The voting is based on a pagerank reputation system. Users earn reputation from getting thier comments upvoted and, in turn, increase the rank of the posts they upvote. This makes rankings sybil-resistant making manipulation harder and moderation easier. So far its been a great way to keep communities focused and resistant to mob mentality.
Each community manages their own reputation system (admins are in the personalization vector).
There is also a prediction-market mechanism overlayed on top of the rankings — you can bet on post's relevance within a given community (this is abstracted from the UX at the moment).
Disclosure, I'm the founder.
Also, TIL of the concept of a Sybil attack, AKA sock-puppeting: https://en.wikipedia.org/wiki/Sybil_attack
> an attack wherein a reputation system is subverted by forging identities in peer-to-peer networks [...] named after the subject of the  book Sybil, a case study of a woman diagnosed with dissociative identity disorder. The name was suggested in or before 2002 by Brian Zill at Microsoft Research
Not that the actual ranking of the post is determined via the pagerank algorithm which is independent of the coins and betting.
If the post has a higher ranking than average it gets a portion of the newly minted tokens — kind of like block rewards. These rewards get distributed to all the users that staked on the post proportional to the amount of shares they own. Everyone also gets their original tokens back. So in reality this it's prediction market 'light'. All users pay a 'tax' on their token balance via the inflation rate and but get a chance to win a portion of the pot by participating in curation.
The betting mechanism is similar to and inspired by Steemit. Except they don't use reputation ranking to resolve the market. Steemit posts are ranked by how many tokens were staked on them, so a whale can basically force the market and get all the rewards — self-fulfilling prophecy. In our case its is more akin to an actual prediction market where the prediction is separate from outcome. At scale the prediction market can actually be a useful way to surface recommendations.