Predictable Outcome Payments — LessWrong