new claimed quasireimann (hexagonmath.org/2610.00200v1): Valiant's conjecture VP != VNP over the complex numbers follows from the boolean-complexity lower bound NP not in P/poly. Burgisser proved this under GRH. (🧵1/4)
25M LOC of Lean by AxiomProver
The competition on LeanEval is fierce. 15 solutions in the last 14 hours. 6 months ago each such solution would be the largest ever made, completely earth-shattering. Now we get one per hour.
Axiom is still in the lead, by three problems. We have 25 million lines of code in our solutions repo, larger than all OpenAI's formalizations, and twice larger than Anthropic's FLT.
Axiom retweeted
AxiomProver strikes back
Following NEAR AI topping putnambench leaderboard, we now have claimed #1 on Lean Eval v1, the official active benchmark maintained by the Lean developers.
We now have a very capable and cheap formal verification agent, and will share how we use it to formally prove NEAR protocol and smart contracts soon.