[cainxinth]

It’s a chain of thought prompt. You literally just tell the LLM to model solve a problem or answer a question by breaking it down into smaller, manageable steps (“think step by step”).