Personally, I think everyone who writes code for fun or profit should have a copy of the CLRS[0] nearby. The first few chapters are enough of have an excellent understanding of the topic of Big O. I would not recommend reading the book cover to cover, but each chapter is pretty self contained and after you get used to reading the book, makes for great weekend reading.

The thing with algorithms is you don't always see how they are useful if you aren't familiar with them, being able to quickly identify a hard problem as hard and have a solution already available is very valuable (even if you have to research the solution). If you code for fun, these are the funnest problems to think about. If you code for profit, even if you almost never need to use sophisticated algorithms in real software, making use of them even just once should give you a pretty good advantage over your competitors (since truth be told, outside of Google and the like most software developers do not have a copy of CLRS on their desk)

[0] http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&...