How we do or don't use something creates the harm or benefit.

Consuming instead of creating has caused.

Passive average skilled prompting will give average results. LLM's can be used to actively work through your own thinking and engage it as quickly and deeply as you like.

I don’t know, to me it’s always fighting the llm to remember as much of the previous context as possible. Instead of moving forward with my thoughts, I seem to fight the llm forgetting what it said five minutes ago. And fact checking. The best example of fact checking, I asked chatgpt once about some istio cli features. It hallucinated available cli commands. But they were very convincing.