Stages of Prompt Engineering

  1. AI Is Worthless: Prompts are too broad, vague and open-ended. “Fix this.”
  2. I Could Do This Myself: Prompts are too narrow, specific and step-by-step. “Update function A to use regex B to check if parameter C matches pattern D.”
  3. Give It Context: Prompts have outlines, context and objectives. “Here are the files I’m working with. Here’s what I’m trying to do.”
  4. Make It Think: Prompts encourage thinking and explanation. “Explain how feature X works. Then modify it to support Y. Verify your work by testing that Z is true.”
  5. What’s next?

This is my journey so far.

If I had to describe my current phase it’s “Wow, this costs a lot!” You know, I feel like I’m finally getting useful output from my AI code assistants, but that extra context and thinking is costing more, and so I’m asking myself, “How can I optimize this?”

Denial → Condescension → Struggle → Breakthrough → Context → Cognition → Optimization

That’s my model for now. We’ll see where it takes me next.