Build One

Make something small this afternoon. The point is not the thing.

What the tool has to do

Anything that builds a working page from a description and gives you a link you can send someone.

If you already use one, use that. If you do not, pick any of them and stop researching. The habits here transfer between tools; the buttons do not, and a page that named three products would be wrong within a year.

One distinction is worth being firm about. A tool that writes text about your idea is not this exercise. It has to run, because a thing that does not work does not argue back, and that is the entire value of building rather than planning.

What to pick

Small. Finishable in an afternoon.

Personal. About something you hold and the model does not. Your programme, your town, your family, the subject you are strangely expert in. The model knows the world in general. You know your world in particular, and this is the one build where that gap is the whole point.

Real. It runs, and somebody who is not you tries it.

Yours to decide. Pick something with no correct answer waiting. A tic-tac-toe game has one and the model already knows it. A page about your grandmother does not.

Before you open anything

Write the one sentence a stranger could use to decide whether this works.

Not "a page about the community garden". Something closer to "a visitor who has never been should leave knowing when it is open and who to ask for a plot".

If you cannot write that sentence, you are not ready to open the tool, and no amount of describing will supply it later. This is the same gate you will put in front of learners, run on yourself first.

Then build it

Describe what you want. Read what comes back against your sentence.

You will get something that does exactly what you asked and not at all what you meant. That moment is what this build is for. Sit in it for a second before you fix it, because it is the thing you are trying to be able to recognise in somebody else's work later.

The test

Hand it to a person and say nothing.

Do not explain, do not stand over them narrating. Watch where they stop. If they need you to explain it, it is not finished.

This costs nothing and it is the fastest evaluation instrument you have.

When you are done

Write one sentence: what did you have to change about your own request?

Not what the tool got wrong. What you had not said. That sentence is what you are taking to Build Two, and it is the thing that lets you hear a weak answer from somebody else.