12 stories tagged compression, newest first.
268 points by stmonty | 94 comments
Richard Feldman's blog post benchmarking roc-deflate, a pure Roc DEFLATE library, against C's libdeflate, finding it under 10% slower, and describing Roc-specific optimizations.
29 points by jxmorris12 | 5 comments
A blog post walking through each step of decoding a PNG file, from chunk parsing through decompression and filtering, using an example image.
77 points by theanonymousone | 33 comments
A personal wiki card explaining how large a Git commit is on disk, covering zlib-compressed loose objects and packfiles.
9 points by surprisetalk | 0 comments
A circa-1995 article by Edward Kmett on The Comonad.Reader explaining the use of wavelets, including the Haar basis, for compression in 3D graphics.
384 points by networked | 149 comments
A blog post by Nathan Barry exploring whether gzip compression can be used as a language model to generate text, with an interactive beam search demo.
392 points by tosh | 111 comments
GitHub repository for bzip3, an LGPL-licensed compression tool described as a better and stronger successor to BZip2.
611 points by nikolay | 252 comments
An ngrok blog post explaining compression basics and how data compressors and LLMs fundamentally solve the same prediction problem.
331 points by georgehill | 90 comments
Blog post by Isaac Liao and Albert Gu describing a method that solves ARC-AGI puzzles using lossless information compression without pretraining.
328 points by tosh | 169 comments
GitHub repository for bzip3, an LGPL-licensed compression tool described as a better and stronger spiritual successor to BZip2.
349 points by todsacerdoti | 226 comments
Evan Hahn's blog post recounting his rejected 2022 proposal to reduce newly published npm package sizes by about 5% with backwards-compatible compression.
511 points by Kubuxu | 185 comments
Blog post describing a trick to compress a website's HTML by encoding it into a lossless WebP image and decoding it in the browser, compared with gzip and Brotli.
341 points by ks2048 | 109 comments
Blog post arguing the "gzip beats BERT" text classification paper reports top-2 accuracy instead of kNN(k=2) accuracy due to a code bug, inflating results.