Willison: "My problem is when this experiment starts wasting the time of people in the real world who had nothing to do with the experiment."

Agreed! And I'll add: Pretty much anything you do with AI in the real is going to run afoul of that principle. When I have to parse a work email that someone delegated to Copilot, they're wasting my time by offloading their part in the work of communication not just onto an LLM, but also onto me. "Don't waste the time of people who didn't consent to AI interaction" is a sound principle, but "AI moderates" aren't applying it consistently.