Design Sprint #2: Can AI Actually Make a Legendary Card Game?

My last two-week design sprint quietly imploded.

Two days after starting it, my wife ended up in the hospital.

Two and a half months later, we’re finally back home. I was with her every day, and everything else went on pause. The sprint technically survived for a few weeks, I even found a great bar near one of the hospitals that unexpectedly helped me refine the game I was developing,  but eventually I just called the whole thing done. My wife has been encouraging me to go ahead and produce the game, called Bar Codes, which I probably will, as a print-and-play game.

Coming back now, I wanted to reboot with something different. Something lighter. Something just as fun, but still a design challenge.

I’ve literally used ChatGPT since the day it launched. At first, it was just out of a wealth of curiosity. Then in grad school, I used it like a multi-tool: filling in placeholder copy on decks, whipping up code I didn’t feel like debugging, sketching paper outlines, etc. For game design, I’ve mostly used it to bounce ideas around or quickly scan the universe of “things that already exist.” It’s fantastic at that task, though. What would take me a day of research, it can bring top the surface in seconds. And it has coded section of my website for me.

But here’s the thing: I’ve never actually trusted it with real creative work. I understand how gen-AI works, and I’m firmly in the “AI doesn’t create, it recombines” camp.

The Two-week Design Sprint

So for this sprint, I’m doing the one thing I’ve never done:

I’m treating ChatGPT like a junior game designer and asking it to build the most legendary card game ever made.

Do I think it’ll succeed?

The odds are about as good as my Broncos making the Super Bowl this year. It’s in the realm of possibility, but really, do we trust them enough to truly Bo-lieve yet?

But that’s the experiment. Here’s my plan:

1. Give ChatGPT a very specific brief: Design the best, most universally fun, most elegant card game imaginable.

2. Let it write the rules. No hand-holding. No “fix this for me.” It's the chef, it comes up with the recipe, I'm just the taste-tester.

3. Build a prototype and playtest it with whatever community I can gather. I’ll share the rules publicly and run playtests to see what actually works.

4. Feed real data back into ChatGPT to see if it can iterate like a human designer. Can it fix pacing? Balance? Frustration points? I have no idea, but we'll see.

5. Once we have something stable, give it the brief to design the launch strategy. I'll want marketing copy, packaging ideas, social media post schedule, the whole kid and kaboodle (which is a great name for a game, btw).

I’m going to document the entire sprint here over the next two weeks. I’ll post progress updates, failures, weird surprises, and maybe, just maybe, an actually-fun card game.

Will this produce something worth playing? Probably not. But it sounds like a hell of an experiment.

Follow along, playtest this game if you want, and let’s see what happens when you hand a design sprint to an AI.