Application-level software engineering hasn't been about hand-writing for-loops for a long time, but there's a tendency for people to feel self-satisfied in what's familiar and what they worked hard to master.

Now that LLMs can write for-loops, too, I've hoped that the field will reconsider the activities that it considers to be valuable. The essential complexity is more about what you described: sit down, design, find a specification. The build to spec part, in the 21st century, can be a tooling step that takes the spec as input.

Perhaps the next step for the industry's state of the art in application-level software engineering is mastering formal spec languages, to seriously tailor the scope of work to the essential complexity of mediating between the agile world of human requirements and the formal world of computing systems. But I won't hold my breath--society is inured to software bugs as a fact of life, and programmers are used to being paid to endlessly patch their own mistakes.