Dax McDonald daxmc99

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

Due to some issues with shipping at the moment my Macbook is delayed in getting to me and I will using Windows as my dev machine. These are the steps I took to set it up

Enable WSL 2 (basically required for good performance in most dev work)

Decide between Scoop & Chocolately

Install Ubuntu 20.04

Enable VS Code remote work

	# Change cpu/memory if required
	cpus: 8
	memory: "16GiB"

	images:
	- location: "https://cloud-images.ubuntu.com/releases/22.04/release/ubuntu-22.04-server-cloudimg-amd64.img"
	arch: "x86_64"
	- location: "https://cloud-images.ubuntu.com/releases/22.04/release/ubuntu-22.04-server-cloudimg-arm64.img"
	arch: "aarch64"

	kubectl get roles,clusterroles -o json \| jq -c '
	.items[]
	\| select(.rules != null) # Exclude entries with null rules
	\| select(.rules[].resources[] == "secrets") # Only resources with "secrets" access
	\| {name: .metadata.name, kind: .kind, namespace: .metadata.namespace, verbs: (.rules[].verbs \| unique)}' \
	\| jq -s 'unique_by(.name, .kind, .namespace)' # Deduplicate based on name, kind, and namespace

	#!/usr/bin/env bash
	set -e

	# Ensure required tools are installed
	if ! command -v glab &> /dev/null; then
	echo "Error: glab CLI is not installed. Please install it before running this script."
	exit 1
	fi

	if ! command -v jq &> /dev/null; then

	apiVersion: v1
	kind: Pod
	metadata:
	creationTimestamp: null
	labels:
	run: cpu-hungry
	name: cpu-hungry
	spec:
	containers:
	- image: ubuntu

	#!/usr/bin/env bash

	set -Eeu
	set -x

	# You need to use a session ticket (https://about.sourcegraph.com/handbook/support/p4-enablement#generate-a-session-ticket)
	# in order to avoid password prompts

	# TO edit this file, you must run `p4 edit create-revision.sh` first

	{
	// Place your snippets for shellscript here. Each snippet is defined under a snippet name and has a prefix, body and
	// description. The prefix is what is used to trigger the snippet and the body will be expanded and inserted. Possible variables are:
	// $1, $2 for tab stops, $0 for the final cursor position, and ${1:label}, ${2:another} for placeholders. Placeholders with the
	// same ids are connected.
	// Example:
	// "Print to console": {
	// "prefix": "log",
	// "body": [
	// "console.log('$1');",


	-- CREATE read-only user when the user was created via the Google Cloud API (ie terraform)
	-- GOOGLE CLOUD by default give fairly broad permissions
	-- Prior reading https://stackoverflow.com/questions/13497352/error-permission-denied-for-relation-tablename-on-postgres-while-trying-a-selec
	-- https://www.digitalocean.com/docs/databases/postgresql/how-to/modify-user-privileges/
	REVOKE ALL ON DATABASE db FROM "dev-readonly";
	-- This next line is important
	REVOKE cloudsqlsuperuser FROM "dev-readonly";
	GRANT USAGE ON SCHEMA public TO "dev-readonly";
	ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT SELECT ON TABLES TO "dev-readonly";

	[alias]
	co = checkout
	cob = checkout -b
	coo = !git fetch && git checkout
	br = branch
	brd = branch -d
	brD = branch -D
	merged = branch --merged
	st = status
	aa = add -A .