I very skeptical that any productive gains were had. There's massive amounts of cherry picking of data, marketing campaigns, misleading metrics, complete disregard for technical debt, and groupthink mania that suppresses critical voices.
I'm inclined to believe LLMs suck for any and all coding and you'd be better off without them, especially due to them being inherently unreliable, but even if they were marginally useful in specific situations, the overall negative effect in trying to leverage the chatbots for all the things are going to be disastrous and ruin many a codebase. It's completely up in the air if any organization will be able to use a coding chatbot responsibly only in situations where it has utility (if it has any) without suffering the negative effects of using it when not called for. Even if someone figures out the perfect use cases without falling into the trap of overusing it, it might still not make economic sense to actually use it due to the cost.