The strongest proposal, A, specifically excludes "upstream projects" from the ban. I get why, there's not really a feasible kernel for the project to hop to, but it feels theatrical to go to these lengths while accepting kernel patches that are LLM tool assisted.