> refuse to engage with breakthrough results

These aren't results yet. They're potential results, but to find out we'd have to get a human mathematician to sift through a metric shitload of LLM hallucinations to see if there's a nugget of truth somewhere in there.