That thing about applicative is that I predicted the idea also years ago. Central in the (IO) monad is that you can inject a value, calculate along with it, and never observe what comes out (the latter makes the IO monad safe.)
And I predicted that any algebraic structure that has those qualities is good enough, or equivalent, to a monad. Moreover, that function composition seems to be a neater fit if you want to compose things outright.
So, now some people want to change from monad to applicative. Which is somewhat neater, but also falls into the category of mindless hogwash scientists amuse themselves with.
Ah well. Another prediction done right, I guess.