SQL is in a sweet spot of declarative, formal, and high level. It's basically a description of what the resulting dataset means, not just how to produce it. For ad-hoc queries maybe LLMs take over in the short term, and that's how I read your example. But for core, repeated business logic, human-written SQL seems to still hold value.