I think this is correct. But the article is also correct (If a bit breathless and overstating on the risk): having to reinterpret neuralese is a layer of indirection which introduces effort and inaccuracy