This confuses OOP/procedural semantics with low-level "CPU" work. The argument seems to be that the programming model of a language does not always expose the efficiency of the underlying implementation. And this is also true for "anthropomorphic" models.