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.

ShawnMcCool

Overview

In this gist, I attempt to communicate techniques for defining explicit boundaries over naturally occuring boundaries (in this case, the boundary between domain and service/data layers) to empower both testing and to improve comprehensibility of the system.

In my view, testing and application code are one-and-the-same and should evolve together. Testing code should not be an after-thought. I believe that haphazardly designing application code and test code leads to lead-time penalities.

If you'd like to reply, consider replying to this twitter thread. Additionally, I am available on Twitter @ShawnMcCool.

Note: I acknowledge that no techniques are right for everyone. But I have a long history with these techniques and I found them quite adequate for large and small-scale systems.

xavriley

# Welcome to Sonic Pi v2.9.0-dev-d0d5c

use_sample_bpm :loop_amen
use_bpm_mul 2

live_loop :foo do
  #sync :dr if spread(1,16).tick(:sync)
  sample :bd_haus if spread(2,4).tick(:bd)
  with_fx :reverb, room: 0.1, damp: 0.1 do
    with_fx :echo, mix: 0.1 do

artzub

#!/bin/bash

apt-get install mc htop git unzip wget curl -y

echo
echo "====================================================="
echo "                     WELCOME"
echo "====================================================="
echo
echo "Hub"

ShawnMcCool

====== PHP RFC: Instance Reference Sugar ======
  * Version: 0.2
  * Date: 2015-03-09
  * Author: Shawn McCool, shawn@heybigname.com
  * Status: In Discussion

===== Summary =====

In order to access instance variables and methods, one must use the `$this->` prefix.

xavriley

# DUBSTEP
# Combines ideas from my other gists
current_bpm = 140.0 
use_bpm current_bpm

# WOBBLE BASS 
define :wob do
  use_synth :dsaw
  lowcut = note(:E1) # ~ 40Hz
  highcut = note(:G8) # ~ 3000Hz

laracasts

<?php

use Faker\Factory as Faker;

abstract class ApiTester extends TestCase {

    /**
     * @var Faker
     */
    protected $fake;

daylerees

My Validation Base Class

I was asked how I deal with validation / create and update validation rulesets. Well here is one method I have used. Don't be afraid to build on top of what the framework has already given you. In my projects I use a base class for almost anything. You never know when you want your classes to inherit some common functionality. My BaseValidator actually has some pretty useful methods and properties in it.

<?php

namespace FooProject\Internal\Validators;

use FooProject\Internal\Sanitizers\BaseSanitizer;

wooorm

/**
 * Author: Titus Wormer <tituswormer@gmail.com>
 * URL: http://wooorm.com/penn-treebank-tokenizer-in-javascript.html
 *
 * The Treebank tokenizer uses regular expressions to tokenize text as in
 * Penn Treebank.  This implementation is a based on both the Sed script
 * written by Robert McIntyre (available at
 * [http://www.cis.upenn.edu/~treebank/tokenizer.sed]), and the Python port
 * by Edward Loper and Michael Heilman (available at
 * [http://nltk.org/_modules/nltk/tokenize/treebank.html]).

dhrrgn

<?php
	public function update($id, $title, $content, $slug, array $tags, $active, DateTime $publish_date)
	{
		$post = $this->find($id);

		if (Config::get('wardrobe.cache'))
		{
			// Forget the old cache
			Cache::forget('post-'.$post->slug);
		}