Interestingly Simon Willison left off the part about "I revisited the problem this summer, along with my partial solutions, when the previous round of stunning proofs came out. Several hours of work with Fable simply convinced me it wasn't yet solvable and reinforced how hard of a problem it was." in his blog post about your comment: https://simonwillison.net/2026/Oct/7/jake-boggan/

Almost as if he didn't want to include the part about you putting your work efforts into the model with an open question as to it's proclivity for stealing mathematical research. I wonder why a pro-AI blogger would do that.

I, personally, am convinced that OpenAI and Anthropic, and probably most other LLM labs cannot be trusted not to launder the work of their users through their own systems. As they did with internet resources for their entire initial training set. There is no telling how many researchers did the same thing with the exact same problem only to have their leads laundered by these companies with a large budget and small morality.