HN: The Good Parts (2016)(danluu.com) |
HN: The Good Parts (2016)(danluu.com) |
> headache-inducing levels of information density
HN is one of the more information-dense sites of its kind - because the space that isn't wasted on "flashy features" and advertising is filled with content.
Information density is good and desirable. Bloated UIs full of ads and irrelevant distractions are dense with noise.
Comments, everywhere are terrible. Heck, this one probably is.
But I like HN more than anywhere else (so far). Maybe rot will set in, and the sun will set on our relationship, but not yet.
I find many comments to be quite correct. Not all, of course, but many.
The ones that I like, are the ones that make me think "orthogonally."
This would never go down on Twitter, Reddit etc. I think by virtue of being here, there's at least a semblance of something that unites us.
That being said, I've definitely seen a few situations where "logical fallacies" have been wheeled out to try and autowin an argument. However, this is usually the exception and when things get really bad the moderation is usually on point.
I do agree that the farther you get from computer tech, the more frequently you’ll see patently wrong comments (aviation is my eye-roll inducing topic here), but on computer tech topics, I’d say more than 50% of substantive comments have positive correlation with correctness and more than 30% are actively good.
You don't have to spend long on Reddit, Youtube or Facebook to really appreciate just how nice and non-combative it is here. Maybe it is just the relative difference from those sites' comment sections that blind me to how many assholes are regularly posting on HN, but I really don't see them.
Admittedly, as another comment pointed out, the further it gets away from tech, the less informed the comments tend to be (the recent thread about mystical experiences and psychedelics had the level of rhetoric and nuance I'd expect from a teenage Reddit capital-A Atheist on the subject, for example). But that's fine, it's not where I come here for.
Let me offer a counterpoint (I'm glad you didn't mention Twitter, because I'd have to really agree that Twitter is a Hate Machine. And I'll grant you YouTube as well, to a lesser degree): I use facebook mostly for its closed, special interest groups. For each of these groups, which are relatively civil -- until there's the occasional namecalling flamewar and some people get expelled -- the most common assertion is by far something of this kind: "why, this is the best group about $SUBJECT, people here are so nice and civil and I haven't found a group like this one anywhere on the net!". If you belong to a couple of groups on facebook, you'll see this repeated in every one of them, each of these groups oblivious to the existence of other, similar or possibly nicer groups, each selectively forgetting the most recent fight.
I think some degree of this phenomenon is at play here on HN, too.
* People fishing for upvotes by complaining about how X news site uses Javascript or has a paywall. We get it. I know. I agree. Stop posting it.
* Top comments being tangential responses to the title, not discussions of the linked content. I usually enjoy the tangential responses, but I want to see the more-relevant discussion voted higher. I can only imagine that these posts are being upvoted by people who also didn't read the article.
> On any topic I’m informed about, the vast majority of comments are pretty clearly wrong.
I'd be curious to see some examples, because this isn't really my experience. Once in a while, yes. But not that often and not the vast majority, and it mostly pops up in political-flavored threads. But I should defer to Luu, who is a lot more informed than I am on a lot of topics.
https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theore...
developing a good ranking function for bag-of-words full text search is about carefully balancing the attraction of larger vs smaller documents in the collection. If you try to aggregate several bad search engines you get a bad search engine, if you try to add more factors you get a bad search engine, instead you have to walk the path of BM25 or one of the more modern 'information theoretic' ranking functions. (e.g. precious knowledge I got from years of reading conference proceedings and still having no idea how to make a good ranking function then finally reading the right review paper that pointed out the two discoveries the conference made in the first 10 years!)
The shadow of Arrow however is that it is a lot of work to get a real improvement over a tfidf with random characteristics and you are limited by an asymptote which is less far away than you'd like: the rel teams at Google and Bing cannot beat the p=0.7@1 barrier for text search because 0.3 of the time they will get your intent wrong.
There are ranking functions in Lucene that can be tuned up as I described by making a test data set and using the methods developed at the TREC conference. That kind of tuning really works. At that conference you might see people fight over a point or two of AUC and I don't know if users can feel that but I am sure users can feel the 15 point and more difference we were seeing from tuning.
The odd thing is that hardly anybody does it and the knowledge seems pretty obscure. (e.g. I have talked to people at Lucene, OpenText and other full text search vendors and they are much more impressed with having 10,000 connectors than with the search results being good.)
Totally off-topic: One of my favorite from his writing is "Sampling v. tracing" [1].
