> the llms write pretty good zig

I doubt this from my personal experience. Every week after a release, I see tweets complaining how AI wrote some depreciated code because Zig is making breaking changes every release. (They are valid in doing so, it's just not AI friendly yet)