S-Tier Prompt Writing

You asked for one thing. You got another.

You wanted a casual summary, it wrote you a formal essay. You asked about the causes, it gave you the consequences. Right ballpark, wrong seat.

And the temptation, right here, is to get annoyed. To decide the thing is broken, or dumb, or not worth the trouble.

Don't.

This exact moment happens between people a hundred times a day. You ask a coworker for the Tuesday numbers and they bring you Wednesday's. You're not done with them forever. You just go "ah, sorry, I meant Tuesday," and you move on. Nobody storms off.

A miss isn't a crisis. It's a misunderstanding. And misunderstandings get cleared up by talking, not quitting.

It's a peer. Not beneath you, not above you.

Here's the mindset that makes all of this work, and most beginners land on the wrong side of it.

They treat the AI as one of two things. Either something dumber than them, so every answer gets a skeptical eye-roll and nothing it says could possibly be worth keeping. Or something smarter than them, a kind of all-knowing oracle, so whatever it says must be true and they write it down without a second thought.

Both are traps.

The AI is neither below you nor above you. It's a peer. A sharp, fast, occasionally wrong colleague sitting across the table. Sometimes it's right and you're not. Sometimes you're right and it's not. You treat it the way you'd treat a smart coworker: you listen, you weigh it, and you say something back.

That third reaction is the whole lesson. Not dismissing. Not swallowing. Engaging.

So push back. And stay open while you do it.

Treating it as a peer cuts two ways, and you need both.

The first way: you're allowed to challenge it. If it says something and your gut says that's not right, say so. "Wait, I don't think that's true." "Are you sure about that? Walk me through it." You do not owe the AI agreement. It is not an authority. Calling it out when it's wrong is exactly what you'd do with a colleague who said something shaky, and it's exactly what you should do here.

The second way, and this is the one people forget: you might be the one who's wrong.

When you push back, the AI might fold instantly and fix it. But it might also explain why it said what it said, and that explanation might be... right. The goal was never to make the AI see things your way. It was to land on what's actually true. And the truth is usually sitting somewhere in the middle, a little to the left of where you started and a little to the right of where it did.

So push, but push to find the truth, not to win. A peer who only ever wants to be proven right is exhausting. Don't be that to your AI, and don't expect it to be that to you.

"No, that's wrong" gets you nowhere

Okay. The answer missed, you've decided it really did miss, and you want to fix it. What do you actually type?

Here's what most people type: "No, that's wrong." Or "That's not it." Or just "no."

And then they're surprised when the next answer misses too.

Think about why. You walk up to that coworker, they hand you Wednesday's numbers, and you say "no, wrong." That's it. No more words. What do they do? They guess again. Maybe Thursday this time. You haven't told them anything.

The fix is to point. Say what was off, and say where to go instead.

One specific sentence and it turns around. Notice you didn't start over. You didn't rewrite the original prompt. You just clarified, the way you'd clarify to a person who misheard you. Because that's all this is.

Sometimes the best fix is a blank page

Now, one honest exception, because I told you in the last lesson not to start over.

That advice holds when the answer is good and you just want more. Restarting then is a waste. You'd throw away everything you built.

But sometimes a conversation gets off on the wrong foot and never recovers. The first prompt was muddled, the AI ran the wrong direction, you tried to correct it, it half-corrected, you corrected the correction, and now the whole thread is a tangle of crossed wires. Three messages deep and it's more confused than when you started.

When that happens, don't keep yanking the knot tighter. Open a fresh chat. Start clean. Write the prompt the way you wish you'd written it the first time, now that you know what went wrong.

There's no shame in it. A fresh page isn't admitting defeat. It's choosing the faster road.

Knowing which situation you're in is its own small skill. Good answer, want more? Stay. Tangled mess from a bad start? Wipe the board.