35

Software industry blogger Gergely Orosz claimed that "Stack overflow almost dead" in a 15 May 2025 post.

Today, Stack overflow has almost as few questions asked per month, as when it launched back in 2009. A recap of its slow, then rapid, downfall.

They base this on a Twitter post citing data from SEDE metrics of number of questions asked per month. The post also summarises a timeline of Stack Overflow actions on moderation, with commentary.

Gergely points to LLM use as a main factor drying up questions. But they also write:

2014: questions started to decline, which was also when Stack Overflow significantly improved moderator efficiency. From then, questions were closed faster, many more were closed, and “low quality” questions were removed more efficiently. This tallies with my memory of feeling that site moderators had gone on a power trip by closing legitimate questions. I stopped asking questions around this time because the site felt unwelcome.

This is interesting context for EmmaBee's summary of Meta's insights on "closed (and potentially useful) questions on Stack Overflow.

My questions: do you agree with the conclusion that "Stack Overflow is almost dead"? Do you agree with that the rise of LLMs, and SO feeling unwelcome, are the cause? Is number of new questions the right metric for "life" or "death" of SO?

My hidden agenda: I think Gergely's commentary will be interesting to Meta Stack Overflow readers, so I want to make it known here.

47
  • 39
    It's just another article about how SO is drying up. Honestly, there's nothing new in that article... I could say I disagree with what he wrote, but the numbers don't lie about SO's traffic. So, I don't really see the point of bringing it up here. Commented May 15, 2025 at 21:49
  • 15
    (Also, that [welcoming] tag leaves a bit of a sour taste. It's consistently been used as a stick to whack at SO's moderation and long-term users) Commented May 15, 2025 at 21:50
  • 22
    I stopped reading after I reached the pay wall, ironically the entire article wouldn’t be possible, without the tools provided by SE to generate the required graphs. I just see it as yet another user upset that their low quality question which can be answered incorrectly by a LLM was not well received. Commented May 15, 2025 at 22:05
  • 12
  • 28
    In short: outside voices like this have no standing to assess whether Stack Overflow is "dead", because they have no understanding of what we seek to accomplish. They impose their own metrics and quality standards, and every other kind of "should", upon us. Commented May 15, 2025 at 22:25
  • 29
    He says "ChatGPT is faster and it’s trained on StackOverflow data, so the quality of answers is similar." Of course it's faster, so is the Magic 8 Ball. But the quality of answer is not similar. If it were true, there would be almost no questions on Stack Overflow, and almost no one here to answer them. It's obvious that he is just another disgruntled former user with an axe to grind. Commented May 15, 2025 at 22:34
  • 26
    "This tallies with my memory of feeling that site moderators had gone on a power trip by closing legitimate questions." Moderators rarely close questions. Most of it is done by normal users. If improved tooling led to more questions closed, it doesn't that the tool is abused but rather that before it a lot of questions that should have been closed weren't. Commented May 15, 2025 at 23:11
  • 12
    We can't deny that LLMs made people post questions less on Stack Overflow, but this doesn't mean the site is dead. Commented May 15, 2025 at 23:13
  • 26
    "ChatGPT is polite and answers all questions, in contrast to StackOverflow moderators" Again, moderators rarely answer questions. However, this statement shows that the author cannot be trusted with what he's saying. The site can only function if people not only ask questions but also answer them. Commented May 15, 2025 at 23:15
  • 19
    It's not entirely a bad thing that the number of new questions is dropping. This is included in the design of the site because there cannot be an infinite number of questions. At some point most of the questions would have been already asked and answered. Commented May 15, 2025 at 23:17
  • 35
    "I'll certainly miss having a space on the internet to ask questions and receive help – not from an AI, but from fellow, human developers." Well then, who's to blame? Could it be people that are claiming the site is dead and are driving developers away from it? Commented May 15, 2025 at 23:18
  • 10
    I still have a problem like the last question that attempted to point out the “Stack Overflow decline”, the data does not take into account all of the low quality garbage that gets deleted. Once deleted the tools that exist only consider the questions that haven’t been deleted. Commented May 16, 2025 at 0:09
  • 29
    I do wonder how someone, that has contributed 10's of posts, hasn't contributed for a decade, and never contributed to Meta, can make these claims, when they've not really experienced the site. As Dharman mentioned in their comments, they don't even understand the difference between curation and moderation, which is fundamental to the site. It's like confusing the people that reported a crime with then police officers that responded... They are completely different people. Commented May 16, 2025 at 12:27
  • 14
    @ThomA I'd argue instead that that's how the vast, vast majority or people experience the site. We on meta are a miniscule, tiny fraction of the contributing population of this platform. We're vocal, we understand it in depth, and we care a lot– but we're small in the face of the overwhelming majority of people that visit, ask, answer, and almost certainly never look at meta or engage with the site on any level deeper than that. Commented May 16, 2025 at 17:00
  • 14
    I've seen communities shrink by more than a factor of 10 and continue on, perfectly healthy and happy, in the long run. Yes, it's uncomfortable. Yes, it should justifiably trigger intense reflection by all parties, and deep change to follow. No, it is not the end of the line. You'll know the end when you see it. We're not there yet. Commented May 19, 2025 at 16:59

12 Answers 12

77

I would argue that Stack Overflow is "dead" in same sense that Wikipedia is "dead": In some sense they are, but in a more important sense they aren't.

