I've only been working on this for a few days so far, but I wanted to get a better feel for what you think is missing in content aggregators. I'll be around for an hour or so – below you'll find the demo, the code, the project blog, and the signup form for the release email. Thanks!

http://demo.fraction.io/

http://code.fraction.io/

http://blog.fraction.io/

http://subscribe.fraction.io/

What will the machine learn?

Is it supposed to find stuff I like that looks like stuff I liked in the past?

Or will it do something more interesting?

It doesn't just base it off of things that you've liked in the past, it will also be based off of the things that are liked by the people who liked the same things that you liked in the past – giving you a way to find new content really easily.

how about making it federated instead of centralized?

I'd absolutely love to, but unfortunately only have about six months of runway. I'm hoping that a centralized version will be enough to show people how great it can be, and that will eventually extend the runway long enough to create a distributed, decentralized network with a public ledger of posts, comments, and votes.

ok, I guess I don't know your motivations. maybe you want to go big commercially. I'm interested in an aggregator 'for the peeps, by the peeps' in the same way bittorrent works.

you might want to look into Open Annotation though.

i guess you don't know getprismatic.com yet.

The topical content aggregator? I couldn't find any mention of machine learning or anything related on their website.