I think gradual typing is definitely likely to become the norm. It allows you to develop things using the dynamic style and then add type annotations where appropriate. It also avoids the problem of mixing the proof into your solution.
I think the main problem with types isn't that they slow you down. In my experience they actually don't and reverse is actually the true. The main problem is that static languages are often too complicated for non professional software developers, i.e. designers, system administrators, data scientists, etc, to be motivated to learn them. They just need to write the code, and when they do it themselves, the stuff is done much faster and much more efficiently.
The inherent complexity of static typing is precisely my argument. You have to demonstrate that the additional complexity actually adds value in the long run. This seems like a prerequisite before you start extolling this approach.
