It's not just this forum and have heard the more nuanced reporting.

So they had a near future good narrative where LLM was going to be a key tool is a long awaited mathematical breakthrough, but they heard about the mathematician effort and felt the need to steal the credit by somehow beating them to the punch.

In the process, they may have effectively stolen the work in progress, though that is admittedly may be less likely than other concerns.

The output was a sloppy mess that usually wouldn't have been considered ready for review, but they went with it anyway to get it out before the more considered work could complete. In pursuit of the narrative that AI alone could out speed and LLM augmented human effort, by skipping the quality that a human supervisor would have added.

They announced victory long before a human could verify the proof. Those working on that believe it likely to verify, but haven't completed that work. Again, very premature.