[correcthorse123]

Agreed. Maybe there are ways to have a single agent and still have internal competition.

An organism (biological or otherwise) could have competing internal parts/models, and not be constrained by the mechanism by which natural selection operates now, i.e. biological reproduction.

So multiple subagents could compete within it, without posing an existential threat to the physical manifestation of the organism itself; somewhat analogous to redundant computers making decisions by consensus in e.g. spacecraft.