Is it possible to infer the waypoints (i.e., the intermediate destinations) of a mobility trajectory in the absence of timing information? By mining a dataset of real mobility traces and fitting a Markovian model, we find that the entropy of conditional Markov trajectories enables us to uncover these waypoints, even though no timing information nor absolute geographic location is provided. The entropy of (conditional) Markov trajectories can be expressed as a linear combination of local entropies at each state (i.e., location) of the trajectory. This makes its computation tractable, thereby leading to efficient algorithms for trajectory segmentation.