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.

Konfekt

" Generate compile_commands.json for clangd.
" Prefer CMake export, else Ninja compdb, else Bear + Make with optional append.

if exists('b:did_c_compdb_ftplugin') | finish | endif
let b:did_c_compdb_ftplugin = 1

let s:inflight = {}

augroup c_compdb
  autocmd! * <buffer>

frostming

# fly.toml file generated for still-snowflake-6351 on 2023-03-30T09:53:40+08:00

kill_signal = "SIGINT"
kill_timeout = 5
primary_region = "sin"
processes = []

[build]
  image = "bayedev/opencatd"

Blankwonder

@import Photos;

    [PHPhotoLibrary requestAuthorizationForAccessLevel:PHAccessLevelReadWrite handler:^(PHAuthorizationStatus status) {
        NSLog(@"PHAuthorizationStatus: %ld", status);
                
        PHFetchOptions *fetchOptions = [[PHFetchOptions alloc] init];
        fetchOptions.includeHiddenAssets = YES;
        fetchOptions.includeAllBurstAssets = YES;

        PHFetchResult<PHAsset *> *result = [PHAsset fetchAssetsWithOptions:fetchOptions];

JamesHopbourn

{
  "key_open_app_link_config": {
    "hosts": {
      "eu_ea": [
        "applink.feishu.cn",
        "applink.larksuite.com"
      ],
      "eu_nc": [
        "applink.feishu.cn",
        "applink.larksuite.com"

mtgto

/*
 * Change from https://gist.github.com/unki2aut/4ac81c33be2e8f121e80a26eba1735d7
 * - Use top level await (Node.js v14.8.0+)
 *   - To use top level await, you need to write a script as ES Modules
 * - Set chokidar options to avoid duplicate building
 * - Define NODE_ENV (For React)
 * - Add API proxy setting by using proxy-middleware
 */
import chokidar from "chokidar";
import esbuild from "esbuild";

codehz

ip rule add fwmark 0x233 lookup 100
ip route add local 0.0.0.0/0 dev lo table 100

ethanhuang13

//
//  FacebookAuth.swift
//  GitHub: ethanhuang13
//  Twitter: @ethanhuang13

import AuthenticationServices
import SafariServices

/*
 Updated: 

zu1k

@echo off


::配置BAT对话框样式
title KMS服务器检查脚本
MODE con: COLS=42 lines=8
color 0a


::检测vlmcs.exe

janlay

Delete and unfavorite tweets

This util deletes your historical tweets from a downloaded archive file, unfavorites tweets from your public profile.

No guarantee, no support, use it at your own risk.

Preparations

1. Apply to become a Twitter developer first.
   Once your Twitter developer account application is approved, You can create a new app and get its Consumer API keys and generate Access token. These keys are used to configure tweet.sh.
2. Obtain archived tweets by requesting to download your Twitter data.
   You can extract tweet.js file from the downloaded .zip file.