He says: "The only thing that knows how to apply a function to a collection is the collection itself." Which is like a monad in the sense that the insides of a monad are opaque; you can only interact with a monad through the functions it gives you.
The "map" function from his "reducers" library has type:
fn * reducible -> reducible
(i.e., it takes a function and a reducible and gives you back a reducible)
while monadic "fmap" is a little higher-order and has type parameters, but it does something analogous:
(t -> u) -> (M t -> M u)
(i.e., take a function from type "t" to type "u", and return a function from "monad of t" to "monad of u"). It's a little different in that Hickey's "reducers/map" applies the argument function itself, while monadic fmap gives you a function that will do that.
Of course, his "reducers" library addresses a bunch of other stuff like parallelism, which isn't something that monads themselves are concerned with. I'm just saying that part of the interface to his new collection abstraction is monad-like.