Very nice - you use AI much like I do, and like you, I always have the final say. It is an odd transition to make - after spending years thinking that computer algorithms, calculators, etc were the most reliable when it comes to certain kinds of things - the most useful tool yet (LLMs) - are highly fallible. I too use GPT and Copilot, and I've seen them make some really dumb mistakes. For example, I once gave it a long list of field names and asked it to follow a pattern I had begun, wrapping them in html tags - and while the html output was ok, it actually changed some of the field names - and hallucinated some more fields that didn't even exist. It's like I have a brilliant apprentice who can do amazing things - but he's a little bit schizophrenic, and I can't ever take my eyes off of him or turn him lose on my project without tight supervision. Still - he's so damned useful that I invite him to work every single day.