Good riddance. The graph peaked in 2014 and started to gradually go down in 2017, with the release of LLMs only adding fuel to the fire. They did this all to themselves with their way of how people are treated there.
Any social organization needs to carefully consider their inclusion-exclusion curve with intentionality.
I think a lot of people might balk at the word "inclusivity" today, but StackExchange had ridiculously high barriers to participation, making it inclusive to the long-time users on the site, but exclusive to the newbie participants who found themselves blocked for asking questions. They slowly killed the site in this manner.
The community might have survived this folly, even with AI, because it was still the best place for people with qualms about AI to ask questions... Except until StackOverflow management alienated those users, too, by shoving AI down their throats in every facet of the site.
Even I had internalized the vagaries and neuroses of the SO community but I had heavy reticence to ask questions, knowing I'd have to consider all the ways a bully eager to use their powers might misunderstand me. I can't imagine asking a question there without having had lurked for longer than a typical Bachelor's + Masters program.
Peak at 207K, minimum at 588. That might be an incomplete date point, so using the next most recent value 1226, StackOverflow has lost 99.41% of its activity.
S/o's somewhat cumbersome scheme to aquire comment points to be able to answer was a awkward kludge in the principal problem of open contribution sites, namely that human slop and gamified tactics tend to kill the site. It probably was effective enough to keep things working for a long time but it could not recruit new users (not just readers) as fast as it needed to. It doesn't seem like the points based system really helped as many devs find jobs as it would have taken for the site to become a recruiting tool. It probably would have had to shift and evolve in several keys ways to survive
It’s moderation policies, or maybe just moderators themselves that killed it, this was apparent before AI, but there was no alternative… until suddenly there was.
SO did that all to themselves when they decided they didn't want a community to form and that only question and answers mattered. The moment something else allowed to have a better way to get your answers, there was no reason to go there, because there was no community.
I still don't understand why anyone would go with that whole "no conversation please"
That basic idea is what made SO attractive in the first place, compared to forums where you had to scroll through pages and pages of in-jokes and tangents and animated-GIF signatures just to try and see if there is an answer.
Where they went wrong, in my opinion, is in the implementation details.
It's mostly death by a thousand cuts: Requiring reputation to gain the ability to post comments, then having one's answers deleted as "this should've been a comment". Overeager marking of questions as duplicates, e.g. despite the equivalence between two situations being non-obvious (e.g. someone asks about data type A, and it turns out that it's a subtype of B for which an answer that applies to both exists; that should not be a duplicate, the fact that it's a subtype is the answer!). Endless other decisions like that, which wouldn't have taken any extra effort to implement correctly.
One feature they could've built that would have taken effort but also greatly helped against the common newbie complaint of "hostility" would've been a "newcomer track", which would've been more forum-like and guided them towards either formulating a good question or seeing that's it's already answered. In the latter case, some of the keywords that came up during this process should've been fed back into SEO so that future newbies would become more likely find the answer via a search engine despite using clumsy terms. I think they tried a simpler and worse version of this idea towards the end with "staging ground" but by then it was too late.
SO had great humans contributing to their platform, even as AI began to serve as the new SO for a lot of people.
Instead of going in the same direction of everyone else adding AI all over the place and trying to eliminate the humans, they could have gone the opposite direction and played to their somewhat unique strength of having a bunch of actual humans and providing a place that actually fostered human and authentic interactions. Instead, for some completely unknown reason (money), they chose to commodify their own platform. Smart.
People were willing to put up with SO in the before times, nobody was excited to ask a question, it was a dreadful experience. It was their UX that doomed them immediately. Reddit and HN are still here after all.
Reddit is on the same track. What I've noticed is that moderators have become increasingly hostile. Reddit's AI moderation, which is designed to remove AI slop, has removed multiple top contributors I used to follow on programming, electronics, bodybuilding, welding, machining, and 3Dprinting subs. And the worst part is that Reddit site admins have no idea about this. I know a guy who reckons he can get any account banned from Reddit by simply mass reporting it using residential IP providers and OpenClaw.
Good luck trying to write any helpful posts in the community anymore, someone will come along and respond with "AI."
Only the new Reddit, old.reddit.com still works for that too.
Though I'm not sure how long that'll last. I would be surprised if old.reddit.com is still functional in a couple of years. When it gets removed, or when it bit rots to the point that it's no longer really feasible to use it, I'm off that site. New Reddit doesn't work for me.
I just noticed that by around 2015 it had got super toxic and snippy.
Usually I’d find answers on SO. Relatively rarely I’d ask questions but, when I did, I’d always try and follow the netiquette rules of yore, and think in terms of, if I was a support engineer trying to help with this, what would I need to know?
Because I have supported products, and we’ve all seen enough bug reports and questions come in that we can tell when someone is going to be easy to help - even if they have a particularly tricky problem - versus someone who’s going to prove more challenging.
So I had this question about Elasticsearch, and it was at a time when the documentation wasn’t great, and you were actively encouraged to go on SO and tag your question to get help.
I wrote out in detail what I’d done, where I’d got stuck, what I’d read and tried to get unstuck, etc. It probably took me 30 minutes or more to pull everything together into a coherent post.
The very first comment was from some insufferable bellend saying, “Oh, so you want us to do your work for you, are you going to pay us too?” or words very much to that effect.
Literally, WTF? Why even post that? If you don’t want to help the option to simply go away without getting involved is always available.
