Yes, as everybody else has said, if you could make this produce graded readers you'd be onto a winner, but it would need to be limited to words from a frequency list. If they could get this to work for Chinese I'd be very happy, it would be amazing to be able to dial up the language complexity so you constantly maintained n+1 comprehensible input. The application they're hinting at here is a bit silly though.