Ryan Heneise ryenski

LLM Wiki v2

A pattern for building personal knowledge bases using LLMs. Extended with lessons from building agentmemory, a persistent memory engine for AI coding agents.

This builds on Andrej Karpathy's original LLM Wiki idea file. Everything in the original still applies. This document adds what we learned running the pattern in production: what breaks at scale, what's missing, and what separates a wiki that stays useful from one that rots.

What the original gets right

The core insight is correct: stop re-deriving, start compiling. RAG retrieves and forgets. A wiki accumulates and compounds. The three-layer architecture (raw sources, wiki, schema) works. The operations (ingest, query, lint) cover the basics. If you haven't read the original, start there.

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.

Disable HTML Form Input Autocomplete and Autofill

Add autocomplete="off" onto <form> element;
Add hidden <input> with autocomplete="false" as a first children element of the form.

<form autocomplete="off" method="post" action="">
    <input autocomplete="false" name="hidden" type="text" style="display:none;">
    ...

You'll need to add a whole slew of hosts for intercom if you've got CSP in place (which you should).

Heres what you'll need to add:

connect-src: https://api-ping.intercom.io https://nexus-websocket-a.intercom.io https://nexus-websocket-b.intercom.io wss://nexus-websocket-a.intercom.io wss://nexus-websocket-b.intercom.io https://api-iam.intercom.io 
script-src: https://widget.intercom.io https://js.intercomcdn.com
image-src: https://js.intercomcdn.com


	# == Schema Information
	#
	# Table name: activities
	#
	# id :uuid not null, primary key
	# user_id :uuid
	# tenant_id :uuid
	# trackable_type :string
	# trackable_id :uuid

	# app/controllers/application_controller.rb

	class ApplicationController < ActionController::Base
	before_action :authenticate_user!
	include CurrentTenant
	end

	<tr data-action="click->table#click" data-controller="table">
	<td> 2018-01-11 </td>
	<td>
	<a data-target="table.url" href="/invoices/1">$7,000.00</a>
	</td>
	</tr>

	class Contact < ActiveRecord::Base

	...

	def after_create
	if Hook.hooks_exist?('new_contact', self)
	Resque.enqueue(Hook, self.class.name, self.id)
	# To trigger directly without Resque: Hook.trigger('new_contact', self)
	end
	end

	# Support for Rspec / Capybara subdomain integration testing
	# Make sure this file is required by spec_helper.rb
	# (e.g. save as spec/support/subdomains.rb)

	def switch_to_subdomain(subdomain)
	# lvh.me always resolves to 127.0.0.1
	hostname = subdomain ? "#{subdomain}.lvh.me" : "lvh.me"
	Capybara.app_host = "http://#{hostname}"
	end

	window.addEventListener "trix-initialize", (e) =>
	Utility.TrixMentions.prepare($(e.target))