Top all-time · #17826
Dipkumar Patel
Personal blog · dipkumar.devaiprogramming
Dipkumar Patel writes Dip, a personal blog at dipkumar.dev about ai and programming. Since 2023, 4 different Hacker News readers have linked it 11 times. It is still writing: the last post was on 2026-06-03, and you can follow it by feed.
Too few sampled comments (2) to say how often HN recommends it.?
Why HN links it
HN readers link this blog as a follow-on to building a GPT-style LLM from scratch, with a post on adding a KV cache implementation. Comments mention it only as a pointer to that topic.
Summarised by Claude from a sample of 2 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Essays and notes on machine learning, LLM internals, AI agents, and distributed systems by Dipkumar Patel.
Most-linked posts?
- /becoming-the-unbeatable/posts/gpt-kvcache/2 people · 2 mentions · first linked 2023-08-22
What Hacker News says?
https://jaykmody.com/blog/gpt-from-scratch/ for a gpt2 inference engine in numpy then https://www.dipkumar.dev/becoming-the-unbeatable/posts/gpt-k... for adding a kv cache implementation
Comment linking /becoming-the-unbeatable/posts/gpt-kvcache/
on “Implementing a ChatGPT-like LLM from scratch, step by step”Submitted Agents Can Reason. They Still Can't Search as a story · 3 points
HN discussionAlso not an expert, but I believe it is a little bit of both for inference. If you are generating token-by-token naively, you do need to pay the n^2 cost since every token must attend to all other tokens. Generating a sequence of 5 tokens starting from 2 (infer 3rd, infer 4th, infer 5th...) will be much faster than starting from 1024 (infer 1025th, infer 1026th, ...) since your n is smaller. But each time time it is…
Comment linking /becoming-the-unbeatable/posts/gpt-kvcache/
on “GPT-3.5/4 response times are linear with output tokens”Submitted GPT-5 Router – Inevitable Future of Chat Interfaces as a story · 3 points
HN discussionSubmitted Essential Database Design: Five Fields Every Table Must Have as a story · 3 points
HN discussionSubmitted Building with Claude Managed Agents – Sharp Edges as a story · 2 points
HN discussionSubmitted Creativity Is a Luxury as a story · 2 points
HN discussionSubmitted Improving Retrieval in RAG (Via Recall, Precision, and NDCG) as a story · 2 points
HN discussionSubmitted AWS BedRock – Converse API – A single endpoint for all models? as a story · 2 points
HN discussionSubmitted Bits-per-Byte (BPB): a tokenizer-agnostic way to measure LLMs as a story · 1 points
HN discussionSubmitted Instruction Aware Embeddings – Why Your Retriever Is Failing as a story · 1 points
HN discussion