In dynamical systems, one the many different ways to understand a flow consists in associating every orbit to a (finite or infinite) code. The process is similar with the one of catching a thief: by installing an alarm saying whether or not a thief is at the position x, we can get a rough idea of the position of the thief; by installing a sufficient number of alarms and by noting the successive alarms that go off, we can more and more precisely trace the trajectory of the thief and hopefully find the stolen diamond. My PhD's diamond consists in understanding the behaviour of a family of flows called Anosov flows and my alarms are called Markov partitions. In this week's talk, I will explain in a non Bourbaki way, how Markov partitions give us almost all the information about the flow and how they can be an important step in the classification of Anosov flows in dimension 3. We will assume no prerequisites in dynamical systems for this week's talk.