I think a cool idea for a browser that could be implemented fairly easily (on macOS at least) is to be a browser that only works with PDFs. That is, you can click links in PDF files to other PDF files that will open, with standard back/forward buttons, location bar as well maybe. With just the browser infrastructure, you could create a whole “new internet” of just PDF documents over http. To make it interesting, it probably makes sense to remap common citation formats like doi to actual PDFs by scraping the webpage. At least until direct PDF linking becomes more common!