High Variance Productivity Advice — LessWrong