I started to edit Wikipedia 22 years ago. At that time many articles I wanted to find were missing, and I actually found myself creating new articles from scratch, on anything from my favorite band, to some small town I happened to visit recently. I made hundreds of edits to Wikipedia every year, and it felt "alive" and I felt part of a community. Today, virtually anything I search for on Wikipedia already exists, so I rarely ever need to edit any article, and I haven't created a new article from scratch in years. I've become more of a "user" of Wikipedia than a member of a "community". But while some might lament that the Wikipedia community is "dead" and blame it on toxic interactions and other things they personally experienced, the simple truth is that Wikipedia is thriving as a mature product used by billions, and by no means "dead". It's more alive than ever, despite the fact that I rarely edit new articles anymore.

Exactly the same thing happened to Stack Overflow. When it started, every question was new, and people (including me) enjoyed answering those questions. But as the years went by, people were mostly using the same systems and programming languages that have been in use for years (consider, for example, that JavaScript, Python and Linux are all 30 years old!), and after Stack Overflow has been in existence for a few years - basically every question was already answered well. So people, including me, discovered that you could find good Stack Overflow answers in a simple Google Search - and the need to write completely new questions, and to answer new questions, became rarer and rarer. This was all before AI, mind you - AI can find answers even better, but the existing answers were already pretty easy to find using Google. On many days, I found myself finding Stack overflow answers multiple times each day, but only ask a new question on Stack Overflow once every few months.

Does this mean that Stack Overflow has died? Of course not - it means it became mature, even "complete" in a certain sense (only requiring minor additions as incrementally new technologies are invented), and super useful. It only "died" if the only thing you cared about was counting new users, new questions, and new answers. Perhaps Stack Overflow the company cares about these metrics, but why should Stack Overflow users care? Does anyone still believe that Wikipedia's success or failure should be measured right now by the number of new articles added every day or the number of new users? At the current point of Wikipedia, the vast majority of users never create any article, and never even bother to create an account. This is a sign of Wikipedia's success, not death. Similarly, the fact that Stack Overflow has amassed a huge amount of good questions and answers and no longer needs many new questions or answers or users that create accounts - is a sign of Stack Overflow's success, not death.

