A List of Hacker News's Undocumented Features and Behaviors
github.com
github.com
Funnily enough, currently the "twos" subject is "Hacker News", with this article and "What I've Learned from Hacker News" both being present in the front page at the same time.
Number of all possible topics is N, number of news on FP is 30. For example if N is 365, you would expect 70% chance that two separate items on FP mention the same topic.
Although, HN recurrence probably a bigger contributor. When a person gets reminded of some previous highly upvoted link and re-submits it. The highly upvoted link gets upvoted again and now you have two items on FP with similar keywords.
Example: reddit.com/r/redditDayOf
The natural place for follow-up links is, of course, the original thread.
Used to be that if a shadowbanned user posted a clearly substantive comment a vouch would always (nearly always?) restore it. Now it never does - I guess vouch is now just a vote that needs x votes.
Suspect that's going to unintentionally have even more people talking to themselves, because they've not yet noticed the shadowban.
So have others noticed a change, or did I accidentally flag my own account somehow?
If you did, the mods won't necessarily tell you, you'll just have to email them and find out.
This has been something I’ve wanted to see in Reddit for a while now. I just recently learned that lobste.rs has this feature, and some other interesting stuff like a public mod-log.
I’d like to add a disclaimer that there may be very good reasons not to do this on Reddit or HN, of which I am not aware.
I used to believe that, honestly, really did. Not anymore, not after watching how many communities actually behave themselves with vote mechanisms. One of those moments where something "looks good on paper" but not as much when the rubber meets the road, I'm finding a very real personal dislike with approval signaling/seeking features in online discussion groups.
"Not all comments are equal"
I don't believe that either. Maybe it's just a matter of my perspectives changing after the last 17 years on the internet, which is fine. Everyone has a different experience with it.
This same fatigue with "approval signaling" on online commentary extends all the way to Facebook and Twitter, which one of the many reasons I'm no longer on either platform. Reddit is probably the next go go, for all sorts of reasons, if I were to list them, "approval signaling" will most undoubtedly be on the list, but probably further down comparatively.
But I should also clarify, I didn't mean to directly advocate a full out feature request and say "get rid of up and downvotes" right here and now on HN, I was kind of thinking through my keyboard there, I full well realize that feature very likely isn't going anywhere here.
It's the right policy, I think. The last thing we need are dozens more one-line "why this sucks" comments on each thread.
The reasons can be semi-hidden. They don't have be part of the default visible thread. A small hyperlink or icon could be clicked on to view them. One doesn't see them unless they want to.
(Rationally speaking, none of us will lie on our deathbeds wishing we'd collected more points in the internet popularity contest, but it's only human to change behaviour in response to external measures)
- One - Two - Three - Four
You can do this but it isn't clear that it is even a bulleted list (particularly with real content):
- One
- Two
- Three
- Four
It does support code blocks but that has strange boxing behavior making them hard to read even for actual code. For example:
- Super long line that will box strangely for some reason... ! Long Line Long Line.
- One
- Two
- Three
- Four
As you move further down a reply thread the box around code blocks shrinks and scrolling increases. If you access the comments with increased font size, a mobile device, or a smaller browser window it also gets worse/happens sooner.There's also now anti-hammering protection on the site (posting/replying/voting too fast) but it is far too easy to trigger. Just voting for a comment and hitting reply to that same comment can easily trigger it.
• one
• two
• three
• four
But which features are missing?
The markdown support is extremely minimal and doesn't support a lot of basics like quotes (which causes many people to "quote" by using a code block, which has issues).
A basic private messaging system would be really helpful in a lot of cases. Right now people always have to post email addresses publicly.
It would be nice to have some built-in visibility into things like title changes, link changes, etc. that mods do, instead of needing them to comment manually about everything they do.
They could implement some scraping to automatically add the year to older stories, instead of users needing to flag it and get the mods to do it manually all the time.
And so on.
> The markdown support is extremely minimal and doesn't support a lot of basics like quotes (which causes many people to "quote" by using a code block, which has issues).
Obvious enough.
An extra button: moderated, is with a comment.
Scraping is different for each website.
One reason we don't do a lot of visible changes is that HN's minimalism is one of its biggest assets. Another is that users don't like things to change, and the internal pressures that cause companies to do arbitrary redesigns are thankfully not in play here.
I wasn't trying to diminish the work being done, but I couldn't find a more recent official update announcement than the one I linked from 2016.
They can, but don't have to. There's no business need being served here like there would be with a startup getting feedback from customers, and HN isn't open sourced, so there's no direct way to do pull requests or report issues.
They could literally just publish changes and ignore people's opinions on the matter, that would be perfectly valid given their small staff and available commitment.
