At this point, is it still worth it to post these manually?
Checking all post submissions, keeping only the ones with sufficient engagement, and a little summary line "7 other submissions ignored, the oldest dating from Oct 2018"
Maybe I'm too lazy, but that really sounds like something I would let a computer automatically do.
You can do exact URL matching that way, but that leaves out a great many related threads. That limitation doesn't show up in this case but see e.g. https://news.ycombinator.com/item?id=36713263 - same story, but very different articles and URLs.
If anyone knows how to write code to build those sorts of lists, I'd love to know about it! the nice thing is that the relevant data is all public so anyone who wants to work on it can.
Technologically the task can be automated by scraping each submission when it’s submitted, calculate the sentence embeddings of the article, and compare the new submissions with other embeddings. However, that’s how you become a Facebook style social network with “more you might like” and whatnot.
Even that is surprisingly hard! Hard enough that we had to stop working on it when we realized that it's a startup, or at least a major undertaking, in its own right.
I agree, hn clearly stays minimal on purpose, which is why I put the last disclaimer line.
>with “more you might like”
In itself that's a very good feature, it becomes a problem with a host of other patterns to transform discoverability into addiction. Most often these 'bad recommendations' are high engagement content (controversial, or low-value but high attractiveness clickbait) that are purposefully known by the platform to only be tangentially related to the content, if at all.
Maybe I'm too lazy, but that really sounds like something I would let a computer automatically do.