I don’t see what this proves or disproves. The article points out that problems that are idly metric with clearly differentiated local data are amenable to parallelization — and potentially only such. Sounds like a lot of game visuals.