I hope you'll appreciate the game as a lot of effort and collaboration with friends went into it to refine it. For me, this side project started as all of mine do, and surely yours. In a race to rapidly complete the side-project before I lost the motivation and enjoyment of development. I threw in haphazard code generated by the o1 model which was surprisingly much more competent than I had originally expected, especially compared to the non-reasoning models.
The AI model unfortunately quickly stalled as the files became too big and complex for it to work effectively, and would get stuck alternating between offering 2 unworkable "solutions". At this point my rapid prototyping came to a screeching halt with tech debt having compounded to dangerous levels. Thankfully I got my first prototype out without too much extra work, and the interest and feedback I received encouraged me to keep iterating and improving.
On the architecture side, while I started with a simple EC2 instance, I ported it to AWS Lambda with ‘scale’ and ‘cost’ in mind, and with a very shameful import of the serverless-http library to wrap the express app. It was way too slow, doubling latency from 40ms to 80ms. I pushed static content onto S3 and used CloudFront to handle caching and compression, but had to admit that the move to Lambda had been naive, and was forced to drop it and re-setup EC2 with proper autoscaling and regional deployments.
So that’s been my one-weekend-turn-two-week side project! I’ve been challenged but never insurmountably, a great experience all up.
Happy to field any comments on the game itself or my development of it, thanks!
