Heuristics for Deciding What to Work On — LessWrong