in fairness, though, grok seems to be more solid on basic arithmetic than some other models.

→ View original post on X — @garymarcus