As a stepping stone, when we add to parametric design the benefits of automated iteration, generative process emerges. The notion of the "parameter" as a tool presented by software and the workflow that results is not novel; the real novelty proposed by parametric design is the first modern presentation of a different "semantics" of design thought. When seen this way, this opens a door to an infinite world of tooling, systems, and models that designers can create and leverage.
Writing code can enable designers to explore design solutions with means that are more akin to the semantics of the design "problem" at hand. Code is language, and like any language, it can express more than instructions and the rote operations of commercial software, but can be molded into modeling the semantics of real-world (and imaginary) objects and systems. Different programming languages are not created equal, and their creators are interested in what they can express and reason about, just as designers are interested in what they can express and render through media and tools.
What are "semantics" in this context? When we consider a design medium like perspective drawing, its semantics are that of station points, vanishing points, converging parallel lines, projection, Euclidean and non-Euclidean geometries, and a mathematical set of construction rules.
These rules are inherently foreign to the design objects themselves. To make perspective images, we execute within these working rules, but those rules influence our decisions. When the image is done, they influence our perceptions, and thus our reasoning. The same applies to software and parametric tools.
Cognitively, as we make design decisions, we actually imagine the design object apart from the medium, employ metaphor to represent the object's form through geometry, and then manipulate that geometry not through the semantics of the object, but through that of the digital medium.
But what if we could express our intent using a language closer to the semantics of the design object? Instead of saying to ourselves, "This NURBS surface represents an enclosure, and I will treat this other NURBS surface as the ground, etc.," we could teach a computer the semantics of enclosure, ground, material, and so on, thereby applying our design ideas as models of growth, organization, flow, evolution, etc.
Boldly speaking, by leveraging the best that computation has to offer (either directly or by partnering with coders), we can, for the first time in history, begin to break down this dependency on design media, and think differently in terms of design semantics. Ultimately, we would operate with automated, semantically rich systems of systems. In this vein, parametric design is less a new paradigm and more a beginning of a revolution of design thought.