What Happens If You Give Claude Code to Kids

Today I taught AI to a group of elementary school kids from our village. Without a doubt, the most fun class I’ve ever given.

We built a game together using Claude Code. The kids got to enter prompts themselves. I had prepared a small starting point: just a green square in the middle of a web page. From there, each child took turns prompting.

Within five to ten minutes, they were literally jumping up and down, eager to be next.

What fascinated me most as a programmer was what they dared to try. The second prompt was: “turn the green square into a sheep.” I thought to myself: this is never going to work. And then it just did.

The green square became a sheep. That sheep went to the barber. Then came strawberries as bullets, a scoreboard, all kinds of game rules. It just kept going.

Five kids, roughly 25 minutes, and they had built a complete game from scratch. I’ve put the result on GitHub so everyone can play it.

What struck me was the energy, the creativity, the curiosity. The complete absence of hesitation. No “is this even possible?” or “how should I do this?” Just trying things and moving forward.

As adults, we’ve learned to be careful. We know the limitations. We’ve been told “no” enough times to internalize it. These kids hadn’t. They just asked for what they wanted, and more often than not, they got it.

Honestly, it was also a bit mind-blowing from a technical perspective. This genuinely wasn’t possible three months ago. The models have improved so rapidly that even I, someone who works with these tools daily, was surprised by what they could handle.

I’m not sure what this means for the future of programming. But watching those kids, I’m reminded that the best way to discover what’s possible is to simply try, without the baggage of knowing what should be impossible.