By skyforbes I've been treating ChatGPT (and Claude, Gemini, whatever) like it's Google on steroids. Just asking questions and hoping for good answers. But after accidentally stumbling into this technique, my results got like 10x better overnight.
Here's what I changed:
Instead of asking "how do I do X", I started having actual conversations. Like:
- I tell it what I'm stuck on
- It suggests something
- I say "hmm that doesn't quite work because…"
- It adapts
- We iterate back and forth
Sounds obvious right?? But most people (including me until recently) just fire off one prompt and either accept whatever comes back or give up.
The other thing – I started treating it like a coworker I'm brainstorming with instead of a magic answer machine. When it gives me something mid, I don't just regenerate… I tell it why it's mid. "This is too formal" or "can you make it sound less like a robot wrote it"
Also been experimenting with giving it a role first ("you're a senior developer reviewing my code" or "pretend you're my brutally honest editor") and that alone changed the vibe completely.
My actual workflow now:
- Set the context properly upfront (who I am, what I'm trying to do, what I've already tried)
- Ask for the thing
- When it responds, I treat it like a first draft and we refine together
- Sometimes I even ask "what questions should you ask me to give a better answer?" and that's chef's kiss
Anyway I'm curious – how do YOU actually interact with these things? Like your real day-to-day approach, not the perfect prompt engineering tutorial stuff. Do you one-shot it? Go back and forth? Use system prompts? Custom instructions?
Drop your methods below cause I feel like I'm still figuring this out 👇