Philosophy as low-energy approximation — LessWrong