Thanks, I know that now, but it's the "almost" part that trips up people like me! And it's not helped by the fact that the same sources will say something like

Thanks, I know that now, but it's the "almost" part that trips up people like me! And it's not helped by the fact that the same sources will say something like "a monad is a monoid in the category of endofunctors", yet if you take Hask seriously there should be a discussion of endofunctors on Hask etc. When details are missing my brain gets stuck on figuring them out (whereas for Haskell the right approach is to take it with a pinch of salt and move on I think). Edit: to be more specific, you need to specify what constitutes a type - all possible ADTs under all possible names? Are equivalent functions considered the same morphism? What about seq? Is there a privileged "IO" type, and if not how is it represented in the set of types? What about other monads hiding different side effects? What about bottom? Many of these choices will of course be equivalent in some sense, but technically speaking you don't have a category until you've nailed this stuff down.