Some of my current curiosities are served only by reading academic papers and contain algorithms that are only documented in patent filings. I see no books.
And another thing, even in computing some of the best printed literature can stink. No matter how many times I read TAOCP or how many lectures I go to I don't think I'm ever going to get an insight in to how Rudolf Bayer came up with red-black trees... an intuition I feel is much more useful to an inquisitive mind than learning the special cases and transformations, taught in CompSci and detailed on Wikipedia, or proving overall asymptotic complexity. TAOCP is a masterpiece but, as a reader, I will never feel the way Knuth feels writing it.
Knuth's TAOCP is an indepth encyclopedia on algorithms. It analyses algorithms that Knuth has learned from the field, but it's purpose was never to develop the "intuition" on development.
Algorithm Design by Jon Kleinberg is an expensive book, but this one teaches algorithm design methodologies. Instead of learning "Linked Lists" or various "Sorting" or "Seminumerical" Algorithms... Algorithm Design teaches approaches to algorithm design.
--------------
Personally speaking, I have more hope in smaller books such as "Effective C++" or "Exceptional C++", which stands head-over-heals over the best "Web Advice" on C++. (IE: The C++ FAQ, which is very good at C++ Advice but doesn't stand up to books).
That said, major discussion of deep programming features and debates are now conducted online. Scott Meyers notes internet discussions on C++11 / C++14 as a major source of material on Effective Modern C++ (his new C++14 book).
Similarly, "Exceptional C++" is basically a collection of Herb Sutter's blog posts... updated based on the comments from the internet. Herb Sutter basically wrote "Exceptional C++", opened himself to internet criticism / discussion of his ideas... and then released the distilled work as a book. The book is a higher quality than his original blog of course (Guru of the Week).
Books will always be superior than the internet, by nature that I'm generally willing to pay authors $30+ to $100+ for their in-depth knowledge. (While the Internet is free). But the internet has forever changed how even bookwriting is conducted.
That said, the majority of books are crap. So its best to preview the author's writing style before buying. Blogs are actually one of the best ways to preview an author.
And another thing, even in computing some of the best printed literature can stink. No matter how many times I read TAOCP or how many lectures I go to I don't think I'm ever going to get an insight in to how Rudolf Bayer came up with red-black trees... an intuition I feel is much more useful to an inquisitive mind than learning the special cases and transformations, taught in CompSci and detailed on Wikipedia, or proving overall asymptotic complexity. TAOCP is a masterpiece but, as a reader, I will never feel the way Knuth feels writing it.