The new Xcode 16 AI autocomplete tech is actually really slick when it has contextual awareness.

I'm converting a file from Go into Swift, and I have the Go language code commented out.

As I type the Swift types (e.g. struct MyStruct) it's actively autocompleting types that match Go definitions that are commented out but with the correct Swift types.

This example just has naive Int types, but the one I actually have has enums that are mapping really nicely

This is *exactly* the type of cool stuff LLMs are great at: useful, focused, transformational mappings / inference.

Unlike the sensationalist nonsense tech bros are praising

@b3ll I’m excited to try it! Any problems with Sequoia that you’ve noticed? Wondering if it’s worth the update
@reftonull the standard issues, Rogue Amoeba apps don't work yet, weird stuff broken with homebrew, etc. but nothing game breaking atm for me
@b3ll Hmm, I wonder if it can do things like quicktype. If I paste a JSON file in as a comment, maybe it can construct a model type?
@calicoding @b3ll I've had good success with copilot in vsc pasting response definitions and it practically generating a matching dataclass representing the response
@alex @calicoding yeah that makes sense, given it's all server-side with tons of GPUs at its disposal. The more impressive bit is how Xcode is doing this purely locally (in near realtime)
@b3ll I’d love to try the Xcode but I’m a bit of a nervous Nelly when it comes to Beta 1 OS’s
@GaryAsh yeah sequoia is no different than any other beta haha
@b3ll Been trialing different local LLMs with Continue and VS Code. Once it gets enough context, it’s crazy how good it can keep you in a flow state.