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.

check_authorization.php

<?php

define('BOT_TOKEN', 'XXXXXXXX:XXXXXXXXXXXXXXXXXXXXXXXX'); // place bot token of your bot here

function checkTelegramAuthorization($auth_data) {
  $check_hash = $auth_data['hash'];
  unset($auth_data['hash']);
  $data_check_arr = [];
  foreach ($auth_data as $key => $value) {
    $data_check_arr[] = $key . '=' . $value;

thanksdanny

vi /etc/environment

add these lines...

LANG=en_US.utf-8
LC_ALL=en_US.utf-8

arubacao

<?php

/**
 * Validates a given latitude $lat
 *
 * @param float|int|string $lat Latitude
 * @return bool `true` if $lat is valid, `false` if not
 */
function validateLatitude($lat) {
  return preg_match('/^(\+|-)?(?:90(?:(?:\.0{1,6})?)|(?:[0-9]|[1-8][0-9])(?:(?:\.[0-9]{1,6})?))$/', $lat);

mrabbani

Mailgun setup

• install Guzzle HTTP library : composer require guzzlehttp/guzzle
• [sign up to mailgun] (http://www.mailgun.com)
• Go to Domains tab and click on domains
• You will find the necessary data for .env setup
  • MAIL_DRIVER=mailgun
  • MAIL_HOST=smtp.mailgun.org
  • MAIL_PORT=587
  • MAIL_USERNAME=postmaster@sandbox------------------------------6f.mailgun.org
• MAIL_PASSWORD=502fd951f7------------------------

Yimiprod

/**
 * This code is licensed under the terms of the MIT license
 *
 * Deep diff between two object, using lodash
 * @param  {Object} object Object compared
 * @param  {Object} base   Object to compare with
 * @return {Object}        Return a new object who represent the diff
 */
function difference(object, base) {
	function changes(object, base) {

realmyst

function declOfNum(number, titles) {  
    cases = [2, 0, 1, 1, 1, 2];  
    return titles[ (number%100>4 && number%100<20)? 2 : cases[(number%10<5)?number%10:5] ];  
}

use:
declOfNum(count, ['найдена', 'найдено', 'найдены']);