Dear Philip,
a central point in your essay is universality. A mathematical structure is more universal - more discovered than invented - when it is instantiated in more structures that superficially appear different from one another; and a physical law is more universal when it describes more physical systems that appear different from one another in some aspects (e.g. different particle types in systems that obey the same thermodynamical laws).
A dense spectrum of universality degrees is envisaged, for which we still lack formal treatment and measure. And yet, the idea to navigate the mathematical universe guided by this compass and reach the top of the hill (or the bottom of the valley), with the idea that the most universal mathematical structure coincide with the meta-laws of physics, is very appealing.
We have another interesting notion of universality, however, more advanced in terms of possibilities of formal treatment, and simply of a True/False type (no spectrum of degrees): that's of course computational universality (Turing universality) - the ability of a model of computation to reproduce any computation of any other conceivable model of computation.
To those who attribute a fundamentally algorithmic nature to the universe (thus justifying its mix of order and disorder better, in my opinion, than any other approach) this would be the first choice for a notion of universality. Then, the democratic idea that all vacua, or all mathematical structures, enjoy some form of existence, could perhaps also apply to the multiplicity of universal models of computation.
There seems to be a very large gap between the idea of an algorithmic uni/multiverse and the scenario that you describe. You place 'algorithms' in your picture below 'games', but still far from the 'point of universality'. I agree that individual algorithms are more invented than discovered, but what about the Universal Turing machine?
Thanks and best regards
Tommaso
P.S. In case the algorithmic paradigm gained credibility in the future, I suggest to replace the elephant with an ant, for its 'minimality', as a metaphor for the head of a Turing machine, and as a reminder of the variety of properties that emerge from the computations of 'turmites' (2D Turing machines), including Langton ant.