Additback.com turns social "likes" into increased traffic for a site. It analyzes the social graph of users and recommends friends to be tagged who are most likely to read, like and discuss the post. Powered by a state-of-the-art recommendation engine, it also generates content recommendations to increase the time spent by a user on a website.

Joe Google Flu Trends is pretty good at tracking geo-mapped disease trends. They track all related queries (symptoms, drugs, treatments etc). Take a look at the nature paper they wrote in 2008: http://www.nature.com/nature/journal/v457/n7232/full/nature0... . The trend are fairly nicely correlated with CDC data. Our system takes these trends and convert them into epidemic alerts.

Supposedly, San Francisco has reached "Epidemic" scale in flu outbreaks... How is that categorized? What's considered an epidemic? I live in San Francisco and have yet to notice any peers with the flu.

Hi, the whole point of the system is to determine early if there is an outbreak. So that the health officials can take pre-emptive measures well before we reach a point where you start seeing sick people around you. If you can see people with flu like symptoms around you then there is no need for the system as you, yourself can tell that it is an Epidemic. An epidemic means higher number of patients with Flu like symptoms than there should be in a particular time period. The system caters for seasonal increase in the number of patients.

Google Flu trends (Nature, Aug'08) tracks the query volume of the symptoms of a disease like Flu; its a pretty neat idea, especially for countries that do not have a CDC-scale disease surveillance system. An important question is whether/how such a surveillance method can be used for early epidemic detection.

In our work, we augmented the capabilities of Google Flu Trends by evaluating algorithms that can translate the raw search query volume produced by this service into actionable alerts and developed Flubreaks, a live system that uses Flu Trends data: http://dritte.org/flubreaks/

Why we don't see any homerun product-based startups in the developing world ...

Thank you for this input. We didn't want to change much in the most popular Bittorrent client (Vuze) that we used for building this. You are right, we will work on releasing a light (perhaps sans-UI) version of this soon ..

Have you considered building it on top of a simpler open-source client, such as Transmission? Azureus/Vuze can single-handedly bring my modern desktop to its knees, so I can't see it working well on the older systems found in developing areas.

Perhaps you could compress it by releasing it as a patch to Vuze? That way people who already have Vuze (which is fairly popular) can get it faster.

Thanks, will do.

I am not sure what you mean by "reverse the changes", but if you want to improve upload contribution for a low-bandwidth node, it is built into BitMate (it improves both download performance and upload contribution of bittorrent). Actually, for low-bandwidth nodes, it could improve upload contribution by as much as 1000%! Please try out the client and let us know what you experience.

By "reversing the changes" I think the OP wants to add this in as a patch for the Transmission, which is a popular open source BitTorrent client.

In fact I'd love to do that as well, given the chance.

EDIT: Just found out that it's open source, details here: http://sourceforge.net/projects/bitmate/develop

Thank you for your interest. If you want I can upload the code as well to sourceforge. The changes in vuze are very simple and minimal for now.

It'd be a hell of a lot easier for the rest of us to deal with if you upload it to a more collaborative code hosting site instead (GitHub being the prominent example). Even Google Code wouldn't be bad, though; SourceForge is dead and gone.

My friends at sourceforge are going to be really sad to hear this after the last re-hall of the entire site. :-)

Just saw the surging traffic on our download site and noticed the link from HN.

Re: Aggressive sharing. Yes and no. If the torrent is encoded as a Merkle tree, you can verify subsets of a piece. For torrents that only include piece-level hashes, a BitMate client can upload an unverified piece. However, in our experience, its rare for pieces to be corrupted (unless the uploader is maliciously uploading corrupt pieces).

Please let us have your feedback regarding the performance and stability of the client since this our first (read: pre-alpha) release.

This has nothing to do with either performance or stability, but the big header saying "Downlowd link" spells download incorrectly.