New Twitter, JTV RFSes; Winter 2010 deadline pushed back 2 days

The internet was pretty closed when DARPA owned it and so TCP was proprietary. Fortunately it became ubiquitous, was written up in an RFC and we can all inter-operate now.

Calling it a 'protocol' is useful.

One reason why I don't use twitter much is that a lot of tweets are uninteresting. Perhaps a Google-style Markov algorithm could work for filtering a feed.

http://blog.directededge.com/2009/10/23/facebooks-news-feed-...

i certainly agree that increasingly, software fortunes will be made by applying ever more intelligent and adaptive criteria to the information that actually makes it to an eyeball.

dave @ davelyon . net

As well as keyword filtering and friend bonus there are a ton of things you can pull into the mix. I was actually working on two ideas, one to tell you whether you should follow someone based on their twitter fingerprint and your preferences, and one to pull out interesting stuff from the stream. As I replied further down this thread, interesting is just damn hard.

Even trying to concretely identify what I found interesting such that it could be expressed as a grammar or decision tree or set of regexps or whatever... yeah, never really got round to it. I'd keep realising new properties of tweets that made them interesting to me.

It's a fascinating problem, though, and very closely related to the "Oh dear gods I have ten thousand unread RSS items" one - that I still haven't solved, nor seen an acceptable solution for! We're drilling down into sentiment right now, but that's not a solution in and of itself; it's an interesting filter though, in terms of only showing stuff with extreme sentiment or stuff that deviates from an individual's normal. It's almost a boringness anti-filter. Kinda works, but is in no fit state to release - again, due to time!

It's not a great system, but even a bad policy is sometimes a step in the right direction.