IIRC I didn’t actually end up finding a solution via SO and instead layered some godawful hack on top of Elasticsearch to get what we needed - because I simply had other work to move on to and I’d already spent a lot of time on the problem.
But I think that was the last question I posted on SO, and maybe the last time I posted anything on the site.
As the years wore on I simply started finding it less and less useful, with often incorrect answers marked as accepted and - if you were lucky - the correct answers marked might be buried further down.
And then there’s what they wanted to charge for job ads versus how effective those ads actually were - again, this was better in their earlier years.
SO started out well - genuinely a breath of fresh air - but as time went on it felt like they thought their model was the last word in online help forums and they didn’t want to evolve to address its flaws, even if that had just been dealing with the toxicity, and the karma farming.
And so this is the result - a site that, like the dinosaur in A Sound of Thunder, is dead but perhaps hasn’t realised it yet - and, at this point, the way I feel is simply good riddance. It’s a shame, but - as you said - they did it to themselves.
I stopped posting there around then. Multiple times in a row I asked a question, got it closed as a duplicate/pointed to another thread/met with "why are you even trying to do that" despite being clear about my problem. It was obvious the people answering/closing the question skimmed it and I already gave a ton of context. It just stopped being useful
Yeah, SO's downfall started looong ago. The community was frankly horribly managed, and its strict "no remotely duplicate-esque questions ever" policy meant that answers to common questions just got more and more outdated as time went on. It's still common to search for something, find an SO thread, and find that the only answers are from 2013. The world has changed since 2013, answers in 2026 would be different, but because the question would be the same, any contemporary attempt at asking the same question would get marked as a duplicate, so the 2013 answer remains SO's only guidance forever.
They also had the problem that easy questions would get downvoted for being too easy, and hard questions would just not get answered because they weren't seen in time by the narrow group of people who could answer so they get buried by the algorithm. Working in something of a less common niche myself (embedded Linux), I never had questions get answered. I believe the question ranking systems and moderation policies really only worked for questions about new, popular web frameworks.
It was ChatGPT which did it in, but it could've been anything. It could've been a new group of people with some clout starting a fresh new knowledge site. People were ready to abandon SO.
It was also the rise of Github and the importance of the software hosted there. More consistent documentation and transparent issue trackers/PRs helped a lot dealing with evolving software.
The graph actually peaked in 2014. That's a decade before AI became a thing while the Software engineering workforce grew a lot since then.
I think Stackoverflow was the last incarnation of romanticized old-school Q&A forums where you first had to earn your badges before being treated with respect. Luckily today's new projects have much better documentation, issue tracker's etc. And apparently AI is able to work with that by now...
Answers from 2013 likely no longer reflect the currently accepted ways to do things, even for technologies which existed in 2013 and earlier. What you describe is a problem they created on their own with their ridiculous duplicates policy which ignored the fact that the world keeps changing.
Not sure I would blame it all on AI though, the incentives of SO only worked while there were worthwhile questions to answer and make you feel smart about. After that well dried up, the only thing left was the stuff AI can do with a prompt; ironically AI got a leg up by scraping SO.
This is similar to the evolution of Wikipedia, except the format of WP allowed it to transform into a feudal dictatorship of nerds who feel like they are deciding what's true, and they can get off on that.
SO did not have that kind of incentive to keep the nerds around.
There still are worthwhile questions to answer! People just don't ask them on SO because SO chased away everyone who wanted to ask questions, long before ChatGPT.
Yeah, I don't understand the HN title. The "downfall" seems to have began in 2018-2020 sometime, what AI was launched and popularized at that point that would have killed SO? LLMs were basically useless until GPT3 which appeared in middle-2020 sometime, after the downfall seemingly already had begun.
I'd call it significant that the number of questions halved within one year following the release of ChatGPT, the biggest relative or absolute rate of decrease in the timeseries.
People love simplistic narratives, i usually don't mind but this is just ridiculous. AI hate is gently overtaking AI hype as the most stupid thing around
It's not really a bell curve. There was obviously a downwards trend from 2016 onwards, but 2023 definitely precipitated the fall to zero. Without AI they might have lasted at least a couple more years, or the activity might have stabilized to a new floor greater than zero.
You mean the fall to a thousand questions per month. Now that the volume is low enough someone has a chance of looking at every single one of them, maybe the StackOverflow community can finally collaborate in peace, safe from the onslaught of questions that could be answered by reading the documentation.
Just because it's approximated by a bell curve doesn't make it a bell curve. There are quite obvious separate phenomena shaping the curve at different times.
Are you discarding the utility of Gaussian functions in analysis simply because the independent variable is time? A Gaussian curve can be used as a descriptive model without claiming that the observations themselves are a probability distribution
The fit does not prove causation, but it does show that the decline was already well described by a trend that began years before generative AI. If the claim is that 2023 created a separate structural break, it's different claim then the title describes
I am not sure. I think SO died way before AI and that graph seems incorrect too.
> Without AI they might have lasted at least a couple more years
Nah, their decline was already readily apparent before AI. You only
need to go through old discussions and other people noticing it.
AI may have accelerated the decay, but the decline happened already
largely prior to AI.
That's a very big word you're using there for what is basically making shapes out of clouds. A bell-curve is the amortised function of a random variable with a mean and standar deviation. What does that have to do with a timeseries dataset?
