Written by: Haotian

After looking at the optimization points of the ELIZA V1 and V2 frameworks, I felt a significant improvement, especially in the unification of Agent wallet management, unified message management, and enhanced scalability capabilities. However, to be honest, the current ELIZA framework is still at the optimization level of basic development architecture and has not yet significantly expanded the 'imagination space'. I personally look forward to its potential for 'commercialization' and 'chainability' in the future.

1) A comprehensive Agent interoperability protocol, including basic message protocols for Agents (format standardization, routing mechanisms, state synchronization, etc.), dialogue interaction protocols for Agents (multi-Agent dialogue management, context information storage and state synchronization, etc.), resource sharing mechanisms for Agents (allocation of computing, storage, and other resources), task allocation mechanisms for Agents (intent understanding, task splitting, progress synchronization, aggregation rules, etc.), and identity and permission management layers for Agents (identity authentication, scoring systems, permission management mechanisms, etc.);

2) A Tokenomics protocol, including governance token definitions (ai16z or ELIZA?), incentive mechanisms for Agent participants (incentives for Agent developers, data contribution incentives, rewards for computing resource provision, verification node rewards, etc.), and a Gas economic system (Agent invocation fees, Memory storage resource fees, cross-chain operation fees, deflationary burning rules, fee distribution mechanisms, etc.);

In addition to these two core 'chainable' essential components, how to design a standard for Agent asset circulation management similar to the ERC20 standard, incorporate a decentralized storage system, decentralized verification mechanisms (memory systems, behavior systems), and decentralized resource allocation and incentive mechanisms, etc., are all issues that ELIZA's long-term direction for 'chainable' reform needs to consider.

Even if the goal of the ELIZA framework is not to chain, how to modularly integrate into the consensus layer of various chains, participating in governance and verification, and other aspects of collaboration is worth exploring.

Perhaps only when a 'chainable' Roadmap is clearly defined can EliZA secure its position as the leading AI Agent at the EVM level. But opportunities and variables also exist in this process.

What will ai16z iterate step by step? Can new frameworks like Arc rise to the top? Which framework will have the clearest route for chainable advanced development first? Who will be the first to find a commercially viable closed-loop path like Virtual? These are all directions to pay attention to. Friends who are concerned about the subsequent evolution of AI Agent framework standards can communicate more.