What worries me more than the bad apps is that the feedback loop never corrects itself, the model keeps saying yes and the person never hits the wall that would normally teach them something.