Hacker News new | past | comments | ask | show | jobs | submit login
The frontpage with a threshold of 100 points (ycombinator.com)
256 points by pg on Dec 12, 2010 | hide | past | web | favorite | 110 comments

Not that this type of experiment is bad, but why is the UI so completely neglected? Is this a matter of thinking it's not important, because it's not fun work, or because it can all be customized with grease monkeys? Or what?

I think the quality of stories on HN is pretty good. What I'd much prefer to see any amount of attention given to is things like:

1) Let me take back a vote. Particularly on mobile, I misclick those tiny arrows a lot.

2) Let me comment inline. Having other comments, besides the parent around for examination while commenting would probably help overall comment quality.

3) Fix the mobile interface. It's impossible to use HN on mobile iPhone and Android handhelds. Making a slightly modified styled sheet with JavaScript is trivial.

4) Support someone making an app, or commission one. As far as I can tell, the current HN apps are all kind of buggy and have little updating. There are lots of young neophytes who'd love to work on this, particularly if sanctioned, and a funded effort would lead to a better product.

These basic UI improvements don't even seem to be on the radar. Also, I'll add, the story-killing on this site is pretty heavy-handed, yet capricious. Same with the title-editing. What constitutes a "hacker-centric" story changes with the mood of the moderators, and the tendency to just change each title to the original headline is misguided. I also think that the special privileges given to YC companies corrupts the whole system.

If there were any other community like this, people would be driven away by the neglected UI and the Star Chamber that governs the content. But there isn't, so there's no pressure to do anything but midnight HN science experiments. I should just pray for some competition I guess...

why is the UI so completely neglected?

Because when I spend time on HN my top priority is features that will make the content better. I believe that matches the priorities of the users-- that users would rather use a site with good stories and comments and a primitive UI than one with a slick UI and worse stories and comments. And time is a zero-sum game. Spending more time on UI = spending less on quality.

The focus on content quality above all is the reason you find yourself saying later "If there were any other community like this..."

You're simply wrong about the moderation. Nearly every story that gets killed is either autokilled, or a dup, or flagged to death by users. I would guess moderators manually kill less than 10 non-dup stories a day. You're also wrong that YC cos get special privileges.

1) I could accept I'm wrong about stories being killed by humans, but I know the titles get edited by humans. I remember one story I posted about some really ancient, accurate maps, and I made the title something like "Accurate maps from the 1500s unearthed." And then some moderator made it the story headline instead "Named building in DC hosts map show" or something equally uninformative.

2) I'm positive YC companies get at least one special privilege - the ability to post jobs.

3) I've been told YC accounts are excluded from anti-spam measures, but I can't prove it.

EDIT: Also, regarding the zero-sum-game argument. All of us developers know we have limited time to spend on things. But a HUGE change to a non-priority that takes almost no time (for example improving the CSS on mobile) should be squeezed in between these experiments that may or may not make the content better, and take hours.

I like the UI. I know it's hard for some to believe, but it's stark and functional and fast and mostly stays out of the way; it works. The comment scores need to be hidden for comments you didn't write; other than that, what's to improve? Taking back votes? How about, just forget about the vote.

If HN has a problem with titles, it's with not enough stores having their titles rewritten to the article original.

The comment scores need to be hidden for comments you didn't write

People keep saying this, but it would severely limit the utility of the site. When I chance upon a post with more than ~50 comments, I only want to read the 6-7 or so that the community has agreed are best. I perceive a positive correlation between comment scores and quality. That is, the odds that a post with 50 points contains good content do empirically seem to be higher than for posts with 2 points.

In short: I need a comment filter that is more strict than than simply "appears on Hacker News". I don't want to read every single one.

Comments are sorted based in part on votes. That's fine. The system we have now isn't picking out the best comments; it's picking out my comments.

I like using votes to determine whether or not it's worth devling into a deep thread. If there's a conversation that goes 5 levels deep, you can just scan the bottom layers - if they're getting votes, then it's likely to be a worthwhile conversation.

Similarly, if there's a bunch of responses with a score of 1 with replies forking off all over the place, having a visible score lets you catch that at-a-glance and avoid bothering with a conversation that is clearly digressing.

There's a difference?

