It's undeniable that there is a significant amount of Rust software still reliant on unstable features, but I have not seen any numbers that would put it in perspective. Having consistent measurements of nightly vs stable adoption would help a lot, inform any conversation, and efforts to move off nightly.
So I think it's premature to say that 'nightly Rust is the only Rust', The article suggests that people are using nightly as the de-facto recommended toolchain, without evidence (though I'm sure somebody has recommended nightly be default). I'd encourage people not to recommend nightly and for Rust library developers to aim for their crates to be compatible with stable. I personally don't have the same sense as the author that Rust application or library developers are using nightly more than they have in the past. Rather I know that developers of key nightly-only libraries are actively involved in ecosystem migration off of nightly.
One of the biggest difference between the 2 / 3 split is that stable is generally compatible with nightly, and is intended to catch up with nightly, whereas Python 2 is not such a strict subset of Python 3 and its development is done. That is, nightly is intended to be a staging ground for stable; Python 3 does not have that same relationship to Python 2.
(Edit: I stated this awkwardly. In this comparison nightly=Python 2, stable=Python 3. The point stands that the relationships between the two branches are different between the two languages. Nightly development feeds into Stable, Python 2 doesn't feed into Python 3, it is not under active development.)
Because of this relationship there's still reason to have confidence this dependency on nightly will shrink over time. But not disappear. We would always expect some people to depend on nightly. That is its intended purpose - as a breeding ground for future features.
Some of the biggest features keeping people off stable Rust are coming, and their biggest users are actively migrating. In the near future there will be fewer reasons to use nightly, not more. And beyond that the reasons to stick to nightly will evolve as Rust evolves. The next demographic of users who will be stuck on nightly will be embedded developers, and that contingent will increase over time. But even as that group increases, the group that is freed up to use stable Rust will grow also. And slowly the stable ecosystem will coalesce for all classes of users and all types of code.
Though I don't believe it's happening to the extent the author does, I'd rather not see Rustaceans who care about Rust's future to adopt a nighly-first mindset. More things will get more stable, faster, by working together. But regardless of people's dispositions to the stable / nightly split, more of the ecosystem is going to work on stable over time - that is how the Rust feature pipeline works - and eventually stable Rust will be the only Rust.
I still expect it to all work out ok.