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.

agners

#!/bin/sh

# Getting Elgato Wave:3 Microphone input "unstuck" on Linux & Pulseaudio
# Replace <card-name> with your microphones card name (check "pacmd list-cards")
# It looks something like "usb-Elgato_Systems_Elgato_Wave_3_<serial>-00"

pacmd set-card-profile alsa_card.<card-name> output:analog-stereo
pacmd set-card-profile alsa_card.<card-name> input:mono-fallback

w0rxbend

set window-title-basename "true"
set selection-clipboard "clipboard"

# Dracula color theme for Zathura
# Swaps Foreground for Background to get a light version if the user prefers

#
# Dracula color theme
#

w0rxbend

# Configuration for Alacritty, the GPU enhanced terminal emulator.

# Any items in the `env` entry below will be added as
# environment variables. Some entries may override variables
# set by alacritty itself.
env:
  # TERM variable
  #
  # This value is used to set the `$TERM` environment variable for
  # each instance of Alacritty. If it is not present, alacritty will