That matches my experience pretty well. The initial sets of results after seeding from my Google social graph / reader subscriptions were mostly about the petty internals of US politics, US professional and college-level sports, unimportant local news from different parts of the US, CSS, and random celebrities. That's about the least interesting thing I can think of, and I can't imagine how it ended up with that kind of article selection from my personal data.
But worse, after 15 minutes of mostly dismissing these kinds of articles (and clicking on anything even remotely interesting), it was still mostly giving results in these kinds of categories. Is this the wrong way to use the site? I have no idea, since there are no usage instructions anywhere.
Right now it needs to be way more aggressive about early negative signals, and introduce more variety rather than try yet another pro-sport article to check whether I'd be more interested in that than the previous 10. Because it's hard to believe that a lot of users are going to persist if the initial results are persistently bad, no matter how pretty the typography.