Chinese feature requests
1) Duolingo team probably already knows this, but they really need to get the the speaking-back pronunciation check thing working for Chinese. I know, however, that this is particularly tricky for Chinese because of tones and their ambiguity in sentences. 2) Because it matters, an additional kind of interface needs to be made for character writing. There is a stroke order, and you can tell when a person has written a character in the wrong order because the strokes don't line up well (one references previous strokes as one makes the next mark, so if you went in a different order with a more complex character, things will line up wrong). I've seen versions of computerized character writing checks that have an "outline" of the character's strokes that you just follow, with then without arrows and numbers indicating stroke order. This would work well on a device with touch interface, but works OK on a screen with a mouse - and works terrible with a laptop touchpad. Totally freehand might be tricky, but with a visual grid and working character by character (not trying phrases or sentences), I bet you could get users' input regularized enough to grade.
Duolingo uses off-the-shelf speech recognition, and it seems they generally don't have access to the best available. In the meantime, you can enter answers using speech recognition on your phone, which you might well decide is preferable even when Duolingo manages to incorporate speech recognition into the course.
I wouldn't hold my breath for stroke order exercises. A whole new infrastructure applicable not even to an entire language's users, but one language's learners using presumably mostly tablets, which aren't all that common, doesn't seem like something likely to find itself high up the priority list. If they managed to license something, who knows, could happen, but if they'd have to develop it themselves, I think it's extremely unlikely. Duolingo isn't all that big staff-wise, and that would seem far afield of their existence competence.
Thanks, incidentally, for spelling out how improper stroke order reveals itself to the reader. I'd been curious about that.
A comment in another thread inspired me to check how Duo's speech recognition is doing these days: https://www.duolingo.com/comment/25768064$comment_id=25781569
In short, it doesn't seem that what they have now even successfully disambiguates syllables, so I'd think something as comparatively subtle as tones would be even more unlikely :(