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.

ttrahan

function sync() {

  var id="XXXXXXXXXX"; // CHANGE - id of the secondary calendar to pull events from

  var today=new Date();
  var enddate=new Date();
  enddate.setDate(today.getDate()+7); // how many days in advance to monitor and block off time
  
  var secondaryCal=CalendarApp.getCalendarById(id);
  var secondaryEvents=secondaryCal.getEvents(today,enddate);

wroman

function sync() {

  var id="XXXXXXXX"; // CHANGE - id of the secondary calendar to pull events from
      
  var secondaryCal=CalendarApp.getCalendarById(id);
  var today=new Date();
  var enddate=new Date();
  enddate.setDate(today.getDate()+30); // how many days in advance to monitor and block off time
  var secondaryEvents=secondaryCal.getEvents(today,enddate);

tomysmile

Firstly install Brew on your MAC

• ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"

Then install PHP

• brew update
• brew tap homebrew/dupes
• brew tap homebrew/php
• brew install php56

gist:1c04bf2423579e9d2dcd

// These two need to be declared outside the try/catch
// so that they can be closed in the finally block.
HttpURLConnection urlConnection = null;
BufferedReader reader = null;
 
// Will contain the raw JSON response as a string.
String forecastJsonStr = null;
 
try {
    // Construct the URL for the OpenWeatherMap query

dergachev

OS X Screencast to animated GIF

This gist shows how to create a GIF screencast using only free OS X tools: QuickTime, ffmpeg, and gifsicle.

Instructions

To capture the video (filesize: 19MB), using the free "QuickTime Player" application:

hubgit

function listFilesInFolder() {
  var folder = DocsList.getFolder("Maudesley Debates");
  var contents = folder.getFiles();
  
  var file;
  var data;
  
  var sheet = SpreadsheetApp.getActiveSheet();
  sheet.clear();
  

ryanj

#!/usr/bin/ruby
#
#   This work is licensed under a Creative Commons Attribution 3.0 Unported License.
#   http://creativecommons.org/licenses/by/3.0/
#
#   With some slight modifications, this script should create
#   a new discount code for each of the 'live' events which are 
#   owned by the user (who is identified by the user_key value).
#
#   See the above license info and Eventbrite API terms for usage limitations.

fredoliveira

1. Achievement

Definition: A virtual or physical representation of having accomplished something. These are often viewed as rewards in and of themselves.

Example: a badge, a level, a reward, points, really anything defined as a reward can be a reward.

2. Appointment Dynamic

Definition: A dynamic in which to succeed, one must return at a predefined time to take some action. Appointment dynamics are often deeply related to interval based reward schedules or avoidance dyanmics.

jmillerinc

#!/usr/bin/python
#
# Monte Carlo simulation of the payoffs to angel investing.
#
# Assume a pool of N different investors, each investing in D deals,
# with a fixed time horizon and a fixed distribution of payoffs.
# Randomly simulate each investor's total payoff, then compute the
# mean and std dev of all IRRs in the overall pool.
#
# This gives an individual angel an idea of what kind of payoff &