We can write complex type narrowing without having to write out the type predict ourselves. I’ve never used narrowing logic in functions since this issue hadn’t been fixed. I’ll use this type predicate inference soon.