Given the advancements of LLMS, have you given thought to automating some moderation to tell the user they're about to leave a predictable repetitive flamewar comment? Ie, a cleverer version of https://blog.xkcd.com/2008/01/14/robot9000-and-xkcd-signal-a...
>>fragme+(OP)
Not yet, but the relevant data for doing this is mostly public, and if anyone wanted to work on it, we'd certainly be interested in what they came up with.
>>fragme+(OP)
Interesting solution. I had some concerns at first but increasing the threshold and manually whitelisting some common valuable but non-unique comments would probably go a long way. Maybe even reduce it to parts of a comment to reduce lazy jokes.