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.

guest271314

Compiling a standalone executable using modern JavaScript/TypeScript runtimes

We have the same code working using node, deno, and bun.

E.g.,

bun run index.js

jsjoeio

/** @OnlyCurrentDoc */

const startDate = new Date("2021-06-16")
// so the first one is D2:K2
function MarkEmptyCellsInYesterdayRowWithHyphen() {
  const sheet = SpreadsheetApp.getActiveSpreadsheet().getSheetByName("Group 2 - B");
  // Activates the sheet 
  SpreadsheetApp.setActiveSheet(sheet);
  
  const today = new Date()

jherr

1. Install @module-federation/nextjs-mf.

yarn add @module-federation/nextjs-mf

2. Add resolutions to package.json:

  "resolutions": {

kmelve

/* eslint-disable no-console */
import client from 'part:@sanity/base/client'

// Run this script with: `sanity exec --with-user-token migrations/migrateValues.js`
//
// This example shows how you may write a migration script that migrates a field value
// on a specific document type.
// This will migrate documents in batches of 100 and continue patching until no more documents are
// returned from the query.
//

tomhicks

/*
Copy this into the console of any web page that is interactive and doesn't
do hard reloads. You will hear your DOM changes as different pitches of
audio.

I have found this interesting for debugging, but also fun to hear web pages
render like UIs do in movies.
*/

const audioCtx = new (window.AudioContext || window.webkitAudioContext)()

Rich-Harris

I've been deceiving you all. I had you believe that Svelte was a UI framework — unlike React and Vue etc, because it shifts work out of the client and into the compiler, but a framework nonetheless.

But that's not exactly accurate. In my defense, I didn't realise it myself until very recently. But with Svelte 3 around the corner, it's time to come clean about what Svelte really is.

Svelte is a language.

Specifically, Svelte is an attempt to answer a question that many people have asked, and a few have answered: what would it look like if we had a language for describing reactive user interfaces?

A few projects that have answered this question:

Calvin-Huang

import { takeEvery } from 'redux-saga';
import { call, put } from 'redux-saga/effects';

// Watcher
function* watchFetchData() {
  yield* takeEvery('FETCH_REQUESTED', fetchData);
}

// Worker
function* fetchData(action) {

NetanelBasal

describe('PostsEffects', () => {
  let runner, postsEffects, postsService;

  beforeEach(() => TestBed.configureTestingModule({
    imports: [
      EffectsTestingModule
    ],
    providers: [
      PostsEffects,
      {

ericelliott

const compose = (...fns) => x => fns.reduceRight((v, f) => f(v), x);