Zach Williams zachwill

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.

Slack System Design: A Grounded Teardown

A reverse-engineered system design of Slack's web application, built from live network traffic analysis of the authenticated Enterprise Grid experience. 200+ API calls captured across boot, search, messaging, reactions, and navigation. Every backend service named.

Architecture Overview

┌─────────────────────────────────────────────────────────────────────────────┐
│  BROWSER (Gantry v2 SPA)                                                    │
│                                                                             │

https://x.com/VictorTaelin/status/2001777678765129832

Opus 4.5 - chat log

 * ▐▛███▜▌ *   Claude Code v2.0.72
* ▝▜█████▛▘ *  Opus 4.5 · Claude Max
 *  ▘▘ ▝▝  *   ~/vic/dev/hvm4-claude

> # Task 1: print unscoped lambdas and floating dups with matching names

Why did you do this?

Relax, I only have one Sunday to work on idea, literally my weekend project. So I tried Deepseek to see if it can help. Surprisingly, it works and it saves me another weekend...

What is your setup?

Just chat.deepseek.com (cost = free) with prompts adapted from this gist.

Does it work in one-shot or I have to prompt it multiple times?

Add it to the body

<body hx-ext="hx-astro-view-transition">

An example

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

	import type { ExtensionAPI } from "@mariozechner/pi-coding-agent";
	import { spawnSync } from "node:child_process";
	import { randomUUID } from "node:crypto";
	import { mkdirSync, readFileSync, writeFileSync } from "node:fs";
	import { tmpdir } from "node:os";
	import { basename, join } from "node:path";

	const PLACEHOLDER = "[Review feedback]";
	const CUSTOM_PENDING = "review-feedback-pending";
	const CUSTOM_RESOLVED = "review-feedback-resolved";

	[
	"fa-classic fa-solid fa-0 fa-fw",
	"fa-classic fa-solid fa-1 fa-fw",
	"fa-classic fa-solid fa-2 fa-fw",
	"fa-classic fa-solid fa-3 fa-fw",
	"fa-classic fa-solid fa-4 fa-fw",
	"fa-brands fa-42-group fa-fw",
	"fa-classic fa-solid fa-5 fa-fw",
	"fa-brands fa-500px fa-fw",
	"fa-classic fa-solid fa-6 fa-fw",

	# /// script
	# requires-python = ">=3.12"
	# dependencies = [
	# "llm",
	# "textual",
	# ]
	# ///
	from textual import on, work
	from textual.app import App, ComposeResult
	from textual.widgets import Header, Input, Footer, Markdown

	import json
	import logging
	import re
	from typing import Any, AsyncGenerator, Optional, Union

	import aiohttp
	import openai
	from azure.search.documents.aio import SearchClient
	from azure.search.documents.models import QueryType