I’ve been asked a few times whether AI is making this game for me. The short answer is no. If anything, AI made me throw away more ideas than it created. I don’t use ChatGPT to generate code, build levels or write complete stories. There are people doing that and, honestly, it’s impressive. That’s simply not how I use it.
What I discovered over the last couple of weeks is that ChatGPT became something completely different. It became the person you call when you’ve been staring at the same problem for three hours and can’t tell anymore whether your idea is brilliant or completely stupid. Sometimes you just need someone to argue with. ChatGPT is surprisingly good at that because it never gets tired and never gets offended when you tell it that its idea makes no sense.
The funny part is that I rarely ask it to invent something. Most conversations begin with me explaining an idea I already have. Then we start pulling on threads. „What if this happened?” „Why would that system exist?” „Wouldn’t this break that other mechanic?” Before I know it, we’re discussing something that wasn’t even part of the original problem. Quite often we end up deleting the initial idea completely because a much better solution appears somewhere in the middle of the conversation.
Looking back, I don’t think ChatGPT has given me many original ideas. What it has done is force me to justify my own. That’s a much bigger contribution than I expected. It’s very easy to fall in love with your own concepts simply because you’ve spent hours thinking about them. Having something constantly asking „Why?” becomes surprisingly annoying… and surprisingly useful. If I can’t explain why a mechanic exists, maybe it shouldn’t exist. If a story element only sounds cool but doesn’t improve anything else, perhaps it’s not worth keeping.
One thing I also noticed is that our discussions almost never stay in one category. I’d start talking about procedural maps and somehow we’d end up discussing the philosophy of an AI character. We’d move from enemy design to world-building, then back to gameplay, then suddenly realise we’d accidentally solved a progression problem without even trying. At first I thought we were constantly getting distracted. Now I think that’s actually how interconnected design works. Good ideas don’t stay in their own little boxes.
Perhaps the biggest change wasn’t adding anything at all. It was changing how I evaluate ideas. A few months ago my thinking was mostly, „Wouldn’t this be cool?” Now it’s more like, „Does this make three other systems better?” If the answer is no, the idea is probably weaker than I originally thought. That single question has probably removed more content from the project than any design document ever could.
I’m still not convinced this is the right way to develop a game. Maybe experienced game developers will read this and immediately point out ten mistakes I’m making. That’s perfectly fine. I’m not writing these posts because I’ve already succeeded. I’m writing them because I’m figuring things out as I go. If this project eventually becomes something worth playing, these blog posts will be an interesting record of how it happened. If it fails spectacularly, they’ll probably be an equally interesting record of why.
One thing is certain, though. ChatGPT didn’t become my game designer. It became my sparring partner. It doesn’t build the game for me. It keeps punching holes in my ideas until either they collapse or become strong enough to survive. Strangely enough, I think that’s exactly what I needed.