Slow neural dynamics are believed to be important for behavior, learning and memory. Rate models operating in the chaotic regime show a rich dynamics at the scale of hundreds of milliseconds and provide remarkable learning capabilities. However, neurons in the brain communicate via spikes and it is a major challenge in computational neuroscience to obtain similar slow rate dynamics in networks of spiking neuron models. This question was addressed in a recent paper by Ostojic. The central claim of that paper is the existence of two states of asynchronous activity separated by a phase transition in spiking networks with fast synapses. We found that the analysis presented in the paper is factually incorrect and conceptually misleading. We provide compelling evidence that there is no such phase transition.