I’ve wanted to build a game for as long as I’ve been playing them. Never did anything about it. I knew it takes forever as it’s a lot of work, so the idea just sat there, way down at the bottom of the list, never a priority.

Then 2026 happened, what a weird year. AI companies kind of forced their way into our workflow whether we asked for it or not. And I thought, fine, I have all these subsidised tokens sitting around, why not finally do the thing? I have all the help I can get.

So I gave it a shot. The concept of the game is simple. You change color by running through color blobs, and you can only touch stuff that matches the color you currently are. That’s it.

The rest is spicing it up with hazards, decent levels, and a difficulty curve that doesn’t feel unfair.

Can AI build a fun game?

Claude helps me a lot in my day to day work, but it also hits walls. The truth is that building a game is not the one-shot magic trick AI bros are claiming on the internet. I had to babysit Claude a lot on this one…

The major issue is art.

Image generation is good these days. It’s also instantly recognisable by most people at this point. I love pixel art, that’s the look I wanted, and pixel art is where the whole thing kinda falls apart. You can generate it, but the pixels don’t sit on the grid properly, or it just doesn’t feel right. There are tools out there that try to fix this. Sometimes they work. Sometimes they really don’t.

There’s also the fact that the gaming community really hates AI art, and honestly, I do too. A lot of artists got screwed over so that someone somewhere could generate more slop. I feel that.

So the deal I made with myself: build whatever I can by hand, generate only the stuff that’s genuinely out of reach, then clean it up with whatever tools I can find until it looks right. Results are honestly not bad.

Level design is an another major issue. AI fails to make a fun level. For every prompt I tried, the results were horrible. Sometimes technically playable, but always boring, or just too easy.

So I’m building every level by hand. It takes a while but there’s a nice hack here: I presented the game to my friends and let them cook. A few levels are 100% their designs, and they’re good. Like really good. Sitting on a Discord call with the boys designing a level together might be my favourite part of this whole thing. One of those nights you remember. Credit where credit is due.

Godot

What a great game engine. Huge respect to everyone who built this thing. Genuinely surprised by how easy it is to work in. I’d been meaning to try Godot properly for ages. Before this I’d opened it maybe twice, clicked around for ten minutes, closed it. This time I stuck with it, and Claude actually did a decent job working with it and scaffolding the game mechanics. Had to fine-tune a few things by hand, but as a starting point? Great. GDScript is rough in some places, though.

Scratching an itch

I’m building this because it’s something I’ve wanted to create for a long time. There’s something special about finally giving shape to an idea that’s lived in your head for years.

I’m genuinely excited.