My thoughts exactly. We use python extensively, so have a large code base laying around, and, even with that, I chose Python 3 for the new project I recently started since it is fairly independent.

I'd word this more like "projects should use Python 3 unless you have an existing reason (a core library, existing code base, etc) to stay on 2.7". A year ago, I couldn't have said this. Six months ago, the possibility of missing, critical libraries was still very large, but, today, it's a reasonable approach (IMO, of course :).

The other bizarre to me part is how Python 2 is called "Today" and Python 3 "The future". Shouldn't they be "Yesterday" and "Today" respectively? Or maybe "The snakes of yesteryear" and "The contemporary Monty".

Most of the guide was written two years ago: https://github.com/kennethreitz/python-guide

It's easy to write Python 3 compatible code with 2.7, you can import most things from __future__. You lose some cool features though.

It isn't, but it is the opinion of the people who run this project.