#Anthropic released an unreleased research version of Claude that tried to solve the Riemann Hypothesis — one of the 7 Millennium Prize Problems worth $1 million. It wasn’t solved, but the result was impressive: it increased the lower bound of zeros of the Riemann zeta function from 41.6% to 67.2% in just two work sessions.
160+ Years Unsolved
67.2% New Lower Bound
2 Session Work
0 Human Intervention
Setup: The Impossible Dream
Anthropic put Claude to the test to try to crack the Riemann Hypothesis — a conjecture by Bernhard Riemann (1859) that has remained unproven to this day. This problem is not just an ordinary "math puzzle." It is one of the most important unsolved problems in pure mathematics, with major implications for cryptography, prime number distribution, and even network security.
But realistically, everyone knew Claude would fail. This was just an experiment to see how AI handles impossible-scale problems.
Twist: Unexpected Synthesis
Now here the plot gets interesting. While attempting RH, Claude found a way to combine two already-published mathematical papers that had never previously been combined. Not brute force, not random guessing — this is a creative connection between existing knowledge.
What does this combination produce? An improvement in the lower bound for the fraction of zeros that satisfy the Riemann Hypothesis. From 41.6% to 67.2%.
Translation: Previously, the math community only believed that at least 41.6% of zeros "behave" according to RH. Now it has been proven that it is 67.2%. Progress verified by human mathematicians.
