Thanks for the extensive response. There are a couple points I'll break it down (I'm also loquacious on occasion):
> One of the central aspects of HN is that humans control it, and they have good taste. I don't think you can take either of those qualities away without sacrificing what makes HN good.
What if the user could "control it", i.e. could pick and choose algorithms of how to present views of the same data? Since everything is in the light, you could also share these algorithms - just like you share a collection of browser extensions or editor color schemes, etc. So you could have the "HN algorithm" which has the characteristics you mentioned with the slow-ban, and upvote/downvotes.
Heavily weight certain user's tags (think of voting up/down in terms of raw tags, and the repercussions of voting as an interpretation of those tags). These users would be who "HN" is - in this case the two that you're mentioning. You could also incorporate AI/ML opinions: Sentiment Analyzer A/B/../N, Some Other ML Analyzer, Bob's Analyzer. The main thing is that you have a way of associating identities with opinions, and then creating a projection algorithm that projects stories based on those identities and opinions.
So interestingly, it would be humans running the platform, but enabling each individual user to create his/her own projection algorithms.
> I think I'm not alone in missing that connection. So my theory is that if another site springs up with similar characteristics to HN, but with actual humans running it – people you can actually strike up a conversation with – then those who are interested in good conversation will flock to it.
At some point, won't the volume of interaction with any individual human exceed the ability of that person to respond? Won't the mentions, responses, etc. eventually be too much to handle? Also, isn't the up/down voting what determines in this site what the definition of what the "good" in "good conversation" is?
> That requires being capable of executing that plan, which is the tricky part.
Darn skippy. The characteristics I mentioned above I've already got well grounded with ibGib and it's taken quite awhile. It's a merkle DAG-based open-data design that I've created. Only I didn't create it to be a blockchain, as I didn't know how blockchains worked and was only vaguely aware of Bitcoin's existence. I created it to be a distributed microservice architecture (again, I didn't know the term microservice...I came up with "autonomous service" 15 years ago - still have the whiteboard). I wanted a SuperMemo-like learning algorithm, but the ability to have all of the aspects of the algorithm measurable (not an easy thing), to maximize the learning process. I've been shaving the yak ever since and it turns out to be like some understand blockchain graph data stores to act.