As someone who writes a _lot_ of Python code (mostly in 3 but with occasional switches to 2), who maintains several libraries which work in both 2 and 3, and as someone who uses a wide range of libraries in stats, machine learning, networking, web development, etc. in my python 3 work...this piece seems totally disconnected from reality. The problems it describes are relentlessly overblown when they're not simply manufactured from whole cloth.
Put another way: the differences between 2 and 3 are not great, the vast majority of libraries you'd want to use (i.e. are actively maintained or of at least good quality) work in 3, and while I don't doubt there's great piles of Python 2 code moldering away big "enterprise" apps and those sorts of places, it's ever been thus in that space, no matter what the language, and doesn't pose any sort of existential threat to Python.
(edit: I used to recommend "Learn Python The Hard Way" to newcomers, and have just kept reflexively doing that over the years because I wasn't aware of a better resource. But if this article has accomplished one thing, it's that it's spurred me to look for a replacement)