Who says that we don’t? The point is that the bounds on the question are completely unknown, and we operate on the assumption that the compute time is relatively short. Do we have any empirical basis for this? I think we do not.