Failed Adventures in Delay — LessWrong