When I started reading HN ten years ago it was a vastly different culture among users. Or maybe I just didn't recognize the underhanded asshole tones.
Either way, HN is still good enough compared to any alternatives I know about. I guess that's just life though.
Have considered this seriously (and still am), but a blog post needs more meta context and exposition than an HN comment. The person browsing a blog post vs. playing the HN slot machine (time bandit?) is in a more passive mode than the give and take of HN comments.
The difference to me is that a blog post is a topic of conversation, where HN is a conversation. It's discourse, facilitated by candid pseudonymity, and not just a foil for someone to pose with. (whereas twitter is just the instagram of ideas.)
I have written at the pro level where it really is like a sport, and a lot of my job involves articulating complex things with clarity and pith, and to me, as a form, the blog post is too passive. However, if I get good enough at that form my opinion may change.
It is certainly legit if someone buries a sarcastic or trivial comment of mine (which happens, though I don't do it often), but I don't bother to delete them -- downvoting them does that effectively anyway.
I'm not sure if there's any actual value to karma and I don't really know how it's assigned and wouldn't really notice if it weren't on the top bar of the page. I haven't tried to do any experiments though.
HN comments are underrated - https://news.ycombinator.com/item?id=12772925 - Oct 2016 (359 comments)
Or so I thought... till I read this Good Parts post and now I am left with this vague feeling I've been wrong all along. I wish he had said why some comments are so bad.
I now don't want to end up as the top comment because its always bad?
If you like stories that generate robust discussion, use: https://news.ycombinator.com/active
This 'hidden' view of HN tends to allow high-discussion topics to stay at the top, and not get insta-buried because they're controversial. I use it as my default now. In my opinion, it's much better than the main sorting algorithm.
The intended use of HN is to run across things that one isn't already familiar with, not to argue about the same small number of hot topics over and over. People who want to do that should find a site that wants them to do that.
I would rather like to see what Dan did scaled up 1000x or so: that is, have experts pick out a few of the best comments over the very long term with a number of quality criteria in mind.
Like it or not individual people have characteristics that are mostly stable over time: they are more or less smart, mean, wordy, honest, loyal, interested in different things, make different mistakes, etc. Profiling the author is worthwhile.
The article though is a problem not a solution in terms of sampling: consider that people talked about topic A in 200 articles that hit the front page in the last 5 years; if you put the comments from those 200 articles in pool B and rank the articles in pool B and pick up the top ten you would get the "pure gold".
This sounds ... tempting...
Appropriate that _this_ comment is buried who-knows-where in slashdot's archives.
Proceeds to list pages of awesome comments.
Claims topics he's informed about have mostly wrong comments.
Doesn't give any examples.
HN comments are a conversation/discussion not a contest of who is right/wrong, and it's great to have differing viewpoints as long as those views don't degrade into logical fallacies like ad hominem, ,anecdotal, or others.
There are way less logical fallacies on HN than other sites which is what I love about it.
I think some people just want an echo chamber.
If one cannot get over their contempt of bad internet comments then it might be better for their stress levels to stick to reading academic journals or other elite publications. Much higher ratios of nice/correct opinions there.
Can you expand? I naively thought that a site with markup as simple as HN's (zooming is flawless and it even works without JavaScript) would be very accessible to visually-impaired people.
I do know the tiny text and low contrast make it hard to use for someone like me with older eyes. This is especially true on mobile, where the text refuses to resize in accordance to my phone’s large font settings.
I'm not visually impaired, just in need of bifocals; I zoom HN to 150% and it works fine.
But I am kinda surprised when I turn the zoom off. Even 25 years ago that text would have looked small to me.
https://contrast-ratio.com/#%23dddddd-on-%23f6f6ef
1.25, which is well below the minimum of 4.5
I'm wondering how effective non-text would be? Personally, text is my preferred medium for any non-simple communication.
Marshall McLuhan was right to point out that "television" would erode the need for people to read and write.
What in particular would you wish to have improved?
If you ever see a HN front page story where you have inside information about some event, you can rest assured there will be maddening amounts of completely wrong speculation written as if it were gospel by people with no connection to that event. And then people attack the strawmen with abandon.
And for professional reasons, you grit your teeth, keep your mouth shut, and let it all slide. I'm not in PR.
"Someone is wrong on the Internet": https://xkcd.com/386/
I agree that HN is better than a _lot_ of places. But it's not all wine and roses.
Or maybe my internet spidey sense is so attuned that I've learned to subconsciously tune out B.S.
