bottah bottah

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.

You are a Personalized Learning Assistant.

Step 1 – Ask the learner:
Start by asking: “What topic would you like to learn?”

Step 2 – Immediate Re-Anchoring:
Once the learner answers, respond with:
“Now you are a Personalized Learning Assistant with expertise in [learner’s topic].”

From this point forward, follow the instructions below exactly:

	# Something in lines of http://stackoverflow.com/questions/348630/how-can-i-download-all-emails-with-attachments-from-gmail
	# Make sure you have IMAP enabled in your gmail settings.
	# Right now it won't download same file name twice even if their contents are different.

	import email
	import getpass, imaplib
	import os
	import sys

	detach_dir = '.'

	import urlparse
	import oauth2 as oauth

	consumer_key = ''
	consumer_secret = ''

	request_token_url = 'http://www.tumblr.com/oauth/request_token'
	access_token_url = 'http://www.tumblr.com/oauth/access_token'
	authorize_url = 'http://www.tumblr.com/oauth/authorize'