I used to care more about the in-community rules for question asking. I totally got it - they want to curate high value question answer pairs. In the old regime this was the best way for the right answer to a question to be on Google.
Now, the rare times I've asked a question, I've violate some rule I've forgotten. Or that's new (they're SO AI allergic).
Reddit, interestingly, has continued to thrive. I think in party because its still largely fun and there are literally millions of communities.
Stackoverflow once benefited from its pedantry, now there's not much value in such pedantry, and its just not a fun place to hang out.
It's a tragedy but I think it's true that SO has lost a significant amount of value given AI's new presence. Having said that, it's not like it's dead - it's a community of real people rather than an amalgamation of knowledge in a single tool. What sets SO apart and will always set it apart is that human aspect. For future problems that AI has no knowledge of, there will always be a group of people better equipped to answer questions through real life experience and deductions that AI simply cannot apply. I suspect in the next few years, SO will evolve into a platform that focuses on that human aspect because ultimately that's the only avenue I can see that would work. It's unlikely SO will die any time soon, it's more likely that the platform will find ways of pivoting to keep business flowing. It's not like SO is struggling for questions right now anyway, it continues to be a massively popular platform. Predicting the future is somewhat of a pointless task but I think it's safe to say, for now, that SO is still alive, and will be that way for a while at least.
The exit of the company by the founders was definitely timely, if having assumption that SO can't be "relevant" anymore at the times of LLMs. Of course there is always value for human-to-human Q&A that goes beyond LLM training set, but that might happen now only at cutting-edge private environments/communities.
First, a disclaimer: I do not speak for other elected moderators, nor for Stack Exchange Inc. My views are my own.
Site traffic has been declining for a long time and that's not a mystery. Empirically, the rise of Large Language Models has sped up the decline or at least did not reverse the trend. This is both good and bad. Good because LLM's capture the vast majority of questions that would be quickly closed — underspecified, unclear, duplicate, not actually about programming, and so on. This increases the signal-to-noise ratio and average usefulness of SO questions. Bad because of all the implications of declining traffic that everyone can imagine.
Is it dead? No. In fact, SO has a great opportunity to specialize in answering questions that LLMs cannot answer (bleeding edge technology, complex debugging problems, emerging issues, you name it.) The community is still alive. Whether Stack Exchange Inc. is able to understand and adapt to this shift is unclear.
Company aside, the biggest challenge we moderators are facing is to identify and remove LLM-generated content. Keeping SO free of LLM-generated content is critical in helping the site maintain an edge against AI tools that provide quick and confident-sounding answers to whatever problem you throw at them. It's an uphill battle though, and one that is probably unwinnable, but the community hasn't given up yet.
A year or more ago some poor guy asked a simple SO or SE question. I studied that for a while, tried multiple carefully crafted Google queries until I found one that gave a good correct AI result, posted a reply NOT POSTING THE AI RESPONSE, that would break their rules, ONLY telling the person what query to do to get an answer and urged them to study and test this carefully, and I posted that. Moments later I got a stern warning that my post had been torn down, likely because their AI hunter bots had seen the word AI in it, almost certainly before the original poster ever saw this. Nobody else ever even tried to help the person. Later, only slightly related, I realized it had been most of a year since anyone posted a question in the little SO and SE corners that I used to watch and so I scanned my last five years of posting little mostly trivial attempts to help users, many of those were "I think you've made a very simple mistake here..", some of those probably resulted in trivial fixes, usually the person didn't even respond to me. SO I thought I'd raise the average quality of posts slightly by going through and deleting my trivial posts. On both SO and SE their hunter bots immediately flagged me for doing this and I got messages from both telling me that all the stuff I had tried to help people by putting up over the last 5-30 years WAS THEIRS (SO and SE) AND BELONGED TO THEM!!! And I believe they put back some or all of my trivial stuff. And I stopped watching SO and SE.
I wanted to do a pause of 1 second in Chez Scheme, but I din't know the spell. I tried Google and Gemini gave me an example. It didn't work because the parameter were in the reverse order (seconds vs nanoseconds). So after looking at the manual and trying both orders, I got the result I wanted. (And reduced the time to .2 seconds, because 1 second was too long.)
