Your criticisms here are too weighted towards an interpretation of her piece as describing, well, copies — which are what have their antecedents in commonplace books and annotated commentary. Her piece is criticisably mushy in that it is not defining the way in which what she’s describing is not embedding, because it definitely is to the extent that she’s saying that authors can fix typos and that the unpermitted insertion of new content would be a problem. Fragmentions don’t allow the text of a page to change without potentially breaking the link and so represent a set of constraints more aligned with the copy than the embed.

Also, quoting bits of things is very Fair Use, but using copyrighted images instead of Unsplash to illustrate an article is a classic example of You’re Going To Get In Trouble, so I’m inclined to say there’s a good reason why people keep trying to crack the monetization egg every six months or so.

Love love love the idea of federated knowledgebases / interwoven wikis / cross-pollinated digital gardens, but I balk at the idea of internal links and webmentions and relations that have meaningful types within the domain of the wiki all being collapsed into one category. Here’s hoping whatever https://multiverse.plus/ is coming up with provokes more ideas!