Warning: I'll be overly harsh in this mail because I assume some people that are new to Haskell might browse this thread now and in the future. I don't mean my comments to be considered personal. On the contrary, I think some of us can learn for their own teaching endeavours.

#!/usr/bin/env stack
-- stack script --resolver lts-10.1 --install-ghc

Imagine a "newbie" running this script just to see what it does. After 15 minutes they'll be disappointed why your small demo took so long to build. Only then do they realize that, what is it today, about half a gigabyte maybe, of their monthly "high speed" bandwidth and of their precious space on the cheap but decade-old hard drive is gone.

Of course you need all that data because there's a ghc in that data, but it should say so on the tin. So please never send a script with this line to a "newbie" without big warnings so they can make an informed decision. Maybe they would want to install ghc at their uni to save on that precious bandwidth at home or something similar.

In fact don't even send scripts with this line to seasoned haskellers. If your code is compatible with the ghc 8.2.1 on my system, why should your script assume I want stack to install 8.2.2 right now?

Of course this behaviour is ingrained in the DNA of stack – after all it values convenience for the professional user more than convenience for the unprivileged newcomer – but AFAIU it can be persuaded to be more inclusive. So especially here, we should maybe try to do that.

  interact
  ( show @Double
  . sum
  . mapMaybe readMaybe
  . words
  )

Also, while I do like the scripting possibilities, interact is possibly the worst way to introduce a newcomer to it because you can't really test it in ghci. Apart from all the other shortcomings it has. It is nice in that it hides IO, but…

Again, not meant as an attack, but since we're talking about giving newcomers a good experience, I thought these two points important.

Cheers,
MarLinn