Is there anyone on hacker news that could explain me secion 6 in this article? Overall the article is very interesting but I am not sure if the waterfall argument is valid.

The waterfall conjecture was stated as : 'Given any chess-playing algorithm A that accesses a "waterfall oracle" W, there is an equally good--chess-playing algorithm A', with similar time and space requirements that does not access W.'

From this we can assume that the article is a metaphor for us as external observer given semantics for computation. Thus by providing the reduction we showed that the computation in and of itself does not have semantics. However, why would we assume there to exists A'? Even if there is A', can't we reduce A -> A', and if we can reduces W to A, then by proxy W -> A'?

Could somebody help me?