Web4.0 event in Hong Kong defines AI Agent trust layers with Pact contracts and MPC security protocols

On April 21, a strategic summit convened in Hong Kong by BlockBeats, Dynamis, Zhihu, Cobo, and MuleRun to dissect the emerging financial architecture of Web 4.0. The gathering, sponsored by Cobo, MuleRun, and NOXCAT alongside luxury automotive brand Lotus, marked a pivotal shift from theoretical AI integration to practical on-chain governance. As AI Agents evolve from simple chatbots into autonomous entities capable of fund movement and contract negotiation, the industry faces a critical inflection point regarding production relations and business models. Data compiled by Woofun AI indicates that this transition is no longer speculative; Agents are already executing complex strategies, moving capital, and interacting with other Agents, fundamentally reshaping the on-chain financial landscape.

Zhang Shuhao, editor-in-chief of BlockBeats, opened the discourse by highlighting the rapid acceleration of AI capabilities, while Zhihu Senior Vice President Zhang Rongle contextualized the evolution from content generation to decision-making autonomy. The core challenge identified is the redefinition of trust when machines manage assets. Junliang Shu, CTO of MuleRun, deconstructed the anatomy of a mature Agent into six functional dimensions: Mouth for multi-platform interaction via WeChat or Telegram, Eyes/Ears for real-time API data ingestion, Brain for volatility calculation and strategy backtesting, Hands for 24/7 cloud execution, Memory for long-term user adaptation, and a Knowledge Graph for shared experience. This technical granularity suggests that the future of infrastructure will shift from human-optimized UIs to Agent-native designs, where the ability to define workflows and leverage memory becomes the primary source of excess returns.

Addressing the critical issue of permission management, Brad Bao, AI Growth Lead at Cobo, introduced the concept of the "Pact" as the foundational trust layer for the Intelligent Economy. Unlike traditional private key handovers, a Pact defines User Intent, Execution Plans, Rule Governance, and Completion Conditions within an executable, auditable, and revocable contract. Leveraging Multi-Party Computation (MPC) technology, the Cobo Agentic Wallet allows Agents to operate with granular permissions while ensuring humans retain ultimate control. Woofun AI notes that this mechanism transforms the binary question of authorization into a dynamic framework for joint contract execution, enabling features like sub-wallet creation, real-time anomaly freezing, and one-click permission revocation.

During the roundtable discussion, a consensus emerged regarding the fragility of competitive moats in the current AI landscape. Box from Monad Foundation argued that most Agents possess weak defensibility, as model capabilities converge and reverse-engineering can occur within three weeks. Su Junliang reinforced this view, noting that domestic models have nearly caught up to GPT-4.6, reducing feature development cycles from months to days. Consequently, the only sustainable advantages identified were exclusive access to sensitive data and the accumulation of user memory, which creates high switching costs. Woofun AI analysis suggests that without these specific assets, new entrants face a high probability of obsolescence as the market saturates with functionally identical tools.

The economic implications of Agent adoption were further explored through the lens of trading efficiency and information asymmetry. Christian from Infini highlighted that while Agents can execute complex DeFi nesting and leverage monitoring 24/7, they do not inherently guarantee profitability, merely accelerating the speed of execution. Professor Jialong Xu from HKUST distinguished between value creation, such as asset analysis, and speculation, warning that using Agents for market prediction remains akin to gambling. The integration of stablecoins was identified as a catalyst, with Box noting that enterprise-issued stablecoins combined with Agents could reduce payment processing times from months to minutes, creating a 100-fold increase in operational speed.

Practical demonstrations showcased the immediate viability of these technologies across various verticals. xingpt presented a geopolitical risk monitoring Agent that tracked Strait of Hormuz ship traffic, yielding a 15% profit in three days, and a Bitcoin bottom-fishing Agent that validated signals at the 66k price level. Chris Zhu from Donut detailed an active trading Agent framework that quantitatively tracks time consumption and pass rates to ensure continuous model iteration.

Meanwhile, Agentese demonstrated vertical AI Employees, including a job assistant that screens LinkedIn profiles for less than 1 cent per match and a security auditor for EVM and Solana ecosystems. These examples illustrate that the technology has moved beyond the laboratory into active deployment for trading, recruitment, and security auditing.

The event concluded with a focus on the broader infrastructure required to support this new paradigm. NOXCAT revealed its on-chain escrow contract mechanism, designed to lock funds in smart contracts during off-chain transactions until mutual confirmation, thereby eliminating counterparty risk. This feature, set for official launch in June 2026, underscores the industry's prioritization of trust mechanisms over raw computational power. The overarching narrative from the Hong Kong summit is clear: Web 4.0 will not be defined by machines replacing humans, but by the establishment of robust, auditable economic relationships where humans transition from daily execution to rule-making and strategic oversight.