In an hypothetical case that it were a SO question, I don't see why my answer would not be good.
It's still good for asking questions, I'm not aware of many more other places where friendly subject matter experts hang around (IRC, Mathstodon, Reddit?).
It would be better if closing questions would cost 1000 reputation. That's one advantage AI has over it - it will at least try to answer your question every time and not just randomly shut you down for its own (wrong) reasons.
I haven't used StackOverflow hardly at all since ChatGPT came out.
Unrelated to AI, I haven't really had a positive experience on StackOverflow in 7+ years. The way they aggressively close questions as duplicates despite the previous questions having incomplete or outdated answers was already making it a much less useful site.
When I was new to AWS and had questions, I found the AWS s subreddit to be a lot better. You have people specifically interested in the topic and you don’t have overzealous moderators.
There is also a higher tolerance for newbie questions and duplicates
Many subreddits are better organized than SO because they funnel newbie questions and duplicates to FAQs, wikis or daily question threads. SO does alert to similar questions during composition, but they're not often what the user is actually asking or not recognizable. subreddit wikis can better respond to that kind of perennial confusion.
If SO comes up in search results, I visit it. But I should also add that I usually skip reading "AI" results or summaries. There's usually nuance in SO questions and answers that "AI" leaves out.
I’ve had many situations where a sourced AI summary has references that don’t show up in the top search results. Clicking through to the source gives me the nuance and context of the full post that I wouldn’t have found without the AI summary. Even if you don’t trust the summary, rightfully so, it can still be good at surfacing good results.
(My experience here was with Kagi, I’m not sure how Google and others are with this)
Yes it's dead, it died at least 5 years ago. Why people keep asking this? Quora is also dead, myspace is dead, Tumblr is dead, so many apps of the past are dead. The healthy thing would be this to be the fate of every app one day, retiring.
I used to care more about the in-community rules for question asking. I totally got it - they want to curate high value question answer pairs. In the old regime this was the best way for the right answer to a question to be on Google.
Now, the rare times I've asked a question, I've violate some rule I've forgotten. Or that's new (they're SO AI allergic).
Reddit, interestingly, has continued to thrive. I think in party because its still largely fun and there are literally millions of communities.
Stackoverflow once benefited from its pedantry, now there's not much value in such pedantry, and its just not a fun place to hang out.
It's a tragedy but I think it's true that SO has lost a significant amount of value given AI's new presence. Having said that, it's not like it's dead - it's a community of real people rather than an amalgamation of knowledge in a single tool. What sets SO apart and will always set it apart is that human aspect. For future problems that AI has no knowledge of, there will always be a group of people better equipped to answer questions through real life experience and deductions that AI simply cannot apply. I suspect in the next few years, SO will evolve into a platform that focuses on that human aspect because ultimately that's the only avenue I can see that would work. It's unlikely SO will die any time soon, it's more likely that the platform will find ways of pivoting to keep business flowing. It's not like SO is struggling for questions right now anyway, it continues to be a massively popular platform. Predicting the future is somewhat of a pointless task but I think it's safe to say, for now, that SO is still alive, and will be that way for a while at least.
Thanks, you just addressed just the points I was curious about the most.
- How will they survive?
- What happens with all those unsolved problems, those that AIs haven't found a source with the solution to scrape from?
The exit of the company by the founders was definitely timely, if having assumption that SO can't be "relevant" anymore at the times of LLMs. Of course there is always value for human-to-human Q&A that goes beyond LLM training set, but that might happen now only at cutting-edge private environments/communities.
I'm a Stack Overflow elected moderator (one of these users: https://stackoverflow.com/users?tab=moderators)
First, a disclaimer: I do not speak for other elected moderators, nor for Stack Exchange Inc. My views are my own.
Site traffic has been declining for a long time and that's not a mystery. Empirically, the rise of Large Language Models has sped up the decline or at least did not reverse the trend. This is both good and bad. Good because LLM's capture the vast majority of questions that would be quickly closed — underspecified, unclear, duplicate, not actually about programming, and so on. This increases the signal-to-noise ratio and average usefulness of SO questions. Bad because of all the implications of declining traffic that everyone can imagine.
