This is why I frame it as a "snippets" plugin, rather than a Code generation tool.

I would be very confused if someone told me that they uncritically used the generated code from a snippet program with no manual input or understanding, and I feel the same with Copilot. At best, it suggests an auto-complete that I read and interpret before accepting.

The closest I come to "code generation" is during test writing, where occasionally I will let the description generate some setup, but only in tests where there are a broad number of examples to follow, and I am still going to end up re-writing a decent chunk of it based on personal example. I would not "let it write the test suite for me" and then trust the green, and I suspect that would easily fail code review (though it would be an interesting experiment...).

Obviously your comment as a good goof and well made, but it does speak to a little bit of the disconnect between what is being touted as an "AI coding tool" and how I, a person who makes react native apps to pay my rent, actually use the dang thing (i.e., "A pretty good snippets plugin"). Is My code 'AI generated'? I wouldn't call it that, but who can say definitively? We're in a fun new semantic world now.