[sam-s]

While I can commiserate with the author's feelings and dispute some of his observations, as others already did, I want to make an orthogonal observation.

Software development differs from other engineering disciplines dramatically: a blueprint/spec for a car or a bridge or a building is not in the same ontological category as a car or a bridge or a building: you cannot drive it or live in it.

However, a spec for a computer program is a kind of a program. It just requires "a truly smart compiler" (an SDE). This conceptual connection might explain the shape of the learning curve for software vs other engineering disciplines.