> the issue is that there are no value adds from humans anymore if the LLM is smarter than everyone

The value adds (from the view of humans wanting it) come from:

1) a human understanding concept X & thus can communicate about X; or

2) finding alternate routes to X that make it easier for humans to get there.

(1) is human-centric & thus non-replaceable, since the requirement of a human is what gives it the value add.

(2) is similar to how there can be many paths to deriving the Pythagorean theorem. None of the existing paths devalue the others, but they each allow those with different perspectives to reach the same destination.