Regarding the "worse is better" discussion: At least its definitely better accessibility-wise. HN is about the last well-known site that allows interacting with it, including writing comments, with plain old Lynx. I am well aware that most web devs do not care anymore these days, and they have their reasons for sure. However, its still nice to see sites that refuse to go for SPA. It makes them so much more useable for people like me (blind). A big THANK YOU to the site maintainers, its one of the last corners of the net where interesting stuff happens which is still accessible.
Accessible UX results in good UX. I use a modern browser and appreciate how reliably Hacker News works. It's a great example of less is more when it comes to UX.
Well, I think proper font scaling would do wonders for making HN more accessible. As things stand, I have to zoom to 120% to read the text. I recall WebKit making a special case in its font rendering logic just for HN.
I can recommend the browser extension "modern for hacker news". Some features are premium (no subscription though) but only the defaults gives SO MUCH ux like font size, spacing between lines, text width.
Another great accesibility feat is the APIs for this web. I use upvoterate listing from quality news as front page.
To sum it up, opening the tab previews under Zen Browser made my experience feel like the coolest SPA, haha.
Hacker News is a perfect example of the "Worse is better" mantra applied to social engineering. I mean, slashdot had more features and functionality in the late 1990s.
What makes HN work is the tight focus and heavy moderation.
For context, "worse is better" refers to Gabriel's observation that products with simple implementations and complicated interfaces tend to achieve adoption faster than products with complex implementations and elegant interfaces.
One of the original motivating examples were Unix-like systems (simple implementation, few correctness guarantees in interfaces) vs. Lisp-based systems (often well specified interfaces, but with complicated implementations as the cost.)
"Incidentally, very few people grasp the amount of effort Daniel Gackle expends running HN now, and what an amazing job he does." -Paul Graham, https://x.com/paulg/status/1282055086433284103
HN may have less features, but do we even need them? I do not think it makes it worse because of that. You could call it minimalistic, which puts it into a more positive light. :)
Edit: or as someone else who has phrased it better: "less is more".
I liked the "friends" and "foes" system that Slashdot had, though I would say generally the "foes" here just get banned which is convenient.
I also thought Slashdot's moderation system was kind of fun. I am not sure it was useful but I enjoyed the annotations (+5 Funny when serious, +5 Insightful when inciteful, etc.) Meta-moderation was also neat?
Dark mode. Sure, Dark Reader exists but many mobile browsers don't support it.
Annoyingly enough it's been talked about for years but it never gets implemented, despite only three colors really needing a swap: background to dark sepia or just dark gray, and text to white and off-white.
I'd say that's the main thing. People hate ads, HN uses unobtrusive text ads. The moderation isn't that a competitive advantage, IMO. Slashdot's was better, mostly because it had measures to stop moderation abuse whereas HN seemingly doesn't. It's just a plain old up/down system with the added filip of a "super down" button, for those who are really committed to banning their opponents. I read with showdead turned on because perfectly reasonable comments are so often greyed out or dead. That used to happen much less on Slashdot because there were far fewer people with moderation rights and the bad ones got filtered out via metamod.
Maybe now it's been ported to Common Lisp it'll be easier to add features.
I'm not sure what you mean? The literal quote from the Wikipedia article on "worse is better" is:
> It refers to the argument that software quality does not necessarily increase with functionality: that there is a point where less functionality ("worse") is a preferable option ("better") in terms of practicality and usability.
For that reason, I think I am applying the term precisely as it was defined.
The irony of my comment, which dang picked up, is that the original idea was a criticism against Lisp, suggesting that the bloat of features was a part of the reason its adoption had lagged behind languages like C.
I've written Python for 14 years and have never seen code like that. It certainly isn't a perfect language, but this doesn't look like a common concern.
People write a lot of Python, because the language is easy to get into for a lot of non computer-science folks (e.g., engineers and scientists) and the ecosystem is massive with libraries for so many important things. It isn't as conceptually pure as lisp, but most probably don't care.
