>What comes after Stack Overflow?
Codidact (https://software.codidact.com), for those who actually care about the Q&A site model (although we're designed explicitly to allow for opening separate "sections" for things that don't work like Q&A). That is, if you're posting a question in order to frame useful, searchable information, and not simply to get help with a project.
Disclosure: I am a moderator there (actual moderator, not just someone entitled to edit posts - which we still do as a community, with the same kind of Creative Commons licensing arrangement).
Github PRs.
I bet Google are kicking themselves for killing Google Code.
https://code.google.com/archive/about
> But alas, in 2015 Google bid farewell to the service
> In fact, from March 2023 to March 2024, the volume of new questions dropped from ~87,000 to ~58,800, a 32% decline in just one year
The decline began before ChatGPT's November 2022 launch. So it would still decline even without the availability of LLMs. For me personally SO has mostly lost it's value because of the bueraucratic and arrogant ivory-tower mentality of the admins.
> "cannibal's dilemma"
Looking at services like Perplexity, I don't think it is/was essential to use SO data for training, since the main point of Perplexity is to use sufficiently intelligent agents (based on LLMs) to search and infer from many sources. The specific knowledge the agent received from the training data is not only static (i.e. not current), but neither essential, as long as the agent has sufficient general knowledge to solve the research request. It's like real experts; the knowlege we were trained on at university is only a small subset of the world and usually not very current; but we have learned how to do research and to solve problems. I'm much more efficient with Perplexity than I ever was with SO, and no longer silly discussions.