If the LLMs run by these people are turning up real bugs then their confidence in touching kernel code seems pretty earned, imo.