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.

driesvints

<?php

// Get an array of all files in a directory.
$files = File::files('path_to_dir');

// Get all of the files from the given directory (recursive).
$files = File::allFiles('path_to_dir');

// Get all of the directories within a given directory.
$files = File::directory('path_to_dir');

maccman

$ = jQuery

queues  = {}
running = false

queue = (name) ->
  name = 'default' if name is true
  queues[name] or= []

next = (name) ->

dfkaye

These are my own commands, shortcuts, configurations for sublime text 2 on windows

There's also a really good one at http://www.sublimetext.com/forum/viewtopic.php?f=2&t=10615.


sublime text 2 documentation at 
  http://www.sublimetext.com/docs/2/


preferences

shostakovich

#!/bin/sh
USERNAME=shostakovich

mkdir ~/tmp
cd ~/tmp

# Install GCC + Git
curl https://github.com/downloads/kennethreitz/osx-gcc-installer/GCC-10.7-v2.pkg > GCC-10.7-v2.pkg
sudo installer -pkg GCC-10.7-v2.pkg -target /

adammw

Node.js for Raspberry Pi

Pre-built binaries

Recent releases have been pre-built using cross-compilers and this script and are downloadable below.

If you have found these packages useful, give me a shout out on twitter: @adammw

koenpunt

<?php

function gfm($text){
	# Extract pre blocks
	$extractions = array();
	
	$text = preg_replace_callback('/<pre>.*?<\/pre>/s', function($matches) use (&$extractions){
		$match = $matches[0];
		$md5 = md5($match);
		$extractions[$md5] = $match;

reusee

import ast
from cStringIO import StringIO
import sys

INFSTR = '1e308'

def interleave(inter, f, seq):
  seq = iter(seq)
  try:
    f(next(seq))

hgmnz

Types of index scans

Indexes

Sequential Scan:

• Read every row in the table
• No reading of index. Reading from indexes is also expensive.

balupton

(function(window,undefined){

	// Prepare our Variables
	var
		document = window.document,
		$ = window.jQuery;

	// Wait for Document
	$(window).bind(function(){
		// Prepare Variables