Have you ever tried to do this? Once you get past fibonacci functions, the amount of time and effort required is measured in PhD-man-years. And even then, all you have done is moved the risk from the code to the proof, because when you work on something that large, the risk that the model does not represent reality (either due to oversight, lack of understanding, or not correctly capturing the interplay between two aspects) becomes an issue.
The issue is that the state space for software is so much larger than the state space for bridge building that we just don't have the tools to effectively model and predict yet. I won't argue that software is somehow an art, because I'm hopeful that someday we will have those tools, but I will argue that it's not engineering yet. The Alan Kay quote about pyramids in the article hit it spot on. We are still piling blocks on each other and hoping it works at this point.