Hello all.
For this game jam, the amount of time and headspace I can dedicate to development is a bit limited because of some personal matters. So I decided to use the jam to experiment with a very different AI workflow.
In my current project, Grid Hackers, I built the foundation and basic systems myself, and then used AI to help me build upon that.
For Worm Tremors, I’m taking that much further: I’m basically letting AI do the programming. I describe very specifically what I want a system to do, how it should behave in certain situations, test the result, and then we keep iterating from there.
The game itself is a 4-player PvP worm game. You dig through the ground, collect objects and bring them back to your nest before the timer runs out.
Moving through an existing tunnel is fast, while digging through solid earth is slower. But there’s a catch: as more and more tunnels are created, parts of the ground can collapse. Those collapses can cause further collapses, creating the chain reaction for the jam theme.
I’m using Nakama for the multiplayer/server backend, while I’m still creating the art myself. The artwork is intentionally pretty basic sprite art, because for me this jam is mostly about experimenting with this development workflow and seeing how far I can get with it.
And so far… it works.
There is an actual game running, and it is increasingly doing what I have in my head.
But I’ve also discovered something I didn’t really expect: I find working this way quite boring.
By letting AI handle so much of the implementation, it feels like I’m skipping a large part of the learning process. I can describe what I want, test it, find problems and tell the AI how I want those problems solved. The end result can be exactly what I had envisioned.
But at the same time, there’s this strange feeling of:
“Yes, this is the game I had in my head… but it doesn’t really feel like I created it.”
That’s probably the most interesting result of this experiment for me so far.
For this jam I’m going to continue working this way. My available time is limited and I really want to get Worm Tremors into a playable state before the end of the jam.
But afterwards, I think I’ll take some lessons from this experiment and find a better balance between using AI to help me develop and actually doing enough of the work myself to keep learning and enjoying the process.
Anyway, that’s my little experiment for this jam. Let’s see how far I can get with it!