26
  • 5
    Very nice read, but I still wonder how you quantify success then? The fewer new questions the more success? How do we really know that we have all the good questions there are? Or is the mere fact that people don't ask here anymore proof that no more questions can be asked? I might want to remain skeptical about it. Commented Jun 24, 2025 at 17:00
  • 6
    As another analogy, the Bible is a very successful book despite nobody having changed or improved it for 2000 years. It's successful because people read and use the information it contains. If people use Stackoverflow's information - not by creating an account or writing things but by Google search or ChatGPT having learned all its info - then Stack Overflow is very successful. Maybe it can't make any money this way, but neither does Wikipedia or the Bible "make" money, and they are still considered successful. Commented Jun 24, 2025 at 18:48
  • 2
    But @NoDataDumpNoContribution to answer more concretely: I consider Stack Overflow successful if questions I have and type into Google get answered with Stack Overflow information. True, with Google it was easier to know - I got back links, and for AI it's harder because it doesn't know where it learned. But I know that Stackoverflow has been in recent year the best source of such technical information, so I'm convinced much of the AI's knowledge was learned from Stackoverflow. So I consider it to be a success for humanity - like Wikipedia is. Commented Jun 24, 2025 at 18:51
  • 2
    Thank you for the comments. I'm not religious, so the bible example may be wasted on me, but I see how you are thinking. For you success is rather a binary thing, either it's there or it's not and StackOverflow is the living proof of a success. For me it's more a quantitative thing. I could imagine a better StackOverflow with more and better questions, that's why I'm not sure about the exact success that StackOverflow enjoys. I agree that one possible measure is how good general search quality is. But I'm not sure I can always estimate it well. Maybe I wouldn't even know what I'm missing. Commented Jun 24, 2025 at 19:38
  • 4
    Thank you for this mature, thoughtful answer that perfectly sums up my feelings. Every knowledge store starts out from empty and therefore grows (in terms of both content and users) extremely quickly initially, but ultimately reaches some level of saturation and the growth drops down to a far more gradual level. That is exactly why SE Inc.'s "line goes up" mentality for new user acquisition is insane to me, because a knowledge store literally does not and cannot work that way! Commented Jun 26, 2025 at 12:02
  • 2
    This answer also touches on the fundamental impedance mismatch between people who curate, and those who don't, which is the most critical issue facing SO at this time. I don't know how to address that issue, and since SE Inc. doesn't even understand the mismatch exists, neither do they. Commented Jun 26, 2025 at 12:06
  • 1
    @NadavHar'El Continuing the analogy: People think that in using an LLM they are talking to God directly ... but for LLM answers "the Devil is in the details". Commented Jul 15, 2025 at 0:38
  • 2
    This answer isn't categorically wrong but it isn't nuanced enough: yes some decrease in new questions makes sense as the most common ones are answered (for established technology). But in technology there are always new topics, new languages, new frameworks. I don't buy that there aren't many more than 442 [zig] questions. There are only so few because it's a young language and people don't ask many new questions anymore. The trend continues 6 months onwards: every 4.2 months the rate of new questions halves. That's definitely not because suddenly everything has been asked already. Commented Jul 15, 2025 at 2:06
  • @CorneliusRoemer I agree, but can't the same be said about Wikipedia (the other example I gave) – since Wikipedia is mature and people find its articles on the top of Google searches and regurgitated by ChatGPT, if a new language "zig" is invented, fewer people are expected to go directly to Wikipedia to look for it and rush to edit the article about it. I don't think this makes Wikipedia "dead" or a less important source of truth for those Google or ChatGPT front ends. Commented Jul 15, 2025 at 8:01
  • 2
    @NadavHar'El Wikipedia is a bad comparison because new article creation rate there has only decreased by a factor of 2 from its peak over a time of 15 years - it's essentially stable since LLMs became a thing. SO collapsed by a factor of 10 in barely 2 years and it keeps decreasing at roughly the same rate. I'm not saying that the argument is logically impossible to be true that "all the questions have been answered", it's just not backed by the data which shows a fast decrease in new questions that doesn't slow down. stats.wikimedia.org/#/en.wikipedia.org/contributing/new-pages/… Commented Jul 15, 2025 at 8:57
  • 2
    @NadavHar'El I wanted to put a link to the wikipedia stats graph but was out of characters, happy to share the useful graph: stats.wikimedia.org/#/en.wikipedia.org/contributing/new-pages/… Data beats personal experience. How can nothing be missing from Wikipedia? What about recent events? New products/companies/concepts? "My involvement with Wikipedia really did go to almost zero." Data show others filled your shoes: stats.wikimedia.org/#/en.wikipedia.org/contributing/… Commented Jul 16, 2025 at 12:11
  • 3
    @NadavHar'El I think the real telltale of deadness is whether SO/Wikipedia are still useful for newly appeared concepts. Wikipedia to me looks healthy for things that only became a thing recently, like LLMs: en.wikipedia.org/wiki/Large_language_model - I don't think the same can be said for SO. There's almost nothing about [ruff] (68 questions), while [pylint] has 1788 questions. Ruff has been around for almost 3 years. It's more popular now than pylint, yet it has barely 3% of the questions. Commented Jul 16, 2025 at 12:20
  • 1
    @CorneliusRoemer Wikipedia is definitely not healthy for anything to do with contemporary political events. Commented Sep 16, 2025 at 23:05
  • 3
    So stack overflow is not dead, it is ... complete. Commented Nov 17, 2025 at 19:25
  • 4
    It feels like lots of people are in denial trying to redefine the meaning of success on the fly to match the current constraints. Before chatGPT there were no questions about it: traffic is growing, more and more people are actively engaging. With chatGPT it plummeted not because all questions were answered, but because people found a better and safer place to ask them. Sorry folks, but the reality had shifted. And I'd argue that toxicity of SO is not among the least contributing factors here. If nothing changes it's gonna be a grave yard bubble with a few core believers sooner than later Commented Dec 24, 2025 at 15:23
55

do you agree with the conclusion that "Stack Overflow is almost dead"?

It has in some senses been dying, and that's fine. It can still carry on for a long time. Being "alive" indefinitely isn't the point. It was never the point. The tour says:

With your help, we're working together to build a library of detailed, high-quality answers to every question about programming.

Everyone who seeks to accomplish anything ought to be prepared for the possibility of succeeding. Stack Overflow has surely not found every question that can support high-quality answers yet; but in also accepting questions that don't support high-quality answers, it has managed to accumulate more than three questions about programming for every Wikipedia article about literally anything. Questions aren't the limiting factor.

Do you agree with that the rise of LLMs, and SO feeling unwelcome, are the cause?

These are absolutely causes of the decline in the rate of incoming questions. However,

Is number of new questions the right metric for "life" or "death" of SO?

Absolutely not.

When your goal is to have a "library of detailed, high-quality answers", you don't die because you run out of new attempts at questions.

You die because people stop trying to improve answer quality (edits), and stop trying to organize the library (identifying duplicates, deleting redundant answers etc.), and because people stop trying to find things there (no good direct measure, but view counts are a start).

If edits and other curation are on the wane, that has more to do with people losing hope, and growing impatient with the company.

If people stop writing site:stackoverflow.com in search queries and instead use an LLM that was trained on Stack Overflow data - well, it's too bad (for them, too) if they trust the LLM without verification, but the LLM is largely working as a cache that the site simply can't account for in the statistics.

