One Update, Twenty-Five Epochs In

Training uses a step decay schedule. The learning rate starts at $\alpha_0 = 0.4$ and is halved every 8 epochs:

$\alpha_e = \alpha_0 \cdot (0.5)^{\lfloor e/8 \rfloor}$

At epoch 25, one parameter currently sits at $w = 2.000$, and the loss gradient at that parameter is

$\frac{\partial L}{\partial w} = 0.6$

Gradient descent updates it with

$w \leftarrow w - \alpha_e \frac{\partial L}{\partial w}$

What is $w$ after this single update?