I like plain C casts. The C++ ones seem unnecessarily verbose and get in the way of my reading comprehension

[cjbacchus]

No description provided.

[PetrilloAtWork]

Maybe we can explain better in the text.
The problem of the C-style cast is that it can do any type of harmful things, from truncating a value to removing constancy to stealing your car.
I would recommend the braces-style cast instead as default: unsigned{ 8 } or (unsigned int){ 8 } (yes, that's not what is written in the guideline).

This default will not work when trying to convert:

• an integer into a floating point; narrowing issues are rare but extremely sneaky; spelling out static_cast() then tells "I know what I am doing"
• a floating point number to integral; here the rules are clear, but I'd still recommend passing via something more explicit (trunc(), or floor()/ceil() — you do still need a cast though)
• an integer into a smaller floating point; narrowing issues are rare but extremely sneaky; am I repeating myself?
• messing with constancy: for that, there is hell const_cast()
• one type of pointer to another (this is what the other three named casts are trying to improve)

I may be charging too much on the details, but if I assume that people by default use the braces cast (the most harmless), then when I see a different type of cast I "know" they must have put some thought in the choice.
This said, I am not a fan of constructs like:

for (gsl::index i = 0; i < static_cast<gsl::index>(v.size()); ++i) { /* ... */ }

and discouraging is not forbidding.