Affordance Widths — LessWrong