x
Goal-directedness via explanations — AI Alignment Forum