A = np.array([[1, dt, 0],
[0, 1, dt],
[0, 0, 1]])
If state is $x$, $\dot{x}$, $\ddot{x}$, then this is an Euler stepper for position and velocity, with acceleration remaining what it was. But that zero in the upper right could be (dt**2)/2, which just takes its Taylor expansion one term further. Not sure this would affect answers much, but maybe worth an experiment.