A bell curve is not an "amortised function." Amortization applies to accounting and algorithmic time complexity, not probability distributions. You're likely thinking of a Probability Density Function (PDF). If you are going to police terminology, it helps to use the correct words. Second, fitting a curve with an R^2 of 0.911 is the exact opposite of "making shapes out of clouds.
You are confusing a Probability Density Function (PDF) with a phenomenological curve fit. No one is claiming that time is a random variable drawn from a normal distribution.
Fitting a mathematical function to a dataset does not implicitly adopt the ontological baggage of probability theory. That is a fundamental misunderstanding of applied mathematics
If that was the case, the graph would never have gotten to the heights it did.
What happened is that as the corpus of useful info increased, the need to pose new Qs decreased. AI much accelerated that decline by making available an 'oracle' trained on that corpus.
The users being generally unhelpful wasn't an issue for them, since they were still significantly more helpful than users anywhere else on the internet. Reddit was and still is filled with completely unvetted answers (on pretty much all topics not just programming), Quora was/is a joke, Yahoo answers had some funny posts I guess but nothing you could actually learn from, what else really was there? Before AI, Stack Overflow was as good as it gets.
It's both. Users tolerated the hostile environment to an extent as long as the site was still the best way to get useful information. When LLMs came out, that was no longer the case.
There was also the pattern of "closing as already answered" with an answer from 6 years earlier which wasn't actually answering the question when you dig into it. Certainly in the code stacks.
So... nothing that it wasn't already doing to itself? There's no one drop where "AI got into the market", SO had been declining steadily for years. I actually expected this post to be about how SO survived by selling its internal organs to AI.
Some of the pre-ai decline in questions might just be that they had filled out of alot of the question space. What might be more interesting is the traffic graph as it would be possible to have a decline in questions but still have traffic rising to the existing ones.
Today I'd expect even when someone is googling and a question is relevant that most people will just read the AI overview.
ChatGPT was released in Nov 2022, and frankly wasn't very good originally. The SO decline started occurring almost two years ahead of that, and was already on a sharp decline before ChatGPT shipped, and certainly before ChatGPT actually became good.
This is revisionist history. People told SO that they were leaving for YEARS because of how incredibly toxic it had become. It was already giving outdated answers before ChatGPT shipped, because new questions/potentially updated answers were [Closed] [Dupe] immediately.
Their answer was essentially "We aren't a Q&A site, we're trying to be a knowledge base! So closing all questions on a Q&A-stylized site, and extremely abrasive moderation, is working as intended."
They entirely did this to themselves. The community was toxic, their policies were toxic, and they didn't listen when warned as such repeatedly - just doubled down.
The graph is not presenting a narrative, did you mean to reply to someone else who is presenting the "revisionist history"?
In 2023, Stack Overflow had already started making unpopular pro-AI moderation decisions, and in 2024, they started mass banning everyone who deleted their questions and answers in protest. I don't think it's wholly incorrect to say "AI killed Stack Overflow" when the death blow came from crazy pro-AI decisions from the admin.
Did they cry about it? No right? Don't apply your own standard then judge them about it, petty people.
Stackoverflow aimed to be a knowledge base. And knowledge base has a ceiling limit. They simply reached the point that almost all questions (regarding the knowledge) were asked for them. You can argue that newer or niche libraries or languages knowledge is still lacking there, but I have never seen them getting closed, just not answered.
The stackoverflow moderation is the reason I do not post on it.
You have middle party with no competence on the technology trying to do useless moderation.
Instead I directly go on the project github page and ask the question directly to the mainteners.
I’m surprised the site had that little activity, relatively, around 2010. That’s when I was using it the most. It seemed plenty large to me at the time… I can’t imagine the experience was that great a few years later. I wasn’t really paying attention by the time it reached the peaks later that decade.
I always found the format of the side obtuse and the culture not very welcoming. My most popular answer ever was something about JavaScript from 2008 or 2009, and to this day, people come in and say “this isn’t the way to do it, this is outdated“. No kidding, but every new question about that gets closed as a duplicate.
AI might've delivered the final blow, but Stack Overflow was in decline LONG before LLMs came on the scene.
I read a great article not long ago outlining the full series of events and changes that led to its downfall. I wish I could find that article, but I've forgotten where it was.
I never understood the point if having the unfathomable churn of thousands of new questions per day. The value of SO to me was always a knowledge base with reputation mechanics, and that did not change. I still default to searching before asking AI.
Honestly, I think that's a good thing. A lot of questions were either duplicates of existing questions, or close derivatives of them. If I had to guess, probably 90% of SO questions already had a solution somewhere on SO. AI surfaces these solutions much quicker, so you don't have to ask. Novel questions or bugs that can't be answered or fixed by AI still get asked, and mods have less spam to deal with. I fail to see the issue here.
SO's downfall started way before AI. A decade or so ago it was always full of interesting questions, people were giving detailed answers, there was sometimes some debate in the answers, etc
And then it started being stupid questions. People who clearly had barely tried anything and just rushed to SO with a half baked question. Answers were just pointing to another thread that already provided the answer. It definitely started before LLMs. I think it lined up with the aggressive "learn-to-code" push.
Yep, SO was dying before the GPTs - in some ways it was baked in the original SO design - to become the canonical source of information about programming stuff.
Many people talk about the negativity, and they are right, but I think the reason more than anything is the waiting. On SO a good question might get answered in minutes (if it was easy and someone was karma farming) but it could be days or weeks for general purpose stuff; compare that to a few seconds for an LLM its a no brainer.
