Wow! It's great to see an Andy Matuschak related post. I've been checking out his content over the last week, and it's really inspiring.
I've been theorizing a similar idea for browsing Wikipedia (or pdfs), except it's only two panels. One is the current page, and the other is a Obsidian-like node graph. As you click on links, words, terms, or phrases in the article, new nodes are created. These would be concepts you're unfamiliar with.
This would build out a tree of ignorance, with an implicitly generated knowledge dependency graph, which you could systematically study.
My goal is to be able to delve into highly advanced topics for which I have little background knowledge, building out a "syllabus" as I go.