I think if it was any other company I bashed, most of them would’ve taken it a bit more deeply ad have the pure-developer response of ‘we made this so we know what we’re doing, you’re wrong, we’re right, so we advice you to take down or edit your post’ where as I’m greatful for the fact that you instead took this kind of a review and discussion in context and were able to teach me and whoever reads this something, and in fact were able to learn something about users yourself. Also, I completely agree with what Elran said about your contribution to the blogosphere as a whole — very few companies do this and I just love those that do.

As for the speed, I’m happy that you are in fact doing something about it, and that’s what counts most.

Do keep in touch and if you have an announcement to make or need someone to test some cool new feature, I’m all ears.

Best,
Sid

I hope that in my previous post I did not sound the least frustrated (since I, honestly, wasn’t) or patronizing (gosh, the least I would like to sound!).

The whole idea was just to clarify a couple of points regarding the learning functionality in Feeds 2.0 in order to help users get the most out of the system with the least effort.

So actually, the fault lies with no one and definitely not with you, since, from your side, you gave exactly the perspective of what users get only from what is up-front.

We are only thankful for that since, sometimes, it is difficult to visualize what other people see in something that you have created yourself. This gives you the impression that users see your application in exactly the same way you see it.

This is obviously wrong, and I think that this is exactly the whole point in the discussion going on in the blogosphere regarding the presentation and review of new products. Without this discussion we would probably make products that would make us (developers) pat each other on the back, but which no one would use!

Best,

Nicholas

P.S. Elran, a new powerful server is coming within two weekds to handle most of database transactions, so please bare with us still being a bit slow for you just a little bit longer

Nicholas – i hope i didn’t sound too frustrated in my previous comment. i should have mentioned that i haven’t been satisfied with any of the feed readers since Searchfox.. and i’ve tried a lot of them.

i will say that i actually liked all the features that feeds2.0 had to offer. i remember thinking that everything was surprisingly intuitive. in fact, i got really excited (and thought i found my new Searchfox) until i imported my 200+ feeds into the system. at the time everything seemed to grind to a halt.

i guess any app has to find a balance between features and speed. but i thought you’re feature set was pretty impressive, you effectively “had” me. i really wanted to use feeds2.0 — i just “couldn’t” because of the speed.

it seems to me that if a service is fast but lacking features, you might still use it (assuming it was still useful to you in some way). anyway, new features get added regularly with mostly any web-based service these days. with an active development team and a CEO that listens to it’s users, it wouldn’t be long before those missing features get addressed in some way. on the otherhand, slowness is slowness. it hurts too much. i can’t take the pain so i have to move on.

but i am definitely willing to give it a second shot – especially if you’ve upgraded the servers to make things faster. it is also really cool that you guys are listening to reactions from the blogosphere, which was definitely one of my all time favorite things about Searchfox (aside from the AI and ofcourse the speed).

well, off i go to play with feeds2.0 ->

I am Nicholas, the CEO of Feeds 2.0. Thank you for your kind review of our service. There are just a couple of things I would like to clarify regarding the way one can use the personalization options.

The system learns that you like a particular post either when you just click on its title to read it or if you mark it by pressing on the “heart” icon. So if you just read posts in the regular way (i.e. by clicking on their titles and have them open in a new window) the system will record these choices that you’ve made and you won’t need to also press the heart icon. The heart icon can be used if, say, you are in hurry or if you don’t care to read the entire article but you still like it.

However learning cannot be complete unless you provide some counter-examples. This is exactly the purpose of the “down arrow”. If you take the time once in a while to mark a few posts by the “down arrow” (uninteresting) this will greatly enhance learning since the two training categories (interesting/uninteresting) will not be hugely inbalanced, e.g. having 600 interesting examples but 0 uninteresting examples. You may never use the “down arrow” and ignore it completely but you will be very much rewarded as far as accuracy is concerned if you do use it.

Feeds that have been marked as uninteresting come with a blue-coloured filled “down icon” which is just a reminder of their status. When you turn “Personalization On” these feeds should be pushed at the bottom of the page or towards somewhere in the last couple of pages if you track many feeds.

I would also like to point out that automatic tagging and Tag Cloud are two features that are not related at all to personalization. By this, I mean that personalization is not achieved on the basis of simply filtering out posts that include tags or keywords of the posts that have been marked by the down arrow and by prioritizing posts that include tags present in posts marked as interesting. So the system does not add or remove keywords from the user’s database but instead it uses advanced statistical analysis of the post’s whole texts to extract semantic features and accordingly classify feeds. So I am afraid that contrary to your suggestion there are indeed some artificial intelligence techniques actually involved in the learning process. For details please see our blog post .

Finally, as far as your example with “google mapsâ€? is concerned I would suggest to all users that it is a good idea to train the system for a while before expecting discrimination abilities between posts with similar thematic subjects since the more you use it the better it gets. This means that given an enough amount of examples it will be able to pick up the fine-grained details that can distinguish between a post about “google maps new version launchedâ€? and ‘google maps still little coverage of africa’ since the extracted semantic features (that I mentioned above) that feed the learning algorithm will become more concrete as the training corpus is getting bigger.

Elran, I can only agree with you on the actual feed read options. However at this stage of development we have decided to shift our attention to the features that make Feeds 2.0 unique among its other Web based RSS aggregators rather than try to implement all the basic reader functions. Obviously we plan to add such features in the immediate future since there are many user requests for them.

I also hope that you will find the speed of the service satisfactory now since we have upgraded our servers and we plan to add even more servers in the near future to handle the increasing load.

Thanks again,

Nicholas

i was also a big fan of searchfox. wrote about it several times on my blog a few eons ago.

anyway, i tried feeds2.com a few times over the last few weeks and found it to be unbearably slow. i thought it had some pretty 2.0 frills – but it’s the actual “feed reader” part that i’m a bit skeptical of.

great site you’ve got going here though..

“Damn it, Bring Back Searchfox!” -elran