Large Language Models Are Bad at Refactoring Code

Test first. LLMs should be able to write tests, as per my experiments. Refactoring is a human task for now.