By then, the fix will be easy. Fire up the latest LLM, point it at your codebase and tell it "rewrite this from scratch. do it well. fix the architecture mistakes"
Same thing happens in other fields. A rich country and a poor country might build equivalent roads, but they won't pay the same price for them.
The system that makes it have an opinion about good vs bad architecture or engineering sensibilities will be something on top of the transformer and probably something more deterministic than a prompt.
"Shit's in the Game!"
"Chunder Everything"
"Maddening NFL 26"
"FIFiAsco 26"
"UFC 26 (Un Finished Code)"
"The Shits 4"
"Battlefailed"
"Need for Greed"
What you're suggesting is a negative flywheel where quality spirals down, but I'm hoping it becomes a positive loop and the quality floor goes up. We had plenty of slop before LLMs, and not all LLM output is slop. Time will tell, but I think LLMs will continue to improve their coding abilities and push overall quality higher.
What I see is that while LLMs can do real tasks, they often produce overengineered unmaintainable slop (plenty of examples where code can be reduced 10x to do the same). I hope this is not a base to continue training LLMs on.
I don't know how long LLMs will hold their position, but in my mind, it makes sense that computers will eventually be the best programmers. Humans will still guide, but rarely write or even read code. I might be dead by then, but that's my prediction for the future.