This paper details how Timnit and her team used Google Street View images of cars, deep learning, and neural networks to (accurately!) predict the demographics and political leanings of a neighborhood based on the types of vehicles detected there.
This is a clear, fun, and fascinating essay that explains how neural networks can get image classification woefully wrong.
This algorithm tells you if a given directed graph is acyclic or not and simultaneously puts its vertices into the most convenient order for processing, all in linear time!
It's soothing to watch the colors on the image gradually pour in. You can get a processing sketch that lets you try it with your own image here: https://www.openprocessing.org/sketch/397165.
Pico-8 is a game design VM with hard restrictions on resources code length, color palette, etc). This forces tradeoffs that boil down a project to its essence, and in my opinion makes it more fun. The game was my first project with this machine - somehow it was both frustrating and enlivening at the same time.
Awesome talk on reconciling OOP vs FP
This project started out as a fun exploration and ended up with a CVE number! https://nvd.nist.gov/vuln/detail/CVE-2017-15298
I like it because it's surprising—seeing applications of technology like computer vision outside of what you would usually expect is pretty cool!
The idea of live-coding music and visualizations is fascinating to me, and Hylogen is a great implementation of it!