Watson can't look things up on the Internet. That would be unfair! Instead, the computer has a fixed, 15-terabyte content store that is "trained" on the equivalent of one million books on diverse topics. The content goes through an analysis stage in which entities such as people, places, things, dates and concepts are marked with meta tags. Search technology retrieves contextually appropriate information quickly, but the real secret behind Watson is its battery of analytic algorithms used to decipher the clue and score the computer's confidence in having the right answer. This application takes two hours to come up with correct answers when running on a single compute node, but that time drops to less than three seconds when running on thousands of nodes available on IBM's massively parallel processing platform.