Real-time robust communication in dynamic and weak wireless network environments is a highly complex system engineering challenge. From the perspective of coding and transmission technologies, existing systems typically adopt a separated architecture of source coding for compression and channel coding for transmission, where effectiveness and reliability are optimized independently. Both coding techniques have already approached their theoretical limits. However, in the face of massive emerging services, the potential for further improvement in transmission efficiency has nearly been exhausted, making it difficult to guarantee the real-time performance and robustness of communication. This calls for a new coding and transmission paradigm to enable highly time-sensitive and robust end-to-end service delivery. This report leverages next-generation artificial intelligence technologies to achieve agile and intelligent adaptation of coding and transmission to rapidly varying environments by analyzing service and environmental characteristics. Furthermore, it exploits semantic contextual distribution of data to realize high-fidelity compression at low bitrates, and employs strong prior knowledge to efficiently compensate for transmission distortions. By suppressing error propagation in decoding and mitigating the “cliff effect” and “saturation effect” caused by dynamic channel variations, the proposed approach ultimately achieves real-time and robust communication.