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.

yayuniversal

#!/bin/bash
BS=64M
ROOT_DEV=/dev/mmcblk0
BOOTFS_BACKUP=${ROOT_DEV}p3
BOOTFS_TARGET=${ROOT_DEV}p1
ROOTFS_BACKUP=${ROOT_DEV}p2
ROOTFS_TARGET=${ROOT_DEV}p4

print_yellow() {
    echo -e "\033[1;33m${1}\033[0m"

purcell

COPY (
WITH
  -- Read the raw log line by line by abusing CSV parser
  raw_log AS (
    FROM read_csv_auto('/tmp/log/access.log-20230904.gz', header=false, delim='\0')
)
, combined_log AS ( 
    SELECT regexp_extract(column0
                         , '^(\S+) (\S+) (\S+) \[(.*?)\] "([A-Z]+?) (.*?) HTTP/(.*?)" (\d+) (\d+) "(.*?)" "(.*?)"$'
                         , [ 'ip', 'identity', 'userid', 'timestamp', 'method'

Aaronius

// Queues multiple asynchronous processes to run sequentially via promises
// Process can be added to the queue while other processes are running

var Sequence = function() {};
Sequence.prototype.enqueue = function(fn) {
  this.tail = this.tail ? this.tail.finally(fn) : fn();
};

// Usage