Methodology of unbounded analysis — LessWrong