"I agree: the WWW Internet is dead, that is your problem. No-one visits websites anymore, everyone has moved to the 10 biggest websites and all data is now siloed there."
That is not the Dead Internet Theory. That's just something anyone can see by looking at the world.
The Dead Internet Theory is that the Internet is already an echo chamber custom fed to you by a collection of bots and other such things, and that a lot of the "people" you think you're interacting with are already, today, faked. You're basically in a constructed echo chamber designed only with the interests of the creators of that chamber in mind, using the powerful social cues of homo sapiens effectively against you.
In particular, those silos aren't where people are communicating. Those silos are where you think you're communicating.
It is obviously not entirely true. When we physically meet friends, sometimes topics wander to "Did you see what I posted on Facebook?" So far, we've not caught Facebook actively forging posts from our real-life friends that we physically know. (Though we have caught them failing to disseminate posts in what seems to be a distinctly slanted manner.)
I am also not terribly convinced that the bots have mastered long-form content like you see on HN. I think we've had some try, and while they can sort of pass, they seem to expend so much effort on merely "passing" that they don't have much left over to actually drive the conversation. HN probably still requires real humans to manipulate things.
Where I do seriously wonder about this theory is Twitter. AI has progressed to the point that short-form content like that can be effectively generated and driven in a certain direction. There's been some chatter on the far-out rumor mills about just how bot-infested Twitter may be, how many people think they have thousands of followers, even having interacted with some of them as "people", and in fact may only have dozens of flesh-and-blood humans following them, if that. Stay tuned, this one is developing.
(Note that while this could be "a big plan", it is also a possible outcome of many groups independently coming to the conclusion that a Twitter bot horde could be useful. A few hundred from X trying to nudge you one way, a few hundred from Y trying to nudge you another, another few thousand from Z trying to nudge you yet another, before you know it, the vast vast majority of everyone's "followers" is bots bots bots, and there was no grand plan to produce that result. It just so happens that Twitter's ancient decision to be dedicated to short-form content, with no particular real-world connection to the conversation participants, where everyone is isolated on their own feed (even if that is shared in some ways) made it the first place where this could happen. Things with real-world connections, things where everyone is in the same "area" like an HN conversation, and long-form content will all be three things that will be harder for AIs to manipulate. Twitter is like the agar dish for this sort of thing, by its structure.)