12
  • 10
    While I agree with you we should be aware that effective site curation can only be done by 2k+ users. With fewer new questions, it's getting successively harder for new users to earn rep and become active curators. In the (far) future, we might run into a recruitment problem. Commented May 23, 2025 at 8:16
  • 1
    @Friedrich well, we can change that standard. It doesn't even have to be reputation-based. And we can explicitly train the skills, too. But personally, I think the time has come to start over, which is why I'm at Codidact. Commented May 23, 2025 at 17:15
  • 9
    👆This comment changes the entire meaning of the answer. One way to read this answer now is that Karl has a vested interest in letting SO die so that more users come over to the place where he is a moderator. If the community is faltering then blaming SO the company seems a strange target. And suggesting that SO is in decline because it has somehow "succeeded" in it's mission seems laughable when outside of the SO bubble folks blame aging content and excessively strong handed gate keeping. Commented Jul 2, 2025 at 8:59
  • 4
    @ARF You imagine multiple contradictions where there are none. I blame the company because they make the site less pleasant to use for the existing community. There is no contradiction between success as a service and decline in interaction; I clearly explained why. Being a moderator at Codidact doesn't give me a "vested interest" in site popularity or traffic; the site is operated by a non-profit foundation and I'm not a member of it anyway. But I would personally like to see the project succeed for its own sake, and to make a point about the model that SE and Codidact share. (1/2) Commented Jul 2, 2025 at 21:51
  • 5
    (2/2) My entire point is that I do not care in the slightest about what "folks outside of the SO bubble" think because they are completely missing the point about what Stack Exchange is supposed to be, and because they have no right to say what it should be. The gate keeping is deliberate to keep them out, because success at the site's goals requires keeping them out if they aren't going to play ball. Again: we get to decide what our goals are - they do not. As for "aging content", this is simply misguided. Many things about computers are equally true today as they were in the 60s. Commented Jul 2, 2025 at 21:54
  • 4
    Yes you want to believe that folks left because "the company made the site less pleasant to use" and absolutely refuse to listen to the millions who left because "they are completely missing the point". And yet those who left don't talk about the company at all. They talk about it's toxic community. The reasons cited for others leaving are never about the company at all, but entirely reasons in the hands of it's community. The idea that anyone who disagrees with you must simply have misunderstood is the biggest delusion of stack overflow. Commented Jul 3, 2025 at 16:29
  • 5
    "Stackoverflow's mission cannot be "complete" if it's content has already aged out beyond usefulness." But it hasn't. "But you know this and want others to go join the thing where you are an admin." This is a bizarre conspiracy theory and I must ask you to cut it out with the personal attacks. Commented Jul 3, 2025 at 18:36
  • 3
    It wasn't intended as a personal attack, nor was it a conspiracy theory. I was merely rephrasing your own comment. You've suggested we should start over on a competetor. This directly conflicts with your assertion of "mission accumplished". Either the mission was a succsess, in which case it doesn't need repeating, or it was a failure in which case we should start over. Commented Jul 11, 2025 at 21:54
  • 5
    @ARF I suspect the reports you've seen are from the unfortunate users that the UI encouraged to post questions without enough guidance on what is needed for a question to succeed on the site. Then, when their question gets closed because it's off-topic or missing kdy details for too long, they think it's a problem with the community when really it's a problem with the onboarding on the website. (I mean, inevitably some users will ignore most or all of the instructions and post bad questions, but it doesn't seem like this is anywhere near becoming the main problem yet.) Commented Jul 20, 2025 at 1:05
  • Could you perhaps suggest view stats we could look at? On SEDE or elsewhere? Commented Jan 11 at 14:58
  • You quoted "Is number of new questions the right metric for "life" or "death" of SO?" and errantly answered with "Absolutely not." Despite what you may believe its charter to be defined as, SO is fundamentally an interaction site and interactions here are initially driven by questions. If questions dramatically die down (they are), then interactions dramatically die down (they are) and thus SO dramatically dies down (it is). Commented Jan 17 at 22:13
  • 1
    I stopped writing on main SO because of the community mainly. Partially because of the parent company - but I think the two are related. Quite a few things would probably be better if the original founders were still in charge. 🤷‍♂️. Now I just pass by, leave a few comments or read some posts .. and then forget about the site for a few weeks :-) Commented Jan 19 at 15:12
32

I've often felt that the issue is a lot more complex than it seems at first glance. Things like attrition amongst core users (and the momentum built up in the early days finally running out), drama in the network causing big drops in engagement amongst the most engaged and such are internal issues that I feel are causes.

Interestingly, one of my personal theories is over time, the company has tried to accommodate folks who want the benefits of a well curated knowledge base without the constraints, or want SE to change without engaging in the community directly. We've carved out pieces of ourself to try to serve folks who turn their nose up at us, perhaps.

I find the stack exchange timeline an interesting reference here - and I'd argue a few key things happened in 2017 and 2018 - or even earlier. The company lost its way with the community, and was both trying to be profitable at any cost and find the next big thing - a combination of downsizing, issues with company culture, and conflicts with key community members kicked off a period of attrition.

Externally, we're in a period where large tech companies, as well as contracting companies are doing large job cuts in the belief that AI can replace a lot of headcount. Incidentally, companies hired heavily during covid and this coincided with a period of user growth.

'Blaming' LLMs is simplistic, and complaints about SE 'moderators' being too quick to close questions is basically our equivalent of eternal september. There are things the company can do, but an excessive focus on the more vocal set of outsiders at the detriment to folks who use and thrive here probably has hurt more than helped.

