Even though this seems to be "amazing" and "innovative" for the folk at Adaptive Path (and I bet they are not the only ones thinking that way), this is actually a very old technique called Set-Based Concurrent Engineering (SBCE, also in software).
This technique is similar to techniques used in brainstorming sessions where participants are encouraged to generate many ideas (broaden the horizon), improve on them incrementally by "using" other people's ideas and enhancing them (improve on other's ideas), and finally to select the most appropriate idea for implementation (narrow and select).
Set-Based Concurrent Engineering is also used to ensure quality when a team (or set of teams) must meet a hard-deadline (as in a deadline that cannot be changed) with a solution that is much better than if you would just go with your first impulse/idea and try to improve on that.
One of the key advantages for Apple in using this technique, is that when they get to the 3 mid-step ideas they actually have syntethized all of the best points of all the other 7 ideas into those select 3. And then they still improve on those!
Good to see that Adaptive Path picked up on this technique, I hope that many other UI/UX people start paying attention to this old, but proven technique!