---
format: "aidr-story-markdown/v1"
id: "b57778213c8ad4afec46ae0f17d4cfdb21b07a9ea82c27449eb4cc7f4fcc6df0"
canonical_url: "https://aidr.today/b5777821?lang=en"
title: "Can gzip be a language model?"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-22T06:08:39.000Z"
category: "Research"
topics: ["gzip","llm","compression","entropy","shakespeare"]
source_urls: ["https://nathan.rs/posts/gzip-lm/","https://news.ycombinator.com/item?id=49797323"]
summary: "A while back I wrote about language modeling without neural networks, where I generated Shakespeare with an unbounded n-gram model: no weights, no training, …"
---

# Can gzip be a language model?

> [Open the canonical story](<https://aidr.today/b5777821?lang=en>)

**Published:** 2026-09-22T06:08:39.000Z
**Category:** Research
**Topics:** gzip, llm, compression, entropy, shakespeare

## Summary

A while back I wrote about language modeling without neural networks, where I generated Shakespeare with an unbounded n\-gram model: no weights, no training, …

## Sources

- [Story source](<https://nathan.rs/posts/gzip-lm/>)
- [Discussion](<https://news.ycombinator.com/item?id=49797323>)