Idealized Factored Cognition — AI Alignment Forum