I think I know what FriendFeed is really all about, and it’s simply brilliant.

For those who have been following the conversations about noise on FriendFeed you know that many people (including me) have been asking for the ability to create filters to solve the problem. Some people have even suggested that the solution will be the Semantic Web or that FriendFeed is already a step in that direction. I think FriendFeed’s creators have even more ambitious plans.

FriendFeed is going to be an intelligent agent.

The setup is perfect. FriendFeed is gathering all of our attention data. It’s just a small leap to analyzing that data and finding patterns of usage for each FFeeder. Then it’s just another small step to apply those patterns to incoming data, both as filters on streams we subscribe to, and as alerts on streams that we don’t. It’s like reverse web site analytics: instead of analyzing what lots of people find interesting, it’s finding what’s interesting to a single person.

FriendFeed’s founders are from Google. They have experience with finding interesting things in large amounts of data. Search, from one angle, is simply gathering data from many sources, extracting what is relevant within a context and then prioritizing the results by ranking. It seems as if the same type of algorithms that are used to determine page rank with search engines could also be used to determine what’s interesting to us in the pool of our attention data.

At FriendFeed is each user is a context and ranking is a result of attention. With page rank, every link to a site is a vote for its contents. With attention rank, there is a chance to be even more precise because there can be negative votes as well. You’re voting on what’s interesting every time you participate and even when you don’t.

Votes Up:

• Contributing content (big)
• Creating or joining a room (big)
• Subscribing to someone (big)
• Creating an imaginary friend (big)
• Commenting (big)
• Liking (minor)

Votes Down:

• Leaving a post in your stream untouched (minor)
• Hiding (big)

The more you participate, the more you share about you and your interests, the better the agent will be able to work for you. All of these votes reveal keywords and context about what’s important to you and what’s not. And you’re constantly generating that information.

So what’s the big deal? It means a pleasurable, efficient, speedy, personalized experience consuming social media with little wasted time digesting noise. The uses could be to filter your stream (no more noise), find people who are similar to you (expand your network), find content you might not have noticed that you would like (information discovery), and even customized advertizing (that you’ll actually appreciate)!

FriendFeed’s founders have most likely learned yet another lesson from their old employer. Google has set itself up to be the gateway for the bulk of the world’s digital information. There’s a lot of power in that. FriendFeed is setting itself up to be the gateway for our attention. We’ll become extremely reliant on it to do the grunt-work of filtering and alerting us to what’s important to us. It will become a service we can’t live without.

And it will be a profitable service as well.

FriendFeed as an intelligent agent can also be a recommendation engine for products and services through tracking how often people with similar interests to you, or people you subscribe to (and generally trust), recommend them. But the most lucrative profits will come from extremely well-targeted advertising. If a FFeeder mentions he’s looking for a new camera, all the sudden he could be presented with offers from camera retailers for exactly the model he’s mentoned. As long as FFeeders participate and announce they are seeking something FriendFeed will deliver advice and ads of relevance. We won’t even mind we’re being advertised to because it will be useful and timely for us. Adding Amazon Wishlists is already a feed option. What if you received a coupon any time something in your wishlist went on sale? Would you be annoyed or appreciative? And what advertiser wouldn’t want to participate and have such a high probability of sales resulting?

It’s even possible that FriendFeed could share revenue with FFeeders based on their influence in resulting sales, giving a cut to the influencer in a model similar to Squidoo. Will they? Probably not, considering there’s possibilities there for abuse by spammers, but maybe. That could even help alleviate the issues many bloggers have with the conversation being diverted from their blogs to FriendFeed. Bloggers could still make money by producing good content that way.

Why do I think FriendFeed is going in this direction? There are already hints in what features they have and what features they don’t (besides the obvious aggregation features):

• Stats for the top 10 people you find most interesting and who find you most interesting are already available. It’s logical to keep expanding on that.
• Only offering Like and Hide on posts simplifies the voting system. Offering a starred ranking system would just confuse things.
• They haven’t offered advanced threaded comments or a full blogging platform and probably won’t. Again, more simple to implement a voting system without these things, and they would be distractions from hooking 3rd party streams into the service. It sets up the impression that FriendFeed is the place to go to consume and attribute information instead of create it.
• Rooms are an awesome way to get people to both cast a vote with their attention and also add context to any content. If content shows up in a room it’s most likely related to the room’s subject.

Maybe this is all just a pipe-dream of mine, but if this isn’t FriendFeed’s developers plan after all then it should be. The potential is there, not only for providing a service that so many people would benefit from, but simply to make a nice profit. I have great expectations.