David Pennington xeoncross

Draw Things

https://drawthings.ai/ provides a closed-source binary that works well with a nice UI as well as a HTTP server you can enable to call different models from other programs/apps.

If you want to manually create images this is the best free choice I've seen so far. It's also pretty low-memory compared to others I've tried. 10GB for their flux.2 [klein] 4b model.

Once you start the Draw Things app pick some models to download, then go to "all settings" and enable the HTTP server. Then call it with CURL:

curl -sS --max-time 600 \

Prompt: "Please review the home.nix configuration and note anything that can be done to simplify the nix configuration"

All the prompts review 2-3 .nix files and recommend 6-10 improvements on them.

The following are raw notes about how each model performs.

Hardware: M1 Max (10 CPU) 64GB RAM Macbook (Tahoe 26.5)

Notes:

https://quesma.com/blog/qwen-36-is-awesome/

It comes in two variants, a mixture-of-experts model Qwen 3.6 35B A3B, and a dense Qwen 3.6 27B - slower, but more powerful. The one I recommend!

Note: I've found the 35B-A3B model to run at least 2x fast which is nice when you just need quick results for focused tasks. For larger or multi-step processes the 27B model is worth the wait. Here is the local setup guide to get it working with VSCode via Cline extension.

1. Install llama on MacOS

brew install llama.cpp

	// ==UserScript==
	// @name LinkedIn Job Description Auto-Expander
	// @namespace https://tampermonkey.net/
	// @version 1.0.0
	// @description Automatically clicks the "…more" link on LinkedIn job descriptions so the full text is always shown
	// @author xeoncross
	// @match https://www.linkedin.com/jobs/*
	// @run-at document-idle
	// @grant none
	// ==/UserScript==

	Latency Comparison Numbers (~2012)
	----------------------------------
	L1 cache reference 0.5 ns
	Branch mispredict 5 ns
	L2 cache reference 7 ns 14x L1 cache
	Mutex lock/unlock 25 ns
	Main memory reference 100 ns 20x L2 cache, 200x L1 cache
	Compress 1K bytes with Zippy 3,000 ns 3 us
	Send 1K bytes over 1 Gbps network 10,000 ns 10 us
	Read 4K randomly from SSD* 150,000 ns 150 us ~1GB/sec SSD

	{
	"theme_overrides": {
	"One Light": {
	"syntax": {
	"attribute": {
	"color": "#000000FF"
	},
	"constructor": {
	"color": "#000000FF"
	},

	EXECUTABLE=executable-name
	WINDOWS=$(EXECUTABLE)_windows_amd64.exe
	LINUX=$(EXECUTABLE)_linux_amd64
	DARWIN=$(EXECUTABLE)_darwin_amd64
	VERSION=$(shell git describe --tags --always --long --dirty)

	.PHONY: all test clean

	all: test build ## Build and run tests

	//Simulation of the Birthday Paradox
	//Find no. files needed for chance of a single collision
	package main

	import "fmt"
	import "math"

	func main() {

	var BITS float64 = 512

	// Run at: https://go.dev/play/p/oFUYPNRZiBp
	package main

	import "fmt"

	// Suppose you have a piece of code you want to split into chunks based on if you're in a string or not.
	// You also want to respect escape characters.

	var input = `fmt.Println("Hello, \"世界\"")`

	package main

	import (
	"bytes"
	"crypto/ecdsa"
	"crypto/elliptic"
	"crypto/rand"
	"crypto/rsa"
	"crypto/x509"
	"crypto/x509/pkix"