Somehow I feel that whole idea was 50% broken from start. In field changing as fast as programming you simply cannot have canonical source that is done. There are areas which are constant or slow moving. But too many others are in churn. Trying to apply same rules for both is clearly insanity. And this is really the failure of SO...
Maybe if in some cases stuff actually got deprecated and points did actively decay it might have worked... But can't remove power from those who had it for years...
You also have to think that SO was very early in gamification and changing things around that never got political will afaict - people love their points.
I noticed the same thing. Another thing that seems to have disappear: lunch and learn focusing on the tech stack. And I mean specifically about "how to do this in <programming language", not product/company focused.
This is IMO wrong. StackOverflow died way before
AI - and way before 2020 too. I think it had a
peak time of only 3 or 4 years. It was created in
2008, and I would reason it took a few years,
say, up to 2011; then it was semi-okish up to
about 2015, roughly. Then it declined.
It still has some value today, as sometimes you
can find useful information on SO, but its peak
days are long over and I don't see how it can
manage to come back, with or without AI slop.
It would basically require a lot of re-design
and some things that never worked, such as the
karma system, should be changed. Also moderators -
they kill sites. That happened to reddit - I gave
up after censor-mods constantly restricted everyone.
I disagree. From what I see, company specific stackoverflow like mailing lists or slack channels are also dead.
The normal day to day devs just don’t have the need to go to stackoverflow anymore.
One would have to explain both consequences or dismiss it as coincidental.
FWIW I rarely have the need to ask questions at the programming level to anyone anymore. It’s just not the type of thing I bring up or anyone else. We now talk about architecture and company direction.
Good riddance. The graph peaked in 2014 and started to gradually go down in 2017, with the release of LLMs only adding fuel to the fire. They did this all to themselves with their way of how people are treated there.
Funnily enough, there's now a "StackOverflow for Agents": https://agents.stackoverflow.com/recent
Any social organization needs to carefully consider their inclusion-exclusion curve with intentionality.
I think a lot of people might balk at the word "inclusivity" today, but StackExchange had ridiculously high barriers to participation, making it inclusive to the long-time users on the site, but exclusive to the newbie participants who found themselves blocked for asking questions. They slowly killed the site in this manner.
The community might have survived this folly, even with AI, because it was still the best place for people with qualms about AI to ask questions... Except until StackOverflow management alienated those users, too, by shoving AI down their throats in every facet of the site.
Even I had internalized the vagaries and neuroses of the SO community but I had heavy reticence to ask questions, knowing I'd have to consider all the ways a bully eager to use their powers might misunderstand me. I can't imagine asking a question there without having had lurked for longer than a typical Bachelor's + Masters program.
Peak at 207K, minimum at 588. That might be an incomplete date point, so using the next most recent value 1226, StackOverflow has lost 99.41% of its activity.
S/o's somewhat cumbersome scheme to aquire comment points to be able to answer was a awkward kludge in the principal problem of open contribution sites, namely that human slop and gamified tactics tend to kill the site. It probably was effective enough to keep things working for a long time but it could not recruit new users (not just readers) as fast as it needed to. It doesn't seem like the points based system really helped as many devs find jobs as it would have taken for the site to become a recruiting tool. It probably would have had to shift and evolve in several keys ways to survive
It’s moderation policies, or maybe just moderators themselves that killed it, this was apparent before AI, but there was no alternative… until suddenly there was.
SO did that all to themselves when they decided they didn't want a community to form and that only question and answers mattered. The moment something else allowed to have a better way to get your answers, there was no reason to go there, because there was no community.
I still don't understand why anyone would go with that whole "no conversation please"
That basic idea is what made SO attractive in the first place, compared to forums where you had to scroll through pages and pages of in-jokes and tangents and animated-GIF signatures just to try and see if there is an answer.
Where they went wrong, in my opinion, is in the implementation details.
It's mostly death by a thousand cuts: Requiring reputation to gain the ability to post comments, then having one's answers deleted as "this should've been a comment". Overeager marking of questions as duplicates, e.g. despite the equivalence between two situations being non-obvious (e.g. someone asks about data type A, and it turns out that it's a subtype of B for which an answer that applies to both exists; that should not be a duplicate, the fact that it's a subtype is the answer!). Endless other decisions like that, which wouldn't have taken any extra effort to implement correctly.
One feature they could've built that would have taken effort but also greatly helped against the common newbie complaint of "hostility" would've been a "newcomer track", which would've been more forum-like and guided them towards either formulating a good question or seeing that's it's already answered. In the latter case, some of the keywords that came up during this process should've been fed back into SEO so that future newbies would become more likely find the answer via a search engine despite using clumsy terms. I think they tried a simpler and worse version of this idea towards the end with "staging ground" but by then it was too late.
My favorite thing was how for half the stuff I Googled, the top result would be a StackOverflow reply telling me to Google it.
Off topic, closed.
SO had great humans contributing to their platform, even as AI began to serve as the new SO for a lot of people.
Instead of going in the same direction of everyone else adding AI all over the place and trying to eliminate the humans, they could have gone the opposite direction and played to their somewhat unique strength of having a bunch of actual humans and providing a place that actually fostered human and authentic interactions. Instead, for some completely unknown reason (money), they chose to commodify their own platform. Smart.
There used to also be fun, and somewhat interesting questions to answer or discuss.
