With collaboration I can't see how is possible there aren't already enough problem-solution sets for pottery much all subjects.

This way the publisher can own copyright on a massive set of problem-solution pairs and only employ one person for a short time? Seems to be financial optimisation rather than educational improvement.

Duolingo (language learning app) has generated materials and so often they're just slightly weird until they've been used, tested live by students, reported 100-times (or whatever), and finally corrected by the app operators.

It's good eventually, but if I were paying I'd be annoyed that there's not a bit more human effort put in.

With Open course materials it seems we'd already have enough materials to go around.