I feel like that doesn't scale well - it works for...
# random
b
I feel like that doesn't scale well - it works for two subclasses, but what if you had 10+? The compiler would potentially have to infer quite a bit, whereas nullability is always a yes or no proposition