There's a difference between not knowing the internals of a dependency you chose deliberately and not understanding the logic of your own product.

When you upgrade a library, you made that decision — you know why, you know what it does for you, and you can evaluate the trade-offs before proceeding (unless you're a react developer).

That's not a fog of war, that's delegation.

When an LLM generates your core logic and you can't explain why it works, that's a fundamentally different situation. You're not delegating — you're outsourcing the understanding, and that makes the result not yours.