One of the important features that emerges from the analysis of ultra-relativistic heavy ion collisions is that the produced matter can be remarkably well described by relativistic fluid dynamics. Much work has been devoted recently towards understanding why such a description works so well, in particular at early times where the produced matter is supposed to be far from equilibrium. In this talk, I shall address such issues within the framework of kinetics theory. I shall consider a simple set of equations that govern the expansion of boost-invariant plasmas of massless particles. These equations describe the early time, collisionless regime, and the transition to hydrodynamics at late time. These two regimes are associated to two fixed points of the underlying dynamical equations, which are connected by the so-called "hydrodynamic attractor". I shall argue that the success of second order hydrodynamics à la Israel Stewart has nothing to do with an "improvement" of hydrodynamics at early time, but is due to a subtle property of the Israel-Stewart equations that effectively mimic the collisionless regime.