We develop a new way to measure evolving production networks in real time using financial news. Combining artificial intelligence with a structural propagation framework, we extract firm-level production linkages and study how changes in network architecture relate to future output. We show that economies with centralized production structures experience slow subsequent growth, while dense and dispersed structures are associated with stronger performance. The framework yields firm-level measures of systemic importance that can be used by regulators to monitor macroeconomic risks. Our results show that the architecture of the production network is a measurable and economically meaningful determinant of aggregate dynamics.