Hybrid Neuro-Symbolic models are nowadays commonly used for intelligent information processing. Nevertheless, no serious theoretical study was performed in the past to settle a solid basis to hybridization. From this idea, a european ESPRIT project, the project MIX, funded by the European Union has been proposed, in order to study hybridization, develop tools and propose strategies to better analyse and emulate hybrid models. We explore in this paper the main results (theoretical and practical) of this project. They correspond to a toolbox for the implementation and communication of neural and symbolic agents and to some taxinomic aspects for the better classification and understanding of hybrid models. Thanks to these results, we were able to propose a general methodology for hybridization and to assess it thanks to real-world applications. Both aspects are also evoked in the paper.