x
Evidence of triple layer processing in LLMs: hidden thought behind the chain of thought. — LessWrong