Using generative algorithms to perform a complex skill leads to deterioration of the skill. The more you rely on the algorithm to perform a given task, the less effective you will be at performing QA of that task, as your grasp of the specifics of the task fades.

Any developer who has not spent time learning to "reliably hand-code" will be completely useless for performing any code QA. If the industry does not provide time, space, and incentives for junior developers to learn those skills on their own, the future will be void of any effective QA.

I know where it frequently fails, I’m very pleased with the output. And, I’ve shipped 4 fully QA’d apps in the past month.

Yes, well, you don't know what you don't know.

I would still be on the first one without AI.

The quality of a product is proportional to the amount of time and (human) attention spent on the product.