[ygra]

This looks homegrown. The source code [1] is fairly readable.

New dots seem to enter the world in the position of the last dot placed but below them. Position and size seem to follow a simple linear movement between frames.

I wonder though, why is this done with canvas instead of SVG. Just to cut down on the number of DOM elements?

[1] http://www.datapointed.net/media/2012/10/factor_min.js

[maqr]

I don't know why an experiment like this would be minified. It would be much more interesting with well-commented source.

[duopixel]

It's canvas. SVG would still need to insert DOM elements and would probably start displaying performance problems at 5000 elements or so.