I don’t see any similarities other than exposing information in function signatures, and if we’re throwing out that baby then why use static types at all? I have to handle all my Maybes at some point? The nerve! I’m trying to get real work done!
Checked exceptions and effect systems address entirely different concerns (except that I suppose an effect system could model checked exceptions but not vice versa) and there are a number of evolving implementations of the latter in multiple languages compared to the one that everyone writes articles about for the former.
It strikes me as a facile comparison and one that bespeaks ignorance of how people actually use these things unless there’s a very deep effect signature floating around to give the argument some teeth. My own use and the examples I steal from tend to expose very small sets of effects and combine them by composition and pretty immediate elimination rather than having to carry around knowledge about the whole universe as one does with checked exceptions.
Local reasoning is the point.