I saw Luis von Ahn speak at the at the LAB in Miami circa 2013. His goal has always been to crowd source translating text like the New York Times everyday into several languages. Professional translators probably translate text better than Duolingo, but not at the price point they offer. One thing he brought up in that talk is that they hired a bunch of respected academics in the field of language learning. He asked them which parts of speech should be taught first which the consultants didn't have an answer so they implemented A/B testing to see which order of words and concepts people learned fastest with. Several years later the thing that surprises me is that with data backed learning they haven't got further. Working through he courses isn't such a bad waste of time for most of the users even though there might be better more expensive ways to learn a language.