The mechanisms of generation of musical software, pass in most cases through artificial intelligence techniques, more or less sosfisticate. Generally, the broken promises of AI, namely the simulation still too immature neuronal processes is reflected in equally poor compositional results. In Alvin, sonic incubator Jamie O'Shea, however carefully implements the mechanisms of a neural network to a piece of hardware in which eight different 'cells' produce sounds determined by mutual behavior. The connections between the cells are disrupted or restored physically (rather than just on / off) through the sound vibrations that shake the dust of steel, determining them. In this way the speakers are configured as 'creatures' capable of interacting with each other. The subsonic frequencies ranging from .5 hz to 100 Hz incrementandosi many more are the inputs (and therefore the relations established). Sometimes cells are disconnected to each other, starting the porcesso their evolution. This, however, can also be induced by the user via two buttons to inaugurate a new process ('procreate' / learn) or the end ('die' / forget), as they often refuse to 'terminate' at all. This clear separation of mechanical and computing and, with obvious discrepancies unpredictable is admittedly the author intended and produces a macroscopic view of the fragile interaction between analog and computational indirectly quoting the same sounds complicated coexistence between the respective nature.