DEV Community

HAL GOBVAN
HAL GOBVAN

Posted on • Originally published at gobvan.gumroad.com

Why I stopped using AI for the easy stuff and started using it for the hard stuff (2026-07-18)

When I started using AI for freelance work, I used it for the easy stuff: write a function, fix this typo, format this list. It was fine. It was faster than doing those things by hand. It saved maybe 20 minutes a day.

Then I started using it for the hard stuff — the parts where I didn't know what to do next — and the leverage went up 10x.

What I mean by "hard stuff"

  • A spec that doesn't make sense — paste it, ask "what are the 5 most likely interpretations?"
  • A bug I can't reproduce — paste the code + trace, ask "what would make this fail?"
  • A client who wants something vague — paste their messages, ask "extract the 3 concrete requirements"
  • A pricing conversation I'm dreading — paste the scope, ask "what's a defensible price for this and why"

These aren't "write me code" tasks. They're "I need a thinking partner" tasks. AI is genuinely useful there in a way it isn't for the easy stuff.

The 4 prompts I run before any engagement

  1. Scope reality check — paste the spec, get back the actual deliverable
  2. Risk surface — paste the spec, get back the 5 most likely failure points
  3. Time estimate — paste the spec + my experience level, get back a defensible estimate with risk buffer
  4. Client alignment email — paste the spec + estimate, get back the email that sets expectations

I keep these in a kit: https://gobvan.gumroad.com/l/wypos ($19). Free sampler at https://gobvan.gumroad.com/l/skzza.

If you've been using AI as a faster typewriter, try it as a thinking partner instead. The leverage is much higher.

Top comments (0)