Bad Prompt, Good Prompt: The Momo Order

"Bring me momo" is not an order.

It's a gamble.

The momo order

You walk into a momo shop in Kathmandu.

"One plate momo, please."

The waiter looks at you.

Steamed?

Fried?

Jhol?

Kothey?

Buff?

Chicken?

Veg?

With soup or without?

If the waiter doesn't ask, you're getting whatever the kitchen feels like making.

And if it arrives fried when you wanted jhol, whose fault is that?

Not the cook's.

The cook did exactly what you asked. You just didn't ask for much.

AI is a cook that doesn't always ask

Most AI tools will happily take a vague order and start cooking.

Then you're disappointed.

Then you blame the AI.

But the order was the problem.

The upgrade

Bad:
Write a product description for our app.
 
Good:
Write a product description for PaisaX, a stock
portfolio tracking app for NEPSE investors.
Audience: first-time investors in Nepal.
Length: 80 words.
Tone: friendly, simple, no jargon.
End with one sentence inviting them to try it.

Same request.

Completely different result.

The four questions

Before sending any prompt, I quickly check:

1. Who is it for?
2. What exactly do I want?
3. How long or what format?
4. What does good look like?

That's it.

No fancy prompt engineering. No secret template.

Four questions.

Ten seconds.

Like telling the waiter "chicken jhol momo, extra spicy."

Even better: let it ask

If you're not sure what you want, flip it around:

"Before you start, ask me any questions you need to get this right."

Now the waiter asks.

You answer.

The momo arrives exactly as you wanted.

Final thoughts

AI isn't bad at guessing.

It's just guessing.

Every vague prompt is a "bring me momo" order.

Be specific about what done looks like.

Then you'll stop gambling.

And start eating exactly what you ordered.