I had the opposite impression.

LLMs can abstract pretty well. They write quality code for obscure frameworks no problem, as longs as you take the time to set up tight feedback loops.