My dissertation examines the consequences of open-science publications for firms in artificial intelligence sectors. I develop and empirically test a theory of the impact of publications on a firm’s ability to shape external knowledge sources, innovate cumulatively, and realize long-term, generative value from their inventions. Publications trigger a “broadcast search” process that helps a firm attract a diverse pool of highly original external follow-up inventions. Moreover, publications serve as a common knowledge interface that facilitate the inventing firm’s learning from the external follow-up inventions. As a result, the focal firm not only creates more internal cumulative inventions, but also inventions from a broader spectrum of technology fields and of higher private value. I further explore various contingencies in the benefits of publications. Firms differ in their ability to benefit from the publication-triggered learning: diversification in a firm’s upstream technological assets facilitates the firm to learn more effectively, while diversification in a firm’s downstream industry scope enable the firm to capture more financial profits from publications. Empirical analyses based on publication and patenting activities of 237 AI-active publicly-traded firms in the U.S. provides support for my theoretical arguments.