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.

mnpenner

#!/usr/bin/env bash
{
  set -euo pipefail

  # --- Re-run inside WSL when invoked from Windows shells ---
  if [[ -z "${WSL_INTEROP-}" && -z "${WSL_DISTRO_NAME-}" ]]; then
    exec wsl -d Ubuntu-24.04 -- /bin/bash -lc "aijail $*"
  fi

  # --- Verify HOME (we bindmount a bunch of $HOME paths) ---

rain-1

How large are large language models? (2025)

This aims to be factual information about the size of large language models. None of this document was written by AI. I do not include any information from leaks or rumors. The focus of this document is on base models (the raw text continuation engines, not 'helpful chatbot/assistants'). This is a view from a few years ago to today of one very tiny fraction of the larger LLM story that's happening.

History

• GPT-2,-medium,-large,-xl (2019): 137M, 380M, 812M, 1.61B. Source: openai-community/gpt2. Trained on the unreleased WebText dataset said to 40GB of Internet text - I estimate that to be roughly 10B tokens. You can see a list of the websites that went into that data set here domains.txt.
• GPT-3 aka davinci, davinci-002 (2020): 175B parameters. There is a good breakdown of how those parameters are 'spent' here [How d

aravindanve

Bypass disable-devtool

(Working as of 2025-02-09)

There are websites that use disable-devtool to prevent you from opening or using devtools. They typically prevent you from right clicking or using the keyboard shortcut to open devtools. Even if you successfully do so, they detect it and redirect you elsewhere. You can bypass this by using one of the following ways.

Opening devtools

If the shortcut F12 on Windows or Option + ⌘ + I on Mac do not work. Press the three vertically aligned dots in the top right corner of your Google Chrome or Microsoft Edge window. Under the section "More Tools", you'll see the option to select "Developer Tools" which opens the toolkit in your window.

voberoi

This is the prompt I use for speaker identification in citymeetings.nyc as of March 23rd, 2024 -- the date of my NYC School of Data talk.

rcarmo

# Drop this into Automator using Python 3 as the shell

from sys import stdin
from json import dumps, loads
from urllib.parse import urlencode
from urllib.request import Request, urlopen

AZURE_ENDPOINT="getyourowndeployment.openai.azure.com"
OPENAI_API_KEY="getyourownkey"
OPENAI_API_VERSION="2023-05-15"

rosslazer

import json
import os
from collections import deque

root_dir = '~/slack_messages'

target_user = "UMQN69N8M"
res = []

for dirpath, dirnames, filenames in os.walk(root_dir):

lleyton

Open Source Business Challenges and Reality

Original Japanese note here.

Original Author: Rui Ueyama (creator of the mold linker)

Translated by @windowsboy111

Minimally edited by @lleyton

yzdbg

Automating Daily Reports, because fuck it, really...

Each day at our company, developers are required to document their activities, painstakingly jotting down their daily work and future plans. A monotonous chore that I just really dislike.

So now, there's a scribe for that :

Code

simonw

{"ISO 693-3":"aaa","Language Name":"Ghotuo","Language Family":"Niger-Congo","Subgrouping":"Atlantic-Congo","ASR":false,"TTS":false,"LID":true,"":"https://www.ethnologue.com/language/aaa"}
{"ISO 693-3":"aab","Language Name":"Arum","Language Family":"Niger-Congo","Subgrouping":"Atlantic-Congo","ASR":false,"TTS":false,"LID":true,"":"https://www.ethnologue.com/language/aab"}
{"ISO 693-3":"aac","Language Name":"Ari","Language Family":"Trans-New Guinea","Subgrouping":"Gogodala-Suki","ASR":false,"TTS":false,"LID":true,"":"https://www.ethnologue.com/language/aac"}
{"ISO 693-3":"aad","Language Name":"Amal","Language Family":"Sepik","Subgrouping":"Iwam","ASR":false,"TTS":false,"LID":true,"":"https://www.ethnologue.com/language/aad"}
{"ISO 693-3":"aaf","Language Name":"Aranadan","Language Family":"Dravidian","Subgrouping":"Southern","ASR":false,"TTS":false,"LID":true,"":"https://www.ethnologue.com/language/aaf"}
{"ISO 693-3":"aai","Language Name":"Miniafia Oyan","Language Family":"Austronesian","Subgrouping":"Malayo-Pol