This isn't solving the problem at all, in my view. It's translating human speech into programmed code; it's sympy that actually solves the problem. As you said,

This isn't solving the problem at all, in my view. It's translating human speech into programmed code; it's sympy that actually solves the problem. As you said, this can be very useful, but knowing how to type the problem into wolfram alpha is not university level math. Problems that can be entered into sympy can be considered "solved", what's remaining (and what they have done) is UX. Also, this problem arguably is a problem of computation, not math. Math is about finding proofs, which is a much harder class of problems (NP-complete at best, undecidable at worst).