[dang]

We try to do that informally but of course the quantity is overwhelming. It's a natural place to experiment with AI classifiers, and we'll eventually get round to that.

So far, the top half of this thread seems to be about the current release - that's after some of the manual moderation I just mentioned. (Basically, we try to downweight generic subthreads until the top subthreads aren't generic any more. There's certainly a place for generic tangents in curious conversation, but they should be lower on the page, and tend to get upvoted a lot higher than that.)

If you (or anyone) sees a counterexample, i.e. a generic subthread in the top half of the thread, it would be interesting to see a link - we can treat the current case as a datapoint.

[consumer451]

Thank you for the reply, and the work.

As a protentional counterpoint to my request, this is just perfect:

https://news.ycombinator.com/item?id=48468156