Stack Overflow has seen a substantial decline in traffic over the last year that appears to be accelerating. https://observablehq.com/@ayhanfuat/the-fall-of-stack-overflow
You must log in or register to comment.
I think this has as much to do with Google being shit at finding stuff lately as it does llms like chatGPT
IDK what shitoverflow gets out of being so fucking toxic. I asked one dumb question and I’m basically banned from posting on the website.
It feels like they’re trying to be a sort of “wikipedia” of every programming problem and solution. The problem is that eventually everything will be posted, and everyone will be banned from the website.
You lack vision, but I see a place where people get blocked and their questions opened then immediately closed as duplicates. Opened and closed, opened and closed all day, all night. Soon, where the internet once stood will be a string of condescending experts, admonitions that “you shouldn’t do that, do Y instead”, pleas for information closed as off-topic. Passive aggression, spiteful ego contests and wonderful, wonderful karma meters reaching as far as the eye can see. My God, it’ll be beautiful.
“you shouldn’t do that, do Y instead”
That’s one of my favorites: ignore the problem, only pick on the scope we can’t change.
I asked for advice on how to express something in UML once:
“No one cares whether you follow the UML standard, just make something up”
“But my company uses waterfall and requires UML diagrams to move onto the next phase of development!”
“That’s an issue with your company then. Ask your boss how to do it. Question closed.”
It isnt even my problem and I still despair reading this.
You have to build Rust from source, then install the dependencies with cargo, then update your node.js because it uses npm to manage it’s configurations and if your npm isn’t at least the current unstable version, the configs will be outdated. This worked for me on Arch, which is what I use btw.
You have to build Rust from source
As someone who actually did out of interest at one point, you’d be surprised how easy this is to do.
x.pyis a godsend.For the rest of your comment, it was immediately invalidated when you said you use Arch. The reality is that more people use Ubuntu, so you should be using Ubuntu too. Don’t use
apt? Figure it out yourself :P
The problem is that eventually everything will be posted, and everyone will be banned from the website.
I don’t think they see that as a problem, that’s the goal
You were able to post on there at all? Don’t they have extremely high barriers to entry for even question comments?
Honestly, I put some effort into get some of their reputation points. Then I asked one question that I didn’t realize was dumb and I can’t post questions anymore. You’re welcome to see my profile and try to figure out how I did it 👍
lmao, how dare you be inquisitive
Not only post, but I have content that still feeds me residual cool-points even now.
I got a nastygram because I was editing the questions to follow a proper style and form (AP) and some people got upset that my comments were more “run on sentence” and " ‘emails’ and ‘helps’ both sound wrong as nouns for the same reason" instead of something like “there-there, Timmy”.
So I said “you can have free editing, or the next guy can be a people person instead.” And they agreed.
So I’m read-only there now too. :-D
I vaguely recall the first time I ever asked something on SO, around 2013, the first reply was “this has already been asked before”. No link to said previous question. Taught me to lurk and search more before asking anything there.
I sometimes also suffer a case of “explaining until I figure the question myself”, where the more details I punch into my question, the more likely I am to find the answer myself.
I bet this is directly related to ChatGPT
People prefer having something generating shitty code and not checking it, instead of asking or searching on internet for a substantially better solution
Because forum posts are always full of accurate and helpful information?
In my experience it still makes good suggestions for most things, and is better than trying to phrase things in a way that Google likes, then trawling through irrelevant forum posts.
It’s only there to make suggestions, so if someone is taking its output without understanding and treating it like gospel then they’re an idiot who’s inevitably going to end up in a world of trouble.
If you take the suggestion, verify it with documentation, then make sure you actually understand it, chatGPT is a great tool.
If I’m honest, stackoverflow was always a shortcut for searching documentation to me.
Simple stuff like how do I turn an InputStream to a String again? I can’t remember it, but I know exactly what to look for, I’m just to lazy.
For that kind of stuff ChatGPT is almost perfect.
Because forum posts are always full of accurate and helpful information?
Not necessarily, but at least there’s much more opportunity for other people to jump in and correct false info or expand upon something. It’s by no means a flawless system, but it’s better than only have one source of information
I didn’t say that people should go on the internet and pick the first forum post either ; that would be like trusting whatever chatgpt is handing you :p
My point was more on the “people are lazy” side of things, but yeah you have to stay critical of both chatgpt and forum posts.
I agree, I just think that those lazy people will do what they do regardless of where they get their info.
To butcher a saying; blame the craftsman, not the tools.
I half expect that, if enough programmers use ChatGPT-written code verbatim, someday it’s going to lead to Skynet. I mean, what’s to stop ChatGPT from inserting bits of extra code to be used for its own distributed processing botnet?
You’ve never written code, have you?
Sadly there are so many people that take its output as gospel and don’t realise it can be wrong. So is a tool that commonly gets abused by people that don’t know how to use it.
You can have it generate shitty code and then compare it against examples it finds online to iterate that code. Also, it was trained on the whole internet, including those good solutions, and can often reproduce them on its own. but you have to tell it, explicitly, to do all this to make better code, rather than just asking for the code.
At least ChatGPT will not flag the question as duplicate.
“I’m sorry, as an AI language model this question has been asked too many times and there is insufficient computer resources to handle your request. You’ve been temporarily silenced for 15 minutes.”
ChatGPT isn’t that good at code generation lol.
Doesn’t need to be good. Just good enough that people need SO less often. If GitHub Copilot gives a code suggestion, I don’t need to look up some syntax or some method I forgot. I’m reminded, and can see that it’s correct. No searching online required.
