The weights ended up edited. It does feel a bit like the long of reasoning that claims LLMs don’t really think, while they’re doing novel mathematical research.