(The above is quote of @swamidass quoting Larry Moran, but I haven’t mastered the forum formatting sufficiently to make that clear. Thus this text, so we all know the actual source of the statements.)
This reminds me of something similar that Stuart Kauffman talks about with regard to autocatalytic sets. An autocatalytic set is a set of molecules (or more generically, entities) that collectively catalyzes the production of its members. These sets tend to be larger rather than smaller, which might seem surprising at first because you might expect it to be easier to make a few molecules than a lot. But basically, as you add molecules (or entities) to a set, the number of things you need to make grows linearly, but the number of interactions that can produce those things grows exponentially. Thus we would expect such autocatalytic sets to be relatively complex.