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.

elithrar

AGENTS.md

Instructions for AI coding agents working in this repository.

Decision-making

• When possible, defer decisions to larger groups. Ensure all stakeholders have an opportunity to weigh in before proceeding.
• Demand precision. Insist on clarifying ambiguities and edge cases before beginning any implementation.
• Reopen previously settled matters if new participants join or if conditions have changed, even slightly.
• Raise questions about whether the current approach aligns with long-term strategy. Suggest forming a working group if uncertainty remains.

JLarky

The tweet

https://twitter.com/JLarky/status/1639832543766941703

Bard

Bing Chat

BaziForYou

-- Lua 5.1+ base64 v3.0 (c) 2009 by Alex Kloss <alexthkloss@web.de>
-- licensed under the terms of the LGPL2

-- character table string
local b='ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/'

-- encoding
function enc(data)
    return ((data:gsub('.', function(x) 
        local r,b='',x:byte()

prologic

Learn Go in ~5mins

This is inspired by A half-hour to learn Rust and Zig in 30 minutes.

Basics

Your first Go program as a classical "Hello World" is pretty simple:

First we create a workspace for our project:

mayneyao

const MY_DOMAIN = "agodrich.com"
const START_PAGE = "https://www.notion.so/gatsby-starter-notion-2c5e3d685aa341088d4cd8daca52fcc2"
const DISQUS_SHORTNAME = "agodrich"

addEventListener('fetch', event => {
  event.respondWith(fetchAndApply(event.request))
})

const corsHeaders = {
  "Access-Control-Allow-Origin": "*",

ryanflorence

import React from "react";
import {
  useRouter,
  Link
} from "@reach/router/unstable-hooks";

function Accounts() {
  let route = useRouter({
    ".": () => <div>boring Accounts</div>,
    dope: () => <div>Dope Accounts</div>

bcnzer

addEventListener('fetch', event => {
  event.respondWith(handleRequest(event))
})

/**
 * Entry point of the worker
 */
async function handleRequest(event) {
  try {
    // Get the JWT

ninoseki

import request from "request";
import puppeteer from "puppeteer";

const API_URL = "https://slack.com/api/files.upload";
const SLACK_API_TOKEN = process.env.SLACK_API_TOKEN as string;
const CHANNEL = "general";

(async () => {
  const url = "http://neverssl.com"
  const browser = await puppeteer.launch();

karolmajta

'use strict';

const pointsToFile = uri => /\/[^/]+\.[^/]+$/.test(uri);

exports.handler = (event, context, callback) => {
    
    // Extract the request from the CloudFront event that is sent to Lambda@Edge 
    var request = event.Records[0].cf.request;

    // Extract the URI from the request