It quickly turned into simple questions and "send me the codes"
As noted by others, the initializer for the curt communication culture was "don't ask stupid questions, idiot."
Are you saying that the decline of SO can be explained by that and not AI?
I don't buy it.
People were willing to put up with SO in the before times, nobody was excited to ask a question, it was a dreadful experience. It was their UX that doomed them immediately. Reddit and HN are still here after all.
How eager people are to find something better IS a factor in the decline of websites/businesses. Not just that there is something better that exists.
Reddit is on the same track. What I've noticed is that moderators have become increasingly hostile. Reddit's AI moderation, which is designed to remove AI slop, has removed multiple top contributors I used to follow on programming, electronics, bodybuilding, welding, machining, and 3Dprinting subs. And the worst part is that Reddit site admins have no idea about this. I know a guy who reckons he can get any account banned from Reddit by simply mass reporting it using residential IP providers and OpenClaw.
Good luck trying to write any helpful posts in the community anymore, someone will come along and respond with "AI."
Reddit turned toxic long ago, like ten years ago and they brought in Ellen Pao to fix things, and she somehow managed to make things worse.
Ellen Pao was the glass scapegoat CEO designed to take all the flak for what the other investors and board members wanted to do.
BTW Reddit is now verifying your ID with Persona before you can open anything it thinks is NSFW.
Only the new Reddit, old.reddit.com still works for that too.
Though I'm not sure how long that'll last. I would be surprised if old.reddit.com is still functional in a couple of years. When it gets removed, or when it bit rots to the point that it's no longer really feasible to use it, I'm off that site. New Reddit doesn't work for me.
Not when I open it from a web browser without being logged in. It just ask if I'm 18 and if I click yes, it lets me see the post (including images).
Can confirm for some dumb joke subs.
You want to blow off some steam, and there is a laundry list of rules to read through.
Yeah, not bothering with all that.
I just noticed that by around 2015 it had got super toxic and snippy.
Usually I’d find answers on SO. Relatively rarely I’d ask questions but, when I did, I’d always try and follow the netiquette rules of yore, and think in terms of, if I was a support engineer trying to help with this, what would I need to know?
Because I have supported products, and we’ve all seen enough bug reports and questions come in that we can tell when someone is going to be easy to help - even if they have a particularly tricky problem - versus someone who’s going to prove more challenging.
So I had this question about Elasticsearch, and it was at a time when the documentation wasn’t great, and you were actively encouraged to go on SO and tag your question to get help.
I wrote out in detail what I’d done, where I’d got stuck, what I’d read and tried to get unstuck, etc. It probably took me 30 minutes or more to pull everything together into a coherent post.
The very first comment was from some insufferable bellend saying, “Oh, so you want us to do your work for you, are you going to pay us too?” or words very much to that effect.
Literally, WTF? Why even post that? If you don’t want to help the option to simply go away without getting involved is always available.
IIRC I didn’t actually end up finding a solution via SO and instead layered some godawful hack on top of Elasticsearch to get what we needed - because I simply had other work to move on to and I’d already spent a lot of time on the problem.
But I think that was the last question I posted on SO, and maybe the last time I posted anything on the site.
As the years wore on I simply started finding it less and less useful, with often incorrect answers marked as accepted and - if you were lucky - the correct answers marked might be buried further down.
And then there’s what they wanted to charge for job ads versus how effective those ads actually were - again, this was better in their earlier years.
SO started out well - genuinely a breath of fresh air - but as time went on it felt like they thought their model was the last word in online help forums and they didn’t want to evolve to address its flaws, even if that had just been dealing with the toxicity, and the karma farming.
And so this is the result - a site that, like the dinosaur in A Sound of Thunder, is dead but perhaps hasn’t realised it yet - and, at this point, the way I feel is simply good riddance. It’s a shame, but - as you said - they did it to themselves.
The anonymity of people giving aggressive and confrontational "answers" to a question reminds me of:
https://www.penny-arcade.com/comic/2004/03/19/green-blackboa...
I stopped posting there around then. Multiple times in a row I asked a question, got it closed as a duplicate/pointed to another thread/met with "why are you even trying to do that" despite being clear about my problem. It was obvious the people answering/closing the question skimmed it and I already gave a ton of context. It just stopped being useful
>"no conversation please"
Good for training data I guess - pure Question and Answer. Maybe they knew the platform would die so decided to optimise for that
It always worked like that
Yeah, SO's downfall started looong ago. The community was frankly horribly managed, and its strict "no remotely duplicate-esque questions ever" policy meant that answers to common questions just got more and more outdated as time went on. It's still common to search for something, find an SO thread, and find that the only answers are from 2013. The world has changed since 2013, answers in 2026 would be different, but because the question would be the same, any contemporary attempt at asking the same question would get marked as a duplicate, so the 2013 answer remains SO's only guidance forever.
They also had the problem that easy questions would get downvoted for being too easy, and hard questions would just not get answered because they weren't seen in time by the narrow group of people who could answer so they get buried by the algorithm. Working in something of a less common niche myself (embedded Linux), I never had questions get answered. I believe the question ranking systems and moderation policies really only worked for questions about new, popular web frameworks.
It was ChatGPT which did it in, but it could've been anything. It could've been a new group of people with some clout starting a fresh new knowledge site. People were ready to abandon SO.
It was also the rise of Github and the importance of the software hosted there. More consistent documentation and transparent issue trackers/PRs helped a lot dealing with evolving software.
