This is mildly interesting: It looks as though current LLMs are becoming at least competitive with expert humans at solving Diophantine equations.

Here a new MO contributor has used GPT 5.4 to find solutions to a couple of Diophantine equations that were posted by Bogdan Grechuk in 2023, and had been unsolved since.

https://mathoverflow.net/a/509413/8217

https://mathoverflow.net/a/509415/8217