Xu Jialong from HKUST: The Agent moat has not yet solidified, and the differences in the model are more reflected in efficiency rather than disruptive breakthroughs According to BlockBeats, on April 21st, during a roundtable discussion on "Cracking Web 4.0: When AI Agents Take Over Chain Permissions," Xu Jialong, Associate Vice President of the Hong Kong University of Science and Technology, stated that there are differences in the model training paths and technical systems that different AI agents rely on, leading to significant differentiation in actual user experience. Recently, some new models and tools have shown better performance in terms of generation quality and execution efficiency, and even demonstrated higher limits in terms of development output. However, he pointed out that at the current stage, this difference has not yet formed a decisive generation gap and is closer to "efficiency improvement" rather than "paradigm breakthrough". In other words, the competition between agents is still in a period of rapid evolution, and stable and insurmountable technological barriers have not yet emerged. The current iteration pace of AI agents and large models is extremely fast, with new products or capabilities emerging almost every week, driving the industry to continue evolving forward. However, from the perspective of practical use and business decision-making, it still needs to be carefully evaluated whether it is necessary to continue to follow up on these changes frequently.