Also, if they wanted to keep feature requests and complaints off of HN proper, having a changelog or request thread might be an effective way to quarantine that content.
I assume there's been lots of work to make the site more efficient/reduce cost. "No [few] internal pressures" sounds like a great recipe for an interesting design.
And an API, practically by definition, is not a benefit to membership or a draw for whatever HN considers a typical user, much less the sort of "normie" non-technical user HN wants to keep away. The quality of comments is not maintained to a high standard by the API being more difficult to query than it could be, nor would it be diluted by the API being easier to use.
And as far as Reddit goes, a lot of content gets posted here from Reddit, and a lot of the userbase here are also Reddit users, making the cultural contempt HN has for Reddit more than a bit hypocritical.
A lot of people here, pg included, seem to assume that the "low tech" nature of the featureset here is keeping some flood of unwanted users away but I submit that the title of the site being "Hacker News", it being on a subdomain and the lack of evangelism on mainstream social media do a far better job of gatekeeping than purposely adding "friction" to frustrate people.
HN is for the intellectually curious. Many of the posts here are technical but there have always been plenty of non-technical discussions as well.
A lot of people here seem to believe the site is supposed to be for technical users and technical content onlym even though that never was the case. The negative impressions people seem to have about "turning into Reddit" or non-technical content diminishing the sites' quality feed into that stereotype, as does the belief that adding "features that reduce friction" would somehow dilute content.
But... the guidelines do use the qualifier "anything that good hackers would find interesting" for determining what's on topic, which does lead people to fill in the blanks for what a "good hacker" should find interesting.
Hacker News doesn't do a good job of avoiding duplicates anyway, might as well make that a feature.
No, don't delete comments! Hide them if you must, but I always like being able to enjoy and send people link to comments from years ago.
- Overhauled voting system. Comment votes are effectively "like/dislike" buttons, and some people use them to help promote/discourage comments central to the discussion. If you get downvoted or upvoted you do not know why, and that helps no one. Slashdot implemented a useful feature for this over 20 years ago.
- Tags on submissions, which would enable very simple filtering by topics.
- New hidden sections to let us dig into dead or controversial stories (maybe /dead and /deadnew?). If you watch /newest, you'll sometimes notice information that is interesting or meaningful, but it's auto-killed. There's no way to only show dead info except to page back through /newest. It would also be neat if we could actually vote on these, even if they remain dead. I don't even need comments to work on them, I'd just like to occasionally look at content that either a bot or overzealous human deems 'inappropriate'.
On the last one, here's an example: I just found this dead article in /newest (https://oilprice.com/Alternative-Energy/Renewable-Energy/The...) talking about the world's largest battery being built in Texas. It was submitted by protomyth, number 55 on the HN leaderboard. Of their past 30 submissions, 5 are dead, 1 has >100 votes, and 1 has >200 votes.
You can't easily find articles like this, because you have to keep paging back and paging back to find 30 or more dead articles. And you can't see the highest voted dead articles, so ones that might be interesting but were killed for some reason are also hard to find.
My particular bug/request: the story submission form will trim page title suffixes for some popular sites, but the form's "too long" JavaScript check measures the original title length, not the auto-trimmed title length. So I often submit a story and have to manually trim the title suffix that would have been trimmed anyways for a story from the same site with a shorter title.
Why on earth is there a karma requirement for this, let alone such a high one?
If you really NEED to change the color, you can simply load a custom CSS in your web browser.
I was disappointed that nothing unusual happened to my "past-life account" with 1000+ karma.
Didn't pg try something like this once?
created: July 21, 2017
karma: 250
You guys don't have to be mean…
I don’t know if there’s a limit to thread depth?
I was looking at the source code, and it seemed like collapsing recursively collapses all comments below it too?
Added it to the doc!
Or, you can select its text to see it in (typically) black on blue.
Anybody known if emailing the mods will let this happen?
An obvious one is brand new comments get a huge boost in the rankings compared to older comments. It is pretty common to see a few minute old comment at the top of a lot of comment threads. It also seems user reputation is a factor in this. A user with a good average karma score (a metric that used to be displayed publicly but is now hidden) seems to have their comments stay near the top longer than other users.
Perhaps such remarks are retaining position by their own merit. An experienced or high quality commentator may simply produce remarks that are better received.
Joking aside, dang specifically denies there is a comment ranking adjustment due to account karma, but not other factors (such as account age, an unseen reputation field, your IP address/choice of web browser/#{1/user.name.size}).
We are the blind to Hacker News's elephant in this.
One feature I've been wanting lately is the ability to see only new comments since my last visit to the thread. I've thought about writing some sort of reader application for this or perhaps a browser plugin.
