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.

bensheldon

# frozen_string_literal: true
require "fileutils"

# Helper for Git worktree-aware behavior
class GitWorktree
  PROJECT_ROOT = File.expand_path("..", __dir__)

  def self.instance = new

  def self.name = instance.name

jjb

For years, people have been using jemalloc with ruby. There were various benchmarks and discussions. Legend had it that Jemalloc 5 didn't work as well as Jemalloc 3.

Then, one day, hope appeared on the horizon. @wjordan offered a config for Jemalloc 5.

Ubuntu/Debian

FROM ruby:3.1.2-bullseye
RUN apt-get update ; \

chrisbloom7

# frozen_string_literal: true

# Basic implementation of a websocket connector for Slack using the slack-ruby-client gem
# 
# Run this rake task in a virtual container that is set to automatically restart on failure; this
# will help deal with disconnects from Slack since socket URLs are temporary.
# https://api.slack.com/apis/connections/socket-implement#disconnect
#
# This rake task can be called in multiple virtual containers to help with resilliancy and rolling restarts
# https://api.slack.com/apis/connections/socket-implement#connections

dentarg

# Need to use ~/.zshenv, ~/.zprofile is not sourced when doing "arch -x86_64 /bin/zsh"
#
# From 'man zsh' (zsh 5.8.1 (x86_64-apple-darwin21.0))
#
#       Commands are then read from $ZDOTDIR/.zshenv.  If the shell is a login
#       shell, commands are read from /etc/zprofile and then $ZDOTDIR/.zprofile.
#       Then, if the shell is interactive, commands are read from /etc/zshrc and
#       then $ZDOTDIR/.zshrc.  Finally, if the shell is a login shell,
#       /etc/zlogin and $ZDOTDIR/.zlogin are read.

simi

class MyEnumerator
  attr_reader :scope, :block, :opened, :headers

  def initialize(scope, headers, &block)
    @scope = scope
    @block = block
    @opened = false
    @cursor = PostgreSQLCursor::Cursor.new(scope.send(:to_unprepared_sql))
    @headers = headers
    @headers_provided = false

simi

# INFO: works well with Ruby 3.0.1
# hangs on any Ruby including https://github.com/ruby/ruby/commit/f9196de1dee6f5ab8b6fe115070b92775a3500fe
# since the yielding code in each method is running in different Fiber (due to be wrapped into Enumerator and next is used)

require "bundler/inline"

gemfile(true) do
  source "https://rubygems.org"

  git_source(:github) { |repo| "https://github.com/#{repo}.git" }

otavioschwanck

Why install in this way?

To fix some problems with the arm version, like puma SSL.

if you problem is only SSL, try it:

Remove gem puma: gem uninstall puma

And run bundle with DISABLE_SSL=1 bundle install

dentarg

#!/usr/bin/env ruby

if %w(-h --help).include?(ARGV.first)
  abort <<~USAGE
    #{__FILE__} <FILE> to count IPs in given file
    #{__FILE__}        to count IPs using latest file on disk
    DL=1 #{__FILE__}   to download the AWS IP ranges
  USAGE
end

dhh

/*
  ERB template chunk from The Feed's display of emails:

  <section class="postings postings--feed-style" id="postings"
      data-controller="pagination" data-pagination-root-margin-value="40px">
    <%= render partial: "postings/snippet", collection: @page.records, as: :posting, cached: true %>

    <%= link_to(spinner_tag, url_for(page: @page.next_param),
          class: "pagination-link", data: { pagination_target: "nextPageLink", preload: @page.first? }) unless @page.last? %>
  </section>