> Stack Overflow would obviously be a much better source of quality training data for the kind of questions you presented.

I dont think we need a llm that only responds with "Your question was marked as a duplicate"