AI Excels in Math but Struggles with Everyday Coding, Sparking Education Debate

Anatoly Yakovenko posted that AI solves many math problems while staying poor at coding, calling this a sign of wasted PhD output over two decades. Salvatore Sanfilippo replied that solved problems can still help users study solution steps. Ben Dickson wrote that AI agents finish expert work quickly but leave users having learned nothing. Other replies highlighted cleaner training data in scientific research versus messier coding or statistics sources.