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.

Molochnikov

$test = @'
namespace Microsoft.Samples.PowerShell.Runspaces
{
 using System;
  using System.Collections.Generic;
  using System.Diagnostics;
  using System.Management.Automation;
  using System.Management.Automation.Runspaces;
  using PowerShell = System.Management.Automation.PowerShell;
    #region GetProcCommand

azidanit

#!/usr/bin/env python3
"""
ventoy-macos-install.py - Install Ventoy on a USB drive from macOS.

This script installs Ventoy on a USB drive without requiring macFUSE,
Linux, or a VM. It writes the GPT partition table, boot code, and
EFI partition image directly to the raw block device.

Usage:
    sudo python3 ventoy-macos-install.py /dev/diskN [--exfat|--ntfs] [--ventoy-version VERSION]

ISawSau

#!/usr/bin/env python3
"""Fix known false-positive MusicBrainz matches + scan for suspicious ones."""
import os, re, sys
from pathlib import Path
from mutagen.mp3 import MP3
from mutagen.id3 import ID3, TIT2, TPE1

# Your corrections: { "Wrong Artist": ("Correct Artist", "Correct Title (optional)") }
CORRECTIONS = {
    "Dio": ("Jordana, TV Girl", "Better in the Dark"),

yasirkula

using System.Collections;
using UnityEngine;
using UnityEngine.UI;

public static class ScrollViewFocusFunctions
{
	public static Vector2 CalculateFocusedScrollPosition( this ScrollRect scrollView, Vector2 focusPoint )
	{
		Vector2 contentSize = scrollView.content.rect.size;
		Vector2 viewportSize = ( (RectTransform) scrollView.content.parent ).rect.size;

jhorsman

#Source www.cze.cz
#This script is tested with "Cisco AnyConnect Secure Mobility Client version 3.1.00495"
# Usage: & '.\Cisco_Anyconnect.ps1' [-Server <server name or ip>] [-Group <group>] [-User <user>] [-Password <password>]


#Please change following variables

#IP address or host name of cisco vpn, Username, Group and Password as parameters
param (
  [string]$Server = $( Read-Host "Input server, please" ),