gregory tomlinson gtzilla

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.

Contract Killer

The popular open-source contract for web designers and developers by Stuff & Nonsense

Originally published: 23/12/2008
Revised date: 15/12/2013
Original post

	# -- mode: ruby --
	# vi: set ft=ruby :

	PASSWORD_PATH = ".password"
	PASSWORD_ID_PATH = ".password_id"

	# Make sure to have installed vagrant-triggers plugin
	# > vagrant plugin install vagrant-triggers

	# After the first `vagrant up` stop the VM and execute the following steps

	exports.arity_decorator = function(fun) {
	function wrapped() {
	if(arguments.length != fun.length)
	throw new Error("Y U NO USE RIGHT?!")
	fun.apply(this, arguments)
	}
	return wrapped;
	}

	> function foo(x,y) {
	if(arguments.length > 2) throw new Error("Stupid no method overloading lameness error");
	... args = Array.prototype.join.call(arguments, ",");
	... console.log('X:' + x + ' Y:' + y + ' Args:' + args)
	... }
	> foo(1,2)
	X:1 Y:2 Args:1,2
	> foo(1,2,3,4,5)
	X:1 Y:2 Args:1,2,3,4,5

	#!/usr/bin/env python
	"""
	Add php-serve.py to your PATH, then, from whatever directory is the root
	of your PHP application, just run:

	$ php-serve.py

	You can optionally specify a port number as an argument. By default,
	port 8000 is used:

	var partial = exports.partial = function partial(fn) {
	var args = arguments,
	func = fn;
	[].shift.apply(args);
	return function() {
	var _args = args;
	[].push.apply(_args, arguments);
	return func.apply(func, _args);
	};
	};

	#!/usr/bin/env python

	### Requires latest boto (cuz I checked in code to boto a moment ago)

	import uuid
	from boto.mturk.connection import MTurkConnection
	from boto.mturk.question import Question, QuestionForm, QuestionContent
	from boto.mturk.question import AnswerSpecification, FreeTextAnswer
	from boto.mturk.question import Overview
	#from boto.mturk.qualification import Qualifications # these exist!

	import pylibmc
	import Queue
	import logging
	import functools

	"""
	This is a transparent pool library that wraps a pylibmc client

	from MemcachePool import mc
	mc.get(key)