Maybe I am just misunderstanding something, but I feel like you might be contradicting yourself here... why can LLM output not be copyrighted, but compiler output can be?
No, that's the point - the compiler output is only copyrighted to the extent that it is derived from the source code. The compiler itself cannot create anything copyrightable, but because there is a deterministic link between the source code and the binary code, and the source code was the product of a human, the binary code is covered by the source code copyright.
It's like a photocopier. If you photocopy a page from a book, that page is still covered by the copyright of the book author, even if the page is 2x larger or otherwise transformed by the machine.
It's like a photocopier. If you photocopy a page from a book, that page is still covered by the copyright of the book author, even if the page is 2x larger or otherwise transformed by the machine.