51WORLD (06651) has announced the development and deployment of 51Claw, an embodied intelligence base system built upon the open-source AI Agent platform OpenClaw. Functioning as a decision-making foundation for physical machines, 51Claw integrates cloud-based multimodal large models with local edge computing, endowing robots with spatial planning and natural language interaction capabilities. Technologically, 51Claw establishes a Real2Sim2Real closed-loop system that bridges the physical and digital worlds. The system collects multimodal spatial data to perform 3DGS/4DGS scene reconstruction, generating a physical world model that incorporates spatial memory. Leveraging the group's simulation training system, it uses reinforcement learning and physics engines to conduct task orchestration and motion training for robot models across vast scenarios, ultimately deploying mature agent models seamlessly to hardware in real-world environments. Currently, the system has been integrated with robotic dogs and humanoid robots, and has been connected to communication platforms such as Tencent's WeChat Work, forming a complete operational workflow loop. The board believes that the implementation of the 51Claw system represents an extension of the group's underlying technology, following core platforms like SimOne and DataOne, enabling its digital twin systems to directly perceive and manipulate the physical world. This advancement further solidifies the group's positioning in the spatial computing technology sector and aligns closely with its strategic goal of becoming the "first physical AI stock" in the capital markets, creating potential commercial growth opportunities for various business lines. For the long-term development of physical AI, the group plans to create an integrated "embodied intelligence training ground." The initial step involves using robotic dogs as carriers to enter specific commercial scenarios. In the long run, the group will combine the virtual training ground provided by its 51Sim platform with real-world physical training spaces to build a "from virtual to real" training and iteration loop. This embodied intelligence training ground aims to break the limitations of single hardware and scenarios, empowering a wide range of multi-category, multi-form embodied intelligence robots in the future, and continuously expanding the group's infrastructure service capabilities and industrial empowerment within the physical AI domain.