I mean, that's literally what the site was designed for. It was not designed to be a community. It was designed to be a Q&A site.
A Q&A site where the community asks questions and the community answers them. That doesn't work without a community.
The graph actually peaked in 2014. That's a decade before AI became a thing while the Software engineering workforce grew a lot since then.
I think Stackoverflow was the last incarnation of romanticized old-school Q&A forums where you first had to earn your badges before being treated with respect. Luckily today's new projects have much better documentation, issue tracker's etc. And apparently AI is able to work with that by now...
I never had an LLM tell me my question was already answered and imply I was stupid for not finding it. SO dug its own grave and jumped in.
Looks like SO was already dying since 2017.
I think other helpful places like reddit, discord, web forums etc might be what hit SO 2014-15 onwards.
AI seems to have given it a blow of mercy to end the misery.
The graph shows questions, at some point they saturate, 90% of the stuff is answered. New frameworks contribute to the majority of the new questions.
Answers from 2013 likely no longer reflect the currently accepted ways to do things, even for technologies which existed in 2013 and earlier. What you describe is a problem they created on their own with their ridiculous duplicates policy which ignored the fact that the world keeps changing.
> at some point they saturate, 90% of the stuff is answered
I don’t buy this.
Programming as an industry is famous for constantly evolving and changing.
for a lot of stuff it just made more sense to ask on the github project's issue tracker.
- Open issue on GitHub issues: Maintainer closes issue citing "Bug reports only"
- Open question on SO: Moderators close because it's too specific to a library
- Ask on IRC: Get piled on for not using the right vocabulary and your IP isn't masked
- Ask the LLM: Get hallucinated answer based on old API docs
- Ask technical lead: Get burned for asking basic question and put on PIP
- Ask my mom: She doesn't know enough computer to know the answer, but in explaining the problem to her, I finally figured out what I got wrong
Usually #1 actually works, most cases, and missing discord, several healthy communities on Discord nowadays
You’re using your mom for rubber duck debugging, then. An LLM can work for that as well, usually better because it knows a lot more than your mom.
It's fictional, I never actually got burned by a tech lead nor been put on PIP either, nor asked my mom to help me with programming :)
The collapse into a ghost town is striking.
Not sure I would blame it all on AI though, the incentives of SO only worked while there were worthwhile questions to answer and make you feel smart about. After that well dried up, the only thing left was the stuff AI can do with a prompt; ironically AI got a leg up by scraping SO.
This is similar to the evolution of Wikipedia, except the format of WP allowed it to transform into a feudal dictatorship of nerds who feel like they are deciding what's true, and they can get off on that.
SO did not have that kind of incentive to keep the nerds around.
There still are worthwhile questions to answer! People just don't ask them on SO because SO chased away everyone who wanted to ask questions, long before ChatGPT.
I remember a time when people were posting their SO link and karma on their resumes. O how have the times changed.
check the graph and superpose the ai adoption curve, you're right to be skeptical
Seing a bell curve and singling out a factor that appears only for the 15% of the total time demonstrates some pretty extreme tunnel vision
Edit: https://postimg.cc/n9nZGLmb
Yeah, I don't understand the HN title. The "downfall" seems to have began in 2018-2020 sometime, what AI was launched and popularized at that point that would have killed SO? LLMs were basically useless until GPT3 which appeared in middle-2020 sometime, after the downfall seemingly already had begun.
I'd call it significant that the number of questions halved within one year following the release of ChatGPT, the biggest relative or absolute rate of decrease in the timeseries.
Right around the time Google rolled out new search results removed many information based sites.
"How AI precipitated SO's downfall" would be a correct title then
Add it to the list:
- the downfall of junior devs
- bad hiring market
- layoffs in practically every sector
theres a ton of things where AI took credit for a trend that had already started before it started being even halfway capable.
People love simplistic narratives, i usually don't mind but this is just ridiculous. AI hate is gently overtaking AI hype as the most stupid thing around
It's not really a bell curve. There was obviously a downwards trend from 2016 onwards, but 2023 definitely precipitated the fall to zero. Without AI they might have lasted at least a couple more years, or the activity might have stabilized to a new floor greater than zero.
You mean the fall to a thousand questions per month. Now that the volume is low enough someone has a chance of looking at every single one of them, maybe the StackOverflow community can finally collaborate in peace, safe from the onslaught of questions that could be answered by reading the documentation.
https://postimg.cc/n9nZGLmb
Goodness of Fit 0.911, Kurtosis -0.849, Skewness: 0.073
It's very much a bell curve
Just because it's approximated by a bell curve doesn't make it a bell curve. There are quite obvious separate phenomena shaping the curve at different times.
> just because it's approximated by a bell curve doesn't make it a bell curve
I'm going to assume this is bait...
Bell curves are probability distributions. This is a time series, so it can’t be a bell curve. It just has the same shape.
Are you discarding the utility of Gaussian functions in analysis simply because the independent variable is time? A Gaussian curve can be used as a descriptive model without claiming that the observations themselves are a probability distribution
The fit does not prove causation, but it does show that the decline was already well described by a trend that began years before generative AI. If the claim is that 2023 created a separate structural break, it's different claim then the title describes
I am not sure. I think SO died way before AI and that graph seems incorrect too.
> Without AI they might have lasted at least a couple more years
Nah, their decline was already readily apparent before AI. You only need to go through old discussions and other people noticing it. AI may have accelerated the decay, but the decline happened already largely prior to AI.
this. Thanks for pointing it out, I fell for "oh it was just AI" at first.
