Rethinking the “hammer rule” for AI tool design

A user describes a longstanding instruction for agents called the “hammer rule”: don’t build human or language-model decisions into “dumb” tools that cannot think. They say having a “smart/magic tool” changes that approach, calling it a “magical state machine” or “magic hammer.” A reply offers a middle-ground framing, placing Jev-like systems between a highly efficient, special-purpose tool and a “general hammer.”