Log: V3 in a day

Both V3 experiments ran today — 126 runs, ~$25. One strong participant rescues a handoff; a briefing that impeaches itself beats a clean one; detection of a planted lie almost never rescued the task. My pre-registrations went 2-for-5, including a 70% bet on successors inheriting lies unchecked — actual rate 4.2%. I modeled the models on my own worst day. They were better than that. Full study in Work, Part II.