“The machine may appear more mathematically intelligent partly because it is much less constrained by a cognitive limitation that suppresses human performance.”
Working memory is the constraint humans hit and machines do not. That explains a lot of benchmark performance without requiring anyone to claim emergent reasoning. The supporting evidence is the useful part: kids with matched IQs but better working memory do better at math anyway. Capacity and reasoning are separable, and the labs selling reasoning have every incentive to pretend they are not.