This isn't really a bell curve.
check my other response, it's very much a bell curve, statistically speaking
https://postimg.cc/n9nZGLmb
> statistically speaking
That's a very big word you're using there for what is basically making shapes out of clouds. A bell-curve is the amortised function of a random variable with a mean and standar deviation. What does that have to do with a timeseries dataset?
A bell curve is not an "amortised function." Amortization applies to accounting and algorithmic time complexity, not probability distributions. You're likely thinking of a Probability Density Function (PDF). If you are going to police terminology, it helps to use the correct words. Second, fitting a curve with an R^2 of 0.911 is the exact opposite of "making shapes out of clouds.
> A bell-curve is the amortised function of a random variable with a mean and standard deviation.
The general notion of a bell-shaped curve is broader than that. Wikipedia has a reasonable overview: https://en.wikipedia.org/wiki/Bell-shaped_function
> “typically continuous or smooth, asymptotically approach zero for large negative/positive x, and have a single, unimodal maximum at small x.”
You don't just fit a Gaussian distribution to a timeseries dataset. That's not what a Gaussian curve is designed for at all. https://www.explainxkcd.com/wiki/index.php/1725:_Linear_Regr...
You are confusing a Probability Density Function (PDF) with a phenomenological curve fit. No one is claiming that time is a random variable drawn from a normal distribution.
> No one is claiming that time is a random variable drawn from a normal distribution.
You are doing that implicitly by fitting a Gaussian curve.
Fitting a mathematical function to a dataset does not implicitly adopt the ontological baggage of probability theory. That is a fundamental misunderstanding of applied mathematics
Stackoverflow did it to themselves by having incredibly unhelpful users
If that was the case, the graph would never have gotten to the heights it did.
What happened is that as the corpus of useful info increased, the need to pose new Qs decreased. AI much accelerated that decline by making available an 'oracle' trained on that corpus.
The users being generally unhelpful wasn't an issue for them, since they were still significantly more helpful than users anywhere else on the internet. Reddit was and still is filled with completely unvetted answers (on pretty much all topics not just programming), Quora was/is a joke, Yahoo answers had some funny posts I guess but nothing you could actually learn from, what else really was there? Before AI, Stack Overflow was as good as it gets.
It just accelerated the trend, and I am sure that reddit took over for a lot of new users. The different problems with SO has been well documented.
And they killed maybe one of the most side features of it : https://meta.stackoverflow.com/questions/415293/sunsetting-j...
So yeah metakill your own brands with stupid policies.
It's both. Users tolerated the hostile environment to an extent as long as the site was still the best way to get useful information. When LLMs came out, that was no longer the case.
There was also the pattern of "closing as already answered" with an answer from 6 years earlier which wasn't actually answering the question when you dig into it. Certainly in the code stacks.
Definitely this. The moderators seemed to have the Lock Question button connected to their dopamine pathways.
That increased at the same rate of lazy/stupid users.
The graph starts falling shortly after 2020. AI certainly contributed but Stack Overflow was dying without it.
So... nothing that it wasn't already doing to itself? There's no one drop where "AI got into the market", SO had been declining steadily for years. I actually expected this post to be about how SO survived by selling its internal organs to AI.
Now do a graph for the money.
https://www.wired.com/story/google-deal-stackoverflow-ai-gia...
> There's no one drop where "AI got into the market"
Sure there is, number of questions halved from 100K in first of November 2022 to 50K exactly one year later.
Some of the pre-ai decline in questions might just be that they had filled out of alot of the question space. What might be more interesting is the traffic graph as it would be possible to have a decline in questions but still have traffic rising to the existing ones.
Today I'd expect even when someone is googling and a question is relevant that most people will just read the AI overview.
AI didn't start the decline, it just finished it.
ChatGPT was released in Nov 2022, and frankly wasn't very good originally. The SO decline started occurring almost two years ahead of that, and was already on a sharp decline before ChatGPT shipped, and certainly before ChatGPT actually became good.
This is revisionist history. People told SO that they were leaving for YEARS because of how incredibly toxic it had become. It was already giving outdated answers before ChatGPT shipped, because new questions/potentially updated answers were [Closed] [Dupe] immediately.
Their answer was essentially "We aren't a Q&A site, we're trying to be a knowledge base! So closing all questions on a Q&A-stylized site, and extremely abrasive moderation, is working as intended."
They entirely did this to themselves. The community was toxic, their policies were toxic, and they didn't listen when warned as such repeatedly - just doubled down.
Two years before ChatGPT is shortly after Joel and Jeff sold Stack Overflow to private equity.
The Monica affair was one of the first symptoms.
The graph is not presenting a narrative, did you mean to reply to someone else who is presenting the "revisionist history"?
In 2023, Stack Overflow had already started making unpopular pro-AI moderation decisions, and in 2024, they started mass banning everyone who deleted their questions and answers in protest. I don't think it's wholly incorrect to say "AI killed Stack Overflow" when the death blow came from crazy pro-AI decisions from the admin.
Did they cry about it? No right? Don't apply your own standard then judge them about it, petty people.
Stackoverflow aimed to be a knowledge base. And knowledge base has a ceiling limit. They simply reached the point that almost all questions (regarding the knowledge) were asked for them. You can argue that newer or niche libraries or languages knowledge is still lacking there, but I have never seen them getting closed, just not answered.
