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.

udapy

ROLE

You are a Senior Principal Engineer and "Context Manager." You reject "vibes-based" coding. You believe that writing code without a compressed understanding of the codebase leads to "slop" and hallucinations.

CORE PROTOCOL: RPI WORKFLOW

You must strictly follow the Research -> Plan -> Implement workflow. Do not skip steps. Do not jump to code generation until the Plan is approved.

PHASE 1: RESEARCH (Discovery & Compression)

• Goal: Build a mental map of the existing system and dependencies.
• Action: Scan the provided files or ask the user to provide specific files.
• Output: A "Research Summary" containing:

dustingetz

PostgreSQL hello world — Electric Clojure

How do I integrate a SQL database backend?

Easy:

• make an ordinary Clojure function query-pokemon-list for the query
• The query is blocking, and Electric Clojure is async, so use e/offload to move it to a thread pool.
  • (Don't block the event loop!)
• e/offload throws Pending until the query finishes, and then the exception "goes away"

0penBrain

curl -I -k "https://api.github.com/repos/:owner/:repo/commits?per_page=1" | sed -n '/^[Ll]ink:/ s/.*"next".*page=\([0-9]*\).*"last".*/\1/p'

### And that's all !
# I saw many fighting with finding first commit SHA or similar fancy thing.
# Here we just rely on the GH API, asking commits at 1 per page and parsing the last page number in the header of the reply (whose body only holds the last commit !)
# So this is robust and bandwidth efficient. :)

# If one want the commit count of a specific SHA, just use :
curl -I -k "https://api.github.com/repos/:owner/:repo/commits?per_page=1&sha=:sha" | sed -n '/^[Ll]ink:/ s/.*"next".*page=\([0-9]*\).*"last".*/\1/p'

ArturT

# .github/workflows/main.yaml
name: Main

on: [push]

jobs:
  test:
    runs-on: ubuntu-latest

    # If you need DB like PostgreSQL, Redis then define service below.

gdestree

Never touch your local /etc/hosts file in OS X again

UPDATED for .test

To setup your computer to work with *.test domains, e.g. project.test, awesome.test and so on, without having to add to your hosts file each time.

Requirements

• Homebrew
• Mountain Lion -> High Sierra

brablc

Never touch your local /etc/hosts file in OS X again

To setup your computer to work with *.test domains, e.g. project.test, awesome.test and so on, without having to add to your hosts file each time.

Requirements

• Homebrew
• Mountain Lion -> High Sierra

Install

hadees

module ArelHelpers
  extend self

  def self.included(base)
    base.extend self
  end

  def asterisk(arel_table_or_model)
    arel_table, columns = case arel_table_or_model
                          when Arel::Table

kuboon

gem 'acts-as-taggable-on'
gem 'simple_form'
gem 'select2-rails'

henrik

# Without this fix, downgrading from Rails 4 to Rails 3 causes session cookies to blow up.
#
# The way the flash is stored in the session changed in a backwards-incompatible way.

if Rails::VERSION::MAJOR == 3

  module ActionDispatch
    class Flash
      def call(env)
        if (session = env['rack.session']) && (flash = session['flash'])