Experimenting with AI in Creative Coding

We’ve been playing with another AI-powered creative coding experiment — this time, a random abstract art generator. It’s not a client project, but rather a stepping stone towards a larger real-world product we're developing here at the studio.

It’s also a little nostalgic nod to those classic Amiga demos from back in the day. We built in real-time controls so you can tweak sliders, shift colour palettes, and play with parameters to your heart’s content. It’s fast, fun, and totally unpredictable.

This took about half a day of hands-on experimenting and prototyping. A fun exercise in between projects — the kind of creative noodling that helps us push boundaries and keep our design muscles sharp.

But the more we explore AI in design and development, the more its limitations start to show. It’s powerful, sure — but far from perfect. You can nudge it in the right direction with well-structured prompts, but at some point, it’ll veer off course, break things, or make choices that just don’t make sense. You reach a place of “yeah, kinda what I had in mind”, but often it’s quicker (and better) to scrap it and start again from scratch.

From a commercial product design standpoint, it’s still a long way off. You need the right design vocabulary to guide it — and even then, it’s more “vibe designing” than anything concrete. Which… let’s be honest… is just code for messing around and seeing what sticks.

Still, there’s something exciting about it. These experiments help us understand where AI fits in a modern design workflow — and more importantly, where it doesn’t.