My 2-year-old knows what he likes. Since before his first birthday it’s been construction vehicles, cars and trucks, anything with an engine that goes. He watches videos of trucks and trains, and any game in GCompris with a vehicle in it — there are a lot — becomes a fast favorite within about thirty seconds.
So when we told him he could make his own game, with any kind of vehicle he wanted, the request wrote itself:
“Make a red car game!”
I explained it the only way that made sense to a two-year-old:
“You type what you want here, and the computer helper will try to make it for you.”
His eyes went wide. “Make a red car game! Tell the computer to make a red car game RIGHT NOW.”
Why I’m Not Teaching My Kid to Code
I’ve spent a decade in software. I’ve written more functions, loops, and algorithms than I’d care to count, and I’m still fairly sure that teaching my toddler to write a proper if…then would be a waste of both our time. Not because he couldn’t get it — he’s already fluent in “If you finish your meatball, then you can have more blueberries.” It’s that the world I learned to code in isn’t the world he’ll grow up in.
That world is gone. By the time he’s working age, learning Python will feel a bit like learning plumbing. Does knowing how the water gets to the faucet, and how to fix the pipe when it bursts, make you more capable and self-sufficient? Absolutely. But 99% of the time, what you actually care about is what you’re doing with the water.
Today’s language models are already good. By the time my son is grown, they’ll be woven so far into ordinary objects that we’ll stop noticing them at all — the watch on your wrist, the television on your wall. (For future readers: a wristwatch was a little device we wore to tell the time, and a television was… you know what, never mind.)
If you want your kids to do well with tomorrow’s technology, don’t teach them to code.
Teach them to build.
My Toddler, the Prompt Engineer
He knows his letters, but keyboard skills are still a work in progress, so we helped him type it in.
make a browser based red car game
VS Code lit up — green lines filling the screen, things happening. He watched like it was a fireworks show. “Can we play it now? Is the computer done yet?”
Eight minutes later we opened the browser, and there it was: a little red car trundling down a sunny road, jumping cones and obstacles every time he hit the space bar. Naturally, this was only the beginning.
“Can we make it a digger? Make a firetruck game. Make an AMBULANCE game!”
(He’s two. The man contains multitudes, all of them with sirens.)
My little prompt engineer has been busy ever since. You can see everything we’ve made so far at madladstudios.com .
Forget the Syntax. Here’s What Actually Mattered.
He thinks we made a little red car game that afternoon. What we actually did was bigger than that.
For most of his life, a computer had been a place where other people’s stuff appeared — YouTube videos of excavators, educational games somebody else dreamed up. After that afternoon, it became something he could hand an idea to and get something back. He has an idea now and simply assumes he can build it, the way he assumes blocks are for stacking.
The deeper thing he’s practicing is the kind of programming that will actually matter in his lifetime: telling a machine what you want in plain language.
I watched him discover that specificity matters. “Make a fire truck game” left a lot up to the model’s imagination. “Make a fire truck game where it sprays water and puts out fires” got him exactly the game in his head. When the red car’s stubby little jumps kept landing it on top of the cones instead of over them, “make it easier for kids” bought him bigger, floatier jumps and a lot less frustration.
He’s learning that making something is a conversation. You say what you want, you look at what you got, you adjust, you go again. It’s the loop I run all day as a developer, just without the syntax standing between the idea and the thing.
The tools have moved. When any kid who can describe an idea can also make it real, the question of what we ought to be teaching them is wide open. I find I care much less about Python than about whether my son can think in systems and say clearly what he means: break a big idea into smaller ones, look hard at the result, imagine boldly, describe precisely, and try again. Those skills were useful long before any of this, and they’ll still be useful long after today’s tools look quaint.
Your Toddler Can Ship Software Today
Want to try this with your own kids?
You need a computer, an AI assistant (we use the Cline extension in VS Code), and a kid with ideas. Setup takes a few minutes; after that, the only real limits are naptime and snack breaks. (Want the specifics? Comment and I’ll walk you through it.)
Start with whatever your kid already loves. Let them describe it in their own words, and type exactly what they say — toddler grammar and all. Then watch their face when the thing appears. The moment they realize their ideas can turn into real things is the moment everything shifts.
My son still asks for the red car game. But now he also asks, “What should we make today?”
I can’t think of a better question to teach a kid to ask.
Have you built anything with AI and your kids? I’d love to hear what you discovered — reply and tell me. My toddler and I read every response, though he’s mainly interested in whether you made any vehicle games.