This is amazing work, lossless image compression in O(n), without all the complexity of PNG/JPEG/HEIC https://phoboslab.org/log/2021/11/qoi-fast-lossless-image-compression…
I'm still blown away that there's so much low-hanging fruit in image compression. I thought most of the innovation would be in tweaking the confirgurable parts of existing codecs.
Great reminder of the value of starting from frost principles
reminds me of other recent work I saw for redoing compression from a machine learning perspective... can't find the link, but compressing to drive learning is different from compressing to match human vision