Python made a choice to have default values instead of default expressions and it comes with positive and negative trade-offs. In languages like Ruby with default expressions you get the footgun the other way where calling a function with a default parameter can trigger side effects. This kind of function is fine in Python because it's unidiomatic to mutate your parameters, you do obj.mutate() not mutate(obj).
So while it's a footgun you will be writing some weird code to actually trigger it.
Ah yes, the ol' default empty list Python gotcha, it bit me I think about 10 years ago, and ever since, sadly I've written code like this so many times it's not funny:
def fun(a = None):
_a = a if a is not None else []
A lot of people claim that but I've never seen evidence of the existence of a vast number of people who would be using Hacker News if only it had more bells and whistles. Craigslist is "ugly" too, and plenty of people use it.
I think it's more likely that most people (even most tech-adjacent people) simply don't know this place exists, or don't care, since no one is sharing links to Hacker News on mainstream social media and nothing goes viral here outside of already established HN-adjacent circles.
I like Hackernews. I like the simplicity. I don't bother with better AI. I prefer it that way and I acknowledge that the look and feel of Hackernews does not suite everyone.
But I don't value the look and feel of Hackernews, because it drives people away -- as if these people are of lesser value. That is just elitist and gatekeeper mentality.
I think HN has some pretty sophisticated automated and human-in-the-loop moderation features that few other sites possess, or throw as much resources at. Because HN is not ad-supported it does not fall victim to tragedy of the commons.
But I think HN built on what Reddit got right (at least old reddit) and also on a context of more online/faster interactions as opposed to Slashdot that brought some of the old forums structure and on a context of slower and more meaningful (ahem, for the most part) interactions. Hence why moderation was more precise, upvotes had color and you still had things like user signatures
In a way, users and posts on HN are "cattle", not pets ;)
Maybe this was tongue-in-cheek in a way that eludes me, but in case any innocent and curious bystanders are as confused as me by your comment, I'm not sure "Worse Is Better" refers to what you think it does. It isn't about "features and functionality", it's about how ease of implementation beats everything else. I can't see how that applies here, or what your comment means in that light.
The genius of Slashdot's moderation system is that it forced you to be fastidious with how your limited mod points were allocated, only using them on posts that really deserved them.
As opposed to tearing through a thread and downvoting any and everything you disagree with.
Slashdot encouraged more positive moderation, unless you were obviously trolling.
The meta-moderators kept any moderation abuse in check.
It's sad to see we have devolved from this model, and conversations have become far more toxic and polarized as a direct result of it. (Dissenting opinions are quickly hidden, and those that reinforce existing norms bubble to the top.)
I believe HN papers over these problems by relying on a lot of manual hand-moderation and curation which sounds very labor intensive, whereas Slashdot was deliberately hands-off and left the power to the people.
I remember slashdot being full of "M$ is teh evill111!!" and other childish nonsense. At the end of the day what matters is the results, and i much prefer the discusions on hn than /.
I don't think that was what made HN prevail against similar sites that were popular in the past. In my opinion, it is the fact that it is tied to Y Combinator and lots of startups/founders that made it stick. Something that is not technical at all.
Not me! For a number of years, I was like "what's with that domain, never heard of ycombinator, oh well, can't be bothered reading up on it right now, anyway, great content here, and nice minimal interface, I'll keep coming back".
I'm still missing being able to read only +5 insightful comments after 20 years.
I'd expect Slashdot's point systems and meta moderation to make a comeback in the LLM slop world we live in currently, but nobody knows about it anymore. Steam kinda rediscovered it in their reviews, perhaps even was inspired by it (I hope...)
Dutch tech news website Tweakers.net basically has this. Comments are moderated on a scale from -1 to +3, and then you can choose to expand only +2 and up.
I don’t think there is heavy moderation in the traditional sense. It’s primarily user-driven, aside from obvious abusive behavior. The downvote and flagging mechanisms do the heavy lifting.
The heuristics that detect a high ratio of arguments to upvotes (as far as I can tell) can be frustrating at times, but they also do a good job of driving ragebait off the front page quickly.
