How creditable would Claude be if it couldn’t answer “1+2=3?”

Worse, this is really human beings trying to pretend that their AI is AGI.