Write prompts like specs
The difference between a vague wish and a clear instruction — and why it changes the output.
A vague prompt gets a vague guess. A clear prompt gets close to what you actually wanted. In real engineering this clarity is called a specification — a precise description of what should happen. You don’t need to write formal specs, but the closer your prompt gets, the better the result.
Vague in, vague out
Compare these two asks for the same feature:
❌ “Make the form better.”
✅ “On the signup form, add a password field below the email field. Require at least 8 characters, and show a red message under the field if it’s shorter.”
The first makes the AI guess what “better” means — it might change colors, add fields you don’t want, or restyle the whole thing. The second leaves little to chance. Same effort to type, wildly different results.
Give it the context it can’t see
The AI doesn’t automatically know your intentions, your constraints, or the parts of your project it isn’t looking at. Tell it. “We use Tailwind for styling,” “keep it consistent with the other buttons,” “don’t touch the database code” — each sentence removes a way for it to go wrong.
Which prompt is most likely to get you what you want on the first try?
Iterate — don’t expect one perfect shot
Even great prompts need follow-ups, and that’s normal. Treat it as a conversation: look at what you got, then refine. “Close, but make the button blue and move it above the total.” Small corrective nudges beat rewriting a giant prompt from scratch.
You asked for a feature and the result is 80% right. What's the most effective next move?
Try it for real
Feel the difference clarity makes:
- Think of a change you want in a real project.
- Write the vague version of the prompt first (e.g. “make the header nicer”) — and notice how many ways it could be interpreted.
- Now rewrite it naming where, what, the rules, and the desired result.
- Send the clear version. Compare it to what the vague one would have given you.