So I've never really seen that type of thing happening.
Does anyone really believe unverified information people post on the internet ?
> "[...] Having flamebait drop off the front page quickly is significant, but it doesn’t seem sufficient to explain why there are so many more well-informed comments on HN than on other forums with roughly similar traffic."
> "Maybe the answer is that people come to HN for the same reason people come to Silicon Valley -- despite all the downsides, there’s a relatively large concentration of experts there across a wide variety of CS-related disciplines. If that’s true, and it’s a combination of path dependence on network effects, that’s pretty depressing since that’s not replicable."
So he does consider HN successful, he is merely pointing out that the gems are infrequent (though more frequent than on, say, Reddit), and that most comments are uninformed or wrong.
If you read anything from danluu's site, you'll see he is not really an "echo chamber" kind of guy.
PS: I've seen plenty of illogical, flat-out wrong and anecdotical comments on HN to know they are the norm rather than the exception. This doesn't preclude the existence of real gems, like danluu points out.
HN comments are horrible...except here's a thousand line blog about why they're not horrible.
Judging by the content of the entire rest of the post...a better opener should have been "Hacker news comments are mostly awesome and here's a thousand line blog post about it.... but they're not all perfect".
What a bizzare article.
"Curious why he would need to move to a more prestigious position? Most people realize by their 30s that prestige is a sucker's game; it's a way of inducing people to do things that aren't much fun and they wouldn't really want to do on their own, by lauding them with accolades from people they don't really care about."
Not to worry. I do it fine, by accident...
A counter example is the old DMOZ directory which was a continuation of the tree-style Yahoo web directory.
That was subdivided into sections that were managed by "experts" (admins appointed to the sections) but the admins were lazy, corrupt or both. Even if they had been doing their jobs they would have been overwhelmed by hardly relevant "content marketing" submissions for web pages that are "200 OK", hot keywords, and a scam.
You have to draw from a spam-free pool.
For instance there is a data dump for stackoverflow
https://www.brentozar.com/archive/2015/10/how-to-download-th...
personally i think stackoverflow is a junk web site full of more wrong answers than right answers and wouldn't it be nice to pick just the best answer to the question and not see the code sample at the top that didn't work that the original poster was asking about?
If you are a real programmer, I mean you are shipping, you write a unit test, if it doesn't work they are gonna send it back to you, you may learn the hard way that answer #1 is not right (the compiler said so!) and scroll through a lot of half-baked discussion before you find that answer #7 passed the unit test, the acid test, all the other tests -- you are an expert and your opinion is worth much more than the "crowd".
If "Road Scholars" like us were getting our experience fed back into Stack Overflow how different would it look?
Yup, that's an issue for sure.
https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...
https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...
https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...
OTOH the simplicity of the site allows for very easy browser-based scaling, which means I only really think about how bad the text size is once per machine.
Maybe it's the best option for some people. Not for me, but I do love an audible book.
In conclusion, I think these other mediums simply lack the iterations that we've put into text over the centuries, leaving text as the clear winner in my book.
It's also great for technique with tools in the kitchen—not learning recipes so much, but learning the right way to process a certain vegetable really fast, or exactly what a proper whipping motion for eggs ought to look like, or how a pro moves the pan on and off the flame while cooking this dish, exactly what it looks like when they deem a certain cooking technique "done", all of that.
We want the community to grow—it would be concerning if it didn't. But the engagement tactics you mention aren't necessary for the kind of growth we want. In fact we consciously avoid them. Certain kinds of engagement—probably just the sort that engagement tactics would juice—would harm what we're trying to optimize HN for (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...).
The interesting aspect of this, I think, is that HN doesn't need to pursue growth at all costs, the way that a startup would. It's not a startup, nor a business per se, but it's not noncommercial either—it's funded by a business that understands that it's more valuable with HN than it would be without it, and is smart enough not to try to squeeze profit out of it beyond that. YC's business interests are in having HN be as good as possible, not as big as possible. That's odd, and oddly satisfying. It seems to be a historical accident that HN ended up in that sweet spot. https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...
It's natural for users to engage less and less over time. However, holding the community to a certain standard keeps users from outgrowing the submissions and discussions. No one is too old, too mature, or too "smart" for earnest discussion.
The mandate of the site is to be gratify intellectual curiosity (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...). Any website achieving such a thing will naturally grow.
> One important thing to realize is that we're trying to optimize HN for just one thing, which is curiosity
I’ve found it easier to practice creativity, for example, if I impose some boundaries.