Mathematicians Solve 19-Year-Old Number Theory Problem Following OpenAI Quasi-Riemann Claim

Mathematicians Frei, Somos, and Widmer released an arXiv preprint unconditionally proving the 2007 Ellenberg–Venkatesh bound for the ℓ-torsion of the class group of a number field, resolving a 19-year-old problem in algebraic number theory without relying on the Generalized Riemann Hypothesis. The sequence began on October 6 when OpenAI claimed its AI had generated a 199-page proof of the quasi-Riemann hypothesis, which number theorist Hector Pasten noted ruled out the Landau–Siegel zero, while Burgisser also proved Valiant's conjecture (VP != VNP over the complex number field). On October 12, mathematicians Cureton and Thorner published a paper on the Hexagon preprint platform applying the AI-polished tools to Hecke L-functions of finite order, establishing that no zeros exist to the right of "1-1/6000". An open-source mathematical effort subsequently expanded that zero-free boundary from 1/6000 to 1/32, with the mathematical agent AxiomProver generating a formal Lean theorem prover verification hosted on GitHub. "We declare that no artificial intelligence tools were used in this work." — Frei, Somos, and Widmer