I've found Dr. Richard Hamming's talk to scientists, "You and Your Research," to have useful advice on the subject. http://www.cs.virginia.edu/~robins/YouAndYou

I've found Dr. Richard Hamming's talk to scientists, "You and Your Research," to have useful advice on the subject. http://www.cs.virginia.edu/~robins/YouAndYourResearch.html http://www.cs.virginia.edu/~robins/YouAndYourResearch.html In particular, he talks about keeping track of a set of important problems worth solving. Then, whenever you find an interesting new technique, compare it against your library of problems, and see if there's a good fit. I expect the reverse is also similarly useful: keep a good library of techniques on hand, and look for one that's applicable whenever you run across an interesting problem. This type of approach isn't necessary for building twitter or facebook clones, but it may well be ideal for looking for new ways to apply machine learning and other types of advanced algorithms, because it's primarily applicable in cases where there's an actual problem to solve.