Are you talking about Half-Life? ;)

Maybe you could do a "I can do data science in a GUI" talk and amaze us all (addressing all points the OP made). Hadley has been working for years on making data science programming with R accessible: development of expressive domain-specific APIs, books freely available online, etc.

You can use environments for hash tables.

It is bad practice because you are hiding the side effect.

What about `a = a + 1`?

The r-value is ephemeral it doesn't exist once the statement ends.

There's nothing weird about closure-based scoping. You find this in Python, R, Javascript, ... It has roots in the lambda calculus.

The point of this kind of UI is that they become an extension of yourself. Saying these UI are old and clunky is exactly like saying a keyboard is old and clunky.

Only when training is lacking. And I don't mean this in a derogatory message, either. People that are trained with a keyboard are pretty much guaranteed to be faster than folks that have only learned to work with a mouse. Discover, though, will be much much faster with an exploratory interface. To that end, the mouse will lead to faster discovery in most cases.

That problem (solved in recent versions) has nothing to do with pipes though.

Unlispy? What about threading macros? The voodoo you're referring to is simply computing on the language, which is quite lispy as well.

That's called a model.