In terms of making an LLM faster but not in terms of meta-cognition. System 1 thinking as defined by Kahneman doesn’t have 100000x more compute than System 2, it is actually the opposite. That completely contradicts your claim