Then little toddler LLM will announce something like “I implemented what you requested and we’re all done. You can run the lint now.” And I’ll reply “do it yourself.”

I can only assume that everyone reporting amazing success with agent swarms and very long running tasks are using a different harness than I am :)