My thoughts on Wolfram Alpha
Whilst I can see what Wolfram Alpha is trying to achieve, (both enviable and interesting), it isn't really that impressive when the search mechanism is restricted to selected databases and feeds.
The general approach is sound - using structured data from vetted sources to build a searchable knowledge base with a computational edge. However, semantic search will make this something that can be very easily replicated and therefore will mean that other search engines can emulate and surpass Wolfram's approach without even having to build the functionality beyond processing semantic tags.
To be fair, Wolfram Alpha isn't claiming to rival Google. So, not a 'Google killer' as has been reported, but a useful experiment in information retrieval. If you search for 'expenses scandal' Wolfram returns no results, whereas Google returns 4 million. When people search they want answers, but often the answers can't be found in vetted sources, so semantic mark-up across all web documents is critical.
A Google killer would need to embrace the semantic web and be instrumental in creating it. It would need to be something that finds a way to factor in reader sentiment and social media, spider the web and also have a marketing budget in multiple millions, with an even bigger tech infrastructure budget. Wolfram has none of these things.
Also worth noting is that Google already has an answer of sorts to Wolfram Alpha in Google Squared, which tries to compile data in a more factual, tabulated format for comparison rather than in the usual text snippet approach, but does it automatically without requiring armies of staff like Wolfram. Wolfram clearly has a long way to go to become useful.