[DanielBMarkham]

Save-for-later is used when the material is too involved for a simple reply, such as if your boss wants your analysis of a report he sent over. Some small percentage of communication involves this type of work, but not much. You'd probably go to a regular browser for this work.

To reply you just hit the reply button and speak. The system will try to do voice-to-text recognition and then it will send out both the initial message and your voice reply to an overseas service that would QA the voice-to-speech conversion. This allows you to train your system without having to do all of that stupid setup work. It also eliminates a keyboard. It's 2010 already. We might not have voice-to-speech worked out, but it's not like we can't just talk and have the system work it out, either. This also provides seamless integration with cellphone devices by use of voicemail.

"Going somewhere good" and "sites" are concepts that are only about 20 years old. You want material that interests you, wherever the source. The internet doesn't have to be some huge virtual city, with the browser taking the place of the tour bus, where you drive from site to site participating or observing. The model doesn't have to work that way, and it's fundamentally broken if only from an efficiency standpoint.

The system can keep track of all the sites you might like, your responses, and even aggregate you in groups with other like-minded folks. You don't go to a site, the material comes to you. Which is as it should be. The brick would handle whatever subscriptions were required behind the scenes, allocating traffic to and from perhaps millions of sites based on your preferences over time.

Look at it this way. Let's say you participate in 3 aggregation sites. Each site has article X. For each site to function correctly for you, you'd really have to consistently upvote or downvote article X on each site. In a P2P system all of that nonsense goes away.