I'm not fully celebrating GATs yet due to all the limitations mentioned in the linked stabilization blog post. [1] I'll be ecstatic if and when I can write a std::iter::Iterator impl that borrows from itself. Fingers crossed that can happen. And I'll be watching to see what libraries implement based on this MVP.
I'm celebrating other things in this release. let else will be great for ergonomics. Stable backtraces in std, split debuginfo on Linux, and const offset_from all address pain points for me. (Backtraces will be more useful though when the provider API lands. Then I think we'll be able to have really great error chains. Right now, there's no obvious way to go from a &dyn std::error::Error to a backtrace without guessing the concrete error type and downcasting, and the same problem applies to other forms of context I'd love to have.)
