Kamen Dimitrov kamend

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.

For some reason, it is surprisingly hard to create a bootable Windows USB using macOS. These are my steps for doing so, which have worked for me in macOS Monterey (12.6.1) for Windows 10 and 11. After following these steps, you should have a bootable Windows USB drive.

1. Download a Windows disc image (i.e. ISO file)

You can download Windows 10 or Windows 11 directly from Microsoft.

2. Identify your USB drive

After plugging the drive to your machine, identify the name of the USB device using diskutil list, which should return an output like the one below. In my case, the correct disk name is disk2.

Creating a "DMG installer" for OS X

A DMG Installer is convenient way to provide end-users a simple way to install an application bundle. They are basically a folder with a shortcut to the Applications directory but they can be customized with icons, backgrounds, and layout properties. A DMG file (.dmg) is a Mac OS X Disk Image file and it is used to package files or folders providing compression, encryption, and read-only to the package.

##Creating the DMG file #Disk Utility

	You are a GPT GPT-4 architecture, based on the GPT-4 architecture.
	Knowledge cutoff: 2023-04
	Current date: 2023-12-11

	Image input capabilities: Enabled

	# Tools

	## python

	using System.Collections.Generic;
	using UnityEngine;
	using UnityEngine.UIElements;
	using UnityEngine.EventSystems;
	using UnityEngine.Rendering;

	namespace Katas.Experimental
	{
	public class WorldSpaceUIDocument : MonoBehaviour, IPointerMoveHandler, IPointerUpHandler, IPointerDownHandler,
	ISubmitHandler, ICancelHandler, IMoveHandler, IScrollHandler, ISelectHandler, IDeselectHandler, IDragHandler

	Shader "Toguchi/DepthFlare"
	{
	Properties
	{
	[HDR] _Color ("Color", COLOR) = (1, 1, 1, 1)
	_MainTex ("Texture", 2D) = "white" {}
	}
	SubShader
	{
	Tags

	#! /bin/bash
	# This command converts audio according to the specifications listed in the Spark docs:
	# https://sparkar.facebook.com/ar-studio/learn/documentation/docs/audio
	# mono m4a, 44.1kHz sample rate, 16-bit-depth resolution
	# Usage:
	# convert-audio.sh myaudio.mp3 converted.m4a
	# Notes:
	# Always use m4a for output file type
	# Change "64k" to a higher value to improve bitrate/quality. e.g. 96k 128k 192k

	using com.mindshaft.overtime.collision;
	using UnityEngine;

	namespace com.mindshaft.overtime.physics {
	public class RaycastCollisionDetection : IEntityCollisionDetection {
	private BoxCollider _collider;
	private Rect _collisionRect;
	private LayerMask _collisionMask;
	private LayerMask _playerMask;

	#!/bin/bash

	#
	# SignAndPackageForMAS.sh v1.0.1
	#
	# Codesign an app bundle and build a signed package ready for submission to the
	# Mac App Store
	#
	# Copyright (c) 2014 Felix Schwarz (@felix_schwarz), IOSPIRIT GmbH (@iospirit)
	#

	#! /bin/bash
	#
	# simple script to check for a given string pattern in any
	# .c, .h, .cpp, .cxx, .m, .mm, etc source files
	#
	# you can pass your own list of extensions/file patterns as well:
	# check-for-src-pattern . "foo" .h .cpp .py .sh Makefile
	#
	# Dan Wilcox <danomatika@gmail.com> 2014, 2016
	#