We're of opposing viewpoints there. If I have a cad file for this, then I print it on some z-printer and assemble, I'm fairly certain the paper airplanes produced by it are encoded in it's very structure even if you don't see the paper or the airplane coded in the design file. I'm thinking you have some arbitrary limits on how you define information encoding?
As for gene to protein to micro structure to complex interaction to knuckle...how terribly wrong you are can be shown with a single event. The second-generation study on it seems to confirm that the blueprints remain intact despite the embryonic issues with expression of first-gen. Are you aware of this body of research?
What a cool Lego machine! There's an issue of the definition of 'coding' here. Your CAD design of the machine would clearly not contain a design of the paper plane, but there would be a direct causative relationship between it and the plane that allows the plane to be predicted on the basis of the machine's design. I agree on that.
Of course, most metabolic pathways involve several components, not just one protein and one substrate, so things get a bit more complicated: the end result cannot be predicted from one design (gene). Potentially though (with considerable extra effort) it could be predicted by looking at all the genes involved. I'm not claiming this not to be the case.
But now imagine the following. You have not one but 10,000 Lego machine designs in a book. You throw this book into a bath tub containing a 3d printer and a heap of raw materials (sheets of paper, bits of wood,...) that the machines described in the book could work on. You're not there to intervene and tell the printer which pages to read, and what we want is not a simple paper plane, but a complex contraption built of hundreds of these little machines. Is the book of CAD designs going to be sufficient to get this done?
It doesn't stop there. Now imagine that we've got millions of these bathtubs, each with the same instruction book and more or less the same raw materials. We want them to make not just one but hundreds of different kinds of contraption, composed of different collections of machines. And then we want the millions of contraptions built to join together in an orderly fashion into one huge entity in which each of them is in just the right place to play its part.
Is the book of CAD designs really going to be sufficient to get this done?