As much as the company sees re-engineering the design of the network and new products as a future, I believe our survival lies in re-energising and rebuilding communities and attracting the sort of people who thrive here rather than disavowing our unique selling point - that we're quality focused and attract experts and trying to turn the network into something else.

I also find that a website - whether its a forum or a Q&A site is easier for finding knowledge artifacts than chat. Stack Exchange might be ailing but perhaps recovery isn't out of the question.

6
  • 4
    "company has tried to accommodate folks who want the benefits of a well curated knowledge base without the constraints," At some point, Google made it very easy to search for Stackoverflow answers directly in its search bar, without logging into stackoverflow. New users did not need stackoverflow accounts, didn't need to "ask a question" or "vote", they just benefited from the quality answers. Some may call this "death" of Stackoverflow, I would call it - success! Commented Jun 24, 2025 at 13:36
  • 2
    @NadavHar'El the original idea of SO was that most users would be landing here from web searches. This isn't an emergent property of the system, it's baked in its design. The site even allows unregistered users to post answers and make suggested edits just so the barrier to participation is lower. An account has never been strictly required. Yet, that's not what JG is talking about. The existence of anonymous users has never really been a problem to my knowledge. Commented Jun 24, 2025 at 13:53
  • 2
    Maybe the idea of Google searches was baked into the design (and they certainly didn't do anything to actively prevent it), but years later people seem to think that "new questions" and "new answers" and "community" and "accounts" and "interactions" are valid metrics for this site. I claim they are not. Stackoverflow isn't "dead" because it "lost its way with the community" (which is what JG said). Commented Jun 24, 2025 at 14:23
  • 1
    Well, I'm not talking about anonymous users - I'm thinking of those folks who claim SE and its model is toxic and want us to move closer to a different model, as mentioned deeper down. And to me the 'hard to quanitify' parts of community, that people hang out here both to do Q&A and to an extent use the social and community parts of it when they arn't is essential. To me, chat and meta being quiet is as much of a sign of an ailing community as poor QPDs for example, as would people leaving the network or roles within it due to disagreements with the community. Commented Jun 24, 2025 at 14:33
  • So to me - SE dying would be cause it lost its way with the community as much as by death by AI or other subsitutes. Commented Jun 24, 2025 at 14:33
  • 4
    @NadavHar'El "years later people seem to think that "new questions" and "new answers" and "community" and "accounts" and "interactions" are valid metrics for this site. I claim they are not." neither does the answer claim that at all. The company (a.k.a. SE) certainly did burn through a lot of goodwill and patience of the community. That's not even debatable - many engaged users have stepped down or even completely left the sites. That by itself is not "death" of the sites. But waving our hands and declaring it a non-factor is also not what we should be doing. Commented Jun 24, 2025 at 14:43
28

With 6 months more data than the original post by Gergely relied on the trend of rapidly decreasing new questions continues. New questions are still dropping exponentially. The current decay rate (exponential fit to last 6 months) is a halving every 4.2 months (a factor of 7 per year). Of course, the decay rate will not stay like this forever but time's kind of running out. (I exclude the current month.)

New questions asked with a log y axis

enter image description here

Update (2026-01-04): The trend continues mostly unabated with December 2025 having only 21% of the questions asked in December 2024 (3862 vs 18029) and that doesn't account for roomba etc deleting some percentage of the still counted December 2025 questions. The exponential fit said it would be down to 14%. At the current rate we'll be at 800 questions for December 2026 and 160 in December 2027. At that point it's 5 new questions day. SEDE Query

0
20

do you agree with the conclusion that "Stack Overflow is almost dead"?

No. Traffic is declining, but I don't think it's at that point say "almost dead". There still are quite a few questions coming in, and users contributing useful content (including new contributors).

Do you agree with that the rise of LLMs, and SO feeling unwelcome, are the cause?

From what I see people say for themselves on Reddit at least, a significant number of people feel like SO is unwelcome, and like that LLMs don't make them feel that way when they have questions. I have related-ish thoughts in https://meta.stackexchange.com/a/384378/997587.

Is number of new questions the right metric for "life" or "death" of SO?

I think it should be part of whatever metric that is. I'd want voting activity to be considered too. Even if we somehow answer every question under the sun, as long as there are people in the field of software development, voting activity gives some picture of how many of those people encounter this platform and find its content useful. "How reusable is our content?" That's part of the core goal here.

For me right now, I think SO is dead when the people who are looking for what we're designed to provide (a community-built library of Q&A) no longer find us able to provide that.

Or otherwise, if nobody finds this platform/library useful for anything. I'm referring to reddit a lot (apologies), but I think it's worth doing since in my mind, they represent the masses, and on reddit threads that discuss declines in SO traffic, it's common to see a highly voted comment saying something like "but if SO dies, who will feed the LLMs?". Of course, we're not the only ones who do, but I suppose even if people hardly use SO directly, what we do can still contribute to making the world a better place... ? I'll leave a question mark there since I don't know if I can really stand behind LLMs as a "frontend" to SO making the world a better place (environmental impacts and all that).