The moderators are also very good at rescuing overlooked stories and putting them in the second chance pool for users to consider again, which feels infinitely better than moderators forcing things to the front page.
It also seems that some times moderators will undo some of the actions that push a story off the front page if it’s relevant. I’ve seen flagged stories come back from the dead or flame war comment sections get a section chance at the front page with a moderator note at the top.
Back in the Slashdot days I remember people rotating through multiple accounts for no reason other than to increase their chances of having one of them with randomly granted moderation points so they could use them as weapons in arguments. Felt like a different era.
> I don’t think there is heavy moderation in the traditional sense.
It seems to be a combination of manual and automated moderation (mostly by dang but he has more help now), using the kind of over/under-engineered custom tools you'd expect from technophiles. I've wondered a lot about the kind of programming logic he and the others coded up that make HN as curious as it is, and I have half a mind to make a little forum (yet another HN clone, but not really) purely for the sake of trying to implement how I think their moderation probably works. If I went through with this, I'd have it solely be for Show HN style project sharing/discussion.
I'm reminded of definitively the most extreme writing on programming I've ever read, here https://llthw.common-lisp.dev/introduction.html, including but in no way limited to claims such as:
> The mind is capable of unconsciously understanding the structure of the computer through the Lisp language, and as such, is able to interface with the computer as if it was an extension to its own nervous system. This is Lisp Consciousness, where programmer and computer are one and the same; they drink of each other, and drink deep; and at least as long as the Lisp Hacker is there in the flow, riding the current of pure creativity and genius with their trusty companions Emacs and SLIME, neither programmer nor computer know where one ends and the other begins. In a manner of speaking, Lispers already know machine intelligence---and it is beautiful.
Has any other language produced such thoughts in the minds of human beings? Maybe yes, but I don't know of one. Maybe Forth, or Haskell, or Prolog, but I haven't found similar writing. Please do share.
I agree, and it gets even better: while low level ML support in Common Lisp does not match Python libraries, now it often does not matter because LLMs are not embedded in applications, then are often accessed via a HTTP request.
That might've been more a reflection on PLT than on Scheme48 (which also had some really smart people on it).
As some point, when I was writing a lot of basic ecosystem code that I tested on many Scheme implementations, PLT Scheme (including MzScheme, DrScheme, and a few other big pieces), by Matthias Felleisen and grad students at Rice, appeared to be getting more resources and making more progress than most.
So I moved to be PLT-first rather than portable-Scheme-first, and a bunch of other people did, too.
After Matthias moved to Northeastern, and students graduated on to their own well-deserved professorships and other roles, some of them continued to contribute to what was soon called Racket (rather than PLT Scheme). With Matthew Flatt still doing highly-skilled and highly-productive systems programming on the core.
Eventually, no matter how good their intentions and how solid their platform for production work, the research-programs-first mindset of Racket started to be a barrier to commercial uptake. They should've brought in at least one of the prolific non-professor Racketeers into the hooded circle of elders a lot sooner, and listened to that person.
One of the weaknesses of Racket for some purposes was lack of easy multi-core. The Racket "Places" concept (implementation?) didn't really solve it. You can work around it creatively, as I did for important production (e.g., the familiar Web interview load-balancing across application servers, and also offloading some tasks to distinct host processes on the same server), but using host multi-core more easily is much nicer.
As a language, I've used both Racket and CL professionally, and I prefer a certain style of Racket. But CL also has more than its share of top programmers, and CL also has some very powerful and solid tools, including strengths over Racket.
The article makes it sounds like Dang also helps with the codebase. There must be others, but Dang is the one I've seen for years at this point.
I've beeing a part of many online communities as both a member and moderator. However, Hackernews is the community that I've been apart of for the longest and the one that brings me the most joy.
Dang, is there anything random people like me can do for you? Can I at least buy you a coffee or something?
Keep in mind Hacker News (formerly Startup News) is effectively a loss-leading advertising arm of Y Combinator, which at this point is one of the most successful investment firms in the world.
