Using reinforcement learning for reasoning is a category error.

→ View original post on X — @pmddomingos