Ask HN: Is reasoning just some iterative context refinement?

context = "you are a useful model. you evaluate your own output for N steps after the initial user input"

prompt = "here comes a prompt by the user"

context += prompt

for _ in range(N): context += evaluate_llm(context)