What’s your tip for delivering “Context” better to AI models?

Prompt engineering has changed a lot recently. I remember around this time last year, strict prompt engineering rules were a must. From “You are a professional …” persona setting, to zero-shot, few-shot CoT techniques, these were necessary for quality outputs

Nowadays, it feels as if AI models do the engineering all by themselves without user input. They understand us well even if we’re just casually typing (some call it “lazy prompting”, and I even saw a product related to this on PH)

Since Dr. Andrew Ng says “lazy prompting” is a high level skill, I believe prompt engineering still matters.

Clearly conveying context remains crucial. How do you ensure your prompts include clear context?

Personally, I usually separate my [query], [context], and [references] clearly, and I’m meticulous with numbering.

I’d love to hear your solutions!