I know the info is there. Somewhere in the pricing pages, docs, or API notes. But for something that can seriously change what you pay in production, it is weirdly under-explained. expeciely for other providers than openai which they do have decent explainer here - https://developers.openai.com/api/docs/guides/prompt-caching So basicly: two prompts can look almost identical, but one can be much cheaper to run just because it is ordered better. Put the changing parts too early, like the...