On April 21, MuleRun's Chief Technology Officer, Shu Junliang, shared insights at an event focused on 'Decoding Web 4.0: When AI Agents Take Over On-Chain Permissions.' According to BlockBeats, Shu emphasized that AI Agents should be viewed as 'personal assistants,' aiming to reduce usage costs and barriers through technological advancements. These agents are structured into modules such as interaction, perception, reasoning, and memory, each supported by different underlying technologies.
Shu highlighted that AI Agents are evolving from traditional text-based interactions on websites or apps to multi-channel communications across platforms like Telegram, Discord, Feishu, DingTalk, and WeChat. This shift aims to provide a seamless interaction experience, significantly lowering user barriers.
MuleRun has proposed an infrastructure solution centered on 'funds permission security' for on-chain scenarios. This includes sandbox isolation, cloud execution, and a fully traceable mechanism to create a trustless operating environment, addressing potential security issues during automatic execution by agents.
In terms of capability evolution, AI Agents will develop self-evolving decision models, continuously learning users' trading strategies and risk preferences to form personalized research and execution systems. Additionally, through a knowledge network mechanism, these agents will enable strategy accumulation and sharing, promoting the reuse and dissemination of on-chain cognition and capabilities.
Shu further noted that as AI Agent capabilities advance, the division of labor in on-chain transactions will be restructured. Agents will gradually take over information processing and execution tasks, allowing humans to focus on higher-level strategy formulation and critical decision-making.