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.

karpathy

"""
The most atomic way to train and run inference for a GPT in pure, dependency-free Python.
This file is the complete algorithm.
Everything else is just efficiency.

@karpathy
"""

import os       # os.path.exists
import math     # math.log, math.exp

SteveHere

// ==UserScript==
// @name         Add 'Hide Civ' Button
// @version      0.1
// @namespace    https://gist.github.com/SteveHere/1a19df5242802df3edcc7d34d5c62523
// @description  Injects a 'Hide Civ' button for each dead civ
// @author       SteveHere
// @match        https://mkremins.github.io/epitaph/
// @icon         none
// @grant        none
// ==/UserScript==

adtac

#!/usr/bin/env -S bash -c "docker run -p 8080:8080 -it --rm \$(docker build --progress plain -f \$0 . 2>&1 | tee /dev/stderr | grep -oP 'sha256:[0-9a-f]*')"

# syntax = docker/dockerfile:1.4.0

FROM node:20

WORKDIR /root

RUN npm install sqlite3

briankung

|Question |ChatGPT Response

moyix

#!/bin/bash

gdb -p "$1" -batch -ex 'set {short}$rip = 0x050f' -ex 'set $rax=231' -ex 'set $rdi=0' -ex 'cont'

sadn1ck

Bakkesmod on Arch with proton-tkg 6.5

Packages Needed

└❯ pacman -Qs mingw
local/mingw-w64-binutils 2.35.1-1 (mingw-w64-toolchain mingw-w64)
    Cross binutils for the MinGW-w64 cross-compiler
local/mingw-w64-crt 8.0.0-1 (mingw-w64-toolchain mingw-w64)
    MinGW-w64 CRT for Windows

elliottmorris

#' Description
#' This file runs a live election-night forecast based on The Economist's pre-election forecasting model
#' available at projects.economist.com/us-2020-forecast/president.
#' It is resampling model based on https://pkremp.github.io/update_prob.html.
#' This script does not input any real election results! You will have to enter your picks/constraints manually (scroll to the bottom of the script).
#' 
#' Licence
#' This software is published by *[The Economist](https://www.economist.com)* under the [MIT licence](https://opensource.org/licenses/MIT). The data generated by *The Economist* are available under the [Creative Commons Attribution 4.0 International License](https://creativecommons.org/licenses/by/4.0/).
#' The licences include only the data and the software authored by *The Economist*, and do not cover any *Economist* content or third-party data or content made available using the software. More information about licensing, syndication and the copyright of *Economist* content can be fou

mgalgs

# Print the name and bounding box (x1, y1, x2, y2) for the active window in
# a loop.

import time
from collections import namedtuple

import Xlib
import Xlib.display

keith

#!/bin/bash

set -euo pipefail

xcrun altool --upload-app --type ios --file "path/to/foo.ipa" --username "$ITC_USER" --password "$ITC_PASSWORD"