It's just a different journey. Now you need to define your architecture and requirements up front in documents, understand models and their cost/capabilities, do detailed reviews of any generated code, basically manage modern CI/CD software developement with a set of new tools. You can certainly write detailed code if you want, and probably should regularly to keep up to speed. Developing software these past few decades has been very inefficient and way more expensive than it should have been. It doesn't surprise me that technology has evolved to solve these problems. Analogy: We can still construct building using manual artisans but leveraging machinery and modern architecture techniques makes sense doesn't it?
I remember learning and writing Cobol with with the Microsoft Cobol compiler on my Tandy as I was training at the local VoTech and working as a night operator on an IBM mainframe flipping tapes. I can categorically say that writing code with Antigravity is worlds different than those early days. It is inspiring, but for me its more about understanding the models and how they do their magic. At 67 I'm refreshing my calculus, linear algebra, and statistics in an effort to be able to read the papers on the subject. In the future, I'm imagining the norm being an automated layer for coding, similar to compilers of today, that take natural language and produce trusted, reliable and performant code all the way down to the machine level. The real work will be developing the models and their layered and optimized machine level interfaces and implementation. It is all kind of amazing.