Tangent: voting really matters. The other complaint I see a lot on reddit is about outdated information being the top answers. Ideally, voting means that the most useful stuff is at the top. Is it that these people aren't voting? If so, why? Is it that they can't vote? Or enough of them can't vote to change the sort order? These are questions I think are important to engage with. I'll take the liberty here to plug some of my thoughts about voting like I usually do: https://meta.stackexchange.com/a/386224/997587, Add an option to sort answers by the viewer's previously cast votes, https://meta.stackexchange.com/a/393604/997587.

Apologies for the rambling.

6
  • 3
    Such a minor part of your excellent answer, but on the "outdated top answers" complaint, I'm really partial to the trending sort order; off the top of my head, I feel reasonably strongly that that should become the default sort order to address precisely that complaint. Commented May 16, 2025 at 16:54
  • 2
    @zcoop98 trending sort order is horrible most of the time, especially in lower traffic tags. If you land on the question with plenty of answers and where there may be newer ones which will surface with trending sort, then it is not hard to switch sort. Most upvoted works the best for the most questions and it is not vulnerable to sock voting. Commented May 18, 2025 at 12:57
  • 2
    @DalijaPrasnikar I can only speak to my own experience with it, which has been nowhere close to "horrible", but it's definitely in primarily higher-traffic tags. The one point I'd argue specifically is that the outdated top answers problem is, in my understanding, most common in higher-traffic tags, which is precisely the place that trending shines because of the higher voting activity. Commented May 19, 2025 at 20:40
  • 1
    "For me right now, I think SO is dead when the people who are looking for what we're designed to provide (a community-built library of Q&A) no longer find us able to provide that." That is also happening right now. LLMs have taken a huge chunk of traffic away from SO. Commented Jan 13 at 14:33
  • Addressing a minor question you pose: "Is it that these people aren't voting? If so, why?" <- perhaps a combination of three factors: 1. If users don't need to ask new questions, they are much less likely to register. I mean, would you really register to vote on questions? 2. Even a user who has registered, but only ever asked a question or two, may not feel it's their place to decide which answers are good or not. 3. ... especially when looking at an old question with a large number of votes which has supposed "decided" the matter. Commented Jan 15 at 21:05
  • SO feeling unwelcoming compared to LLMs is a strong point. It's arguable whether the high standards of SO is a pro or con, but its certainly part of why LLMs can be more approachable. Questions on SO can be rejected for reasons that are unintuitive to the average user. *Posts get closed as being not about programming. *Questions get closed as duplicates if the answer can be found elsewhere, even if the linked dupe isn't an exact match. *Unanswered questions with a score of 0 get auto-deleted. Commented 2 days ago
17

Sadly, I believe it is somewhat true. It's not quite dead yet, but it's on a fast track.

While the number of new questions dropping isn't concerning on its own, it creates a negative feedback loop. No new questions means contributors have less incentive to stick around and answer questions. Fewer answers mean there is less incentive for people to ask questions.

More than the number of new questions, the number of visits is a more important sign of how well Stack Overflow is doing. This has been pretty good until ChatGPT came along. LLMs have taken a huge chunk of traffic away from Stack Overflow, together with search providers offering summaries. These cause Stack Overflow to lose traffic and income from ads. Without revenue the site cannot stay alive.

As much as we always wanted to build a repository of information and we compared ourselves to the likes of Wikipedia, we were always very far from this goal. Unlike Wikipedia, people don't write an article together. They reply to a problem statement. And as such, the Q&A format is something between an encyclopedia and a forum. It's easier to find an answer here than on a forum, but there is too much noise and clutter compared to an encyclopedia. LLMs excel at solving this problem as they can summarise and quickly provide an amalgamation of all answers without the fluff. If they can do it reliably and accurately close to 100% of the time, then people will have no reason to come to Stack Overflow. But of course, Stack Overflow enabled that as it provided LLMs with the training data.

1
  • This is a key point. Less Qs and As aren't a problem, less search engine eyeballs to roll on ads is. I don't think the company can increase the number of active users in any impactful way, but I do hope they somehow manage to redirect some of the passive traffic to pay the bill. Commented Jan 26 at 12:43
10

I want to talk not specifically about SO, but about all websites like it.

We had, have, and will continue to have sites like SO. Why? IMHO:

  1. We always need teachers to tell us what is "good" and what is "bad."
  2. We need a way to acquire new knowledge.

As far as I understand, the current mechanism for training AI creates a "knowledge loop." Since AI outputs combined data from old sources, it generally doesn't produce truly new data — yet the AI will continue to learn from its own generated content. Because of this, I’m sure that real humans are needed to contribute original data that hasn't been discovered or documented before.

There will always be business-specific problems that haven't been solved yet, or solutions that exist but aren't open source. How will AI deal with that?

In one of my previous jobs, I had a task to create functionality that generated barcodes for 10,000 different items — complete with names, prices, and the ability to adjust almost every setting with a live preview. All of this had to be generated in under 10 seconds across Firefox, Chrome, and Safari. Good luck asking an AI to architect that from scratch!

I’m not saying AI will never be able to handle tasks like this, the point is that human engineers are still necessary. We need to talk to each other, help each other, and add new knowledge to the "global database" so that others can eventually find those answers via AI.

