karpathy

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

johnloy

Electron apps

• Marp: Markdown presentation ecosystem.

• muryoimpl/slippr: markdown presentation app by Electron

• joe-re/cafe-pitch Markdown-driven presentation tool built on Electron.

Slide deck frameworks supporting Markdown source

jesseschalken

My preferred code style is 2-space K&R. This is intended to provide a justification for this style.

Why K&R?

K&R style has the following properties:

1. Provides symmetric size (in terms of screen space consumed) between the opening and closing syntax of a clode block.
2. Forces no empty or meaningless lines, thereby avoiding artificial distance between related things that should be together.
3. Consumes the minimum vertical space while keeping the opening and closing syntax of a block on separate lines from the content.