Live data from Hacker News

The Fall of Stack Overflow

observablehq.com

101–110 of 808 posts

Re: The Fall of Stack Overflow

#101
post #56

"How to do A?" Comment: "You shouldn't do A, it's better to do B" Closed, duplicate of "How to do C". Yeah sure, in the beginning, there were many more basic questions (how to increase number by 1, how to get division remainder, how to check if file exists,...), and if you're coding in Perl, you can still find all the answers... if you're working in python, you'll find answers for python2, some for python3, some spec…

This is just another anecdote. But here goes. My experience was that there was a long period where your complaint was quite true. I believe that they have heard this complaint enough and have lightened up a little. There are still more and stricter rules than most sites, but the obsession with pruning duplicates seems to have cooled.

And my opinion as someone with high reputation is that letting in silly questions has ruined the site and made me stop using it. SO is not a support forum, it's not your teacher, it's a place for interesting knowledge sharing and if the knowledge is drowned in noise, well it's just noise then.

Re: The Fall of Stack Overflow

#103
post #43

Even though StackOverflow in the common use case has been taken over by ChatGPT, I sincerely hope it keeps operating, stays strict (even if it causes collateral) and keeps ban on LLM-generated content. The wheels of this kind of stuff turn slowly, but obviously ChatGPT was trained partly with data only gainable from a healthy StackOverflow-kind of site with users actively asking unique questions and enough people ans…

> Even though StackOverflow in the common use case has been taken over by ChatGPT Claim asserted without evidence.

Certainly true for me and the people I know.

Re: The Fall of Stack Overflow

#104
The most help I ever got from SO for questions not already there, was because of their (perceived) strictness. The process of writing a high-quality question, with a minimally viable example, clearly lined out thought-processes, and other things tried, solved the question for me in most cases without me ever having to post it.

Re: The Fall of Stack Overflow

#105

Where were these data sourced from? As far as I know, StackOverflow does not publicize internal analytics with such granularity. If these figures are real, are they leaked?

Stack Overflow used to release their data archives quarterly on BigQuery. Looking at the BQ datasets, they were last updated Nov 2022, which doesn't have the latest 2023 info in the submission.

FWIW, I now analyze the Stack Overflow dumps on Snowflake

https://medium.com/snowflake/how-to-load-the-stack-overflow-...

Re: The Fall of Stack Overflow

#106
post #93

Earlier quoted context omitted.

> Even though StackOverflow in the common use case has been taken over by ChatGPT Claim asserted without evidence.

It has for me, and several developer friends, and considering the fame ChatGPT has gotten, and that StackOverflow's fall has accelerated, it's obvious the milkshake's migrating. Not all of it of course, as I stated, "for the common use case"

To contrast, I've not once used it where I would have previously used SO (professional developer).

Re: The Fall of Stack Overflow

#108
I tried asking a legitimate question before on SO about physics education and it got shut down because it wasn't a question with a definitive answer.

What non-trivial question has straightforward answers? It's just a homework help site. No serious discussion is allowed.

Re: The Fall of Stack Overflow

#109
The problems I’ve noticed with Stack Overflow are a few and hard for me to narrow down but basically:

- google used to return really relevant results for SO, and it stopped doing so at some point a while ago

- moderation on SO has gotten progressively more horrible. can’t tell you how many times I found the exact, bizarre question I was asking only to see one comment trying to answer it and then a mod aggressively shutting it down for not being “on topic” enough or whatever.

- because of the previous bullet, oftentimes the best answer is buried in comments and has very negative feedback despite answering the exact question

Due to a combination of these things, filtering against the noise for what I wanted became increasingly more difficult and often the solution to my problem was easier found searching github comments or random blogs.

Re: The Fall of Stack Overflow

#110
As long as OpenAI and others in the LLM space are for profit companies should start charging them heavy fees for their data. Surely there will be more novel knowledge in the world produced after 2023, and these companies will need that
Post reply on HN