Three different papers pushed to ArXiv proving same 1986 conjecture; all use AI

Three independent papers posted to ArXiv within days prove a 1986 conjecture, with all three acknowledging extensive use of ChatGPT. The papers include one posted Friday (arXiv:2608.13553) and two posted today (arXiv:2608.14507 and arXiv:2608.14438), raising questions about priority and credit assignment in AI-assisted research.

Today two different independent papers were simultaneously posted that prove a conjecture from 1986, both acknowledging extensive use of ChatGPT to get there. And both acknowledging that they had this independent of another proof of the same result posted on Friday. Steve McCormick on X: "Today two different independent papers were simultaneously posted that prove a conjecture from 1986, both acknowledging extensive use of ChatGPT to get there. And both acknowledging that they had this independent of another proof of the same result posted on Friday." - The timing is astounding, and makes me more convinced that AI is good at finding proofs if all the right things align in the literature. It also raises a serious question about priority and the assignment of credit for a result, I think.Oh right, the links: Friday's paper: arxiv.org/abs/2608.13553 https://arxiv.org/abs/2608.13553 One of today's: arxiv.org/abs/2608.14507 https://arxiv.org/abs/2608.14507 The other: arxiv.org/abs/2608.14438 https://arxiv.org/abs/2608.14438 Also, hey I just realised I know one of the authors 😅 I met Gioacchino earlier this year. - The same happened in physics, quantum information theory, a few weeks ago. A distinguished problem was solved by 4 different groups, all using AI, within a few days. Werner state un-distillability. We were one of the groups. I post the corresponding entry on vibemathed, although