Infini Founder Christian: The "personification expression" of large models has become the core threshold, and the moat of agents is shifting from models to the application layer.

BlockBeats News, April 21 — During the roundtable discussion “Decoding Web 4.0: When AI Agents Take Over On-Chain Permissions,” Infini founder Christian discussed the topic of “Agent Moats,” stating that from the perspectives of product and growth, the core competitiveness of AI still primarily depends on the language processing capabilities of large models, especially the “human-like” quality of content output.

He pointed out that compared to earlier models with mechanical and template-based responses, current users are more concerned with whether AI can provide expressions and content quality close to real human conversations, which directly impacts its practical value in external communication and team collaboration.

Christian mentioned that as model capabilities continue to evolve, the basic information processing ability is becoming more homogeneous. Once large models can handle about 90% of standardized tasks, relying solely on the models themselves will no longer be able to form a long-term moat.

The competition for 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 understanding complex scenarios such as finance and trading, enabling higher-level task execution and decision support through deep understanding of business processes.