It's funny that this guy claims fitting the problem to suit FP is a bad thing. But fitting the problem to suit OOP is seemingly a good thing. There is no difference really. All problems have to be made to fit your tooling and practices in same way. The difference is how much squashing is required and the two or three second decision it takes to select the right tool. Only performance optimisations really warrant falling back to mutation of state, other than IO of course. The default should always be immutable. Don't let some brain dead OOP-only troll deter you from seeing the light.
He is speaking is riddles, like pseudo academics love to. Seriously, he is suggesting you can't have an editable text box on a GUI, written in a language like Scala, F# or OCaml? What a moron. (This link will prove particularly embarrassing to a certain person here: http://www.scala-lang.org/api/2.10.2/index.html#scala.swing....) He is arguing an argument that doesn't even exist here, but one that only exists in his own head.
Pure FP, much like pure OOP, is utter shit and painful. The sweet spot is reached by mixing the two and using a multi-paradigm language, ala F#, Scala, OCaml, etc.