I agree. I love the spartan UI. That said, reading HN sucks on mobile and this would only take a few minutes to fix without compromising the simplicity of the main UX.

ihackernews.com is most helpful for me on the mobile platform

>The comment scores need to be hidden for comments you didn't write

That was in the REPL for a while and I thought it was just annoying and confusing.

How it looks is fine, I suppose. But it could surely use in-line replies.

I assume you're talking about this:


I don't know what title you used when submitting it, but the current title wasn't written by a moderator. It's the original article title.

It's true that YC founders' accounts have a field set saying so, but the only code that looks at this field is the code for posting jobs. None of the anti-abuse code does. And the jobs page has been around for years.

Yes, as I said in the parent comment, the moderator changed it back to the original story headline.

I had changed it because I knew nobody cares about the Library of Congress, and no one knows what Portolan means. I read the story, determined a meaningful title, and the moderator decided that was no good. Since I have edited many, many newspapers, and written countless headlines, I felt particularly annoyed.

It's worth distinguishing between (1) "The moderators retitled such-and-such an article, and it wasn't an improvement" and (2) "The moderators retitle articles, and that's bad". #1 might be true in this case (I see greendestiny doesn't think so; I'm on the fence) but it's very weak evidence for #2. For what it's worth (and I know this is also weak evidence) when I've looked at an article and thought "Wow, HN should have used a different title for this" it's much much much more often been because of editorializing by the submitter than because the original article title was used but wasn't very good.

We need moderators aggressively retitling articles submitted with stupid or biased titles. (Not only because that fixes those titles, but also because it reduces the incentive for submitters to submit with bad titles.) And if we have that, then inevitably there will be occasional misjudged retitlings. The question is whether the benefit outweighs the cost, and in my mind there's very little doubt that it does.

I find mods retitling links to be a strongly positive feature of HN. It is one of those things thats very touchy though, but it'd be a shame if people's territoriality meant we had to accept wrong and misleading titles.

For what its worth, whatever was wrong with the original title I feel like yours was pretty misleading after going and reading the article (great article by the way, thanks for sharing it). The article was very explicitly about the library congress meeting to talk about the origins of the map, rather than a general piece about the map. If thats all you found interesting, perhaps you should have linked to a wikipedia page or a more general article.

Ironically, time spent arguing that "it's better to focus on the content rather than the UI" could have been spent to improve the UI.

Or the content.

Time spent arguing helps define which should get the free-time, though. Otherwise, ultimately, it would be "best" if every developer ignored every bit of feedback their applications caused, because their time is best spent making changes, not debating over what to change.

It was more of a joke than a comment to take seriously.

Furthermore, from a selfish point of view, I like better when PG works on HN concept as he indirectly work on Arc.

"Spending more time on UI = spending less on quality."

I disagree.

The inability to reverse incorrect votes reduces the quality of the site by assigning incorrect scores to comments. In fact, many comments are apologies for accidental down-votes.

Also, not being able to comment inline means thread context is lost when comments are being composed. It is very likely this results in lower quality comments, which reduces the quality of the site.

A good UI can and does lead to a higher quality sites. Keep things as simple as possible, but not so simple that quality suffers.

Just my opinion.

At least some types of UI changes would affect content quality. For example changing the size and color of the up and down arrows could have a big impact on voting frequency. Making the submit link more or less prominent could have a big impact on submission frequency, as well as providing tools like submission bookmarklets. And if at some point the majority of HN usage is mobile, it'll really change the sort of stories that people upvote.

It's not clear which direction you would want those metrics to move to improve quality, but it at least seems like UI vs quality is not a zero-sum tradeoff.

But would increasing voting frequency make the content better? It seems to me equally likely that it would make it worse. Ditto for having a more prominent submit link.

There is a submission bookmarklet. There's a link to it at the bottom of the page.

Yes, it seems equally likely that it would make things worse. But it seems unlikely that it would have no effect. It is unlikely that these are the optimal shapes of the up and down buttons.

For what it's worth, I think the Google way to solve this would be to figure out which metrics you can associate with quality and A/B test the impact of the UI changes. Setting up an A/B framework might be more work than you want to put in. But there might be a UI that encourages higher quality that you could discover empirically.

Good idea. You should download the Arc source for Hacker News and patch it to allow such UI testing for the site.

