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

# -----------------------------------------------------------------------------
# AI-powered Git Commit Function
# Copy paste this gist into your ~/.bashrc or ~/.zshrc to gain the `gcm` command. It:
# 1) gets the current staged changed diff
# 2) sends them to an LLM to write the git commit message
# 3) allows you to easily accept, edit, regenerate, cancel
# But - just read and edit the code however you like
# the `llm` CLI util is awesome, can get it here: https://llm.datasette.io/en/stable/

gcm() {

wch

library(shiny)

# Example usage:
# lmGadget(mtcars, "wt", "mpg")
#
# Returns a list with two items:
#   $data: Data with excluded rows removed.
#   $model: lm (model) object.
lmGadget <- function(data, xvar, yvar) {
  library(miniUI)

tbates

The Difference Pool

1. Subjects: Population sample

• N must be over 500, preferably 2,000
• Ideally genetically informative

6. Recruitment

• Students
• Public advertisements

2. Content: There will be a set of basic features that will be tested at the outset

• Need to compromise on this. Perhaps 2-hours of testing in the first instance?

dmarcelinobr

# rm(list = ls())  # clear objects
# graphics.off()   # close graphics windows
 plot.new() # call new plot window 
x = seq(-5,5, length=250)
y = dnorm(x)
plot(x,y, las=1, ylab='dnorm', type='n', yaxs='i', ylim=c(0, 0.5))
x2 = seq(qnorm(0.95), 5, length=50)
y2 = dnorm(x2)
polygon(c(x2[1], x2, x2[length(x2)]),  c(0, y2, 0), border=NA, col='grey')
lines(x, y)

brandonb927

#!/bin/sh

###
# SOME COMMANDS WILL NOT WORK ON macOS (Sierra or newer)
# For Sierra or newer, see https://github.com/mathiasbynens/dotfiles/blob/master/.macos
###

# Alot of these configs have been taken from the various places
# on the web, most from here
# https://github.com/mathiasbynens/dotfiles/blob/5b3c8418ed42d93af2e647dc9d122f25cc034871/.osx

jeromyanglim

This post examines the features of [R Markdown](http://www.rstudio.org/docs/authoring/using_markdown) 
using [knitr](http://yihui.name/knitr/) in Rstudio 0.96.
This combination of tools provides an exciting improvement in usability for 
[reproducible analysis](http://stats.stackexchange.com/a/15006/183).
Specifically, this post 
(1) discusses getting started with R Markdown and `knitr` in  Rstudio 0.96;
(2) provides a basic example of producing console output and plots using R Markdown;
(3) highlights several code chunk options such as caching and controlling how input and output is displayed;
(4) demonstrates use of standard Markdown notation as well as the extended features of formulas and tables; and 
(5) discusses the implications of R Markdown.