Raising AI Kids: Issue 8

Building Games With Your Kids (And Forgetting To Write This Newsletter)

Date: March 2026

My son looked at me last week and said, "You spend all your time with your agents and not with us."

I was sitting at my desk working on this newsletter. About AI parenting. The irony didn't miss me.

So I closed the laptop. And we built a video game together instead.

My daughter had already been messing around with a prompt — something about dinosaurs, survival blocks, explosions. By the time I came over she'd written out a full game concept. Defend your base against creatures that get smarter each day. Place blocks to build walls. Press T for TNT. She'd never written a line of code in her life.

My son asked what we were doing. I said "let's find out." And I hit "go" on her prompt. The code landed. We opened it. The game loaded. And instead of reading through it and explaining what she'd done, I just... played it with her.

For eight hours.

We'd play for 30 seconds, hit a bug or a janky mechanic, and one of the kids would say "what if we change this" or "that's dumb, let me ask Claude to fix it." We'd go back to the prompt, tweak one sentence, generate fresh code. Sometimes it broke things. Sometimes it made things better. Mostly we'd just laugh and iterate again.

My daughter discovered that if you place blocks in the sky, they fall and stack — so you can build stairs to climb. She spent 20 minutes just figuring that out, trying different shapes, taller and taller. My son realized the enemy gets stronger each day, so he started planning ahead: "we need more blocks by day 10" became his whole strategy.

The game never shipped. It never got polished. Half the features were janky. But somewhere around 2 PM, both kids were sitting on the floor with me, actually thinking about game design — not because I was teaching them, but because they wanted the thing to work.

That's when I realized I wasn't writing the newsletter that day. I was living it.

The lesson wasn't theoretical anymore. It wasn't me explaining how AI is a tool for iteration, or how the best part of making something with your kid is the making, not the finished thing. It was my kids showing me that the moment where you break something and fix it is the moment you actually understand how it works.

We didn't finish the game. We probably never will. But Friday wasn't about finishing — it was about the fact that my 10-year-old is now comfortable talking to an AI, asking it to make a thing, playing the thing, deciding what's wrong with it, and asking it to change. Without me explaining anything. Without a tutorial. Just — "hey, let's build this together" — and her understanding that "build" included iteration, failure, and the weird joy of breaking your own creation because that's how you learn what it can do.

When I sat down to write this issue, I realized the best thing I could offer wasn't more parenting advice. It was permission to do exactly what I did: put work down, follow your kid's curiosity, and build something together that you have no idea how to finish.

If you want to try it this week, here's the exact prompt my daughter used. Paste it into Claude or ChatGPT with your kid sitting right there. Don't explain what code is. Don't warn them it might not work. Just hit go and play it together. Then let them tell you what to change.

Do Now: Build a Game With Your Kid This Week

Copy the prompt below. Where it says [INSERT YOUR NEW GAME DESCRIPTION HERE], replace it with whatever your kid wants — zombies, robots, dinosaurs, pizza monsters, anything. Paste the whole thing into Claude or ChatGPT. Hit generate.

You'll get a complete, playable game in one HTML file. To open it: save it to your desktop, then double-click it (or drag it into Chrome or Firefox). No server, no setup, no install. It just works.

Create a complete single-file browser-based 2D survival building game using Phaser 3 (via CDN) in one HTML file.

Game Core Concept:
[INSERT YOUR NEW GAME DESCRIPTION HERE]

Target Audience: Casual gamers, kids and parents playing together.

Technical Requirements:
- Use Phaser 3.x (latest via CDN)
- Canvas size: 800x600
- Support both keyboard (WASD + Arrow keys) and touch swipe movement
- Click/tap to interact (build, break, attack)
- Keep total assets under 500KB (use graphics and shapes, no heavy images)
- Must run smoothly at 60 FPS even on mobile

Core Mechanics to Include:
- Player movement with smooth controls
- Building system: Click to place blocks (fall realistically, stack on other blocks)
- Block limit of 50 (when placing the 51st, the oldest disappears)
- Click on existing blocks to break them
- Enemy AI: tries to path around, climb over, and slowly break blocks
- Attacking enemies by clicking directly on them
- Player health and enemy health with visual feedback
- Realistic death animations (scale down + spin + fade)
- Day system: enemies get stronger and bigger each day
- After day 50, spawn a second enemy
- TNT: Press T to place explosive TNT (5 second fuse)
- Pause with P key | R key resets to Day 1
- Minecraft-inspired pixel-art style visuals

UI Elements:
- Day counter, Block count (current/50), Player health, Enemy health
- Status messages for win, game over, and day reset

Make the code clean, well-commented, and easy to modify.

Then play it together. Something will be janky. Your kid will say "can we make the enemy slower" or "I want it to be nighttime." Go back to Claude together and ask. That iteration — breaking it, changing it, laughing at the weird bugs — is where the real learning is.

Don't overthink it. Just put it in front of your kid and say: "What should our game be about?"

Then get out of the way.

— A dad who should've been writing this newsletter