Is it dead? No. In fact, SO has a great opportunity to specialize in answering questions that LLMs cannot answer (bleeding edge technology, complex debugging problems, emerging issues, you name it.) The community is still alive. Whether Stack Exchange Inc. is able to understand and adapt to this shift is unclear.
Company aside, the biggest challenge we moderators are facing is to identify and remove LLM-generated content. Keeping SO free of LLM-generated content is critical in helping the site maintain an edge against AI tools that provide quick and confident-sounding answers to whatever problem you throw at them. It's an uphill battle though, and one that is probably unwinnable, but the community hasn't given up yet.
A year or more ago some poor guy asked a simple SO or SE question. I studied that for a while, tried multiple carefully crafted Google queries until I found one that gave a good correct AI result, posted a reply NOT POSTING THE AI RESPONSE, that would break their rules, ONLY telling the person what query to do to get an answer and urged them to study and test this carefully, and I posted that. Moments later I got a stern warning that my post had been torn down, likely because their AI hunter bots had seen the word AI in it, almost certainly before the original poster ever saw this. Nobody else ever even tried to help the person. Later, only slightly related, I realized it had been most of a year since anyone posted a question in the little SO and SE corners that I used to watch and so I scanned my last five years of posting little mostly trivial attempts to help users, many of those were "I think you've made a very simple mistake here..", some of those probably resulted in trivial fixes, usually the person didn't even respond to me. SO I thought I'd raise the average quality of posts slightly by going through and deleting my trivial posts. On both SO and SE their hunter bots immediately flagged me for doing this and I got messages from both telling me that all the stuff I had tried to help people by putting up over the last 5-30 years WAS THEIRS (SO and SE) AND BELONGED TO THEM!!! And I believe they put back some or all of my trivial stuff. And I stopped watching SO and SE.
I wouldn't want an answer like that, if I wanted AI replies of any kind I would ask AI myself instead of writing a question on StackOverflow.
Anecdote time:
I wanted to do a pause of 1 second in Chez Scheme, but I din't know the spell. I tried Google and Gemini gave me an example. It didn't work because the parameter were in the reverse order (seconds vs nanoseconds). So after looking at the manual and trying both orders, I got the result I wanted. (And reduced the time to .2 seconds, because 1 second was too long.)
In an hypothetical case that it were a SO question, I don't see why my answer would not be good.
It's still good for asking questions, I'm not aware of many more other places where friendly subject matter experts hang around (IRC, Mathstodon, Reddit?).
It would be better if closing questions would cost 1000 reputation. That's one advantage AI has over it - it will at least try to answer your question every time and not just randomly shut you down for its own (wrong) reasons.
I haven't used StackOverflow hardly at all since ChatGPT came out.
Unrelated to AI, I haven't really had a positive experience on StackOverflow in 7+ years. The way they aggressively close questions as duplicates despite the previous questions having incomplete or outdated answers was already making it a much less useful site.
When I was new to AWS and had questions, I found the AWS s subreddit to be a lot better. You have people specifically interested in the topic and you don’t have overzealous moderators.
There is also a higher tolerance for newbie questions and duplicates
Many subreddits are better organized than SO because they funnel newbie questions and duplicates to FAQs, wikis or daily question threads. SO does alert to similar questions during composition, but they're not often what the user is actually asking or not recognizable. subreddit wikis can better respond to that kind of perennial confusion.
It will be used to answer the harder questions that AI models can't answer.
If SO comes up in search results, I visit it. But I should also add that I usually skip reading "AI" results or summaries. There's usually nuance in SO questions and answers that "AI" leaves out.
I’ve had many situations where a sourced AI summary has references that don’t show up in the top search results. Clicking through to the source gives me the nuance and context of the full post that I wouldn’t have found without the AI summary. Even if you don’t trust the summary, rightfully so, it can still be good at surfacing good results.
(My experience here was with Kagi, I’m not sure how Google and others are with this)
Yes it's dead, it died at least 5 years ago. Why people keep asking this? Quora is also dead, myspace is dead, Tumblr is dead, so many apps of the past are dead. The healthy thing would be this to be the fate of every app one day, retiring.
If not, I bet everyone is waiting for it to die, and take hn with it.