I started this project back when I was in college (nearly 10 years ago!) after I came across Webgazer.js on Hacker News. I used to play BrickBreaker on my Dad's BlackBerry as a kid and thought it would be a cool gimmick to try out. After putting together the base gameplay, I realized that the eye-tracking library was too jittery to functionally control the paddle. It somewhat worked, but was impossible to play.

Fast-forward to now, I figured AI could solve some of the smoothening logic needed to make it work. While it was able to get the movement to be less jittery, the library seems to lack the eye-tracking granularity to effectively control the paddle -- it is too inconsistent and inaccurate. So instead, I've resorted to using more broad face tracking as the control scheme which still feels somewhat magical.