On the other hand, it is probably a good thing that AI can answer basic questions. We don’t need a 100th thread on "how to deep clone an object" or "how to change a button color." This allows us to spend our time on more complex and interesting problems.

People almost always choose the easiest path. For example, a student will usually ask a teacher a question to get a quick, understandable answer because a teacher can provide tailored examples. Does that mean we don't need books or libraries anymore? Of course not!

The way we share information has simply evolved, and it will continue to do so. We shouldn’t be afraid of these changes, instead, platforms must evolve to adapt to them.

For me, SO is not dead. It is just becoming a more specialized hub for real engineers to tackle the complex tasks AI can't handle yet

3

do you agree with the conclusion that "Stack Overflow is almost dead"?

Strictly speaking, it's hard to agree with a statement that allows for that many interpretations. Yet, my guts tend to agree. Or, I would try to put it in a bit more strict way: Stack Overflow, the way it was conceived by founding fathers, failed to reach its goal, and - assuming the trend - hardly expected to reach it in the future.

Do you agree with that the rise of LLMs...

This looks like an established fact, no matter if I agree or disagree

... and SO feeling unwelcome

Although this fact drives gatekeepers crazy, but if anyone would try to impartially consider the following incomplete list of closure reasons,

  • Duplicate
  • Needs details or clarity
  • Needs more focus
  • Opinion-based
  • Needs debugging details
  • Seeking recommendations for books, tools, software libraries, and more
  • Not reproducible or was caused by a typo

they'd have but to conclude that it covers up to 90% of newly asked questions.

And however noble (and as of now - unreachable!) goal dictates these rules, a closed question makes any feeling than welcoming.

Still, the cause isn't lost yet. Although the traffic is declining, Stack Overflow still has a unique competitive advantage: the expertise of human experts. All needs to be done is to obsolete the aforementioned excuses. And

let people ask.

18
  • 16
    Yeap, let's just open the floodgates and stop closing questions. Sure, that'll increase traffic... For a while. Commented May 19, 2025 at 16:44
  • 16
    @Cerbrus First, there is not much flood, which this post is exactly about :) Second, there are people willing to answer. Right now they just huddle in the comments under the closed question, making it look ridiculous. Commented May 19, 2025 at 16:49
  • 8
    @Clive Truth be told, even before LLM, Stack Overflow sucked a big one in regard of quality. It claims being "not a forum" but any popular question looks exactly like a forum thread! Instead of just a single "detailed, high-quality answer" there are dozens, sometimes arguing with others, each with a heated discussion on its own! Only people that don't have any other choice would tolerate that mess. No wonder everyone and their uncle turned to LLM. Compare that exodus with that of Wikipedia, were you can see what a quality detailed answer is. Commented May 19, 2025 at 17:36
  • 6
    " It claims being "not a forum" but any popular question looks exactly like a forum thread!" - Okay, but that is precisely a consequence of people not having done enough (especially in 2008, but still now) of what you now propose to stop doing. As much as the question seems well-posed (and phrased in a highly searchable way), it reflects distracting ideas about what the terms ought to mean (by casting everything in terms of the calling semantics instead of also the variable semantics, and allowing arguments about whether the term "pass by value" is misleading in that context). Commented May 19, 2025 at 17:50
  • 3
    @KarlKnechtel not at all! If you don't like this specific question, you can choose any other. They all just spammed with answers. This is Stack Overflow, it's the way it meant to be! And no, it's not lack of closures. Rather, this self contradicting approach is a direct result of ill-conceived "gamification", which put quality WAY beyond quantity. Nobody gets a single reputation point, let alone a virtual trinket, for making existing answer better. While for a hasty new answer you'd easily get a hundred. This is how it's started. Commented May 19, 2025 at 18:13
  • 2
    @KarlKnechtel not to mention that such gamification is exactly like a stick in the famous bike meme. As much SO aspires to be that trove of high quality answers, it's doomed to be that enormous garbage dump of millions hasty answers, with a few selected quality answers that make anything but a rule. Commented May 19, 2025 at 18:17
  • 1
    On the other hand, if you consider this particular question controversial and insist that such questions must be banned from Stack Overflow, than where all these 2.8m people should go for the answer? Commented May 19, 2025 at 18:49
  • 3
    I agree the gamification was ill conceived. I disagree that the existence of that gamification reveals an intent for how Stack Overflow was supposed to be. I think Atwood and Spolsky just made a mistake, or else didn't have a better idea for handing out curation privileges, or were too influenced by existing sites despite their desire to make something different. Commented May 19, 2025 at 18:57
  • 1
    And I don't think such questions should be banned. I think they should be asked better, and that we should have better answer deletion mechanisms. But deleting answers is the same spirit of curation as enforcing standards on questions. Commented May 19, 2025 at 18:58
  • 6
    @KarlKnechtel I think they should be asked better. Isn't it quite illogical to expect a well-rounded question from someone who is already asking? Wouldn't it be much more helpful if knowledgeable people shaped the question after it gets some traction? You don't have to answer - it would. Just that Stack Overflow won't let it be. It would punish community effort and defend selfishness and greed. Commented May 19, 2025 at 19:09
  • 8
    "Asked better" includes the result after editing. Commented May 19, 2025 at 21:08
  • 1
    Stack overflow focused so much on 'quality answers and quality' questions that drifted beginners away from the platform. Basically, shooting their own foot. Maybe a community that would accept any sorts of questions and not judging people for lack of explanation or knowledge about the problem would help bring beginners back instead of them going to chatgpt to ask that. Commented Jan 13 at 14:59
  • 2
    @HQSantos some time ago I pondered heavily on a proposal, how to meet both goals. Of course, it was meet by a fierce backslash. But yes, since then, the "quality answers" goal became obsoleted, and now only one way is left - just answering questions, whatever they are Commented Jan 13 at 15:15
  • 1
    I saw the graph on Reddit and came back to SO and it’s the same old arguments about quality vs quantity. As an experienced SO user it was hard even for me to ask a question without copping the wrath of someone, so I left. Seems that one of the biggest gripes talked about on reddit was questions closes because duplicate that weren’t actually duplicates. Commented Jan 14 at 20:29
  • 2
    Gatekeeping as an issue... it gets toxic. But .. oh well... it is what it is. If only many of the gatekeepers in Meta would both to answer/ask in main stackoverflow. Commented Jan 15 at 10:42
2

