WolframAlpha ... is like a roomful of idiot savants.
But I do have two things to say about the events surrounding the launch. Far be it from me to offer Stephen "brain the size of a planet" Wolfram advice, but here goes:
Mistake 1: Allowing Wolfram Alpha to be pitched as a search engine. Clearly, it isn't. It's much more like a front end for Mathematica.
Mistake 2: Allowing the "Google-killer" hype to get out of control for the sake of publicity. Cuil anyone?
Why should I care about this? So that we learn from these mistakes and don't let the same things happen to our SM@LL project.