"internet users in Europe"

Top link is:

http://www.internetworldstats.com/stats4.htm

One more click, but I think at least as much information in that page as the Wolfram results.

"weather oakland"

Google has cute little icons for rainy/cloudy/sunny, current temperature and conditions, and hi/lo for next 4 days. Probably more of the data a "normal" person wants, in a more concise format.

"oakland"

First link is Google maps, more close in view. If you don't already know that Oakland is on west coast of U.S., the Wolfram map could be more helpful. The other information that Wolfram has likely saves some clicks to the Wikipedia or official city of Oakland site.

"uncle's uncle's brother's son"

This is the classic kind of query that Computer Scientists obsess over answering correctly, that is utterly useless to any normal human being. If they think the kind of inference necessary to answer this query is also needed for actually useful queries, they should have demonstrated the actually useful query instead.

"water 550C 3 atm"

OK, if you want an engine to answer your Physics questions, looks like you want Wolfram Alpha. Google did not have anything useful here.

"integrate x^3 sin^2 x dx"

Again, nothing interesting from Google. I wonder if it would be hard for them to tack this on, if they wanted to. But then, they would likely get scolded by all of the Calculus teachers that they are making it too easy for students to cheat on their homework. :)

"bob"

I like how Google finds various common "bob"s, and divides the page to make it clear which is which. Bob Dylan, Bob Strollers, B.O.B. as acronym for Bank of Baroda. This is the kind of thing that is difficult to replicate without watching the kinds of things the world actually clicks on when they search for "bob."

So, if your question is physics or math (maybe other sciences?) use Wolfram Alpha. Anything else, stick with Google.

EDIT: I see Techcrunch has a similar take away:

"The engine looks awesome for science students and researchers."

http://www.techcrunch.com/2009/04/30/the-wolfram-alpha-demo-...

You pretty much just summed the point of alpha in your last sentence, only people who don't understand label it as a google killer as soon as they see that it recognizes common speech.

But well see as it improves maybe it will replace "just google it" functionality.

Wikipedia articles are often condensed information that is on google, but the condensing it's self is useful.

(some company) revenue - and I hope to get graphs for past years, and a pie chart for the current year, given that all that info is in a [public] company's 10k filing.

Napoleon relationships - hopefully I should get a tree showing positive affiliations with Josephine & others, negative ones with Wellington & others, and ambiguous ones with with Talleyrand etc.

I can think of a lot of others, but I'm keeping my mouth shut in case they have commercial mashup potential :-) I hope there will be an API or licensing terms which offer a low barrier to commercial entry.

Looking forward to playing with it whenever it comes out.

This is a completely separate service, and from what I can see, it's going to absolutely rock in its domain.

I'm always confused by familial relations. The diagram seems biased to maternal grandparents (assuming square/circle is boy/girl) - an uncle could also be on the father's side. Also, your "uncle's brother's son" could be you.

I'm pretty sure they discussed these alternatives (and probably deliberately chose this very example as an excuse to do so.)

Seriously, I can see why they're so cagey about when it'll go live, if they let The Internets onto six racks of machines doing un-cacheable, potentially expensive computations they'll pull a Cuil.