Sanity-checking in an age of hyperbole — LessWrong