Adam and RMSProp on the Second Update

Two optimisers are run on the same parameter from the same start, with $\alpha = 0.02$ and $\epsilon$ small enough to ignore. Both receive the identical gradient $g_t = 0.5$ at every step, and both begin with all accumulators at zero.

RMSProp
$s_t = \beta_2\,s_{t-1} + (1-\beta_2)\,g_t^2, \qquad \Delta\theta_t = -\frac{\alpha}{\sqrt{s_t}}\,g_t$

Adam
$m_t = \beta_1 m_{t-1} + (1-\beta_1)\,g_t, \qquad s_t = \beta_2 s_{t-1} + (1-\beta_2)\,g_t^2, \qquad \Delta\theta_t = -\frac{\alpha}{\sqrt{s_t}}\,m_t$

with $\beta_1 = 0.9$ and $\beta_2 = 0.99$.

On the second update ($t = 2$), what is the size $|\Delta\theta_2|$ of each optimiser's step?