On Doing the Impossible — LessWrong