And HN founder and original author Paul Graham is (at least on paper) billionaire, not merely the decamillionare he used to be.
Though it's still good for it to be a self-funding project even if that means accepting donations.
Modern CPUs are crazy fast. 4chan was serving 4 million users with a single server, a ten year old version of PHP and like 10000 lines of spaghetti code. If you do even basic code quality, profiling and optimization you can serve a huge number of users with a fraction of a CPU core.
I/O tends to be the bottleneck (disk IOPS and throughput, network connections, IOPS and throughput). HN only serves text so that's mostly an easy problem.
I still can't wrap my head around how the conventional wisdom in the industry to work around that problem is to add even more slow network I/O dependencies.
4chan is a special case, because all of its content pages are static HTML files being served by nginx that are rewritten on the server every time someone makes a post. There's nothing dynamic, everyone is served the exact same page, which makes it much easier to scale.
Modern CPUs are stupid fast when you use them the right way. You can take scale-up surprisingly far before being forced to scale out, even when that scale out is something as modest as running on multiple cores.
Based on context, you are insinuating that a discussion board like HN _can_ be hard on the CPU alone? If so, how? My guess would be _also_ be that the CPU would have little to do by itself, but that I/O would take the brunt?
I was going to reply that this is pretty common for web apps, e.g. NodeJS or many Python applications also do not use multi-threading, instead just spawning separate processes that run in parallel. But apparently, HN ran as 1 process on 1 core on 1 machine (https://news.ycombinator.com/item?id=5229548) O_O
HN is not really that much of a workload. Links with text only comments, each link gets a few hundred comments at most, and commenting on stories ends after they are old enough.
Probably everything that's current fits easily in RAM and the older stories are candidates for serving from a static cache.
I wouldn't say this is an astounding technical achievement so much as demonstrating that simplicity can fall out of good taste and resisting groupthink around "best practices".
In fairness, HN wouldn't show more than what, twenty-ish thread roots at a time, requiring you to click "more" to bring in more... which could contain the same set of thread roots you'd been looking at, depending on upvote activity.
(I assume that this update has removed that HN restriction, but haven't bothered to go look to verify this assumption.)
Good, sbcl it's great for CL. And now with current CLX from QuickLisp (the one with daily releases, I can't remember it's name) MCClim runs snappy even under Intel n270 ATom machines. Under ECL it almost runs snappy, but the performance gain it's astronomical. From a really laggy UI to instant rendering.
Readit News
Another great accesibility feat is the APIs for this web. I use upvoterate listing from quality news as front page.
To sum it up, opening the tab previews under Zen Browser made my experience feel like the coolest SPA, haha.
What makes HN work is the tight focus and heavy moderation.
One of the original motivating examples were Unix-like systems (simple implementation, few correctness guarantees in interfaces) vs. Lisp-based systems (often well specified interfaces, but with complicated implementations as the cost.)
Edit: or as someone else who has phrased it better: "less is more".
I also thought Slashdot's moderation system was kind of fun. I am not sure it was useful but I enjoyed the annotations (+5 Funny when serious, +5 Insightful when inciteful, etc.) Meta-moderation was also neat?
At least, that's how my bash pager has it in the manpage.
Annoyingly enough it's been talked about for years but it never gets implemented, despite only three colors really needing a swap: background to dark sepia or just dark gray, and text to white and off-white.
Deleted Comment
Maybe now it's been ported to Common Lisp it'll be easier to add features.
Dead Comment
"worse is better" is people putting up with footguns like this in python, because it's percieved easier to find a python job:
HN is very much "less is better", not "worse is better".> It refers to the argument that software quality does not necessarily increase with functionality: that there is a point where less functionality ("worse") is a preferable option ("better") in terms of practicality and usability.
For that reason, I think I am applying the term precisely as it was defined.
The irony of my comment, which dang picked up, is that the original idea was a criticism against Lisp, suggesting that the bloat of features was a part of the reason its adoption had lagged behind languages like C.
