I just wrote up a slightly different take on the exact same situation you described. As a fellow "advanced" gray beard with decades of experience and broad knowledge of all the various "cool architectures du jour" that have popped up and gone away in that time, I feel that I am (as you are) fully qualified in writing code that others consider to be of "superior" quality.
The hands of such a person, these tools can build our knowledge. But only because we have all these learned system architectures upon which we can base our read and understanding of whatever new thing we're looking at.
Yeah, in a way we're "novices" in that field, but dude...we're novices with a vast knowledge library to build upon. Even without these tools, we can dive into an area we're entirely unfamiliar with and come up to full competency FAR faster than most true "novices".
The novices of concern here and being referred to by the author you replied to are those that aren't yet able to, as you wrote, "review the tests the LLM had written and direct it to improve them" in the same way you and I can. *Those* are the novices I am / we are concerned about never advancing beyond "novice" level.
Now...what we do not know is the long term playout of this. There was a time, I'm sure because I was there as I suspect you were as well, where higher level compilers were considered suspect. I've had to review the assembly they spit out to find floating point bugs I *knew* weren't in my code. But now? In 2026? I haven't reviewed compiler output in at least 2 decades, maybe almost 3...certainly not to the extent I was 4 decades ago when C compilers for the 65816 were first coming out, for example...
Is that where we are now in this genAI spool up? Perhaps. Perhaps the newbies now that seem like novices because they don't know the details of the underlying libraries and APIs like the back of their hand are just the new users of the new "compilers" and at some point in the future they're all going to look back and say "remember when we used to have to write actual code and memorize APIs and read through piles of documentation....OMG things were so painful back then".
That's what I'm hoping is happening now. I suspect it's more of that than what some are claiming that "software is doomed" and "we're going to lose all experienced coders". Nah...I suspect we're just changing the type of coder that's going to be considered "experienced" and the domain we're going to consider them experienced in.