Even nicer would be something that takes into account comments I've actually read by using what my viewport has had open, but that may be less useful.
implement some kind of keystroke scheme to to basic page steerage in emacs/vi mode respecting ways, and ? to show what the keystrokes are
So .. thank you kind HN reader.
I think using front languages which "compile" to JS is a fine foonly thing to do. ghc-js and elm come to mind.
Is that true? I may misremember, but I thought I had comments in the minus a dozen or so.
https://slatestarcodex.com/2018/10/30/sort-by-controversial/
When voting is applied with no real guidelines or standards, it ceases to be an effective filter for anything but the immediate emotional zeitgeist, and this goes against HN's stated purpose of bias towards quality.
I know people have strong feelings about downvotes, but those feelings are rooted in something other than discussion quality. Everything I think I know about how to keep HN from deteriorating too badly tells me that the downvotes here serve a critical purpose—even though they're a crude weapon with a lot of downside.
How do you know? People have different opinions about what constitutes and contributes to quality. From what I've seen, a lot of people assume Reddit's standard should apply here, and they seem surprised to learn that it doesn't. Those people do seem to be concerned about discussion quality.
>Everything I think I know about how to keep HN from deteriorating too badly tells me that the downvotes here serve a critical purpose—even though they're a crude and imprecise weapon, with a lot of downside.
Fine... but why is it wrong to have a standard for downvotes?
The argument being made here is that arbitrary and excessive downvotes are a symptom of deterioration.
Taking the last point first: HN downvotes aren't arbitrary and excessive. In most cases, not 99% but maybe 90 and certainly 80, it's easy to see why a comment has been downvoted—except when you agree with it on a topic that pushes your buttons, in which case you will always think the downvote was unfair, but then your opinion can't be trusted. (That applies to all us; we just have different buttons and agree with different things.)
That does leave a margin where the statistical vote cloud converges on a negative score unfairly. But how often does that really happen—maybe 10%? Once you account for the many factors of randomness, e.g. in who happens to see a comment, there's not much room left to make outcomes more precise. Certainly a feeble "rule change" wouldn't do it; if you think it would, try running an internet forum and telling users how to behave. You will quickly know how King Canute must have felt. We'd be better off hiring someone fair-minded to go through all the comments, find those 10%, and upvote them. But what a fate to subject a human being to.
> a symptom of deterioration
HN downvote behavior has been stable for a long time, so whatever's going on, I don't think it's deterioration. But if it is, then I go with what Voltaire said about coffee being a slow poison: it must be very slow.
> a lot of people assume Reddit's standard should apply here,
People assume Reddit's standards apply on HN because Reddit is so much bigger than HN, and therefore much better known even among HN users. This is a simple consequence of size. It has nothing to do with what Reddit's standards specifically are or how high its quality is. I respect Reddit—Reddit is an amazing achievement—but it is not where HN should be taking lessons in discussion quality.
> How do you know? People have different opinions about what constitutes and contributes to quality.
I don't know, but I'll tell you why I say it. The emotional dynamic in downvoting is very strong. It stings to get downvoted—it feels like you've been downvoted. It sucks for me as much as anybody. From observing this reaction in myself, and how people's statements about downvotes are connected to their feelings in thousands of cases, I believe that this emotional dynamic accounts for most of what people say on the topic. That's not a criticism; it's just how we are. But given that, it's easy to see how the common belief about downvotes arises: it's not that there was anything bad about my comment (impossible!)—it must rather be that some schmuck disagreed. Therefore, to make the world a better place, people shouldn't be allowed to downvote for disagreement. This is wishful thinking.
I can tell you for sure that, whatever beliefs we have about it, people overwhelmingly downvote based on how they feel about a comment, probably in the first 5 seconds. They're not following any "guidelines". Most don't even know what the guidelines are. It's just lizard-brain like/dislike. Suppose we changed the rules to ask users only to downvote under more refined conditions. Whose behavior would that modify? Not most people's—only that of the very most conscientious users. But those are precisely the users whose instincts should be trusted in the first place.
That is why I don't think we should set up such a rule: first, it's wishful thinking; second, no clear upside. And third, HN's origins are in a kind of counterintuitive minimalism that I think is worth something, and that it takes a certain stubbornness to preserve. Everyone disagrees with the specific acts of stubborn preservation, but somehow people end up liking, or like/hating, the sum it all adds up to.
Have you lot thought much about ways of reducing that, without changing the effect votes have on ordering? I subscribe to the 'emotional dynamic' theory of this and it seems like it's empirically testable without fiddling with the way the site much.
I'd link to it, but won't out of respect for dang. Look for it on my GitHub.
I can't help but notice that it's still missing `noprocrast` and related settings.