A simple model of the Löbstacle — LessWrong