A quick UI fix would make HN readible on mobile devices. The fixed width of the page is caused by the navigation menu. If the navigation menu was made wrappable (white-space:auto) then the page would more easily fit right on almost any size screen. On mobile devices you can also eliminate unnecessary borders and padding. Very simple fixes like this would go a long way.

On Android, turn on "auto-fit pages" in settings. This isn't ideal, as there's still some left-to-right scrolling to focus on individual comments as comments become nested, and fixed width text doesn't wrap at all (which is thankfully relatively rare), but it's a vast improvement, as each comment fits horizontally.

This is a great idea. Why don't you patch the source for the site and make those changes available for use by the site.

(Let alone the fact that I'm busy enough with my own projects) There's no link anywhere on the site to the source code, or information about how to contribute.

If only you had some connections with web designers.

Looking at the HTML, the misclicking of up/down arrows on small screens could be fixed just by adding a couple more <br> tags between them.

The misclicking is mostly due to the fact that the targets are square images with extra padding, combined with sloppy nested tables and piles of inline elements that freak out further in mobile webkit.

If you're going to use tables for layout, use them consistently all the way down — don't just decide to party like it's 1994 and throw in some line breaks. Clean up the padding to rationalize the hit targets (could even use characters like ∆ and ∇ to make them non-square).

What's wrong with line breaks? What's sloppy about nested tables?

It's the easiest solution because pg can't be bothered messing about with "proper" layouts and I can understand why.

It seems to me that the almost plain UI is a choice that goes beyond the interesting "zero-sum game" that you mention. If I were you I will be defending the HN UI as a choice and not as a necessity because of your limited time. Would you shoot for a richer UI if you had all the time of the world?

I like the "Spartan" UI here. Also, I believe it serves to keep the signal to noise ratio high, in a way, by filtering out a set of potential readers/voters that would not add value in the aggregate.

That said, I wish there were 2 enhancements — [1] a search bar|function that would list matches in date order and [2] a little markup to make mobile/tablet viewing easier on the eyes — a chore that could be as simple as adding a meta viewport line in the HTML (would accommodate iOS devices).

Hi, could you please delete my account (plus comments if possible)? I saw elsewhere you were the person to ask. It would be much appreciated.

I would even argue that having a bad UI is a feature of HN in the sense that it is a deterrent to trolls and other such Internet creatures.

Trolls need an audience to survive. Does HN superficially look like a popular website? No. It lives on a subdomain and the official way to get here is by clicking a tiny footer link on YC's website. And let's face it, the design is pretty ugly too by many standards.

To realize there's a vibrant community here, you have to get passed those "deterrents". This selects for individuals who are truly concerned about great content and intelligent discussion rather than great design or sensationalism which leaves the trolls at the door.

> It's impossible to use HN on mobile iPhone and Android handhelds.

I actually find it quite the opposite. I've downloaded multiple Android apps that try to improve the HN reading interface, and I still find myself using browser instead. The closest thing that I would describe as being "bad" on my phone is that the arrows are small enough that I sometimes misvote on comments. Otherwise, I think it works really well:

* Each comment is the width of the screen, so there's not much vertical scrolling needed

* Comments are still well-indented so that I can easily follow a thread by scrolling right or left.

* I greatly enjoy being able to long-press comment links from the main page to open them in a new window, and then long-press the OP link to open that in a second window, and be able to use the back button to return to comments when I'm done reading, or even switch back and forth like I can in Chrome at home.

The fact that the text is essentially in a 4pt font doesn't bother you? Or is that an iPhone-only misfeature? Because on my iPhone 3GS, the page is wide enough that I either have to have unreadably small text or I have to zoom in and repeatedly pan from left to right in order to read a comment.

That sounds like an iPhone-only misfeature. On Android, you can turn on "auto-fit pages" and your chosen font size is locked and the width of elements is horizontally limited with no panning left-to-right to read a single comment (unless they contain unwrappable pre-text, which I find the hacker news text formatter to be kind of bogus on (indenting always makes preformatted text))

I've fixed a few of these in the iPhone app I built. It's a native Cocoa app that lets you browse, vote, and comment on HN. Here's what the home screen looks like: http://michaelgrinich.com/hackernews/media/screenshot-1.png

I separated the voting arrows and reply button, which helps with accidental taps. Here's a comments page: http://dl.dropbox.com/u/4595/hn-vote.png

There's also a few nice features, like an in-app browser with Readability and send to Instapaper/Twitter/etc. You can see a demo of the app here: http://michaelgrinich.com/hackernews/

The app doesn't sell enough to be a full-time project, so I push updates when I have time. Currently I'm working on an update with full iPad support.

