Game Postmortem - Cambridge Game Jam 2025
This weekend, the game dev group I run in Cambridge hosted a game jam! Rob and his fiancé Lindsey did at least half the prep work, which reduced some of the stress that normally goes into doing that. We had 9 jammers, which was less than usual, but enough. Organisationally everything went fine. But - this is a portmortem for my game, not the jam.
My goal going into the jam was to make a classic arcade game. Something elegant with tight controls, instant mechanical appeal and a decent amount of depth. I also had it in my head to have a unconventional element to help me avoid building something overly derivative (I'm only really interested in making things with their own identity). I wrote down a bunch of traits I was after when I started (I won't cover these here, but I might write some thoughts about this design space another time), and decided to work in screen resolution that was as low as my design could enable. I also decided to deliberately use a very tall or wide aspect ratio, as my unconventional element.
I chatted through some ideas with AI (first lesson - AI is excellent at rapidly throwing out thought starters. I tried to avoid leaning on it in my design though, since I enjoy the process) quickly landed on the idea of a game where you'd bounce between two walls on either sides of the screen that were closing in on you. I wanted the size of the play space to be something the player had some control over, and where the size of the space affected your risks and opportunities. I still think this is a pretty solid foundation, although I never quite managed to make it click.
Simultaneously I was exploring engines. I knew I wanted to offload the programming to AI. I almost exclusively use AI to program in my day job, and I'd never really tested the same approach for game development. I enjoy programming, but I've implemented the same code thousands of times at this point. Collision detection, entity management, game loops - implementation details for these are rarely interesting and just eat up development time. Game engines already reduce the burden for developers here, but AI takes it a step further. I shan't miss it. Between that, my project goals, and wanting to avoid using Unity (it's historically been my tool of choice, but for various reasons I'm pretty sick of it at this point), I decided to use Pico-8. That didn't work out because it's not a signed application and my computer refused to open it, so I swapped to LOVE2D.
In late 2025, AI coding for simple games works fantastically well. It set up my project and I had something playable within about 10 minutes. It enabled me to stay in designer-mode, rather than getting tangled up in implementation details. I'm genuinely excited to build games in this manner again.
This freed up an incredible amount of time, which I spent either playing my own game, or taking breaks playing other games on my Switch. Delightful. Despite this, I never quite worked out how to make my game click. It had various elements that had good interplay, but just as many that never clicked. It had some fun in it - I played it myself for hours, fairly happily - but I never felt like I had a solid design. I should have made better snapshots so I could more easily revisit previous designs. Broadly, the game improved over time, although the final design explorations I made probably made it worse than it had been on Sunday morning. Knowing what to do when a tight game design isn't working is always difficult and I'm sorry to say I didn't find the solution this time either. Being able to use AI to rapidly test new ideas was incredibly valuable, but I feel like I probably need some sort of progress to help push through design problems. It might have been useful to just cut some elements I liked but weren't working. Starting over can feel like the wrong decision, especially in a game jam. I've done this in jams before and it's never served me well in the past, but perhaps AI changes this dynamic a bit. Talking things over with other jammers was always helpful, but it's hard to solve problems you can't express clearly.
I didn't emerge with a game I'm especially happy with, but I'm not un-proud of it either. The process was enjoyable (something I've not felt in a jam for a while) and I'd like to try it again in the future. AI development using code-first engines was a complete success, at least for the sort of game I wanted (and still want) to make. It's reminded me of how much I enjoy thinking on game designs, and I'm happy that I don't feel toooooo strong a pull to return to the idea. I'd like to kill the design and try a completely new take on the same goals next time.











