With flow typing, an expression's type can change depending on it's position in the control flow. Scala does not naturally support this, but can it be imitated?

Is pattern matching just better?

Could it learn something other than just an expression's type as the scope is narrowed?

In this talk, we'll explore a technique for modeling type-safe flows in which an expression's type stays the same throughout the control flow, but its set of allowed operations does indeed change.