The average programming book is pretty awful - many seem to be sold by weight. If you skew towards those recently published, there's a very high concentration of instantly obsolete books that cover surface details of the current incarnation of rapidly changing technology. There are real incentives for the programming book publishing industry to churn out tedious garbage: People buy 'em, instant obsolescence means the same people replace 'em.

Older books aren't necessarily better, but those remaining have been through a generation or two of garbage collection already, making it easier to find the good ones. Excellent books have been published recently, of course (http://www1.idc.ac.il/tecs/, http://landoflisp.com/), it's just easier to find them in hindsight.

Books about parsing techniques* rather than ANTLR (or whatever) specifically, language semantics and implementation techniques rather than an awful 200 lb. Wrox book with a mug shot on the cover, etc. tend to stay relevant for much longer. Rather than reading yet another API guide, check out _The Art of the Interpreter_ (http://repository.readscheme.org/ftp/papers/ai-lab-pubs/AIM-...). Also, CiteSeer (http://citeseerx.ist.psu.edu/) is a gold mine.

I haven't gotten _Hackers Delight_ yet, but it's on my list.

* Recommended: Dick Grune's _Parsing Techniques, a Practical Guide_. Lots of deep content, but would still be worth it for the bibliography alone.