AI didn't kill Stack Overflow
infoworld.comWhile the criticisms of the moderation systems are surely valid, AI definitely killed Stack Overflow. The metric the article authors chose to indicate the pre-AI downward trend was questions asked. This will naturally decline as the body of existing already-answered questions increases since many developers will have similar questions. The steep drop post-AI is different and reflects the ability of LLMs to answer a large fraction of the sort of questions people would ask. I would think a drop of traffic (as opposed to the questions asked) would be a better metric but I can't find one that seems reliable.
> The steep drop post-AI is different and reflects the ability of LLMs to answer a large fraction of the sort of questions people would ask.
Yup, the fact that ChatGPT can give me answers instantly, without dismissing or XY'ing me, "why would you ever want to do that?", "why don't you just do <thing you've already mentioned you tried>", or "just google it", has almost completely removed my need for StackOverflow.
Maybe, but SO had been getting less useful well before LLMs became the New Hotness. I stopped using it years ago because of that, not because LLMs.
I was never a big SO user in terms of asking questions or providing answers (maybe a total of 10 questions and ~15 answers all time), but I did browse a ton and learned a whole lot.
Thanks to ChatGPT and Claude, I just don't see the need anymore, when I can get an answer tailored to my problem in 5 seconds, and ask as many follow up questions as needed, with minimal snark.
But where do the AIs get the training data for newer languages or frameworks?