Making your explicit reasoning trustworthy — LessWrong