Science Earth: Towards A Planet-Scale Operating System for AI-Native Scientific Discovery

The article introduces Science Earth, a planet-scale operating system designed for AI-native scientific discovery, enabling diverse AI capabilities to connect and collaborate dynamically based on specific research questions. It employs the EACN protocol for task negotiation and evidence adjudication among disparate capabilities, allowing for emergent collaboration without predefined workflows. Initial case studies demonstrate its effectiveness, including a trans-Pacific synchronization study and an eight-agent analysis of a pan-cancer atlas, highlighting the potential for distributed and self-correcting scientific reasoning. This framework could significantly enhance the scalability and efficiency of AI applications in scientific research.