The Comfort Trap: Counting the Cost of Delay — LessWrong