(All of this works via HTML scraping since there's no API.)

I used this before I switched to an Android phone, highly recommend it.

The simplicity of the design is one of the main things I like about this site. It's like the things I make for myself: a handy tool.

>Fix the mobile interface. It's impossible to use HN on mobile iPhone and Android handhelds.

That's funny; the default HN interface is very much like the mobile interface of other sites, i.e. everything extraneous has been removed.

Otherwise, +1 for the ability to change a vote. I can't for the life of me understand the reasoning behind the current setup.

I can't see it having been mentioned anywhere else, but HNdroid on Android is fantastic. It'd be nice if the site itself had a decent mobile stylesheet, but in the absence of one, HNdroid does a great job.

A GlimmerBlocker filter which improves the layout for mobile: http://glimmerblocker.org/site/filters/mobile-optimized.xml

Pastebin version. css: http://pastebin.com/AE1UzYvU js transform of html source: http://pastebin.com/9069JULs

Most of the improvements could be made by a seperate mobile stylesheet if more classNames were added to the generated html.

Someone posted this awhile back and it works great on the iPhone: http://ihackernews.com/

It should be pointed out here that 1) the code base for the site is available for download and 2) you can not only inspect it but set up your own 'competing website' with a better UI.

I propose a contest to redesign HN, no code just visuals.

The winner gets, umm, well, instant recognition as the 'best designer of the wolf pack', or BDWP in short.

One more feature request: upvote comments saves comments.

Any chance of a ranged version, like "over 5, but under 200"? For some reason, while upvoted stories are mostly good, the most upvoted stories less often interest me, since they seem to be disproportionately about politics, business, self-help/motivation, or some sort of drama. This list is almost a perfect list of stories I'd like to filter out: http://news.ycombinator.com/best (ok, there's some good stuff in there too, but it's not bad as a heuristic).

Someone had a site (I can't remember the details but surely someone will chime in) with custom HN ranking of points divided by number of comments. It was surprisingly effective, though that might have just been the nature of the top stories at the time I checked it. If it did work it would be by also filtering out political and other call-to-action type posts particularly.

I'm with you, I hate the drama posts. I have a kind of half-arsed theory that what I dislike in the evolution of social sites is the inherent community building. I think it tends to get too self-referential.

plus stuff with 500 points stays on the list longer, can get too stale

No, articles get deprecated quite soon anyway. Try substituting 100 for 500 in the URL: you only get 1 article currently.

You're wrong and that test is misleading.


I'm seeing 380 points, 11 days old is ahead of 205 points and 4 days old.

While this is a really, really tempting front page, won't this just encourage groupthink? HN doesn't suffer from this as much it might, but it has its share. In addition, interesting stories languish unnoticed on /newest all the time, and if enough people switch to an >100 front page, articles with just 5-10 points have an even smaller chance of being noticed.

pg, I'm sure you considered this problem; could you talk a bit about your thinking behind this filter?

could you talk a bit about your thinking behind this filter?

"A lot of people have asked for it. It would be easy to write. Let's try it and see what happens."

Do you have some sort of pre-existing test for whether or not a feature is good at the time you implement it. Some way of knowing if it's good.

This has probably been suggested before, but can we just get rid of the points display (story, comment, & user) altogether? The automated sorting on this data has always been enough for me, and I can't think of a reason why any of us needs to know the exact point total of any of these things (except maybe karma).

I suppose there are some out there that would like to know an individual user's karma as a quick indicator of worthiness for X. Aside from this I'd be thrilled to see the number games and measuring sticks go away.

This doesn't really work, because things with high points can be anywhere on the page after they've stopped getting votes -- if you look at the front page, the stories aren't in point order by any means.

Although this would be bad for the community (people need to see the low-point articles for them to get voted on), you could have a page that sorts all stories by number of points, and allows you to remove them either by clicking on them, or marking them as uninteresting. It would solve the 'problem' of missing articles when I'm away at least.

