One thing I love to do is to block all the garbage of websites with uBlock Origin by adding my own filters. That does not give a result as good as this project but it quicker to do.

Yes. One example where this worked particularly well for me is the Youtube homepage, where I blocked all the content so only the search bar and sidebar are still visible. They've since changed the html and I have since made two workarounds for the homepage, but for years that saved me from getting distracted by all the crap people watch and forgetting what I was going to search for.

For this, you could use uMatrix (by the same author).

It would be easier to "make" these with semantic web. Maybe your browser would just have some generic viewer so that you would not have to make any at all.

I'm writing a small JavaScript engine that converts websites to JSON, and displays JSON as websites. Since an entire site is described in the same JSON file, the whole site is a single download, and it rebuilds pages in the browser instead of calling the server - much faster, especially on mobile.

If the browser had this small engine, it could render these sites natively.

Neat. But my comment is about the exact opposite: having a standard way to represent data in place of HTML and custom formats.

Your project sounds very interesting, is it hosted somewhere?

The site is in alpha, at https://www.sparational.com. it has links to my GitHub, and to the site engine on S3. It started as a simple Node project to host a blog and a few random pages. Now, the goal is a single-page application, in more than one sense.