Ask HN: What did you build this winter?
I often take a vacation around December (my company has a use it or lose it policy) and use it to hack on personal projects. With the double limits on Claude Code, and Codex, I got A LOT more done than normal. I'm curious what people built the past few weeks. I finished off a browser-based modern take on Asteroids, and the game engine to support it. The music synthesiser is definitely still buggy, if I go back to this I'll replace "synthesise this MIDI file" with "loop a pre-rasterised sound file", but TBH it's almost certainly more sensible for my next project to just use an existing engine, this was originally just an experiment to see what GenAI could even do. Everything else besides the music is basically as I'd like it to be: gameplay progression, achievements, particle effects, game controller support, collision detection, etc. The art assets are also AI-generated. I also updated my Raspberry Pi webcam interface. It's now 30fps, not 4, and has an acceptable UI. Also made progress with some automated motion detection in there, but I have yet to test that, I won't speak for it actually working until I do. Couple of other projects too, but now's not the time to reveal them. One might even turn into from project into product. you are right - december is a good time to disconnect from the daily chaos and do some focussed work. I worked on this: https://github.com/shudv/deltasort Its a more efficient way to incrementally repair sorted arrays. Feedback welcome! What did you build? That's interesting, what was the original problem that inspired you to build it? If you have worked with React before - you would know that its actually not straightforward to correctly memoize computation. That is specially true for operations on collection data like filters and sorts. That pushed me to first build this - https://github.com/shudv/memotable which is a way to build a "read-optimised" in-memory collection so that re-renders in the ui layer do not run filters and sorts on the data again and again unnecessarily. And this was sufficient to solve my original problem. DeltaSort was just something that came to me while I was thinking about taking memotable performance to the next level - make the sorting as fast as possible. Thats when it hit me - if I am able to somehow use the information about the delta since the previous sort - that might help me re-sort much faster. And it turns out, it actually is much faster (at least for the Rust based implementation)