标签：#hackernews

I built an experiment that uses an overfitted transformer and arithmetic coding to compress individual files. Instead of training the model to generalize, I train a 900KB transformer to memorize a single file and predict the next byte. Those predictions are fed into an arithmetic coder to produce the compressed output. On a 100MB NYC taxi CSV, it compresses to about 7MB (~0.5 bits/byte). On a 100MB slice of enwik9, it compresses to about 21MB (~1.68 bits/byte). It's pretty slow right now (roughl

Shumai is an open source platform for uploading creative files, managing projects, collecting precise feedback, sharing work, and collaborating with AI agents, all in one simple creative-first workspace. I’ve always liked the product design of Frame.io, and I wanted to build an alternative that feels just as polished, while being open source and easy to self host. You can deploy Shumai with docker compose in just a few minutes. It can also be installed from npm, though you'll need to provide you