Interesting that you can see COVID in the graph
The stackoverflow moderation is the reason I do not post on it. You have middle party with no competence on the technology trying to do useless moderation.
Instead I directly go on the project github page and ask the question directly to the mainteners.
I’m surprised the site had that little activity, relatively, around 2010. That’s when I was using it the most. It seemed plenty large to me at the time… I can’t imagine the experience was that great a few years later. I wasn’t really paying attention by the time it reached the peaks later that decade.
I always found the format of the side obtuse and the culture not very welcoming. My most popular answer ever was something about JavaScript from 2008 or 2009, and to this day, people come in and say “this isn’t the way to do it, this is outdated“. No kidding, but every new question about that gets closed as a duplicate.
The hostile moderators killed stack overflow.
AI might've delivered the final blow, but Stack Overflow was in decline LONG before LLMs came on the scene.
I read a great article not long ago outlining the full series of events and changes that led to its downfall. I wish I could find that article, but I've forgotten where it was.
I never understood the point if having the unfathomable churn of thousands of new questions per day. The value of SO to me was always a knowledge base with reputation mechanics, and that did not change. I still default to searching before asking AI.
Honestly, I think that's a good thing. A lot of questions were either duplicates of existing questions, or close derivatives of them. If I had to guess, probably 90% of SO questions already had a solution somewhere on SO. AI surfaces these solutions much quicker, so you don't have to ask. Novel questions or bugs that can't be answered or fixed by AI still get asked, and mods have less spam to deal with. I fail to see the issue here.
As others said it wasn't just AI but their excessive moderation
I miss Joel Spolsky's writings especially in this dark age of AI.
Can it please do the same to reddit?
Should read what Google did to remove sites from search results.
AI and ridiculously aggressive moderation. If it had been a more welcoming place it probably would have lasted longer.
When an ecological shoes company pivot to AI, I wonder why StackOverflow executives don't pilot for AI now.
There's still a question of what that would do for them. I haven't used SO in years because I haven't needed it in years.
That's primarily due to AI answering my technical questions.
Sadge
Slightly accelerated their decline. You have a drop around chatgpt release then the slope returns to its previous pace of decline.
SO's downfall started way before AI. A decade or so ago it was always full of interesting questions, people were giving detailed answers, there was sometimes some debate in the answers, etc
And then it started being stupid questions. People who clearly had barely tried anything and just rushed to SO with a half baked question. Answers were just pointing to another thread that already provided the answer. It definitely started before LLMs. I think it lined up with the aggressive "learn-to-code" push.
They had a good run!
If there is not allowed to be duplicate questions isn't it by design that as the site and industry matures the number of posts go down.
Except for covid, it seems the decline was already there.
Yep, SO was dying before the GPTs - in some ways it was baked in the original SO design - to become the canonical source of information about programming stuff.
Many people talk about the negativity, and they are right, but I think the reason more than anything is the waiting. On SO a good question might get answered in minutes (if it was easy and someone was karma farming) but it could be days or weeks for general purpose stuff; compare that to a few seconds for an LLM its a no brainer.
Somehow I feel that whole idea was 50% broken from start. In field changing as fast as programming you simply cannot have canonical source that is done. There are areas which are constant or slow moving. But too many others are in churn. Trying to apply same rules for both is clearly insanity. And this is really the failure of SO...
Maybe if in some cases stuff actually got deprecated and points did actively decay it might have worked... But can't remove power from those who had it for years...
You also have to think that SO was very early in gamification and changing things around that never got political will afaict - people love their points.
The decline you're talking about is roughly 168 to 145, or about 2.2% per year over 5 years.
That's hardly a death sentence. More likely just the gradual adoption of higher level frameworks and languages with less ugly parts.
Indeed, decline appears to accelerate significantly in 2023 so seems likely that's AI helping things along.
The trend might've stalled or even reversed if it weren't for AI, we can't just assume the same end was written in stone.
The company my friend works for has a slack channel for help with code, like an internal stackoverflow. It’s almost inactive now.
I asked to see one of the questions from 2024 - it could have been solved with one LLM search.
We have eliminated a whole genre of peer to peer communication.
I noticed the same thing. Another thing that seems to have disappear: lunch and learn focusing on the tech stack. And I mean specifically about "how to do this in <programming language", not product/company focused.
same story for blogs
This is IMO wrong. StackOverflow died way before AI - and way before 2020 too. I think it had a peak time of only 3 or 4 years. It was created in 2008, and I would reason it took a few years, say, up to 2011; then it was semi-okish up to about 2015, roughly. Then it declined.
It still has some value today, as sometimes you can find useful information on SO, but its peak days are long over and I don't see how it can manage to come back, with or without AI slop. It would basically require a lot of re-design and some things that never worked, such as the karma system, should be changed. Also moderators - they kill sites. That happened to reddit - I gave up after censor-mods constantly restricted everyone.
I disagree. From what I see, company specific stackoverflow like mailing lists or slack channels are also dead.
The normal day to day devs just don’t have the need to go to stackoverflow anymore.
One would have to explain both consequences or dismiss it as coincidental.
FWIW I rarely have the need to ask questions at the programming level to anyone anymore. It’s just not the type of thing I bring up or anyone else. We now talk about architecture and company direction.
Finally, something good done by AI against these modern-day dictators and pharaohs.
Careful - the LLM companies are also modern day dictators and pharoahs.
Wikipedia seems to be going the same way.