karpathy

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.

dvf

If you're using a high-end bluetooth headset on your Macbook Pro it's likely your mac is using an audio codec which favors battery efficiency over high quality. This results in a drastic degradation of sound, the SBC codec is the likely culprit, read more about it here.

Find out what codec you're using

1. Play a song on your headphones
2. Option (⌥) click the Bluetooth button at the top of your screen
3. If you're using AAC or aptX, you can stop here—those are the highest quality codecs.

Change your codec to AAC or aptX

r35krag0th

#!/usr/bin/env bash

# NOTE: This requires you to have Homebrew installed
#       --> https://brew.sh/

# NOTE: You need to have Homebrew's GCC installed
#       --> brew install gcc

wget ftp://ftp.cpc.ncep.noaa.gov/wd51we/wgrib2/wgrib2.tgz
tar xvzf wgrib2.tgz

bradwilson

# Disables touchscreen emulation
touchscreen.vusb.present = "FALSE"

# Enables support for nested hypervisors
hypervisor.cpuid.v0 = "FALSE"
vhv.enable = "TRUE"
vpmc.enable = "TRUE"
mce.enable = "TRUE"
vhu.enable = "TRUE"