Interestingly, I see a parallel with the "vertical stripe" nature of "stories" in XP. A lot of teams will spend considerable amount of effort making an MVP (Minimum Viable Product), but when you look at it nothing works properly. It can be used, but you would never want to use it if you had any other choice. As the product grows, those pain points never quite disappear and they complicate the design forever. Even if you are successful, your competitors can learn from your mistakes and build something coherent that takes considerably less time -- second to market can have a big advantage.

In contrast, with a vertical stripe, you are encouraged to build something that is not just done, it is done done. With the information you have now, you would not change it (NB: this is considerably different than perfect). Each stripe should be as good as you can make it allowing for imperfect knowledge.

So in the same way you are limiting scope, but trying to make the result "final quality". From there it is a lot easier to iterate your design. I'm not sure I think it is necessary to nail down all your decisions quite in the way the article suggests -- good ideas are good ideas even if you have them later on in the process. Building on a solid foundation is just dramatically easier than trying to reinforce a Jenga tower.