Great anecdote, thanks for sharing. Forcing ChatGPT to go back and sense check its overarching initial interpretation is such a good way of doing things in these types of uses cases.