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.

lelandbatey

Description

This simple script will take a picture of a whiteboard and use parts of the ImageMagick library with sane defaults to clean it up tremendously.

The script is here:

#!/bin/bash
convert "$1" -morphology Convolve DoG:15,100,0 -negate -normalize -blur 0x1 -channel RBG -level 60%,91%,0.1 "$2"

Results

adamwiggins

It's with a heavy heart that I announce that Friday, May 31 2013 will be my last day at Heroku.

How can I possibly put into words what Heroku has meant to me these last six years? I can say it was a tremendous experience; or the opportunity of a lifetime; or the greatest thing I have ever been a part of. I can say that Heroku has been my life's work, as I did recently in a public blog post. All of those things are true, but none seem to capture the enormity of what's transpired these past six years.

I tend to focus on mechanical elements of a company: product, code, design, process. But what has surprised me the most at Heroku is that none of these things is the best part. The best part is the team.

I've never had the chance to work with a more singular group of people. Talented, passionate, skilled, dedicated. Most of all, sharing a set of values: elegance, craft, maniacal focus on simplicity; and an uncompromising belief that the future will be made of software, and how that software gets made will shape