Earlier quoted context omitted.
It's true that I could be lacking imagination. Because to me the portal proposal reads like it's simply moving the implementation of Google's AMP carousel into the browser in order to solve the address-bar problem. Your suggested use cases aren't terribly convincing, I'm afraid. Zooming UIs are perfectly possible using existing web technology. It would be interesting to have a standard way to embed Twitter/Facebook-s…
Sorry about that, I didn't mean you have little imagination, but rather that any reading of a tech spec requires a bit of imagination. It doesn't seem all that unreasonable for a web page to show a preview if the window size is small enough. You could do that with a media query. And then when it goes full-screen, it would already be loaded and that would just be a resize. This would take cooperation from web sites.
Embedded youtube videos could support a navigation to the video page where the video just animated from the embed to its in-page position while continuing to play.
Embedded twitter posts could be clicked on to do the same sort of transition. Really this sort of thing applies to all kinds of media embeds.
Amazon ads could do the same sort of transition to the advertised item.
Basically, this could allow multiple different domains to coordinate to get some of the fluidity that you can build on a single page. Ex. you could develop a federated social network of many different providers (or video sharing or whatever) while retaining the UX of interacting with a single thing.