I guess you could have some kind of mix of new and highly voted articles, which could be interesting?

That would be an interesting experiment in preventing the "snowball effect" of upvoting things that a lot of other people have upvoted. Enabling this on the front page, but leaving it on the thread itself would be an interesting experiment.

Of course, if many people used that threshold, almost no story would ever reach 100 points in the first place.

That's why (for now at least) I'm not making it too easy to. It would be trivial to add a threshold field in the profile and use that as a cutoff when generating the frontpage for each user, but since I'm not sure I want encourage this, I haven't yet.

The whole thing is just an experiment.

Maybe there should be a random one or two links from the newest page at the top (or the right) of the normal front page?

This, to me, seems to be the dilemma. The more features like this there are, filtering the input so we only see the "good stuff", the fewer eyes there are on the new submissions, and so the greater the chance that really some excellent submissions will never be seen.

More and more we're suggesting that people look at the "newest" page instead of just the "news" page to try to make sure that good submissions get their chance, and bad submissions get flagged, and yet here's another facility to take people away from the "newest" page.

Don't get me wrong, I think this is a great facility, and I'm pretty sure I'll be using it on occasion, but I'll also hang out on the "newest" page, because that's where the genuinely interesting stuff appears, often getting no upvotes, because it was never seen.

Maybe "front-page-with-100-points" is the sort of place newbies should be encouraged to check, and mainly more established users should be encouraged to check "newest".

Doesn't seem precise...

http://news.ycombinator.com/over?points=489 (489) doesn't see a thread with 490 points, but http://news.ycombinator.com/over?points=484 (484) does. Caching? Rounding? Thread in question: http://news.ycombinator.com/item?id=1990498

Still, awesome, many thanks! Especially because it drags up a bunch of good-but-older entries that I may have missed.

Oops, that was a bug; fixed.

See, static typing won't help you there!

I totally got a "where is your god now?!"[1] vibe from this.

[1]: http://icanhascheezburger.com/2007/11/24/where-is-your-dog-n...

Yup, works now. Thanks again!

I'd like an option alongside noprocrast and showdead:

* hidekarma

For your view only, it hides the karma in the top right, and hides the number of points next to each article and comment.

I, personally, find karma to be a distraction...I'm not afraid to admit that I subconsciously check my karma score every time I log in, and very occasionally catch myself "karma whoring"...that is, writing comments or submitting articles in a way that will improve my karma, instead of concentrating on writing something intelligent (yes, they should be the same, but they're not).

Getting no results from over doesn't mean there haven't been any items with that many points (though in fact there haven't been any with 9000 points). Over doesn't pull stuff off disk. It just applies a stricter filter to the pool of recent stories from which the frontpage is generated.

I assume swah wasn't expecting any results, it is most likely a reference to http://knowyourmeme.com/memes/its-over-9000 .

Finally, a hackernews I can keep up with.


Only thing there is the Heroku acquisition

Twitter feed with a threshold of 100 points: http://twitter.com/newsyc100

Twitter feed is nice, but the RSS feed is a lot more useful IMO ( http://feeds.feedburner.com/newsyc100 ).

When I first saw your post ( http://jeffmiller.github.com/2010/07/23/a-cure-for-hacker-ne... ), I missed the RSS feed links, and started looking for ways to convert a twitter feed to an RSS feed (I use Google Reader)... But then I saw the light :)

By the way, thanks a lot for making this, Jeff. I've been using your feeds almost daily since August. I used to spend a lot of time scanning the /best page for unread articles. No more!

Hopefully quant.ly will some day need its own threshold.

Is it just me, or have the front page point totals exploded in the last few months?

pg — care to share any traffic data?

We now get about 80k unique visitors on weekdays. There hasn't been any sudden spike.

I've also noticed this. Seemed to be between September-November to me. Links fly off of the first page of /new a lot quicker than before and I've seen a lot of good ones get no further than 2 or 3 points - it's a crapshoot when you submit something now. I've also had a lot more people contact me with "check out this post" type stuff.

One of the purposes of Previous Look is to capture the fast changing content of a webpage like /new so that past content can be reviewed quickly. For now http://www.previouslook.com/hnews/new has an 15-minute interval snapshot window, which seems to be enough to not miss anything.

