It should burn N + len(answer), because you have to re-cache the whole answer without the prompt stack.
Perhaps more persnickety, it pushes the LLM out of distribution - if it’s unnatural for it to write in plain language without the prompt stack, your prefix will be an unnatural conversation which can reduce intelligence in hard to measure ways, especially over long conversations.
Not saying don’t do it, clarity is perhaps worth the intelligence hit, but it’s not going to be a free lunch.
Perhaps more persnickety, it pushes the LLM out of distribution - if it’s unnatural for it to write in plain language without the prompt stack, your prefix will be an unnatural conversation which can reduce intelligence in hard to measure ways, especially over long conversations.
Not saying don’t do it, clarity is perhaps worth the intelligence hit, but it’s not going to be a free lunch.