This first generation of AI powered tools is so terrible, that it is most likely hype wave will come to end quite quickly before semantically deeper approaches appear.

It is very simple. Creating software is quite different from typing “hello how are you” in a messager. Repetitive tasks simply mean code is not well thought of.

Older CASE tools even from 1990s were covering software starting from high level model. Making software is much more than selecting some statistically learned chunks of code.

So this new tool may be fun, but hardly that useful as it is, without support for “programming in large”. Some modern languages like Python allowed professionals to program with the speed of thought for decades. Good IDEs like those from JetBrains automate refactorings. This means AI-powered assistance should be so much more useful. Not that it’s totally impossible, but yet to be seen how to empower already powerful programmer so that those suggestions were more than nuisance.