It’s a little more decent than you give credit for. I use it all the time for easy generic subroutines and functions. It struggles a bit with specific, complex requests but is generally pretty versatile as a miniature code assistant. It’s good at catching human errors like loops starting or ending at the wrong specified integer, so I use it as a debugging tool.
I thought chatgpt is kinda shit now since the newest updates
I think it’s overblown tbh.
In my experience it still makes good suggestions for most things, and is better than trying to phrase things in a way that Google likes, then trawling through irrelevant forum posts.
It’s only there to make suggestions, so if someone is taking its output without understanding and treating it like gospel then they’re an idiot who’s inevitably going to end up in a world of trouble.
If you take the suggestion, verify it with documentation, then make sure you actually understand it, chatGPT is a great tool.
ChatGPT has been a great tool to help me teach coding. It lets my students with a few months experience write better code, as if they had a few extra months experience, but like you say it’s very easy to get in trouble with it. We had it generate some code to interface a web app with some cloud triggers, and chatGPT suggested we put all the API keys / creds right there in the front end where anyone with “view source” could see them. It made for a really good lesson, actually, on the need to gain experience, understand what code does , and to validate with documentation.
trawling through irrelevant forum posts.
This makes it worth it from just a time savings perspective. Also, describing it as trawling is very accurate lol. It takes a lot of trawling to get the answer you need, and even then sometimes it isn’t right because you’re relying on a single individual’s answer.
Most of the comments here seem to be arguing whether it’s better to get help now from SO or ChatGPT, but this is a pretty short-sighted mindset.
What happens when the next new standard comes out that ChatGPT hasn’t been trained on? If SO tanks and dies, where will you go?
I’m not saying use a lesser resource, I’m saying this is kinda tragic and I hope they can sustain themselves; AI is propped up by human input and can’t train itself.
Hey, if people are going to go back to reading manuals like we’re in the 1980’s again is it such a bad thing? /s
It’s insane how a single tool managed to completely destroy the value collectively created by people in over a decade.
That single tool is still propped up by that collective decade of knowledge. ChatGPT would be nothing without sites like stackoverflow
Yeah but will people still care about contributing that information if they’re not going to be compensated for it in any way? Like people get something out of contributing to stack overflow, even if it’s just recognition. This is gone with ChatGPT.
very good point! I find myself using ChatGPT more for references and I am also afraid what will happen if there isn’t enough “human generated content” to train on. I can picture an edge case a chunk of the internet is AI generated content (with even users at the wheel). The the next wave of AI will train on previous gen AI output
AI should be trained by itself though. I just wouldn’t call LLMs “AI” as a term
Also, it shall be possible in the future to just feed it the documentation and have answers. Obviously we are still nowhere near yet
Crazy idea, what about a “federated” search. Hook up the websites’ internal search engines to an aggregator. Stop allowing random indexing spiders to scrape.
We go back to expertsexchange
SO is a shithole, just like Reddit. All the work is done by volunteers. When it was time to cash out with the platform, they also did several things to fuck with their community. I’ve contributed quite a bit to the trilogy sites, and served as a moderator. I regret every second of it. But at least a few people got rich in the process.
I am not sure when this started, but google searches now sort by paid content first rather then relevant content first, so Stack Overflow started to drop down into page 2 or more.
I start my search string with
stackoverflowas a workaround.
Is there a fediverse alternative yet?
Also, if you are a technical person I urge you to start a blog where you document problems you solve. It’s a great ressource for others and a resumé for you.
This doesn’t tell us much without also including the quality of the posts. Are we sure this isn’t just idiots who ask stupid question that can be found on Google over and over not doing that now that they have chatgpt
I really like using code.whatever.social as an alternative frontend to Stack Overflow. It has way less distractions and allows me to only look at the question and the answers and nothing else.
I really like this, never saw it before. Thanks!
No problem. You can use extensions like LibRedirect in order to make it automatically change SO to this one.
Annnnd bookmarked. Thanks, this is really cool!
I actually go there more often now that I try to avoid reddit in my search results. Sometimes valuable posts have been edited or deleted.
Oh wow, thanks. I didn’t realize that making this an image post got rid of the link
Yeah, I don’t know if it’s a bug or a feature. I got a similar problem before with one of my posts. I think a workaround would be to post it as a link and paste the image in the
Body.
All questions have obviously been answered.
Not arguing with the other possible reasons given, but it can be really hard to get started with SO as anything other than a reader. Gaining enough points to comment, answer, or even answer a comment feels really hard now that so many questions are already answered well.
A lot of my answers I get answered with ChatGPT. And I can always ask ChatGPT to tell me where I can look to verify the answer. I find myself on stack overflow for very specific or very technical topics.
It’s funny how if everyone just went and “read the documentation” like they tend to obnoxiously tell you to do… stackoverflow wouldn’t exist. Personally I go and look for things I can answer if someone asks a question that I know will get obliterated but I can tell they just need some help. I’ll try to answer it before it gets downgraded and they get banned
Have you ever wanted to do something from the uncharted area? Encountered bad documentation? This is what it’s supposed to be for, not handholding.
Uncharted is quite subjective. I used SO most when I was starting out in SE. Looking back through the questions I posted, most of them were very much beginner questions that I would just know nowadays or know where to look for. That was what I used SO for. Beginners asking veterans for help. The least of them were due to bad documentation or exploring uncharted territory. As I grew more confident in the field, I stopped using SO more and more. The latest only for best practices on simple problems I don’t want to reinvent. And exactly those cases GPT now solves faster and I’d be surprised if not even better than SO posts.