Top all-time · #12991
Aman Arora
Personal blog · amaarora.github.ioaiprogramming
Aman Arora's personal blog at amaarora.github.io covers ai and programming. Since 2020, 6 different Hacker News readers have linked it 6 times, most often “The Annotated GPT-2”. It is still writing: the last post was on 2026-09-20, and you can follow it by feed.
Too few sampled comments (1) to say how often HN recommends it.?
Why HN links it
The linked posts appear to cover vision transformers and how image patches become tokens. The one comment does not cite or evaluate the writing itself, only discusses ViT token embeddings.
Summarised by Claude from a sample of 1 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- The Annotated GPT-22 people · 2 mentions · first linked 2020-03-11
What Hacker News says?
Yeah but we're not talking about LLMs here but vision transformers, which don't use the same type of token vocabulary to produce embeddings from the input as the LLMs do. The pixel data is much more dense than a few characters is, per token. I looked it up - the original ViT models directly projected for example 16x16 pixel patches into 768-dimensional "tokens". So a 224x224 image ended up as 14*14=196 "tokens" each…
Comment linking /posts/2021-01-18-ViT.html
on “The killer app of Gemini Pro 1.5 is using video as an input”Submitted The Annotated GPT-2 as a story · 3 points
HN discussionSubmitted The Annotated GPT-2 as a story · 3 points
HN discussionSubmitted How LLMs Scaled from 512 to 2M Context: A Technical Deep Dive as a story · 2 points
HN discussionSubmitted Sliding Window Attention as a story · 2 points
HN discussionSubmitted The Annotated CLIP (Part-1) as a story · 1 points
HN discussion