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.

clhenrick

node_modules/
.DS_Store

imliam

<?php
if (! function_exists('take')) {
    /**
     * Run functions consecutively by piping through the result of one
     * into the next.
     *
     * @param mixed  $value    A value
     *
     * @return object
     */

tylermcginnis

import React, { Component } from 'react'
import { TransitionGroup, CSSTransition } from 'react-transition-group'
import {
  BrowserRouter as Router,
  Switch,
  Route,
  Link,
  Redirect
} from 'react-router-dom'

james2doyle

<?php

use Slim\Http\Request;
use Slim\Http\Response;
use Slim\Http\Stream;

$app->get('/stream', function (Request $request, Response $response, array $args) {
    // a 100mb file
    $path = '../public/files/document.pdf';

nicolashery

node_modules
dist

haampie

<?php

final class User
{
    private $name;

    public function __construct(string $name)
    {
        $this->name = $name;
    }

realyze

const defer = (res, ms) => new Promise(resolve => setTimeout(() => resolve(res), ms));
                                                                                        
const makeDough = () => defer({add: () => {}}, 500);                                    
const makeSauce = () => defer({determineCheese: () => 'cheddar'}, 250);                 
const grateCheese = () => defer({type: 'cheddar'}, 150);                                
                                                                                        
                                                                                        
async function makePizza(sauceType = 'red') {                                           
                                                                                                               
  const pDough = makeDough();                                                           

mindplay-dk

You can have a linear workflow with the array functions.

The following is unreadable:

$output = array_reduce(
  array_map(
    function($n) { return $n ** 2; }
    array_filter($input, function($n) { return $n % 2 == 0; })
 ),

spion

function translateError(msg) {
  var newErr = new Error(msg); // placed here to get correct stack
  return e => { 
    newErr.originalError = e;
    throw newErr;
  }
}

async function asyncTask() {  
    const user = await UserModel.findById(1).catch(translateError('No user found'))