I thought fine-tuning allows you do "inject the prompt" only once. https://platform.openai.com/docs/guides/fine-tuning

Ah, the rub is that a fine-tuned Davinci costs 6x a standard Davinci. Interesting. https://openai.com/api/pricing/