I personally think I’m going to avoid copilot.

I can see the benefits, but I consider it like going to live in another country where you don’t speak the language and never putting down the dictionary. Too many people are going to use this to avoid thinking through problems. I just feel like there was already enough of that problem in the field.

I think that Copilot will used much like we use Intellisense today. We will use Copilot to “autocomplete” some of the stuff we’re doing, but Copilot will never be able to understand or even come up with novel solutions for domain specific implementations.

I signed up for the preview but then I read a bunch of articles on questionable choices from GitHub when they built copilot - such as including GPL code as a source for learning - and decided to skip on it when I got an invite.

See this article:

Furthermore this goes over how easy it is to prompt copilot to generate subtly buggy code, which arguably is one of the worst possible things with a tool like copilot.

If I’m forced to scrutinize every piece of code copilot generates for me, is it really helpful?

No, copilot or tools like it will become a tool in our toolbelt down the road, but just like a hammer is no good for a screw so copilot is no good beyond generating simplistic boilerplate.