In July 2025, AI models solved five of six problems at the International Mathematical Olympiad, sparking a shift in how mathematicians work.

The success drew figures like Terence Tao and Daniel Litt into experimentation. They found that large language models could formulate and prove conjectures with minimal human input, accelerating discovery from weeks to days.

Teams such as DeepMind’s AlphaEvolve used Gemini‑driven Python scripts and genetic algorithms to tackle 67 problems; 23 saw improved solutions, 36 matched the best known.

Mathematicians now view AI as a new research partner rather than a replaceable tool, reshaping practice and culture while keeping rigorous proof standards intact.