Károly Zsolnai-Fehér explains that an unreleased Claude model did not solve the Riemann hypothesis, but it did improve a related mathematical bound beyond the previous human record. He reports that mathematicians found the result surprising and significant while keeping the larger unsolved problem in perspective.
The breakthrough followed roughly 650 unsuccessful attempts. A non-mathematician mainly encouraged the model to continue, while Claude explored wrong approaches, recovered from them and produced the first crucial result after about 37 minutes of uninterrupted work.
The technical paper includes an AI-generated explanation and a formalized proof that can be checked automatically. Zsolnai-Fehér notes that Claude itself initially treated the result as suspiciously strong, while clarifying that its apparent surprise reflects learned language patterns rather than a human emotional state.
Watch on YouTube