ZTX to Launch AI Agent Gaming Ecosystem on Blockchain

- ZTX is the blockchain initiative of ZEPETO, the largest metaverse in Asia, which has over 430 million lifetime users. Founded in 2022 by Rudy Lee, Chris Jang, and Daewook Kim, ZTX raised $13 million in a seed round led by Jump Crypto to build out its Web3 platform. - The China AI agent market, valued at USD 577.0 billion in 2025, is projected to grow at a compound annual growth rate of 50.8% to reach USD 14,796.0 billion by 2033. Beijing is a major hub for AI startups in China, with companies like Zhipu AI and Moonshot AI attracting significant investment. - For orchestrating multi-agent systems, open-source frameworks like LangChain, AutoGen, and CrewAI are prominent. LangChain provides a modular toolkit for building LLM applications, while AutoGen, developed by Microsoft Research, excels at multi-agent collaboration and conversation-driven automation. CrewAI is designed for quickly prototyping multi-agent behaviors for linear tasks. - A key technical challenge in scaling multi-agent systems is managing coordination and communication overhead, as the number of potential interaction pathways grows exponentially with each new agent. Other significant hurdles include state synchronization failures, resource competition for memory and context windows, and the risk of cascading failures triggering retry storms that can overwhelm the system. - Recent AI research focuses on agent architectures that enhance reasoning, planning, and tool use. A survey paper from April 2024 outlines key themes in agentic architecture, including leadership impact on agent systems, various communication styles, and distinct phases for planning, execution, and reflection to build robust AI agent systems. - Effective user experience design for AI agents emphasizes transparency and user control to build trust. Emerging UX design patterns include providing clear explanations for an agent's actions, offering proactive "nudges" based on anticipated needs, and ensuring users can easily override or correct agent behavior. - In China's competitive landscape, major technology companies like Alibaba and Tencent are developing their own open-source agent frameworks, such as Qwen-Agent and Youtu-Agent, respectively, to compete with Western projects. Recently, a new Chinese AI agent system named CodeBrain 1 reportedly outperformed Anthropic's Claude Opus on the Terminal-Bench 2.0 coding benchmark. - Architecturally, multi-agent orchestration platforms typically include a planner to break down complex goals, an orchestrator to assign tasks and manage execution, specialized agents for specific functions, and a shared memory for maintaining context. This structure allows for greater task specialization, which can lead to cost efficiencies by using smaller, more focused models for individual tasks.