Sounds like you were inspired by the Layar link with this one. So, without harping on augmented reality...

What if mobile devices had an internal "user profile" similar to those found on Social Network websites. (Of course, this info could just be retrieved from Social Network sites via the user entering various service logins, but it would work the same regardless.)

Anyway, the device then, based on keywords, favorite activities (hobbies, books, movies, etc.), could initiate content retrieval and display based on a persons proximity to various POIs. Likewise, a person could setup certain rules to where specific types of media would be tied to certain locations (think "Personal Soundtrack"). Also, based on media that was tagged by the user, notifications could be triggered when the person was in proximity to items/areas (POIs) that related to those tags.

This could also extend into manual searches that a person initiated on their device. What if Web searches could be tied to a persons location and specific profile keywords. Example: A person searches for pizza and the search results pulls up several places that are local to the persons physical location (no need to search with city if you have a GPS on hand), as well as pizza places that are near to soccer fields (since soccer is on the of the person's interests), as well as a couple of not-so-close pizza places inside of movie theaters (because the person has indicated that they like movies), and lastly, there is a nearby store that sells Italian cook books (because the person enjoys cooking). All of this would then be mapped out in conjuntion with the person's current location.

Lastly, everything needs to be in real-time. Why wait for the person to manually input anything when the device knows exactly where the person is. And, if the device also knows what the person likes, everything should just become available whenever and wherever appropriate.

Tim