This was really interesting! It's great to see someone taking a quantitative approach to design. Art schools would never teach an algorithmic approach to kerning, but for the programmers/artists who are smart enough to learn it, a breakthrough permitting effective use of this approach would be invaluable.