Next, Byte trained a neural network on publicly released datasets of the original architects’ speech and handwriting. After thousands of iterations, the model produced a synthetic “signature” that, when fed to the verification system, produced a soft acceptance—just enough for the AI to grant limited read access.