I think people are the higher order ramifications of functions - but your point about the initial instructions needing to circumscribe the ramifications, I'm still tossing that around, trying to understand it.
I think the only thing that would need to be circumscribed by the initial instructions is some kind of general reflexivity, rather than the particular manifestation of some reflexive process. Some kind of taking-prior-output-as-next-input, but perhaps the discreteness of that algorithmic concept can be taken too far. Not sure.
I think you can get it just right, or right enough, so as to have some base unit operate in such a way that leads, if even rarely, to drastically higher order complex systems. And I don't think the particular complexity of the ramified - and ramifying - system needs to be accounted for in advance. I think what needs to be accounted for is the general dynamic/mechanism of reflexivity.
And then if that is accounted for, there may well be no discernible limit to the arabesque and fanciful things that emerge.