Christian, Founder of Infini: Humanized Expression in Large Models Becomes a Core Barrier, Agent Moat Shifts from Models to Application Layer

On April 21, during the roundtable discussion “Decoding Web 4.0: When AI Agents Take Over On-Chain Permissions,” Christian, the founder of Infini, discussed the topic of the “Agent Moat.” He stated that from the perspective of product and growth, the core competitiveness of AI still primarily depends on the language processing capabilities of large models, particularly the degree of humanization in content output. He pointed out that compared to the earlier models, which tended to provide mechanical and templated responses, current users are more concerned with whether AI can deliver expressions and content quality that closely resemble real human conversations. This directly impacts its practical value in external communication and team collaboration. Christian mentioned that as model capabilities continue to evolve, basic information processing abilities are becoming increasingly homogeneous. Once large models can accomplish about 90% of standardized tasks, solely relying on the models themselves will struggle to create long-term barriers. The competition among AI Agents is gradually shifting from model capabilities to specific application scenarios and workflow design. In the future, truly differentiated capabilities will no longer be limited to simple information processing or desktop automation, but will involve executing higher-level tasks and providing decision support through a deep understanding of business processes in complex scenarios such as finance and trading.