Because there seems to be a fundamental misunderstanding producing a lot of nonsense.
Of course LLMs are a fantastic tool to improve productivity, but current LLM's cannot produce anything novel. They can only reproduce what they have seen.
But they assist developers and collect novel coding experience from their projects all the time. Each application of LLM creates feedback to the AI code - the human might leave it as is, slightly change it, or refuse it.
Of course LLMs are a fantastic tool to improve productivity, but current LLM's cannot produce anything novel. They can only reproduce what they have seen.