If you follow the "personalization" field for very long, you learn that the hype comes in waves. No one must know this better than Walter Bender who has run the "News in the Future" project at MIT for more than 20 years. Bender, along with fellow MIT maven Nicholas Negroponte, could likely be tagged as the owner of the tagline "The Daily Me". Much of the latest hype around personalized news and personalized search is a replay of where we've been in this space for a long time now.
Personalization and news have a long history, for good reason. First, news is a broad genre and a popular one: I've heard news described as the thrid most popular online activity behind email and search. And it is not hard to imagine that the news reading experience could be improved by including more news the reader is uniquely interested in. Some would argue, what is unique about the whole online medium is that something like a personalized newspaper could be published to each user. Second, online news is a popular area for information retrieval experimentation and lends itself to investigation of areas like personalization. Why is that? Because the scale of news is so much smaller than broader areas (like the whole web) but the user interest is likely higher than the rest of the web (so you can get people to give you feedback about your experimentation). People can tell you if the news is personalized, timely, well-clustered, etc. where they can't always judge in the whole field of IR. It's not surprising, for this reason, to see features like Google's "News Alerts" start in news and move over to the whole web.
If you take a broad church attitude toward personalization (and I do) you can make the case that at some level, all of the popular activities online (search, email, reading news) are personalized activities. How is that? Well, you and I don't get the same email, we don't search for the same things, and chances are that we don't read the same news. But this isn't to say that the systems we use (for searching, emailing, and reading news) take account of our differences at an architectural level. If they did, we'd be further along toward a 1:1 future.
The sort of personalization I'm talking about here is what I call "cheap user modeling" (hat-tip to mookie). The best case of this in action in an old atom based medium is the good old newspaper. You and I might get the same newspaper, but chances are we don't read it the same way. You and I would get the same search results on most sites - but even that is changing. I'd be willing to bet that the personalization approach that makes use of click through and redirects that is going on over at Google News displaces the "customization" through checkboxes in this early experiment (check the topic of Krishna Bharat's PhD research to understand he's been thinking about this for a long time, and don't believe that Sept. 11 story he tells).
You can see much of this exploration in Walter Bender's "News in the Future" project. Going back to 1981, Bender was unleashing the ominously named "News Peek" which was exactly the sort of "cheap user modeling" I'm talking about. A user would type in a request and see some news that relates to their input. Sound familiar? Its basically what we think of as a search engine today. But it was far from a primitive system - it tried to bring a users desktop (their calendar, email, and location) to bear on creating a user profile to help select the news. Nevertheless, by pivoting news around the readers interest, we had an early prototype for personalized news, with search at its center.
Early on, Bender saw the shortcomings of having few news sources in his system (likely, because it was hard to match queries with timely news from a small pool of publishers). This system he called "Network Plus" was an early news aggregator, with multiple publishers including user created content. You can see way back in the early 80s that personalization presupposes universal selection to really improve the user experience. At amazon.com, when we were building the personalization systems – it would have mattered far less if we had recommendations, but only limited selection. You wouldn’t have trusted the recommendations because they’d only be recommendations of the books we carry. But with universal selection (whether in books, or news publishers) not only does personalization improve the user experience – it is arguably necessary to make the selection useful rather than overwhelming.
In the early 90s, Network Plus was supplanted by “Fish Wrap”. What Fish Wrap brought to the personalized experience was deeper community participation. In his retrospective on 20 years of personalized news (PDF), Walter Bender describes this period as when rather than being the "filtering" process decried by critics of the "Daily Me" was transformed into a community driven news experience, where readers turned into content creators. We see this experience happening at massive scale today through blogging and the recycling of news across hundreds of thousands of blogs.