The Apprentice Experiment — LessWrong