In the earliest days of the Web, before there were search engines, and when computers and servers were much slower than they are now, the thinking of internet visionaries was that you'd need a "personal information agent" – a software entity that would understand your interests, roam the Web looking for stuff you might like, and deliver its discoveries to your computer. Gradually it would learn more about you, both from your explicit input and from noticing which items you accessed and which you didn't, and through this process it would improve what it found for you.
Before long, this concept went on a back burner as search engines were developed that could take a user inquiry and return matches from all over the Web in a fraction of second. But today, about 15 years after the earliest search engines, Web users have nothing that does what the personal information agent concept promised. Search can still be a frustrating experience, RSS readers require some technical competence to set up and are not widely used, bookmarking and visiting aggregation sites is time-consuming, and all of these tools require work on the part of the consumer.
An unknown student in a focus group famously said (responding to why he didn't actively read newspapers or news sites): "If the news is that important, it will find me." A "hyperpersonalized news stream," as proposed by Google's Marissa Mayer, and as being developed by CircLabs, will do just that – once a consumer installs Circulate, it will go to work with some basic user-expressed demographics and preferences, begin delivering a stream of news tailored to that consumer, and gradually, it will learn more about the consumer (always with the consumer's full permission and control), and will evolve into an intelligent personal information agent.