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.

argyleink

:root {
  --ease-in-quad:       cubic-bezier(0.55, 0.085, 0.68, 0.53);
  --ease-in-cubic:      cubic-bezier(0.55, 0.055, 0.675, 0.19);
  --ease-in-quart:      cubic-bezier(0.895, 0.03, 0.685, 0.22);
  --ease-in-quint:      cubic-bezier(0.755, 0.05, 0.855, 0.06);
  --ease-in-expo:       cubic-bezier(0.95, 0.05, 0.795, 0.035);
  --ease-in-circ:       cubic-bezier(0.6, 0.04, 0.98, 0.335);
  --ease-out-quad:      cubic-bezier(0.25, 0.46, 0.45, 0.94);
  --ease-out-cubic:     cubic-bezier(0.215, 0.61, 0.355, 1);
  --ease-out-quart:     cubic-bezier(0.165, 0.84, 0.44, 1);

davidgilbertson

function logColor(color, args) {
    console.log(`%c ${args.join(' ')}`, `color: ${color}`);
}

const log = {
  aliceblue: (...args) => { logColor('aliceblue', args)},
  antiquewhite: (...args) => { logColor('antiquewhite', args)},
  aqua: (...args) => { logColor('aqua', args)},
  aquamarine: (...args) => { logColor('aquamarine', args)},
  azure: (...args) => { logColor('azure', args)},

darrenscerri

var Middleware = function() {};

Middleware.prototype.use = function(fn) {
  var self = this;

  this.go = (function(stack) {
    return function(next) {
      stack.call(self, function() {
        fn.call(self, next.bind(self));
      });

davideicardi

const fs = require("fs");
const { Transform, PassThrough } = require('stream');

function concatStreams(streams) {
  let pass = new PassThrough();
  let waiting = streams.length;
  for (let stream of streams) {
      pass = stream.pipe(pass, {end: false});
      stream.once('end', () => --waiting === 0 && pass.emit('end'));
  }