Whether the people involved, or invested in it, like it or not, Yes it is.

I have answered a lot of the "Blazor" tagged questions since it launched. The number of questions now asked are a trickle compared to say 4 years ago. The context has also changed: most are now very specific, niche or custom security setups i.e. unanswerable without a lot of context, some specialist odd framework knowledge, or a view of the project.

Where once I answered several a day, I'm down to say one every two weeks.

On the "Toxic" question: it still thrives. I had a perfectly valid answer marked down by lurkers last week. I've seen valid questions closed by people who didn't have the knowledge to judge the validity of the question in the first place. I've reported stuff, but never received a reply. It's quite a turn off even for a seasoned contributor who doesn't give a **** about points: I'm within a cat's whisker of joining the exodus and leaving the building.

New contributor
MrC aka Shaun Curtis is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
0

Since I have been using ChatGPT, I no longer need Google as much as I did before, and therefore, I no longer use Stack Overflow either. But I think ChatGPT learns solutions from Stack Overflow, so it is not dying. Without external learning, ChatGPT is not able to replace Stack Overflow.

2
  • 5
    Problem is that if SO stops getting new questions and new answers, GPT does not need to learn from it any longer and can replace is with datadump from 2026/01. In order to stay relevant in the future, SO needs to stay afloat with new quality content. Commented Jan 19 at 10:40
  • @Kromster yes, this is a reason it won't die (unless it is replaced by some smarter SO that for example solves the problem of duplicates better). Commented Jan 20 at 15:12
-2

Does the graph show reasonable data? Is it representing a proof of death?

  • In the beginning there was just StackOverflow for all programming-related Q&A. Some time ago it split in multiple sites under then new StackExchange. How does the sum of performance metrics on StackOverflow, TeX, UX, and other specialized sites show compared to only SO metrics? Isn't the SO's decline emphasized by focused questions being asked on dedicated sites?

  • Is the "New questions per day" a good metric of a Q&A server, because the larger the pool of existing questions (and answers) is, the harder it is to find a question that hasn't been asked.
    How does the "weighted questions per day" metric look like? Say we give a question a weight based on its score N days after asking (discarding ones below a threshold). Or estimate the weight as a sum of question's reputation times 1, top M answer's reputation times 1/m (for m-th most favourite answer).

  • Can we expect change in the question targets? Say from "give me code for task X" to more abstract and general questions? The graph can be read as well as having a platteau between 2012 and 2022 with question count 10000-20000 qpd. Can we expect that offloading some questions to LLMs and saturating the LLM hype (LLMs are known for being very confident liars) is leading to lower qpd counts but the metric will hit new platteau (reach different equilibrium)?

New contributor
Crowley is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
1
  • 3
    1: no, all tech stacks have seen a similar drop off. 2. it's a good metric of user engagement, but not a good metric of how much good future-useful content is being created. There's certainly some correlation between the two, but it's surely not 1:1, given the majority of questions we are no longer getting are ones that have been answered. Commented 2 days ago
-3

Yes, it is definitely dying. Mainly because of AI as most people turn to LLMs and even Google AI summary now.

And many of the policies have also contributed to long time veteran contributors and mods leaving. I find also quite a lot of conflict so whatever incentives there were to participate... have all but disappeared now that there are alternatives. Add to that, new users being bulldozed... and it doesn't do much now that there are faster alternatives.

Unfortunately, it also may seem that there is a lot of activity on Meta but not so much in main Stack Overflow - especially for answering and asking.

And especially with the gatekeeping … quite a lot of active gatekeeping though unfortunately there is not the same enthusiasm with actually providing content to the website… in terms of answers and questions.

Also, I was quite impressed with the low volume of new questions/answers for the last hour. 30 questions? 30 Answers? Nothing in comparison to 2017...2018...

enter image description here

0

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.