[tedsanders]

In my own testing of single-turn instructions with GPT-4, I got basically the same performance putting it in a single system message or single user message. Possible that this changes for future models, though.