It gets worse if you imagine a future AGI which just tells us new novel implementations of previously unknown physics but it either isn’t willing or can’t explain the rationale.