The problem with that is that GPT often hallucinates things that look reasonable that a human would never make up. Like this:
That one was easy to spot, others less so. Takes longer to “fine-tune” the idiot robot sick than to write from scratch. Look at that garbage about setting various states on a binary sensor.