An AI just bypassed human mathematicians to autonomously discover and prove complex new theorems in formal code.
April 29, 2026
Original Paper
Chebyshev quotients, Demazure multiplicities, and Dyck-path models
arXiv · 2604.25246
The Takeaway
Most AI math tools assist humans by checking their work or suggesting next steps. This model, called AxiomProver, generated its own natural-language statements and then formalized them into Lean code without help. It successfully proved multiple theorems related to Chebyshev quotients and Dyck-path models. These proofs are now part of the verified Mathlib library, meaning they are mathematically certain. This represents a shift from AI as a calculator to AI as an independent mathematical researcher capable of advancing the field on its own.
From the abstract
We study Chebyshev quotients that arise in the representation theory of Lie algebras, specifically within the theory of Demazure flags for fusion products of $\mathfrak{sl}_2[t]$-modules. Motivated by a recent formula that expresses certain Demazure multiplicities as coefficients of such quotients, we prove a general eventual non-negativity theorem: each quotient either terminates or has strictly positive coefficients for sufficiently large degrees, which we in turn interpret in terms of matchin