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.

karpathy

"""
stable diffusion dreaming
creates hypnotic moving videos by smoothly walking randomly through the sample space

example way to run this script:

$ python stablediffusionwalk.py --prompt "blueberry spaghetti" --name blueberry

to stitch together the images, e.g.:
$ ffmpeg -r 10 -f image2 -s 512x512 -i blueberry/frame%06d.jpg -vcodec libx264 -crf 10 -pix_fmt yuv420p blueberry.mp4

edi33416

# gdb configuration file
# place contents in ~/.gdbinit
source /home/fawkes/ws/pwndbg/gdbinit.py

# auto-load directory in which we run gdb, so we can have a
# .gdbinit per project directory
add-auto-load-safe-path .

# path to your library. Multiple dirs can be separated by ':' as in the comment below
set directories /home/fawkes/ws/glibc/glibc-2.31/stdio-common/

gwerbin

"""
Export a Conda environment with --from-history, but also append
Pip-installed dependencies

Exports only manually-installed dependencies, excluding build versions, but
including Pip-installed dependencies.

Lots of issues requesting this functionality in the Conda issue tracker,
e.g. https://github.com/conda/conda/issues/9628

cbilson

;;; ob-powershell.el --- org-babel functions for powershell evaluation

;; Authors: Chris Bilson
;; Keywords: literate programming, reproducible research
;; Homepage: http://orgmode.org

;;; Commentary:

;; Org-Babel support for evaluating powershell source code.

monperrus

#!/usr/bin/python3
"""
Creates an RSS feed from DBLP JSON data. 
Meant to be used as CGI script but can also be used as command-line.
Feedback and pull request welcome.
Martin Monperrus
March 2016
Licence: public domain
URL: https://gist.github.com/monperrus/978079d39c1cc7b4cbae78ddd1b8ed99
Operation: https://www.monperrus.net/martin/dblp-rss.py?search=venue:FASE:

savanovich

define pyp
set \$\_unused\_void = PyObject\_Print($arg0, stderr, 1)
printf "\n"
end

netj

#!/usr/bin/env bash
# memusg -- Measure memory usage of processes
# Usage: memusg COMMAND [ARGS]...
#
# Author: Jaeho Shin <netj@sparcs.org>
# Created: 2010-08-16
############################################################################
# Copyright 2010 Jaeho Shin.                                               #
#                                                                          #
# Licensed under the Apache License, Version 2.0 (the "License");          #