This seems like the very definition of solving the wrong problem, especially for something like Medium.
Really, you don’t need Javascript to load a page of text.
Youtube videos already have their own bar, why is the rest of the page so bloated that it needs bar? All the thumbnails? Maybe it's time to stop loading so many by default. Too many widgets for comments, ratings, favorites, etc.? Again, time to cut down on all the widgets.
The problem is that you can never rely on a load being fast. If you trim the fat on the app side and get a pageload down to 100ms, it might load almost-instantly for you on a 100Mbit/s connection.
But it still won't for someone on a 768Kbit/s DSL connection. In that case, you can't just let them sit there for a second or two with no feedback. A loading indicator is necessary for those instances.
I agree that sites should trim the fat and make pages load as quickly as possible so that a loading bar isn't really necessary. But you have a) instances where you're not in control of load times, like when you're relying on an external API (like Facebook, for example) to return data, and b) even if your app is blazing fast, you're still subject to slow connections that necessitate a loading bar.
But the page itself should not necessitate a page wide loading bar. That's a design flaw. Your page is bloated. Even a 768Kbit DSL should be able to load the page itself rapidly. If you're beyond that just for the basic page itself then you've failed in your job as a designer/developer, IMHO.
And I've designed web apps that never actually send you to a different page (just update the DOM), and in my experience they are extremely fast. Even making many API calls retrieving tons of records to a very modest server. That design is not in and of itself a reason to have load bars.
It sort of puts paid to the idea of progress.
There's also a related "law": “What Andy giveth, Bill taketh away.”
I used to think this was true. For me, the perceived inflection point where perceived speed of software started improving was somewhere around 2008. This may be because I switched to a mac in 2006 and the releases around that era had a big performance focus.
I'd imagine that, until broadband becomes truly instant, you will have a need to ensure users that the page is still loading. A user's patience will never grow. It will only shrink. The least we can do is buy ourselves precious seconds by using a progress bar. It is the difference between "Your call is important to us, please hold" and "You are now 5th in line to be answered. Please hold." The illusion of progress goes a long way in fooling the user's brain into having just a smidge more patience.