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.

khalidx

The experience of using Node.JS with TypeScript, ts-node, and ESM is horrible.

There are countless guides of how to integrate them, but none of them seem to work.

Here's what worked for me.

Just add the following files and run npm run dev. You'll be good to go!

package.json

lizthegrey

default['sshd']['sshd_config']['AuthenticationMethods'] = 'publickey,keyboard-interactive:pam'
default['sshd']['sshd_config']['ChallengeResponseAuthentication'] = 'yes'
default['sshd']['sshd_config']['PasswordAuthentication'] = 'no'

3n21c0

package main

import (
	"context"
	"flag"
	"fmt"
	"log"
	"net/http"
	"os"
	"os/signal"

kyle-eshares

from __future__ import unicode_literals

from django.db import models
from django.db.models.fields.related_descriptors import ForwardManyToOneDescriptor  # noqa


class RelationNotLoaded(Exception):
    pass

mikelehen

/**
 * Fancy ID generator that creates 20-character string identifiers with the following properties:
 *
 * 1. They're based on timestamp so that they sort *after* any existing ids.
 * 2. They contain 72-bits of random data after the timestamp so that IDs won't collide with other clients' IDs.
 * 3. They sort *lexicographically* (so the timestamp is converted to characters that will sort properly).
 * 4. They're monotonically increasing.  Even if you generate more than one in the same timestamp, the
 *    latter ones will sort after the former ones.  We do this by using the previous random bits
 *    but "incrementing" them by 1 (only in the case of a timestamp collision).
 */

ericelliott

Essential JavaScript Links

I no longer mantain this list. There are lots of other very comprehensive JavaScript link lists out there. Please see those, instead (Google "awesome JavaScript" for a start).

Ceasar

"""
REST client over requests.
"""
import logging

import requests


logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

staltz

The introduction to Reactive Programming you've been missing

(by @andrestaltz)

---

This tutorial as a series of videos

If you prefer to watch video tutorials with live-coding, then check out this series I recorded with the same contents as in this article: Egghead.io - Introduction to Reactive Programming.

---

LucasRoesler

class JSONMiddleware(object):
    """
    Process application/json requests data from GET and POST requests.
    """
    def process_request(self, request):
        if 'application/json' in request.META['CONTENT_TYPE']:
            # load the json data
            data = json.loads(request.body)
            # for consistency sake, we want to return
            # a Django QueryDict and not a plain Dict.