Vomer: A Human-Bionic Fundamental Architecture for Next-Generation General Autonomous AI Systems

This paper presents Vomer, a full human-bionic fundamental AI architecture designed to replace the traditional Transformer framework, breaking the long-standing fundamental bottlenecks of static von Neumann architecture in modern AI systems. Traditional Transformer-based AI systems face inherent limitations in infinite scalability, real-time fault tolerance, end-to-end security, autonomous resource scheduling, and closed-loop self-evolution. To solve these problems, Vomer fully replicates the physiological structure and operating mechanism of the human body, building a complete bionic AI system with core modules including a bionic sensory input system, translation core, Shu core for token scheduling, distributed execution units, hierarchical authority system, and biological self-healing & evolution mechanisms. This work redefines the underlying operating logic of general autonomous AI systems, shifting from static one-way inference to dynamic, closed-loop, autonomous biological operation. It provides a new fundamental paradigm for next-generation AI systems, which can be widely used in distributed AI clusters, industrial control systems, smart city management, aerospace systems, and other high-demand scenarios. The full open-source implementation of the Vomer architecture is available at:https://github.com/speciadres-ux/Vomer-ArchitectureThe permanent DOI of this preprint is reserved upon official publication.