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.

ifree

from pywinauto.application import Application
import time
import math
import re
from collections import Counter

WORD = re.compile(r"\w+")
# cosine similarity https://stackoverflow.com/a/15174569/309846

def get_cosine(vec1, vec2):

goldengrape

// WARNING：此脚本仅做学习和演示用途，在不了解其用途前不建议使用
// 本脚本的用途是将输入内容分页，每次提取一页内容，编辑第二条消息，发送，然后收集结果
// 使用前，需要有两条消息，参考模板 https://chat.openai.com/share/17195108-30c2-4c62-8d59-980ca645f111
// 演示视频： https://www.bilibili.com/video/BV1tp4y1c7ME/?vd_source=e71f65cbc40a72fce570b20ffcb28b22
// 
(function (fullText) {
  const wait = (ms) => new Promise((resolve) => setTimeout(resolve, ms));

  const groupSentences = (fullText, maxCharecters = 2800) => {
    const sentences = fullText.split("\n").filter((line) => line.trim().length > 0);

JimLiu

// WARNING：此脚本仅做学习和演示用途，在不了解其用途前不建议使用
// 本脚本的用途是将输入内容分页，每次提取一页内容，编辑第二条消息，发送，然后收集结果
// 使用前，需要有两条消息，参考模板 https://chat.openai.com/share/17195108-30c2-4c62-8d59-980ca645f111
// 演示视频： https://www.bilibili.com/video/BV1tp4y1c7ME/?vd_source=e71f65cbc40a72fce570b20ffcb28b22
// 
(function (fullText) {
  const wait = (ms) => new Promise((resolve) => setTimeout(resolve, ms));

  const groupSentences = (fullText, maxCharecters = 2800) => {
    const sentences = fullText.split("\n").filter((line) => line.trim().length > 0);

goldengrape

from langchain.callbacks.base import BaseCallbackHandler
import azure.cognitiveservices.speech as speechsdk
import os
import base64
import time
class StreamDisplayHandler(BaseCallbackHandler):
    def __init__(self, container, initial_text="", display_method='markdown'):
        self.container = container
        self.text = initial_text
        self.display_method = display_method

ninehills

ninehills

IanColdwater

Mute these words in your settings here: https://twitter.com/settings/muted_keywords

ActivityTweet
generic_activity_highlights
generic_activity_momentsbreaking
RankedOrganicTweet
suggest_activity
suggest_activity_feed
suggest_activity_highlights
suggest_activity_tweet

Piasy

brew options ffmpeg

brew install ffmpeg \
--with-chromaprint \
--with-fdk-aac \
--with-fontconfig \
--with-freetype \
--with-frei0r \
--with-game-music-emu \
--with-libass \

rbnpi

#Program to drive Sonic Pi 3 visualiser written in "processing"
#by Robin Newman, September 2017
#see article at https://rbnrpi.wordpress.com
#set up OSC address of processing sketch
use_osc '127.0.0.1',5000
#select shapes to show
osc "/viz/shapes","e"  #"s" "e" "r" Star,Ellipse, Rectangle or combination
sleep 0.1

live_loop :c do