Agree that everything must be contextualized. My point is, you saw this presentation and liked it because it seems like a good idea, and the performance is apparently good relative to something else. For an engineer (or a theoretician) to reason about that claim of improved performance there needs to be a systematic way to understand it. All I'm saying is that way this was done doesn't allow us to understand it very well - precomputation being a big part of that.