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.

carlosfgti

1) Se tiver o Docker instalado, remova!

2) Habilite o WSL no Windows 10
dism.exe /online /enable-feature /featurename:Microsoft-Windows-Subsystem-Linux /all /norestart
dism.exe /online /enable-feature /featurename:VirtualMachinePlatform /all /norestart

3) Habilitar o WSL para a versão 2
wsl --set-default-version 2

4) Instalar o Ubuntu na Microsoft Store

skwid138

<?php

/**
 * The table Printer class has no dependencies and generates an array of strings
 * out of an array of objects or arrays and the child properties that should be printed given by an array of strings
 *
 * Should result in something like the following:
 * +------+---------------+----------------+---------------+
 * |  id  |     name      | col_name_three | col_name_four |
 * +------+---------------+----------------+---------------+

laravel-shift

<?php

use PhpCsFixer\Config;
use PhpCsFixer\Finder;

$rules = [
    'array_indentation' => true,
    'array_syntax' => ['syntax' => 'short'],
    'binary_operator_spaces' => [
        'default' => 'single_space',

PARC6502

Backend as a Service

Note on sizes: a lot of those might be inaccurate as there might be many microservices required aside from the core release/docker image. I haven't tested these recently so I'm not sure

Supabase - ~94.9K stars

• Designed explicitly as an open source Firebase alternative
• Typescript based
• Docker support
• Full Postgres database with Realtime subscriptions

PocketBase - ~54.5K stars

redestructa

<?php

/**
 * The table Printer class has no dependencies and generates an array (strings)
 * out of an object array / countable and the child properties that should be printed given by an array (strings)
 */
class TablePrinter
{
    /**
     * @param \Countable|array|object[] $historyItems

crissilvaeng

Styleguides

Mensagens de commit styleguide

• Usar modo imperativo ("Adiciona feature" não "Adicionando feature" ou "Adicionada feature")
• Primeira linha deve ter no máximo 72 caracteres
• Considere descrever com detalhes no corpo do commit
• Considere usar um emoji no início da mensagem de commit

Emoji | Code | Commit Type

liruqi

var run = 0;
var mails = {}
total = 3000; //滚动次数，可以自己根据情况定义

function getEmails (cont) {
	var friendbutton=cont.getElementsByClassName("_ohe");
	for(var i=0; i<friendbutton.length; i++) {
		var link = friendbutton[i].getAttribute("href");

		if(link && link.substr(0,25)=="https://www.facebook.com/") {

pbojinov

Two way iframe communication

The main difference between the two pages is the method of sending messages. Recieving messages is the same in both.

Parent

Send messages to iframe using iframeEl.contentWindow.postMessage Recieve messages using window.addEventListener('message')

iframe

cowboy

var stringify = function(obj, prop) {
  var placeholder = '____PLACEHOLDER____';
  var fns = [];
  var json = JSON.stringify(obj, function(key, value) {
    if (typeof value === 'function') {
      fns.push(value);
      return placeholder;
    }
    return value;
  }, 2);