Top all-time · #8600
Nathan Barry
Personal blog · nathan.rs/posts · on HN as nathan-barryaiprogramming
Nathan Barry's personal blog at nathan.rs covers ai and programming. Since 2024, 8 different Hacker News readers have linked it 16 times, most often “Can gzip be a language model?”. It is still writing: the last post was on 2026-06-27, and you can follow it by feed.
Too few sampled comments (5) to say how often HN recommends it.?
Why HN links it
HN commenters link Nathan Barry's posts on text diffusion from BERT (roberta-diffusion), gzip as a language model, compression as prediction, and fixing iOS Safari code blocks. Links are used as sources or see-also references.
Summarised by Claude from a sample of 5 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
This is the website of Nathan Barry
Most-linked posts?
- Can gzip be a language model?4 people · 4 mentions · first linked 2026-06-16
- BERT is just a single text diffusion step3 people · 4 mentions · first linked 2025-10-20
- Text Diffusion Models Are Faster at Writing Code2 people · 2 mentions · first linked 2025-12-13
- Fixing iOS code-block font-size rendering quirk2 people · 2 mentions · first linked 2024-02-06
What Hacker News says?
Found a solution to this recently (this is not my blog): https://nathan.rs/posts/fixing-ios-codeblocks/ CSS needs some tweaking for iOS Safari, I guess.
Comment linking /posts/fixing-ios-codeblocks/
on “You already have a Git server”Submitted BERT is just a single text diffusion step as a story · 455 points
HN discussionSee also: Can gzip be a language model? https://nathan.rs/posts/gzip-lm/
Comment linking /posts/gzip-lm/
on “Compression Is Prediction”Submitted Can gzip be a language model? as a story · 414 points
HN discussionYou can actually generate surprisingly coherent text with minimal finetuning of BERT, by reinterpreting it as a diffusion model: https://nathan.rs/posts/roberta-diffusion/ I don’t see a useful definition of LLM that doesn’t include BERT, especially given its historical importance. 340M parameters is only “small” in the sense that a baby whale is small.
Comment linking /posts/roberta-diffusion/
on “Microgpt”Submitted Running GPT-2 in WebGL: Rediscovering the Lost Art of GPU Shader Programming as a story · 157 points
HN discussionMasked language modeling has been compared loosely to text diffusion [1], so the paper's title claim may be loosely true in some sense even if it's misleading. [1] https://nathan.rs/posts/roberta-diffusion/
Comment linking /posts/roberta-diffusion/
on “Introspective Diffusion Language Models”Submitted Can gzip be a language model? as a story · 5 points
HN discussionReminds me of https://nathan.rs/posts/roberta-diffusion/
Comment linking /posts/roberta-diffusion/
on “Introspective Diffusion Language Models”Submitted Generating Shakespeare Without Neural Networks as a story · 4 points
HN discussionSubmitted Text Diffusion Models Are Faster at Writing Code as a story · 3 points
HN discussionSubmitted Can gzip be a language model? as a story · 2 points
HN discussionSubmitted Language Modeling Without Neural Networks as a story · 2 points
HN discussionSubmitted Intro to Autograd Engines: Karpathy's Micrograd Implemented in Go as a story · 2 points
HN discussionSubmitted Text Diffusion Models Are Faster at Writing Code as a story · 1 points
HN discussionSubmitted Fixing iOS code-block font-size rendering quirk as a story · 1 points
HN discussion