Physicists Giorgio Parisi and Francesco Zamponi of the University of Rome La Sapienza used Claude’s AI model to solve a multi-year mathematical puzzle in the theory of conspiration. The problem concerned the relationship between two parameters, which always add up to one when describing systems that suddenly become rigid. The researchers hoped that a fresh look at AI would help uncover a new understanding of equations hidden from human experts.

Claude’s initial proof was flawed, but the underlying approach turned out to be the right one. Parisi and Zamponi developed the basic assumptions of the model and came to a surprisingly simple solution, setting out their reasoning in a new article. According to the scientists, the answer was "right in front of them," but they were looking for something deeper and missed a conceptually simple case.

The growing use of AI in mathematics is generating mixed feelings among experts: models are effective in finding patterns, but do not necessarily generate fundamentally new ideas. The mathematician Will Soyin has noted that AI helps to detect connections that people may not have noticed, but testing and interpreting results still requires human expertise. This case demonstrates the potential of AI as a "co-author" tool, not a substitute for the researcher.