Skip to content

Instantly share code, notes, and snippets.

@corelon
corelon / llm-wiki.md
Created May 14, 2026 07:48 — forked from karpathy/llm-wiki.md
llm-wiki

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.

@corelon
corelon / vim_install.sh
Created December 1, 2025 14:42 — forked from pirafrank/vim_install.sh
Install vim from source (+python3 +ruby and more) on Ubuntu 20.04 Focal Fossa (GUI options enabled)
#!/bin/bash
# Remove old vim installs
sudo apt-get remove vim vim-runtime gvim vim-tiny \
vim-common vim-gui-common vim-nox gvim
# install vim-gtk deps
sudo apt-get build-dep vim-gtk
# install some more deps
@corelon
corelon / Kafka-MirrorMaker-Set-Up.md
Created July 26, 2018 10:37 — forked from rajkrrsingh/Kafka-MirrorMaker-Set-Up.md
Kafka Mirror Maker - from source non-kerberized cluster to kerberized cluster

Kafka Mirror Maker - from source non-kerberized cluster to target (kerberized) cluster

Env:

source cluster:
HDP242
un-secure
hostname: rksnode1

destination cluster:

Keybase proof

I hereby claim:

  • I am corelon on github.
  • I am corelon (https://keybase.io/corelon) on keybase.
  • I have a public key whose fingerprint is FFFD 4041 7D56 5747 D694 01A9 1FF1 D094 37AE E7E6

To claim this, I am signing this object:

@corelon
corelon / build-emacs.sh
Created October 1, 2015 21:03 — forked from favadi/build-emacs.sh
Compile latest emacs version (24.5) in Ubuntu 14.04
#!/bin/bash
# Build latest version of Emacs, version management with stow
# OS: Ubuntu 14.04 LTS
# version: 24.5
# Toolkit: lucid
set -e
readonly version="24.5"
@corelon
corelon / ipy_repl.py
Last active March 16, 2020 19:16 — forked from MattDMo/ipy_repl.py
SublimeREPL ipy_repl.py for running IPython 4/Jupyter in Sublime Text
import os
import sys
import json
import socket
import threading
activate_this = os.environ.get("SUBLIMEREPL_ACTIVATE_THIS", None)
# turn off pager
os.environ['TERM'] = 'emacs'
13:15 <xQuasar> | HASKELL IS FOR FUCKIN FAGGOTS. YOU'RE ALL A BUNCH OF
| FUCKIN PUSSIES
13:15 <xQuasar> | JAVASCRIPT FOR LIFE FAGS
13:16 <luite> | hello
13:16 <ChongLi> | somebody has a mental illness!
13:16 <merijn> | Wow...I suddenly see the error of my ways and feel
| compelled to write Node.js!
13:16 <genisage> | hi
13:16 <luite> | you might be pleased to learn that you can compile
| haskell to javascript now
#!/bin/sh
# Some things taken from here
# https://github.com/mathiasbynens/dotfiles/blob/master/.osx
# Set the colours you can use
black='\033[0;30m'
white='\033[0;37m'
red='\033[0;31m'
green='\033[0;32m'
Failed to run fat jar
java.lang.NullPointerException
at org.vertx.java.platform.impl.FatJarStarter.go(FatJarStarter.java:134)
at org.vertx.java.platform.impl.FatJarStarter.main(FatJarStarter.java:53)
# Word and Character Count service for Mac OS X
#
# Adds a Word and Character Count option to the text selection context menu in all apps
#
# Use Automator.app to create a new service, and then select the Run AppleScript action.
# Paste this code in to the text box, and save as Word and Character Count. Now switch to
# a new app, select some text, and open the context menu to find the new option.
on run {input, parameters}
tell application "System Events"