Zachary Robertson comments on Sufficiently Advanced Language Models Can Do Reinforcement Learning