Distilling the Internal Model Principle — AI Alignment Forum