You’re thinking way too hard. The solution is in the instructions: They say to step out of frame first.
Here’s my guess on how it works:
The moment when you are out of frame the computer memorizes every pixel that it sees. From then on, it only records pixels that are DIFFERENT from the ones in its memory. That’s it!
As you can see in the demo, however, it’s not a very “clean” key. It’s certainly good enough for chatting with your friends, but I don’t think you’d be using that method in After Effects anytime soon. It also makes me think that if you do this on a laptop you’d have to re-set the background key every time you adjusted or moved the laptop.