The Dual-Helical Evolution of Network Computing: Toward Autonomous Intelligence Over Computing Power Networks

Driven by explosively growing application demands and rapid technological advances, network computing paradigms have continuously reshaped how computational resources are organized and allocated. However, computing infrastructures have become increasingly heterogeneous, distributed, and administratively fragmented, leading to significant resource silos. In response, the concept of Computing Power Networks (CPNs) has emerged to interconnect ubiquitous computing resources across domains, and CPNs are now undergoing rapid development with an urgent need for high-level intelligence. In this paper, we review the evolution of network computing along a dual-helical trajectory and provide a taxonomy of representative network computing paradigms, analyzing their characteristics and limitations. Drawing from this trajectory, we further identify the key challenges and development directions toward intelligent CPNs. As a pioneering exploration within this context, cabin computing demonstrates distinct advantages in coupling of intelligent computational elements, maneuverability, cross-domain scalability, and system autonomy. To further enhance the autonomous intelligence of CPNs across diverse application scenarios, we introduce Computing Agents by Intelligence Networks (CABIN*). In this framework, computing agents are interconnected as intelligence networks, collaboratively interacting with users to deliver personalized services. Therefore, each computing agent can automatically generate schemes to reconfigure coupled resources, including data, algorithms, and computing power. Subsequently, these schemes will be executed via cabin entities. Overall, this work offers a structured perspective on the evolution of network computing paradigms and lays a conceptual foundation for intelligent CPNs.