Your post might be partly applicable to philosophy book. Some books just can not be summarized (for example David Dennet's books - they are too good to let me interrupt the reading), but some books is just not readable to me (most of XX-century philosophy) and in this case LLM works like a charm.

Your post is kind of applicable to fiction books, but anyway I consider them useless. Last time I have read that book was "Hyperboloid of Garin the engineer" because my friend has recommended me to read with drawing a parallel to Elon Musk's business and the most possible endgame for this person. In this situation AI has no value, there is no other way to feel Garin's way, not to peek at Garin's end.

Same with any book of good fiction - peeking at the end do not make this book as finished for you. Asking LLM to draw that parallel instead of you might get rid of reading for you, but your consciousness will not have the person of Garin no matter what prompt you will feed at megacorp.