I have an alternate view there. The data is already crawled and indexed by ChatGPT. It probably has all the building blocks for the code bases. The large part of the traffic on Stack overflow was to search and get the solution. The contributing traffic was always much lower. Models like ChatGPT can build on top of the building blocks that stack overflow has. I don't see this as a problem. ChatGPT is no longer relying on more solutions posted on stack overflow for it to be effective. It has already entered the realm of synthetic code generation and already has the critical mass of code from stack overflow.

On the topic of new languages and libraries, that was always a low end work. Why calling it low end is because I come from the YACC coding days where it was a laborious work to build a new language, and also new libraries. That part is now pretty much automated and now you need to think of new libraries and post them on prompts than code new libraries. This is what evolution looks like and this will lead to new problems and demand for human ingenuity to solve those problems.