2 ms·
I'm currently building a solution to this which is still in very early beta. Using machine learning, I'm working on solving the Goldilocks problem of community
by ChristianBundy 13y ago
I'm currently building a solution to this which is still in very early beta. Using machine learning, I'm working on solving the Goldilocks problem of community affinity (not too close, not too disparate), aiming to fix the hivemind problem that reddit has and the strictness that topical communities like HN have.
I'd love to hear your thoughts if you have a few moments to spare!
Code: http://code.fraction.io http://code.fraction.io
Beta: http://beta.fraction.io http://beta.fraction.io
- 3rd3 13y agoThat looks great. I think they experimented with this idea on reddit aswell at an early stage. However it didn’t turn out to be beneficial. I unfortunately don’t know the details, I read about it in some comment on HN a while ago (I unfortunately don’t remember where exactly).