I presume, people who can code, as opposed to people who can only prompt an LLM.

In the real world, you hit problems that the LLM doesn't know what to do with. When that happens, are you stuck, or can you write the code?

Id be seeing if the candidate actually understanding what the llm is spitting out and pushing back when it doesn't make sense vs are they one of the "infinite monkeys on infinite typewriters"

IF (and a big IF) LLMs are the future of coding, this doesn't mean humans don't do anything, but the role has changed from author to editor. Maybe you don't need to create the implementation but you sure better know how to read and understand it.