osman - オスマン osbm

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.

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

Pre-Transformer Models

ChatGPT Resources

Context

ChatGPT appeared like an explosion on all my social media timelines in early December 2022. While I keep up with machine learning as an industry, I wasn't focused so much on this particular corner, and all the screenshots seemed like they came out of nowhere. What was this model? How did the chat prompting work? What was the context of OpenAI doing this work and collecting my prompts for training data?

I decided to do a quick investigation. Here's all the information I've found so far. I'm aggregating and synthesizing it as I go, so it's currently changing pretty frequently.

	let
	budgetIsDefined = value: (builtins.tryEval value).success;
	in

	import <nixpkgs/nixos> {
	configuration =
	{ config, lib, ... }:
	{
	imports = [
	{

	"""Hello world, with a genetic algorithm.

	https://twitter.com/matthen2/status/1769368467067621791
	"""

	import random
	import time
	from dataclasses import dataclass
	from itertools import chain
	from typing import Iterable, List

	{
	config,
	pkgs,
	options,
	...
	}: let
	hostname = "oatman-pc"; # to alllow per-machine config
	in {
	networking.hostName = hostname;

	#!/bin/bash

	# Colors
	RED='\033[0;31m'
	GREEN='\033[0;32m'
	NO_COLOR='\033[0m'
	BLUE='\033[0;34m'
	YELLOW='\033[0;33m'
	NO_COLOR='\033[0m'

	{ pkgs, ... }:
	{
	# ...
	nix.settings.system-features = [ "expose-cuda" ];
	nix.settings.pre-build-hook = pkgs.writers.writePython3 "nix-pre-build.py" { } (builtins.readFile ./nix-pre-build-hook.py);
	# ...
	}

	var geometry =
	/* color: #ffc82d */
	/* shown: false */
	ee.Geometry.Polygon(
	[[[36.803785364345735, 37.6022824651851],
	[36.84292415829105, 37.569634979372736],
	[36.8896160528223, 37.548134235962486],
	[36.96171383114261, 37.54377889953015],
	[36.99741939754886, 37.565280899470345],
	[36.97819332333011, 37.58759286852694],

	# study stream aliases
	# Requires https://github.com/caarlos0/timer to be installed. spd-say should ship with your distro

	declare -A pomo_options
	pomo_options["work"]="45"
	pomo_options["break"]="10"

	pomodoro () {
	if [ -n "$1" -a -n "${pomo_options["$1"]}" ]; then
	val=$1