My thesis asked how to generate software from a formal model of a problem domain and a formal model of the software meant to solve it. A decade in industry answered part of that empirically. Modern AI reopened the theoretical half — that's where my independent research time goes now, without a university behind it.