1. https://en.wikipedia.org/wiki/Worse_is_better
People write a lot of Python, because the language is easy to get into for a lot of non computer-science folks (e.g., engineers and scientists) and the ecosystem is massive with libraries for so many important things. It isn't as conceptually pure as lisp, but most probably don't care.
So while it's a footgun you will be writing some weird code to actually trigger it.
I think it's more likely that most people (even most tech-adjacent people) simply don't know this place exists, or don't care, since no one is sharing links to Hacker News on mainstream social media and nothing goes viral here outside of already established HN-adjacent circles.
But I don't value the look and feel of Hackernews, because it drives people away -- as if these people are of lesser value. That is just elitist and gatekeeper mentality.
It's like they know somewhere deep inside that "mo tech" is not helping anyone.
- technologists and startup wannabes feeling like HN is "underground" because of the stripped down aesthetic and weird tech stack
- out of touch VCs who are successful because of money and connections but want to cosplay as technical
- the end users of the startups, who are fed the enshittified products funded by the VCs and created by the technologists
But I think HN built on what Reddit got right (at least old reddit) and also on a context of more online/faster interactions as opposed to Slashdot that brought some of the old forums structure and on a context of slower and more meaningful (ahem, for the most part) interactions. Hence why moderation was more precise, upvotes had color and you still had things like user signatures
In a way, users and posts on HN are "cattle", not pets ;)
Here's the original essay -- https://www.dreamsongs.com/RiseOfWorseIsBetter.html
This is a good little overview entitled "Worse is Better Considered Harmful" -- https://cs.stanford.edu/people/eroberts/cs201/projects/2010-... -- in which the authors argue for "Growable Is Better".
In summary - it's about ease of implementation trumping all else. C and Unix are memorably labelled "the ultimate computer viruses".
As opposed to tearing through a thread and downvoting any and everything you disagree with.
Slashdot encouraged more positive moderation, unless you were obviously trolling.
The meta-moderators kept any moderation abuse in check.
It's sad to see we have devolved from this model, and conversations have become far more toxic and polarized as a direct result of it. (Dissenting opinions are quickly hidden, and those that reinforce existing norms bubble to the top.)
I believe HN papers over these problems by relying on a lot of manual hand-moderation and curation which sounds very labor intensive, whereas Slashdot was deliberately hands-off and left the power to the people.
unsure why precisely it descended so much
not crazy about HN's approach but the quality of the discourse here is so high through whatever mechanism, I don't much care
I'd expect Slashdot's point systems and meta moderation to make a comeback in the LLM slop world we live in currently, but nobody knows about it anymore. Steam kinda rediscovered it in their reviews, perhaps even was inspired by it (I hope...)
https://m.xkcd.com/1019/
https://m.xkcd.com/2159/
I don’t think there is heavy moderation in the traditional sense. It’s primarily user-driven, aside from obvious abusive behavior. The downvote and flagging mechanisms do the heavy lifting.
The heuristics that detect a high ratio of arguments to upvotes (as far as I can tell) can be frustrating at times, but they also do a good job of driving ragebait off the front page quickly.
The moderators are also very good at rescuing overlooked stories and putting them in the second chance pool for users to consider again, which feels infinitely better than moderators forcing things to the front page.
It also seems that some times moderators will undo some of the actions that push a story off the front page if it’s relevant. I’ve seen flagged stories come back from the dead or flame war comment sections get a section chance at the front page with a moderator note at the top.
Back in the Slashdot days I remember people rotating through multiple accounts for no reason other than to increase their chances of having one of them with randomly granted moderation points so they could use them as weapons in arguments. Felt like a different era.
It seems to be a combination of manual and automated moderation (mostly by dang but he has more help now), using the kind of over/under-engineered custom tools you'd expect from technophiles. I've wondered a lot about the kind of programming logic he and the others coded up that make HN as curious as it is, and I have half a mind to make a little forum (yet another HN clone, but not really) purely for the sake of trying to implement how I think their moderation probably works. If I went through with this, I'd have it solely be for Show HN style project sharing/discussion.
