It seems to me the main problem with all these devices is that the handling an edge-case tends to disproportionally affect the nature of the community. (The edge-case is trolls, correct?) It also seems to me that using heuristics and machine learning should be limited to communities that need to scale up dramatically. The human touch -- specially if it is transparent -- certainly can work for a smallish community that intends to grow responsibly. (And here you can parse responsibly in context of balancing issues such as censorship, quality of conversation, etc.) Anyway, you can still try experiments if you expose a UI for members to opt in on trying a new algorithm/look-n-feel. > Personally, it is the conversations that have most interested me. We've had some good ones, you and I, and it is the potential for those conversations that I find most interesting. That is something that is too difficult to come by, and that is a good measure IMHO. Thanks, btw. Thank you! (Sentiment is shared.)