A prompt is not a command. It is a lasso.

Ten windows open. 46 agents. At the same time.

I am not typing. I am speaking. The screen fills up anyway, line by line, as if someone else were driving it.

Next to me sits an intern, sixteen years old. He says nothing. He does not know where to look.

Then comes the question. The one everyone asks who sees this for the first time.

„Does AI still need the human at all?"

I do not answer with a technical term. I reach across the table. There is a glass standing there, and a cable next to it. I take the cable and throw it over the glass like a lasso.

The first throw rarely lands straight away

This cable around the glass, that is a prompt. Not a click that lands perfectly straight away. The first throw does not always miss, but usually the loop sits far too loosely. The glass could run off at any moment, like an animal you have only half caught.

And now comes the part most people skip: with every further prompt you pull the loop tighter. And tighter. Until in the end exactly what you wanted sits inside it. Not some result, but your idea as a finished product.

Anyone who believes AI is a vending machine, where you drop an instruction in the top and the perfect result falls out the bottom, has thrown the lasso once and wondered why the glass is not in it yet.

Where the line „AI makes you stupid" comes from

There is a lot of half-knowledge about AI, and one of the most stubborn lines is: AI makes you stupid. I understand where it comes from. But it is only true under one condition: when you say a single thing and immediately take the first thing that comes back.

That is like kicking the ball once, missing, and concluding that football is not for you.

AI does not make you stupid. It only rewards the person willing to make the second and third and fourth throw. The first prompt gets you close. The work, the real craft, is the tightening.

Why I know this from sport

I come from sport, and I see the same pattern here as in every training hall. Most people want the perfect kick straight away. They see a pro who makes it all look easy, and they have no idea how much technique sits behind it. How many repetitions. How many throws that missed.

With prompting it is identical. The ease you see in someone who works fluently with AI is not talent. It is muscle memory. The more often you work with the models, the faster the throw lands, the fewer rounds you need until the loop is tight.

And there is no way around it. You do not learn boxing by watching videos. You do not learn basketball from the sofa. You show up and put in the repetitions. Working with AI is exactly the same, only nobody calls it that.

What this means for your work

Very practically, for Monday morning:

Do not expect your first prompt to land. Plan the second and third from the outset. Not as a sign that something is going wrong, but as the normal path to the result.

If the result is not right, that is not a dead end. It is a throw that was still too wide. Tell the AI what is missing, what is too much, what you actually meant, and pull the loop tighter.

And do not stop after the first attempt. That is exactly where most people give up, and that is exactly where the whole difference lies.

The fastest way to pull the loop tighter

And now comes the part that turns a nice picture into a real tool. Because the real question is this: how do I pull the loop tighter faster, without prompting my way through ten rounds?

The answer is uncomfortably simple, and hardly anyone arrives at it on their own: do not pull alone. Let it help you.

Because the problem with the tightening is often you yourself. In your head there is a clear picture, but you only say a fraction of it out loud. The AI guesses the rest, and that is where the throw goes too wide. Instead of guessing and correcting, turn the direction around. Give the AI the task of questioning you.

One sentence is enough:

„Before you start, ask me three to five targeted questions to understand me better."

And often something different happens. The AI asks you about things that were in your head and never made it onto the screen. Who is this for? How long? What tone? What is the actual goal? Not every question lands, and it does not have to. But every answer you give pulls the loop a little tighter, usually in one step instead of five blind attempts.

A monologue becomes a conversation. And a conversation lands faster than any set of instructions.

The real point

When the intern was later asked what he was taking away, he said something more mature than I would have said at 16. Not „the AI needs to get better." But:

„I need to work on myself so that I can even operate it."

That is the whole reframe in one sentence. The bottleneck is not the tool. The bottleneck is the willingness to make the second throw. And that is good news, because willingness and technique can be trained. You need no talent for it. You need only the repetitions, and the patience to pull the loop tighter, until your glass sits inside it.

A prompt is not a command. It is a lasso. And you only learn to throw by throwing.