Consuming Hacker News

I think aggregators like SkimFeed miss an important point, often made on HN: for a lot of people, including me, a lot of the value of HN is in the comments. In fact, going to the extreme, some posts have little learnable content other than TIL. First example that comes to mind is a post I recently upvoted: Reply of the Zaporozhian Cossacks (https://news.ycombinator.com/item?id=19946989) which had a small number of information-rich comments. Or there are pots where the content is obvious from the title, where I go directly to comments to see people's assessment, e.g. Federated Learning (https://news.ycombinator.com/item?id=19944510). On average I would say value of comments to post is 80 to 20 percent for me.

So, rather than tools for submission-based skimming/alerts, I'd like to see advanced tools based on comments: Some ideas:

* Number of comments divided by submission points is a rough measure of contentious topics in the HN community, esp. if this is > 1. Recent example: https://news.ycombinator.com/item?id=19970544. Sort front page by this score. Another obvious measure would be the number of grayed out and flagged comments.

* Identify comments leading to discussions and highlight them. Since comment scores are hidden this has to be done trough analysis of replies (maybe keywords "good reply" or number of replies).

* Identify comments that are informational, simple measure would be to count the number of links; harder to do would be to analyze the "factfullness".

Just some random thoughts.

P.S. I offer myself as a cautionary example of getting too much into HN: This is the first site I skim on my phone when I get up (to get awake, you know) and I easily spend more than an hour every day (who am I kidding, more like two hours). This is idiotic! Don't let your idea debt (https://news.ycombinator.com/item?id=11027684) accumulate.

Sometimes I do not even read the article at all. Either I read a few comments, and if it seems interesting I go to read the article. And sometime I just read all the comments without even putting an eye on the article. So yes, I strongly agree. HN value is in the comments.

PS: I just notice I did it again here... replying to a comment without having seen the article. I know in general it is better to have read the article before commenting, for sure for top levels comments at least, but in this case my answer is independent of the original content.

I think a really useful interface / AI project would be to write something that categorizes all the comments being made to an article.

That way I could dive into specific comments on topics that interest me, and also decide which comment branches to follow.

I have no idea how this might be implemented though.

My process with HN articles is

1. Find a topic that looks interesting or important based on the title

2. Read the top two comments.

3a. If they are incredibly polarized, read the article.

3b. If they are in agreement, just read the comments.

I use an old FF version running greasemonkey just to browse HN. My GM script concats two pages of posts, filters out any posts that have less votes than x/hr since being posted, and highlights ones that have a list of keywords I'm interested in.

I love FF, but it's a constantly shifting target esp wrt greasemonkey! I gave up trying to convert my old scripts.