To be spartan is to philosophize.
Dead Comment
Though, Reddit eventually realized that javascript: URLs - in Markdown - were an XSS risk.
I'm reminded of definitively the most extreme writing on programming I've ever read, here https://llthw.common-lisp.dev/introduction.html, including but in no way limited to claims such as:
> The mind is capable of unconsciously understanding the structure of the computer through the Lisp language, and as such, is able to interface with the computer as if it was an extension to its own nervous system. This is Lisp Consciousness, where programmer and computer are one and the same; they drink of each other, and drink deep; and at least as long as the Lisp Hacker is there in the flow, riding the current of pure creativity and genius with their trusty companions Emacs and SLIME, neither programmer nor computer know where one ends and the other begins. In a manner of speaking, Lispers already know machine intelligence---and it is beautiful.
Has any other language produced such thoughts in the minds of human beings? Maybe yes, but I don't know of one. Maybe Forth, or Haskell, or Prolog, but I haven't found similar writing. Please do share.
Originally on MzScheme, then later PLT Scheme. It was ported to Racket by the great kogir, IIRC.
Also, I believe pg started implementing Arc on Scheme48 based on mailing list activity at the time. I've always been curious about the switch to PLT!
As some point, when I was writing a lot of basic ecosystem code that I tested on many Scheme implementations, PLT Scheme (including MzScheme, DrScheme, and a few other big pieces), by Matthias Felleisen and grad students at Rice, appeared to be getting more resources and making more progress than most.
So I moved to be PLT-first rather than portable-Scheme-first, and a bunch of other people did, too.
After Matthias moved to Northeastern, and students graduated on to their own well-deserved professorships and other roles, some of them continued to contribute to what was soon called Racket (rather than PLT Scheme). With Matthew Flatt still doing highly-skilled and highly-productive systems programming on the core.
Eventually, no matter how good their intentions and how solid their platform for production work, the research-programs-first mindset of Racket started to be a barrier to commercial uptake. They should've brought in at least one of the prolific non-professor Racketeers into the hooded circle of elders a lot sooner, and listened to that person.
One of the weaknesses of Racket for some purposes was lack of easy multi-core. The Racket "Places" concept (implementation?) didn't really solve it. You can work around it creatively, as I did for important production (e.g., the familiar Web interview load-balancing across application servers, and also offloading some tasks to distinct host processes on the same server), but using host multi-core more easily is much nicer.
As a language, I've used both Racket and CL professionally, and I prefer a certain style of Racket. But CL also has more than its share of top programmers, and CL also has some very powerful and solid tools, including strengths over Racket.
(This conversation has turned unexpectedly ontological!)
HN runs now on SBCL, which is much faster and also multi-threaded.
I've beeing a part of many online communities as both a member and moderator. However, Hackernews is the community that I've been apart of for the longest and the one that brings me the most joy.
Dang, is there anything random people like me can do for you? Can I at least buy you a coffee or something?
And HN founder and original author Paul Graham is (at least on paper) billionaire, not merely the decamillionare he used to be.
Though it's still good for it to be a self-funding project even if that means accepting donations.
Oh I have been on HN since 2008 and didn't know that.
This was all running on a single core??
I/O tends to be the bottleneck (disk IOPS and throughput, network connections, IOPS and throughput). HN only serves text so that's mostly an easy problem.
Probably everything that's current fits easily in RAM and the older stories are candidates for serving from a static cache.
I wouldn't say this is an astounding technical achievement so much as demonstrating that simplicity can fall out of good taste and resisting groupthink around "best practices".
Spawning new processes for every user is possible but would probabaly be less scalable than even thread-switching.
Deleted Comment
https://news.ycombinator.com/item?id=5229548
https://news.ycombinator.com/item?id=28478379
https://news.ycombinator.com/item?id=27452276
HN is an island of sanity in a sad world.
(I assume that this update has removed that HN restriction, but haven't bothered to go look to verify this assumption.)
EDIT: UltraLisp for QuickLisp.