The Slope of a Loss Curve at One Weight

A model has been stripped down to a single trainable weight $w$. Its loss — the number training tries to push downhill — is

$L(w) = 2w^{3} - 9w^{2} + 12w + 4$

Gradient descent never uses the loss itself. It uses the slope of the loss, $L'(w)$, to decide which way to step.

Training currently sits at $w = 3$. What is $L'(3)$?