I'm going to pivot my answer: it's not the best thing I've found on HN, it's how I use HN to find things.
- Ask questions about things you're unsure of, even if it's just TLAs.[1] I also like to ask for references, and those can occasionally be real gems.
- Respond to others. Especially, respond in a "yes, and..." rather than a "no, but ..." style, where that's possible.
- Avoid tendentious arguments. I'll point out corrections, but try to keep those short where possible.[2] Changing minds is ... difficult at best. Leaving clues for other readers may have value.
- Skip the main page and hit the lists: <https://news.ycombinator.com/lists>. "Best", "Invited", and "Pool" especially are worth exploring.
- If you find someone interesting, check to see what they've submitted. Often individual's submissions are interesting curations of their own, though often these fail to survive the HN queue.
- Use search. I rely on HN somewhat to search terms or individuals of interest, and to look for commentary on articles I turn up to see if there's any illuminating relevant discussion (especially older articles). This isn't always successful, but it's virtually always worth the effort, particularly using DuckDuckGo's !bang search capability: "!hn <search terms>". A blank search with a date range serves as a "best of day/week/month/year" feature (see comment below).
- Try to read the article first, or at least early in your perusal of the comments thread. Comments ... often ... deviate from and/or are only very vaguely grounded in the article itself.
- There are a lot of people shooting from the hip. There are also some absolute domain experts and Internet legends who drop by. It's quite an eclectic crowd, though the gems may be well hidden.
- You can read an individual's comments and posts from their profile. Reading the dang's (HN's moderator) comments is a good way to familiarise yourself with HN's culture and norms: <https://news.ycombinator.com/threads?id=dang> You can also email mods (hn@ycombinator.com) with questions or concerns about the site and/or your account and activity.
- Karma and votes count for far less than you'd think, though they tend to filter out obvious crud pretty well.
If you're interested in things I've found interesting on HN, I've favourited far too many submissions. There might be some gems (and embarrassments) amongst those: <https://news.ycombinator.com/favorites?id=dredmorbius>
I've also been looking at HN's historical front-page activity and have a sense of the topics which are most covered (based on classifying the submitted site). The top 20 (+ "UNCLASSIFIED") of those represent 97.6% of all front-page posts, and are:
(Data through 21-6-2023, code is sort of an evolving situation, and the tallies above fix some errors in other recent similar posts, though overall magnitude shifts are fairly small.)
Edits: A few late tweaks and additions partially based on my follow-up comment below.
________________________________
Notes:
1. Three letter acronyms. See also ETLAs (extended TLAs) and DETLAs (double-extended ...).
2. There are ... many ... exceptions to this. It's a goal and aspiration, not an accomplishment.
One more trick: Use Algolia search with an unspecified term but a defined date range.
What you'll get back are the highest-voted stories for that period, sort of a best of day / week / month / year thing.
With some more fiddling you can search specific date ranges (Algolia's date-selector UI/UX is a bit fugly).
Note neither HN's voting nor its karma are especially good indicators overall quality (they tend to reflect popularity and/or posting/commenting frequency far more), but they do tend to filter out the really low-value stuff. Capturing and promoting true quality is far harder.
There are a lot of really interesting members with relatively low karma, often barely into three digits.
(I spend far too much time on the site, as I suspect most of the other leaderboard listees could claim.)
The question of what quality is is the pursuit of one HN perennial favourite, Zen and the Art of Motorcycle Maintenance, though that book's treatment might be slightly too enigmatic.
I've been narrowing in on notion of quality as, usually and/or generally, fitness to some purpose (which might simply be enjoyment or aesthetic response, though it's often functional). One corollary of this is that quality is dependent on the applied purpose, by which point, different assessments of quality based on different purposes will come up with different values.
Again, online voting and response / interactions metrics (votes, comments, flags, reshares where such features exist) are only a very thin view on overall quality. As the meme says, when you've got an airliner and an incapacitated pilot, one skilled pilot is worth far more than a million untrained passengers. Translating this to content moderation introduces other issues: are the skilled / trained experts operating within their inherently limited areas of expertise, are they free of any compromising influences or corruption? But in general the view that "the crowd is always right" does seem to hit real limits.
In practice, HN does put its thumb on the scale in all sorts of ways, some nuanced, some not. As of 2009 there was a list of well over 1,000 banned sites, there are permanent and temporary penalties put against sites and topics, member flags, moderator actions, banned accounts, and a few (mostly slight) privileges of karma (mostly at fairly low levels). A few accounts may benefit from name recognition or be crowd favourites (pg comes to mind).
- Ask questions about things you're unsure of, even if it's just TLAs.[1] I also like to ask for references, and those can occasionally be real gems.
- Respond to others. Especially, respond in a "yes, and..." rather than a "no, but ..." style, where that's possible.
- Avoid tendentious arguments. I'll point out corrections, but try to keep those short where possible.[2] Changing minds is ... difficult at best. Leaving clues for other readers may have value.
- Skip the main page and hit the lists: <https://news.ycombinator.com/lists>. "Best", "Invited", and "Pool" especially are worth exploring.
- If you find someone interesting, check to see what they've submitted. Often individual's submissions are interesting curations of their own, though often these fail to survive the HN queue.
- Use search. I rely on HN somewhat to search terms or individuals of interest, and to look for commentary on articles I turn up to see if there's any illuminating relevant discussion (especially older articles). This isn't always successful, but it's virtually always worth the effort, particularly using DuckDuckGo's !bang search capability: "!hn <search terms>". A blank search with a date range serves as a "best of day/week/month/year" feature (see comment below).
- Try to read the article first, or at least early in your perusal of the comments thread. Comments ... often ... deviate from and/or are only very vaguely grounded in the article itself.
- There are a lot of people shooting from the hip. There are also some absolute domain experts and Internet legends who drop by. It's quite an eclectic crowd, though the gems may be well hidden.
- You can read an individual's comments and posts from their profile. Reading the dang's (HN's moderator) comments is a good way to familiarise yourself with HN's culture and norms: <https://news.ycombinator.com/threads?id=dang> You can also email mods (hn@ycombinator.com) with questions or concerns about the site and/or your account and activity.
- Karma and votes count for far less than you'd think, though they tend to filter out obvious crud pretty well.
If you're interested in things I've found interesting on HN, I've favourited far too many submissions. There might be some gems (and embarrassments) amongst those: <https://news.ycombinator.com/favorites?id=dredmorbius>
I've also been looking at HN's historical front-page activity and have a sense of the topics which are most covered (based on classifying the submitted site). The top 20 (+ "UNCLASSIFIED") of those represent 97.6% of all front-page posts, and are:
(Data through 21-6-2023, code is sort of an evolving situation, and the tallies above fix some errors in other recent similar posts, though overall magnitude shifts are fairly small.)Edits: A few late tweaks and additions partially based on my follow-up comment below.
________________________________
Notes:
1. Three letter acronyms. See also ETLAs (extended TLAs) and DETLAs (double-extended ...).
2. There are ... many ... exceptions to this. It's a goal and aspiration, not an accomplishment.