By Matteo Wong
Publication Date: 2026-02-24 12:55:00
In recent months, several researchers have begun making the same provocative claim: They used generative AI tools to solve a previously unanswered mathematical problem.
The most extreme promises—AI-powered solutions to some of the toughest problems in mathematics—may well turn out to be empty hype. But a number of AI-written solutions, albeit for far less lauded problems, have proven successful. These were answers to a series of the Erdős Problems – more than 1,000 mathematical questions by Hungarian mathematician Paul Erdős – written using generative AI models, including ChatGPT. OpenAI quickly scored a victory: “GPT-5.2 Pro to solve another open Erdős problem,” OpenAI President Greg Brockman wrote on X in January. “This is going to be a wild year for mathematical and scientific progress!” (OpenAI and The Atlantic have a corporate partnership.)
Much of the excitement surrounding the news comes from the juror of this AI-written evidence: Terence Tao,…
