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.

OrionReed

// 3D Dom viewer, copy-paste this into your console to visualise the DOM as a stack of solid blocks.
// You can also minify and save it as a bookmarklet (https://www.freecodecamp.org/news/what-are-bookmarklets/)
(() => {
  const SHOW_SIDES = false; // color sides of DOM nodes?
  const COLOR_SURFACE = true; // color tops of DOM nodes?
  const COLOR_RANDOM = false; // randomise color?
  const COLOR_HUE = 190; // hue in HSL (https://hslpicker.com)
  const MAX_ROTATION = 180; // set to 360 to rotate all the way round
  const THICKNESS = 20; // thickness of layers
  const DISTANCE = 10000; // ¯\\_(ツ)_/¯

RazhanHameed

import re

def resolve_ae(text):
    """
    This function takes a text input in Central Kurdish (Sorani) script and performs a series of character replacements
    to standardize variations in the script. Specifically, it addresses cases where the character 'ە' (Arabic letter
    AE) may be used in different contexts.
    """
    # First replace all occurrences of 'ه' with 'ە'
    text = re.sub("ه", "ە", text)

Justasic

#include <std/mem.pat>
#include <std/io.pat>
#include <std/sys.pat>

enum ConstructorIDType : u32 {
    CID_TLSTYPE = 0x12eb4386,
    CID_NATCONST = 0x8ce940b1,
    CID_NATVAR = 0x4e8a14f0,
    CID_EXPRNAT = 0xdcb49bd8,
    CID_EXPRTYPE = 0xecc9da78,

DavideGalilei

Idea from https://youtu.be/9IULfQH7E90

Elschnagoo

import crypto from 'crypto';

export type SemElement = [string, (value: any | PromiseLike<any>) => void];
export class Semaphor {
  private curQueue: Map<string, SemElement>;

  private readonly max: number;

  private callbackQueue: SemElement[];

alemidev

#!/bin/bash
#
# 	Commit Convention Hook | alemi <me@alemi.dev> May 2022
#
# A simple git hook to enforce (sort of) Commit Convention.
# 		www.conventionalcommits.org
# 
# This script uses bash builtin regex matching, so it's not sh compatible.
# 
# Commit message will be stripped of comments (lines starting with #) and matched 

allekok

import hashlib

blockchain = []
difficulty = 3
mine_pattern = '0' * difficulty

def add_block(data):
        block = make_block(data)
        block['prev'] = '' if blockchain == [] else blockchain[-1]['hash']
        mine_block(block)

MidSpike

CVE-2022-23812

---

RIAEvangelist/node-ipc is malware / protestware

The RIAEvangelist/node-ipc module contains protestware peacenotwar.

Excerpt from RIAEvangelist/node-ipc:

as of v11.0.0 & v9.2.2 this module uses the peacenotwar module.

painor

import { StoreSession} from "telegram/sessions";
import { TelegramClient } from "telegram/client/TelegramClient";

import { Api } from "./tl";

import { generateRandomBigInt, readBigIntFromBuffer, readBufferFromBigInt, sha256 } from "telegram/Helpers";
import bigInt from "big-integer";


const apiId = ;