I see this as proof of how bad LLMs actually are. You have an AI trained on essentially humanity's collective programming library. Languages of machines and computers. The result should be ungodly and near perfection. If there was any semblance of understanding in AI, it should be revealed in it's capability to produce code.

Although... I can definitely see Microsoft thinking that their code is the example of perfection and training copilot on that rather than github.