The One Skill AI Can't Replace: Knowing What to Build

Answers got cheap. Questions did not.

For most of computing history the bottleneck was building. You had an idea, and then you spent six weeks making it exist. Ideas were plentiful and the making was slow.

That has flipped. The making is fast now. So the scarce thing is no longer the answer, it is knowing which question was worth asking in the first place. This is the centre ring we described in why kids should still learn to code, and it is the one AI has not touched.

Picture a field covered in solutions, thousands of them, all available instantly. That field is not valuable any more, because everybody has the same field. The valuable move is walking into it and picking the one that matters.

Where good ideas actually come from

Adults tend to look for ideas by brainstorming. Children do something better without being taught: they complain.

Every complaint a child makes is a problem statement in disguise. The homework list keeps getting lost. Nobody remembers whose turn it is to feed the cat. The cricket scores take too long to add up. Each one of those is a project.

So the most useful thing you can do is not to suggest a project. It is to notice the complaint out loud. "That does sound annoying. Could you build something for it?"

This is also why we resist handing children a list of approved project ideas. A project somebody else chose teaches typing. A project the child chose teaches everything else, including the stubbornness to finish it.

Every complaint a child makes is a problem statement wearing a disguise.

Small, useless, wonderful projects

Please let the projects be useless. A program that prints your chores in a slightly nicer way is not going to change the world, and it is not supposed to.

What it does is give a child the full loop in one sitting: I noticed something, I made a thing, the thing exists now. Three lines is enough to close that loop, and the skills to check it are the ones we care most about.

chores = ["dishes", "homework", "feed the cat"]

for job in chores:
    print("[ ] " + job)

Then ask the only question that matters

Once it runs, ask your child: who is this for?

That question does more work than any tutorial. If the answer is "me", ask whether they would actually use it. If the answer is "Amma", ask whether Amma can understand the output without help.

Suddenly the child is thinking about a real human being who is not them. They start adding numbers to the list, or a tick when something is done, or a nicer word than "homework". None of that is coding. All of it is the job.

This is the thread that runs through everything we teach. A child who can read code but never asks who it is for ends up as a very fast typist. A child who asks that question first will find the code they need, or direct a tool to produce it, which is the subject of your child won't compete with AI.

What you can do this week

You are looking for one complaint and one small build. That is the whole assignment.

• ✓Write down three things your child complained about this week. Do not comment on them.
• ✓Read the list back to them and ask which one they could build something for.
• ✓Whatever they build, make them show it to the person it was built for.