Yes. The deliverable is not the work done. Imagine you a patch submitted to your code base, you have to decide whether to start deeply reviewing it.

You are given an envelope, with a message inside revealing whether the patch came from a human contributor with expertise in the field, or whether it was prompted with "make stuff better" out of an LLM.

Do you throw the "useless" envelope away without opening it? Authorship indirectly tells us things about other work that happened, like checking alternative approaches, checking for edge-cases, and considering future work.

This is a pretty lazy post. Sure, the quality of code that AI produces has gone up. "Sloppy" means so much more than just "incorrect", though; it's bloating tests with additions that don't meaningfully add anything or actually test the code. Code used to be slow to write, quick to read. With AI, it's quick to write, slow to read. Thing is, that has never been the bottleneck. Often, it's a feature- you want to think before committing to an approach. If there is no barrier to putting pen to paper, you get... Slop.