Operationalizing timelines — LessWrong