I feel that there is the danger here of a blind leading the blind effect. I tried it out for giggles. Not to pick on my reviewer, but here:
There are alot of words on the landing page [of BCC]. Look at some of the ycombinator pages and you will notice the average site has about 45 words and 4-9 links. Don't give them so many options and so much to read. Otherewise, looks good!
The quality of your users' interactions with FeedbackRoulette is directly related to the quality of the advice they receive. To recruit and retain the best users who give the best advice, you have to be giving out advice which is worth their time. The above quoted advice is well-intentioned but does not convince me that I am likely to realize business value by continuing to be engaged with the service.
To see another take at resolving this problem, take a look at StackOverflow: by breaking the 1:1 nature of the communication, you can use people's desire to peacock to get hypercontributing members of the community to be responsible for most of the good advice on the site, and let the 90% who do no writing merely vote the good stuff to the top. Hopefully they pick the right good stuff. If, on the other hand, StackOverflow picked two random programmers with a question and connected them, the average experience of using it would be getting your query on Rails ActiveRecord validation answered by a junior Java programmer from India. That would be a much less effective StackOverflow. It would likely not maintain the interest of somebody like John Skeet (who, if you are not familiar with the SO community, is something of a local legend).
(Sidenote: if this reviewer was right and there are in fact YC companies with 45 words on the front page... how do I put this gently... there must be someone in your alumni network who you trust on SEO, right? Talk to him about whether that is a good idea. It will be a very brief conversation.)