How I Built Starblaster 3000 by Directing AI Like a Studio Team
Creating Starblaster 3000, my retro-style browser game, started as a simple design experiment. I had a vision: a fast-paced, arcade-inspired space shooter that felt like something you’d find on a dusty arcade cabinet in the back of a pizza joint. I knew what I wanted it to look like. I knew how it should play. But I didn’t want to get buried in code.
Instead of hiring a programmer or spending weeks experimenting and back-engineering JavaScript, I brought in some unconventional help. I used AI tools like ChatGPT and Grok to handle development. I treated them as if they were members of my team. I described what I wanted the game to do, and they wrote the JavaScript code with me handing the HTML/CSS and any Adobe creative suite elements.
This setup let me focus entirely on creative direction. I was in charge of designing the player experience, from the way enemies attack to the timing of boss battles and power-ups. Every decision about pacing, style, and layout came from me. Meanwhile, the AIs translated those directions into functioning JavaScript – and helped debug problems along the way.
What surprised me most was how collaborative it felt. I’d explain a feature, like a warning message before a boss battle, and ChatGPT would give me a few ways to code it. If something broke or didn’t work how I imagined, I’d describe what went wrong and it would suggest a fix. The loop was tight and efficient. I wasn’t just asking for snippets. I was “running a team”.
Creating a Database-Driven Global Leaderboard
One of the more advanced features I added was a global scoreboard. I wanted players to be able to compete worldwide and see how they stacked up. With help from the AIs, I built a database-driven leaderboard that tracks scores in real time. Setting that up without AI would have taken me much longer, but I was able to walk through the logic and let the AI write the backend code for it while I concentrated on how it should display and feel in the game flow.
The beauty of working this way is that I could rapidly prototype and iterate. I didn’t need to constantly reference documentation or second-guess syntax. If I wanted to experiment with a new enemy type or tweak player movement, I just said so. The AIs handled the rest.
Using ChatGPT and Grok as my developers didn’t make the process impersonal. If anything, it gave me more control. I was free to push the creative side as far as I wanted, knowing that the technical support was always a message away. Starblaster 3000 was built faster and more exactly to my vision than it would have been if I had tried to do it all myself.
This wasn’t just a side project. It became a real-world case study in using AI as a production tool. And it’s highly likely I will build every future game this way too.
