From an article in Wired in 2013:
“It turned out that nonlinear reading spaces had a problem: They were incredibly difficult to write. When you tried to make an argument or tell a journalistic story in which any individual section could be a starting or ending point, it wound up creating a whole host of technical problems, the main one being that you had to reintroduce characters or concepts in every section. Feed did manage other interesting hypertext experiments: We annotated important documents or passages from new books, and we held multithreaded hypertext debates. But we never managed to publish a true branching-path narrative. This wound up being true across the early web and remains true of our hypertext today. At last count, there are somewhere in the neighborhood of 30 trillion web pages, all connected through the axons and dendrites of hypertext. How many of those pages involve real nonlinear storytelling? Almost none—the rounding error of a rounding error.”
Read the rest here.
And check some pseudo-nonlinear storytelling here.