🎮: https://night-shift.magician.dev
My youngest came to me and said,
“Let’s build a game, Dad.”
And in under two hours, we had a game built and published for others to enjoy. This blog post goes through the experience and points out some of the reasons that helped us pull it off.
It’s important to know that I try to spend one-on-one time with each of my kids doing whatever they want to do. Sometimes that’s just playing games. Sometimes it’s building something. This time, he wanted to build a game. It’s a request all of my kids have asked me to do at one point or another. They’ve seen me create games from a VR Jurassic Park adventure to small web games. So of course, they wanted to join in!
Before AI, this usually meant spending most of the time on scaffolding. Picking an editor. Finding assets. Wiring things together just to get to the point where you can even start thinking about the game itself. The lesson often ended up being that game development is hard, which is true, but it didn’t do much to spark imagination. Oftentimes, they’d lose interest before we could get to the game mechanics, or we’d wind up prototyping something in Scratch only for it to disappoint when they wanted to take it to the next level.
That part has changed. AI has made it easier, especially with low-code tooling that can lower the barrier to focusing on logic, game mechanics, etc, instead of scaffolding and keeping the kids engaged.
Lowering the Barrier Without Lowering the Bar
My youngest already had a clear direction. His older sister had instilled a love of FNaF-style “survive the night” games, and he wanted to build something in that vein, but make it his own.
We started on his phone using Gemini. Not coding yet, just breaking the idea down:
Once we had a solid description of the core components, we had Gemini generate a playable version directly in Canvas. It was surprisingly good. Good enough that we knew we were onto something.
That early win matters. Keeping his excitement high meant he’d be more likely to make it through some of the harder parts I knew were coming, and it kept the focus on ideas instead of tooling.
Iteration (and a Real Failure)
As we added features, we eventually asked it to build a settings menu for a custom night.
Instead, it overwrote the project with a blank game.
This was almost a catastrophe, but that led to a useful lesson… We’d hit the context window limit. He didn’t know that to him his game was just gone… No amount of prompting was going to recover that version. So we adjusted.
Knowing the basics of the game now we were able to reprompt a new chat and recreate it. Once we got to a stable point, we then moved the project into AI Studio, where we could manage the codebase more intentionally.
From there, the process became very familiar:
My son made most of those decisions. He’d play, find something that didn’t feel right, and we’d talk through whether it was worth fixing now or saving for later. When he wanted to add a feature, I’d ask if it was necessary for the game to work.
The conversations with him was the real value, he was so excited to add features but very pragmatic about not taking on too much at a time. It was pretty cool to be a part of.
Teaching the Agile Way (Without Calling It That)
At no point did we talk about “Agile,” but we practiced it the entire time:
Start with a core idea
Break it into small, achievable pieces
Build just enough to test
Iterate based on what actually happens
Ship when it’s playable, not perfect
This is the same way I approach professional work. The difference is that AI removed enough friction to let my nine-year-old stay engaged the whole time.
A good example of this was the custom night.
Me, not knowing about FNaF, had to ask him what things we should customize, what that meant to the game and if we needed to or not. His reasoning was sound.
“It’s in the main game dad (Sarcasm implied)… also it will give people a way to test the settings”
So we added it!
Shipping It
Once the game felt solid, he asked the best possible question:
“Hey dad, can we publish this so my friends can play it?”
Of course we could. So I pushed the project to GitHub and then began using Antigravity to add the bits we’d need to host it on GitHub Pages. Mainly, this was a workflow for publishing the built project to a branch, setting up a devcontainer so we could dev on any machine. I also took the opertunity to add some mobile optimizations we didn’t worry about in the core build.
And Night Shift was published! https://night-shift.magician.dev
After it was live, he tried it out on all of our devices :) so we had some bugs to fix from that and then he wanted to add some easter eggs (Tell me if you find them)
So yeah, from idea to published game took about two hours… That’s right, TWO HOURS!
TWO HOURS
Is the game perfect? No.
Does it have:
a gameplay loop
working interfaces for mobile and desktop
audio (some are directly from code!)
level progression
and a concept entirely driven by a nine-year-old
Yes… Yes it does and it’s out there ready to be played by his friends.
More importantly, he’ll always know that we built something together and shipped it. That matters far more than polish.
AI didn’t replace the work. It reduced the barrier to entry enough that the work could stay focused on creativity, iteration, and learning.
So what are the takeaways?
AI removed the scaffolding overhead
We focused on ideas instead of tools
Iteration and playtesting happened early
MVP decisions were intentional
Shipping was part of the lesson, not the finish line
Do this with your kids!
If you’re a parent, teacher, or mentor, this is a moment worth recreating. You don’t need a full weekend or a deep background in game engines. Start with a simple idea, break it down, and build just enough to test.
And if you’re doing this for yourself, the same applies. AI is at its best when it helps you think and build, rather than trying to replace either.
Sometimes the most important thing you ship is the experience itself. My son has inundated me with idea lists and bug reports, so I’m pretty sure what we’ll be spending our next 1:1 time doing :)
