Live data from Hacker News

What AI did to stackoverflow in a graph

data.stackexchange.com

331–340 of 615 posts

Re: What AI did to stackoverflow in a graph

#331

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 t…

The graph proves the cause of their decline was AI, and not aggressive question moderation. Ask yourself: in what year did it become difficult to ask questions on Stack Overflow? 2014? 2016? 2018? 2020? Aggressive question-closing was part of their design from the very beginning. Their high barriers to question-asking was the cause of their rise , as their primary user was never question writers: it was Google, and a…

Ouch, great, you should show that protocol you used to get causation out of an observational study to the entire scientific community! It's unclear what area will give you a Nobel Prize for that, but several of them will rush to do it first.

Re: What AI did to stackoverflow in a graph

#333

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 t…

In the age of gratuitously encouraging LLMs and constant slop I sometimes miss the way in which stack exchange punished people for putting in low effort. I recently asked a LLM a question simply out of newfound habit. I realized while reading the line that this command seemed very familiar. I had bookmarked the exact same command in stack exchange many years ago. Of course the fact that stack exchange traffic has dra…

> Taking personal offense to someone insulting your question is something you should learn to get over.

I did get over it when I stopped bothering with SO. Sorry, but a community whose ethos is "you need a thick skin to belong" is not one most people even want to belong to.

Re: What AI did to stackoverflow in a graph

#334
post #313

Earlier quoted context omitted.

Their rules, (I believe unintentionally) give iron-fisted fiefdom rulers a toolbox of justifications to control and alienate under the guise of protecting the quality of the site data. I honestly don’t even think most of the control freak mods could objectively judge the propriety of their actions because it was all encouraged by the rules. (And I do not think this was universal among the mods, but it was certainly e…

I remember having to fight a mod for him to restore a reply penned by Mike Pall, to a LuaJIT question. Mike Pall is the author of LuaJIT. The reply had been either deleted or edited to the point of being wrong (memory is foggy), because Mike Pall wasn't an expert at SO, and had somehow not used the site exactly as intended. The mod was very dismissive and patronizing.

I was going through some answers on a stackoverflow thread, and noticed that every single one had been edited by the same guy, just.. adding his own personal opinions and 'corrections' to them, and in the process making them universally worse and less correct

The idea that answers should be editable, and the gamification of stackoverflow, was an absolutely terrible combination

Re: What AI did to stackoverflow in a graph

#335

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 t…

The graph proves the cause of their decline was AI, and not aggressive question moderation. Ask yourself: in what year did it become difficult to ask questions on Stack Overflow? 2014? 2016? 2018? 2020? Aggressive question-closing was part of their design from the very beginning. Their high barriers to question-asking was the cause of their rise , as their primary user was never question writers: it was Google, and a…

It's the same thing. Why did AI compete so aggressively with them? It's because their system was one that produced confident misinformation from hobbyists while gatekeeping actual experts.

The same thing is not yet happening to wikipedia as you can see with the pageview tool. You may be confusing a covid bump. At most any drop is within an order of magnitude.

Re: What AI did to stackoverflow in a graph

#336

Earlier quoted context omitted.

At least 50% of the time on SO, the best answer is 3rd or worse, and I'm always thinking, "Why is this not the top answer?"

Because the person who asked the question has accepted the first answer that looked decent enough and moved on. Usually the most superficial, not that it's always a bad thing. Oh, also on SO there's this kind of exchange: Q: "I want to do X because of this and that - or because I simply fucking want to". A: You should never do X, do Y or Z instead.

The "that's a XY problem" guys always seem to think they're the smartest in the room.

Re: What AI did to stackoverflow in a graph

#337
post #293

Earlier quoted context omitted.

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 real…

> aquire comment points to be able to answer I thought it was the opposite, you need answer points in order to comment (which resulted in people using answers as comments because they had no other option).

The rules changed, a lot.

You created an account one day and the only things you could do were commenting and asking questions; you created it some other day and the only things you could do were asking and answering questions; some other day the only thing you could do was asking questions.

Any day you signed up, asking any question first was a sure way to be downvoted bellow the threshold that would ban you from the site.

Re: What AI did to stackoverflow in a graph

#338
post #89

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.

But the LLM will confidently give you a wrong answer and then waste three hours of your time going down tangential rabbit holes trying to sort it out.

I never had an account, never asked or answered a question on SO, but found the answer I needed there plenty of times and got on with my work.

Re: What AI did to stackoverflow in a graph

#339
post #314

Earlier quoted context omitted.

Does it matter? A site and community that refuses to allow or accommodate anyone to grow and learn as a new user will quickly run out of users. The gatekeeping on SO is a blockade, not a teaching mechanism.

How else do you help people learn if not by giving feedback? As an avid helper in a programming related discord channel I need you to understand that a lot of the people who come to these places for help are incredibly stupid. It's not uncommon to have to converse with a person for 15+ minutes just to get them to give you enough information to understand what their problem is. Like they will just paste an error messa…

You just described the complete opposite of the StackOverflow experience.

Re: What AI did to stackoverflow in a graph

#340

Earlier quoted context omitted.

I don’t think we need to blame that sale. The same steady decline had been going since 2016. Itms clearly there in the graph, but everyone trumpeting the ChatGPT angle conveniently ignores the preceding 5+ years of continual decline.

The sale is relevant. The spike in activity right before the sale already shows the management hand and how they actively ran the org to optimize for this sale. The rebound right after is hard, and it was (right) before chatgpt They accelerated the downfall with this and then chatgpt came over

Are we talking about the large Covid spike in early 2020? I don’t think that has anything to do with the sale.
Post reply on HN