As a fellow geek, I like this idea but I've checked out similar systems over the years and found I could never fully get on board.. why? Because I ended up checking the source anyway to make sure I didn't miss anything that I might find interesting but that the filter filtered out anyway.

I can trust filtering for e-mail, since it wipes out non-content (usually). But filtering out content I "might not like" is tricky - I learn more about the world by being exposed to things I don't (initially) appreciate so I like to see a little noise in the mix. HN provides a sort of filtered noise that strikes just the right balance for me.

So instead of filtering, I'm starting with Bayesian ranking. I agree about the pop-back-over thing too, but I wonder why? I'm really curious to see why I would pop back over to the regular site. As I identify these reasons, I can decide whether or not these are good reasons or not.

In my case I'm configuring canto(http://codezen.org/canto/) to use crm114. It's not working properly yet, but you have my repo here: http://gitorious.org/foobarnews

Basically I'd like to eventually have a stand-alone app. No point in making a web app if people can just like on another tab. So I've picked what I'm most comfortable with in stand-alone apps: Microsoft stack with F#. I'm using Webkit as a browser and doing some persistence in a MS Access database for now, although that will probably change.

I think I'm pretty far along for just a solo effort. It pulls all the main articles from HackerNews, Digg, Reddit, and a couple more places. You can view the target article and vote them up or down, all without leaving the app. Of course, like any project, still a long ways to go.

Would love to hear how things go for you.