The Five Experts Hiding in One Sentence

The Simple Task

I had what I thought was a simple prompt. It said: take a number the client enters, find its square root, and use that as a base price for a set of properties. I read it and thought it would take five minutes. One sentence. One idea. How hard could it be.

Then I thought about it in a different way. What would it take to build this if I were hiring real people instead of typing into a chat box?

The Count

I would need a math expert for the logic. Not just "square root" but all the tricky cases. What do you do with a negative number? How many decimal places? What rounding rules apply? How does the math between input and price actually work?

I would need a senior engineer for the software design. Where does the input go? Where does the math run? How does the result get saved and shown? What happens when something goes wrong? What types are allowed?

I would need someone to manage both of those people. They would make sure the math formula and the code match each other. They would make sure every tricky case is covered in both places.

I would need a web designer for the screen. What does the input box look like? How is the price shown? What does a person see and do when they use it?

I would need backend engineers for the server work. API endpoints. Login checks. Rate limits. Database tables. Putting it all online.

Five different kinds of experts. For one sentence.

What This Told Me

I was expecting one sentence to replace all five people. I typed "do this thing" and expected the model to be a math expert, an architect, a project manager, a designer, and a backend engineer all at once. I never said which role to play, or when, or how.

The model can act like a human expert for a specific role at a specific step. It can be a Python engineer. An electrical engineer. A sales strategist. But only if I tell it to. I have to say which expert it is, for which part of the work, and with which limits.

When I did not assign a role, the model tried to be everyone at once. That is when the output fell apart.

The Missing Role

If something comes out broken or incomplete, I missed a role. I did not give the right job to the right step. I left a gap, and the model filled it with its best guess. That guess is not the expert I needed. It is just a general-purpose stand-in.

Now I count the roles before I write a single word. I ask myself: how many different experts would this need in the real world? Then I make sure every one of those experts appears in the prompt. One at a time. In the right order. With the right limits.

The sentence was never simple. I just was not looking closely enough.