So I completely removed this possibility from my example - the constraints are provdied by the user.
well, no. not completely. you're still matching it to a predetermined design. the word itself is being artifically selected by a master intelligence (the user) who is manipulating the system to create something.
evolution runs on its own, with no user, and no pre-design to match to.
Also, you say this as if it were different in principle from what happens in nature. It isn't. Just as nature provides constraints in the form of temperature, weather, water supply, food supply, predators and so forth, the program provides constraints, too. These constraints are what provides selection, which is one of the two key components of evolution: descent and with modification and natural selection.
well, that's not what i have issue with. i have issue with the pre-conception and matching process, and that it doesn't run on its own.
Following the rules of grammar isn't going to produce a story, or even very many rational sentences. I can see a lot of sentences like, "The tinted cognition drove to the rusty iota."
and evolution does not produce perfect organisms, either.
What would be your selection mechanism for choosing the "winners" in each generation?
evolution rarely choose winners. it just chooses LOSERS. so to duplicate the model, all you'd really have to do is throw all of the ones that are unfit according to grammatical rules and parts of speech.
if you want to throw in actual selection, then we could add a user or users in, to vote for their favourites.
and as someone pointed out, darwinian poetry works quite well.
אָרַח