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.

karpathy

"""
The most atomic way to train and run inference for a GPT in pure, dependency-free Python.
This file is the complete algorithm.
Everything else is just efficiency.

@karpathy
"""

import os       # os.path.exists
import math     # math.log, math.exp

trandaison

Note: This is the guide for v 2.x.

For the v3, please follow this url: https://blog.csdn.net/sam_shan/article/details/80585240 Thanks @liy-cn for contributing.

For the v6, please follow the comment below: https://gist.github.com/trandaison/40b1d83618ae8e3d2da59df8c395093a?permalink_comment_id=5079514#gistcomment-5079514

StarUML

Download: StarUML.io

leommoore

File Magic Numbers

Magic numbers are the first bits of a file which uniquely identify the type of file. This makes programming easier because complicated file structures need not be searched in order to identify the file type.

For example, a jpeg file starts with ffd8 ffe0 0010 4a46 4946 0001 0101 0047 ......JFIF.....G ffd8 shows that it's a JPEG file, and ffe0 identify a JFIF type structure. There is an ascii encoding of "JFIF" which comes after a length code, but that is not necessary in order to identify the file. The first 4 bytes do that uniquely.

This gives an ongoing list of file-type magic numbers.

Image Files