I jumped straight into an in depth conversation with Ton Zijlstra, a long time "blog buddy" who I met for the first time in person last year at BarCamp Amsterdam.
Somehow, we quickly dived into feed reading strategies. Ton uses essentially the "Blink" strategy (my words, not his) to capture hilights of the content that he reads. And that was our first point of interest. He doesn't read via content or by information/topic slice, but rather primarily reads via people. Ton uses Lektora, and tends to read in disconnected mode on the train.
Aside: I need to write up some thoughts on the use of media by "commuters"...I have a virtually non-existent commute, so my media usage is very different than those that do commute. Silicon Valley/the Bay Area has very long commutes...does this have an effect on innovation?
Ton's strategy of skimming/browsing to pick out the high level themes that he is looking at, combined with his offline status in reading, means that he picks out themes as well as links to follow and captures them on a wiki that is locally installed on his laptop. These are entered as kind of "to do" or "followup" items.
I find this mode of operating fascinating. I think I will need to hack my feed reading to become more efficient like this...it seems to encourage better/richer/more intense feedback loops, too. Right now, my feedreader is a bit like an email inbox (with the attendant "worry" about unread status as well...) where each individual item isn't critical and/or directed at me...except for some of my notification feeds, which are in fact directed and personalized just for me.
Do we need two modes for feeds? I know that my "to do" and "project status" feeds -- with time critical and/or followup information necessary -- are perhaps better suited to delivery via a pubsub infrastructure (there's Jabber again...), coupled with a way to pin/flag them for my attention. Coupling this with the Growl notification system might work.
In any case, Ton would like a stronger focus on the people to which feeds are attached. In his head, his social network of blogs that he reads is quite clear. In his feed reader, it is a list of blog titles in a list. Not quite the format to which a human mind can easily pattern match to a group of social connections.
I pointed to the 43 People subscriptions page as an example that is quite a bit closer than what we have in feed readers today. You subscribe to people, and then attach feeds to them. If they aren't in the system, you can add them yourself as placeholders, seeding their identity within the system with feeds you specify. In fact, anyone can add feeds to people, although the people themselves can remove them. You can additionally choose to subscribe/follow only a subset of the feeds defined about a person.
While not designed as a feed reader, I have been using my subscriptions page as a back up feed reader whenever I am away from my laptop, which is the centralized home of my subscriptions and feed reader.
Even if the Robots don't evolve subscriptions into a full fledged reader (e.g. read and unread status), their APIs are quite open, so one can imagine building such a reader based on this information.
As well as having a 43 People entry for myself which has many feeds attached, I run my own "person aggregator" at http://personal.bmannconsulting.com for my non-business persona. This is composed of Flickr pictures, All Consuming entries, etc. -- all feeds from the many systems that I interact with and create content at. Subscribing to that aggregate feed is quite a complete picture of that non-business persona.
After our direct discussion (which continued...lots more to write), we decided to do a session on how we would like to see social software tools to evolve to add/display more of a shared context, based on information that we post and the interlinking (both semantically and via http) between them.
But most importantly, as per the title of this post, subscribing to people and their content, and deriving connections and new content that helps inform and filter what you are reading.