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.

pfreixes

The numbers claimed by this benchamark about Gevent [1] comparaed with the numbers got by Asyncio with the uvloop and even with the default loop has left me a bit frozen. Ive repeated a few of them : gevent, asyncio, asyncio-uvloop and go for the echo server and these are the numbers roughly:

For gevent

$ ./echo_client
685393 0.98KiB messages in 30 seconds
Latency: min 0.04ms; max 4.48ms; mean 0.126ms; std: 0.048ms (37.68%)
Latency distribution: 25% under 0.088ms; 50% under 0.122ms; 75% under 0.158ms; 90% under 0.182ms; 99% under 0.242ms; 99.99% under 0.91ms

steinim

FIELDS = ['cmd', 'command', 'start', 'end', 'delta', 'msg', 'stdout', 'stderr']

def human_log(res):

    if type(res) == type(dict()):
        for field in FIELDS:
            if field in res.keys():
                encoded_field = res[field].encode('utf-8')
                print '\n{0}:\n{1}'.format(field, encoded_field)

Kartones

PSQL Cheatsheet

Magic words:

psql -U postgres

Some interesting flags (to see all, use -h or --help depending on your psql version):

• -E: will describe the underlaying queries of the \ commands (cool for learning!)
• -l: psql will list all databases and then exit (useful if the user you connect with doesn't has a default database, like at AWS RDS)

lazarofl

#!/bin/bash
 
#Force file syncronization and lock writes
mongo admin --eval "printjson(db.fsyncLock())"
 
MONGODUMP_PATH="/usr/bin/mongodump"
MONGO_HOST="prod.example.com"
MONGO_PORT="27017"
MONGO_DATABASE="dbname"
 

adrienbrault

#!/bin/sh

# Credits to:
#  - http://vstone.eu/reducing-vagrant-box-size/
#  - https://github.com/mitchellh/vagrant/issues/343

aptitude -y purge ri
aptitude -y purge installation-report landscape-common wireless-tools wpasupplicant ubuntu-serverguide
aptitude -y purge python-dbus libnl1 python-smartpm python-twisted-core libiw30
aptitude -y purge python-twisted-bin libdbus-glib-1-2 python-pexpect python-pycurl python-serial python-gobject python-pam python-openssl libffi5