>This work shows a potential pathway to intepretability, which is actually part of what will give the labs that have it the sort of economy-swallowing capability

And how would you argument that? From what I see, if anything, it's the opposite.