When I find something really interesting in a book that I want to remember later, I write the page number on the end page. A good book might get 5 to 10 page numbers so noted. My copy of On Lisp has 25. Now, not every one of the 25 contains something I agree completely with, but they all made me think. If you're experienced at Lisp, you'll find a similar number of thought-provoking ideas, and if you're new to Lisp, you'll find a whole new way of looking at programming (and you'll find that you can apply the new ideas to other languages as well).
Looking at Graham's code felt like reading my own code masterfully translated, say, from Danish to Swedish. A lot of the ideas are the same, some of the old friends had new names, and there were some new friends that I had never bothered to abstract and name, but recognized instantly once Graham did so (e.g. mapcars, fn).
Along with the ideas, I admire the many handy turns of phrase that make the book a real page-turner:
"It used to be thought that you could judge someone's character by looking at the shape of his head. Whether or not this is true for people, it is generally true of Lisp programs."
"The classic Common Lisp defmacro is like a cook's knife: an elegant idea which seems dangerous, but which experts use with confidence."
"Lisp is not inherently about processing lists any more than Polo shirts are for Polo" (Note that the sentence would have been a little confusing if Graham had written "list processing" rather than "processing lists". In Graham's prose, like his code, every word counts. Time and again I can only say "I wish I had thought of that.")