Marc Puig mpuig

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.

System Design Cheatsheet

Picking the right architecture = Picking the right battles + Managing trade-offs

Basic Steps

Clarify and agree on the scope of the system

User cases (description of sequences of events that, taken together, lead to a system doing something useful)
- Who is going to use it?
- How are they going to use it?

This is a post by Joel Spolsky. The original post is linked at the bottom.

This is such a common question here and elsewhere that I will attempt to write the world's most canonical answer to this question. Hopefully in the future when someone on answers.onstartups asks how to split up the ownership of their new company, you can simply point to this answer.

The most important principle: Fairness, and the perception of fairness, is much more valuable than owning a large stake. Almost everything that can go wrong in a startup will go wrong, and one of the biggest things that can go wrong is huge, angry, shouting matches between the founders as to who worked harder, who owns more, whose idea was it anyway, etc. That is why I would always rather split a new company 50-50 with a friend than insist on owning 60% because "it was my idea," or because "I was more experienced" or anything else. Why? Because if I split the company 60-40, the company is going to fail when we argue ourselves to death. And if you ju

Install Python

$ brew install readline sqlite gdbm --universal
$ brew install python --universal --framework
$ python --version
Python 2.7.3

Symlinks...

	""" To use: install LLM studio (or Ollama), clone OpenVoice, run this script in the OpenVoice directory
	git clone https://github.com/myshell-ai/OpenVoice
	cd OpenVoice
	git clone https://huggingface.co/myshell-ai/OpenVoice
	cp -r OpenVoice/* .
	pip install whisper pynput pyaudio
	"""

	from openai import OpenAI
	import time

	#!/bin/bash

	echo "This script prints the filenames of any dylibs in your /usr/local/ that depend on the System Python"
	for f in `find /usr/local/lib`; do
	otool -L "$f" 2> /dev/null\| grep Python \| grep System &> /dev/null
	status=$?
	if [ $status -eq 0 ]; then
	echo "$status: $f"
	fi
	done

	#If you don't have Postgres installed yet:
	# https://github.com/tsaleh/tammer-saleh/blob/master/views/posts/installing-postgresql-for-rails-3-1-on-lion.html.textil

	#1. Install postgis 1.5.3. Latest is 2.0.0, so using url old formula on github
	brew install https://raw.github.com/mxcl/homebrew/8a04a43763906e6a9bef68881acf997f3a6f6687/Library/Formula/postgis.rb

	#2. Create a template to be used on creating GIS-enabled databases
	createdb template_postgis

	#3. Import Postgis Data

	$ wget http://people.centos.org/tru/devtools-2/devtools-2.repo -O /etc/yum.repos.d/devtools-2.repo
	$ yum install devtoolset-2-gcc devtoolset-2-binutils
	$ yum install devtoolset-2-gcc-c++ devtoolset-2-gcc-gfortran

	// Use Gists to store code you would like to remember later on
	console.log(window); // log the "window" object to the console