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.

mnguyenngo

import scipy.stats as scs

def min_sample_size(bcr, mde, power=0.8, sig_level=0.05):
    """Returns the minimum sample size to set up a split test

    Arguments:
        bcr (float): probability of success for control, sometimes
        referred to as baseline conversion rate

        mde (float): minimum change in measurement between control

simondahla

  if (typeof jQuery != 'undefined') {
      jQuery(document).ready(function($) {
          var filetypes = /\.(zip|exe|pdf|doc*|xls*|ppt*|mp3)$/i;
          var baseHref = '';
          if (jQuery('base').attr('href') != undefined)
              baseHref = jQuery('base').attr('href');
          jQuery('a').each(function() {
              var href = jQuery(this).attr('href');
              if (href && (href.match(/^https?\:/i)) && (!href.match(document.domain))) {
                  jQuery(this).click(function() {

simondahla

/* Add this in the top of your advertising script */
_adblock = false;

jonathanmoore

Share Counts

I have always struggled with getting all the various share buttons from Facebook, Twitter, Google Plus, Pinterest, etc to align correctly and to not look like a tacky explosion of buttons. Seeing a number of sites rolling their own share buttons with counts, for example The Next Web I decided to look into the various APIs on how to simply return the share count.

If you want to roll up all of these into a single jQuery plugin check out Sharrre

Many of these API calls and methods are undocumented, so anticipate that they will change in the future. Also, if you are planning on rolling these out across a site I would recommend creating a simple endpoint that periodically caches results from all of the APIs so that you are not overloading the services will requests.

Twitter