I think so too. If unclear, I don't use LLMs for coding at the moment and was just commenting on what I've seen from others who do in computational fluid dynamics.

Edit: Let me add that while I think it would be easy to instruct a LLM to do what I'd like, LLMs don't do these things by default despite them being recognized as best practices, and I'm not confident in LLMs getting the data or references right for validation tests. My own experience is that LLMs are pretty bad when it comes to reproducing citations, and they tend to miss a lot of the literature.