The project proposes a new strategy for creating evolving VR structures (3D audio and graphics) based on the idea of adaptation to a dynamically changing environment and with the use of advanced AI methods (transformer model). The project investigates a topology of oscillations, gradients and fluctuations where each 3d manifold potentially hosts new manifolds, balance of forces between tension and relaxation, expansion and contraction. It is a constantly evolving assembling and unfolding mechanism. There is no division between the performer and the environment. The "inner" space is topologically in contact with the "outer" space. By significantly expanding existing research, the project creates a machine learning model useful for testing various aspects of adaptation to a complex dynamic environment and sound spatialization.