veekaybee

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

Pre-Transformer Models

wojteklu

Code is clean if it can be understood easily – by everyone on the team. Clean code can be read and enhanced by a developer other than its original author. With understandability comes readability, changeability, extensibility and maintainability.

---

General rules

1. Follow standard conventions.
2. Keep it simple stupid. Simpler is always better. Reduce complexity as much as possible.
3. Boy scout rule. Leave the campground cleaner than you found it.
4. Always find root cause. Always look for the root cause of a problem.

Design rules

jvns

A lot of these are outright stolen from Edward O'Campo-Gooding's list of questions. I really like his list.

I'm having some trouble paring this down to a manageable list of questions -- I realistically want to know all of these things before starting to work at a company, but it's a lot to ask all at once. My current game plan is to pick 6 before an interview and ask those.

I'd love comments and suggestions about any of these.

I've found questions like "do you have smart people? Can I learn a lot at your company?" to be basically totally useless -- everybody will say "yeah, definitely!" and it's hard to learn anything from them. So I'm trying to make all of these questions pretty concrete -- if a team doesn't have an issue tracker, they don't have an issue tracker.

I'm also mostly not asking about principles, but the way things are -- not "do you think code review is important?", but "Does all code get reviewed?".

tartakynov

import numpy as np
import pylab as pl
from numpy import fft
    
def fourierExtrapolation(x, n_predict):
    n = x.size
    n_harm = 10                     # number of harmonics in model
    t = np.arange(0, n)
    p = np.polyfit(t, x, 1)         # find linear trend in x
    x_notrend = x - p[0] * t        # detrended x

ttesmer

{-# LANGUAGE TypeSynonymInstances #-}
data Dual d = D Float d deriving Show
type Float' = Float

diff :: (Dual Float' -> Dual Float') -> Float -> Float'
diff f x = y'
  where D y y' = f (D x 1)

class VectorSpace v where
  zero  :: v

mattdenner

I found a post about suspending and then going into hibernate that included a really clever script. Turns out that with NixOS this is even esaier to coordinate as you have systemd so can have a before and after service. I just include this in my /etc/nixos/configuration.nix file and nixos-rebuild; then a systemctl suspend or a close of the lid will cause the hibernate timer to be set.

mandrean

# Instructions for fresh install

$ sh <(curl -L https://nixos.org/nix/install) --darwin-use-unencrypted-nix-store-volume --daemon
# reboot
$ source /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh
$ echo 'export NIX_PATH=darwin-config=$HOME/.nixpkgs/darwin-configuration.nix:$HOME/.nix-defexpr/channels${NIX_PATH:+:}$NIX_PATH' | tee -a ~/.zshrc
$ echo 'source $HOME/.nix-profile/etc/profile.d/hm-session-vars.sh' | tee -a ~/.zshrc
$ nix-channel --add https://nixos.org/channels/nixpkgs-unstable
$ nix-channel --add https://github.com/LnL7/nix-darwin/archive/master.tar.gz darwin
$ nix-channel --add https://github.com/nix-community/home-manager/archive/master.tar.gz home-manager

simonw

How to recover lost Python source code if it's still resident in-memory

I screwed up using git ("git checkout --" on the wrong file) and managed to delete the code I had just written... but it was still running in a process in a docker container. Here's how I got it back, using https://pypi.python.org/pypi/pyrasite/ and https://pypi.python.org/pypi/uncompyle6

Attach a shell to the docker container

Install GDB (needed by pyrasite)

apt-get update && apt-get install gdb

prakhar1989

Rich Hickey on becoming a better developer

Rich Hickey • 3 years ago

Sorry, I have to disagree with the entire premise here.

A wide variety of experiences might lead to well-roundedness, but not to greatness, nor even goodness. By constantly switching from one thing to another you are always reaching above your comfort zone, yes, but doing so by resetting your skill and knowledge level to zero.

Mastery comes from a combination of at least several of the following:

defuse

<?php

/*
 * This code is copied from 
 * http://www.warpconduit.net/2013/04/14/highly-secure-data-encryption-decryption-made-easy-with-php-mcrypt-rijndael-256-and-cbc/
 * to demonstrate an attack against it. Specifically, we simulate a timing leak
 * in the MAC comparison which, in a Mac-then-Encrypt (MtA) design, we show
 * breaks confidentiality.
 *
 * Slight modifications such as making it not serialize/unserialize and removing