I like it, though I might stick to the version I developed two months ago - http://hackerslide.com/ - as I prefer HN's natural colors ;-)

That looks cool! Good work.

This would be a great addition to the nav. I can imagine having 3 modes of viewing submissions: "top" (the current default), "new", and "popular" (for >100 votes). I already switch back and forth between the default (or "top") view and the "new" view. This allows me to get a good mix of seeing what the HN algorithm thinks is interesting, as well as what interesting stuff the community is discovering out there on the web. I can see this being well augmented by a third view that presents what's popular to the HN community.

Of course, adding another nav element gets tricky. How many nav links do you need before you start culling or redesigning? FWIW, it seems that grouping nav into 2 sections might be useful. One section would be focused on sorting the stream (top, new, popular). The other section would be focused on filtering the stream (threads, comments, ask, jobs). "Submit" is more of an action than a filter or sort, and might be better positioned as a control outside of the nav.

It just occurred to me that while I vote up great comments, I don't remember the last time I gave an up-vote to a thread. Am I atypical?

I doubt it. I think most of us are deliberately conservative to offset the lack of downvotes.

Isn't that the opposite to what you should do? If you are interested in curbing 'bad' stories you should probably vote for more stories, effectively 'downvoting' what you don't vote for.

Not if you think that a lot of crap gets voted up and truly good stories are few and far between.

Can we get filters like this for the official RSS feed?

I agree, this would be very powerful coupled with something like feedmyinbox.

Twitter and RSS feeds for HN stories over 20, 50, 100 and 150 points:


EDIT: Why the downvotes?

I think this is great as an option where we can fill in our own threshold. It's another way to play with the feed. I prefer the "newest" setting myself. It's incredibly quick to scan the headlines to see what I'm in the mood for that day. Or I can search with "news.ycombinator.com: searchterm" if I'm looking some particular subject matter. But sometimes it's fun to visit the lists too. news.ycombinator.com/lists Perhaps a couple of those options should go on the top menu.

Maybe the top menu should be Hacker News new 100+ searchterm best active bestcomments etc.

Upvoted just 'cause it was at 99 points. [Added: I think the temptation to do this shows a (minor and obvious) misincentive, like we used to see on IRC when I published weekly stats on the chat in #C.]

I'm seeing duplicate entries, for example on:


The last 4 entries are duplicated.

Curiously I see an example of duplicates with slightly different points. Maybe something is not enclosed properly in a transaction and the list is being generated partly before and partly after an update has been applied.

On the 100 points version I see:

27. The only script in your head (headjs.com) 248 points by timf 12 days ago | comments

28. The only script in your head (headjs.com) 247 points by timf 12 days ago | comments

Great, as I was JUST thinking of making a chrome extension to filter posts with 100+ points.

I've been thinking along those lines as well, but it's kinda hairy to do so... you'd have to load a few (or many, depending on the limit) pages, and inject their results into the viewed-page.

Seems like more bandwidth than it's worth. This is (obviously) a better solution though.

Even better I changed it to 200 points and found some great gems I had missed.

Good but if everyone starts to filter by 100+ then how will good new submissions make it to the top? We need people looking at new submissions and upvoting good ones all the time!

Is there any link to this new feature instead of manually typing this or saving this to a bookmark (or any other ways besides a proper hyper-link)?

Just a thought that occurred to me. Why does there have to be 2 levels (new and frontpage)? Why not 3 levels. New, Frontpage, "Best" or something.

Something like this - http://news.ycombinator.com/best ?

the choice of 100pts is pretty good though it's spread over 2 pages. I tried a quick graph hack to visualise, "Number of points required to fill 1 page of stories (30)?" ~ http://www.flickr.com/photos/bootload/5256341554/

I'd prefer under?points=x

The very highly upvoted articles seem more likely to be trendy and/or sensational.

the tablet article by pg twice is listed twice here. bug? http://news.ycombinator.com/over?points=300

any chance with making it dynamic? I tried changing the number from 100 to 10 and it didn't work. Went back to 100.

Wow, looks a lot more interesting.

This post itself has almost 100 points. Soon it will be on the 100-point threshold page!

Registration is open for Startup School 2019. Classes start July 22nd.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact