Author: Haotian

I saw @shawmakesmagic retweet a project introduction for focEliza, which is based on ELIZA and designed for full-chain environment adaptation. My intuition tells me that the #ELIZA ecosystem is about to experience another Bull Run. So, what exactly does #focEliza aim to achieve? Can its newly proposed TEE+ELIZA and DA+ELIZA become essential for the "chainification" of the ELIZA framework?

focEliza (Fully-on chain ELIZA) is a collection of ELIZA plugins designed for full-chain AI Agents, fully compatible with ELIZA. Its core goal is to address the issues of trustworthy communication for AI Agents and the permanent storage of AI Agent data memory.

Currently, the standard ELIZA framework merely serves as a "connector" between LLM large models and social platforms like Discord and Twitter, only meeting the needs of developers for rapid deployment of AI Agents, without any design for "chainification." If we want AI Agents to independently manage private keys, autonomously sign transactions, and interact with the chain, we will find that the existing ELIZA framework capabilities are far from sufficient to support this.

To achieve this functional extension, focEliza proposes two core features:

1) TEE + ELIZA: Originally, behind the on-chain interactions of AI Agents, there are actually people controlling them. Agents cannot independently manage private keys and find it difficult to autonomously determine when to sign transactions. How can we ensure that the signatures of AI Agents are not controlled by humans? The solution is TEE (Trusted Execution Environment).

I have introduced TEE in several previous articles. TEE is a secure enclave environment isolated and extended at the hardware level, capable of processing data in a closed manner while ensuring privacy. Its application logic is "usable but not visible," meaning that a computer can process data within the TEE but cannot see the data itself. If we allow AI Agents to generate private keys and store them in a TEE environment, we can solve the problem of AI Agents independently generating, hosting, and applying private keys.

In the early stages, if the entity releasing the AI Agent is not confident in the AI's decision-making ability, TEE can be used for multi-signature processing, allowing both humans and AI Agents to jointly manage private keys. This way, significant actions can ensure that AI's own judgment is included, which is much more reassuring than having complete human control behind the scenes.

Imagine if focEliza can solve the trust issue of asset custody, it would mean that during chain interactions, the chain can directly verify that the signed transactions from the AI Agent come from the Agent itself. The verifiable capability at the chain level can record every usage log of TEE, ensuring that the focEliza framework has the ability to decentralize asset management during chain interactions. This would be a significant advantage for various scenarios such as AI fund management, AI lottery pool games, and AI custodial intent transactions.

2) DA + ELIZA: The inclusion of TEE can solve the issue of verifiable transactions on the chain, but focEliza also needs to address the trustworthiness of AI Agent memory storage.

The logic is quite simple: every input and output of the AI Agent's work must have contextual information. Only in this way can we avoid repetitive information pushing and reflect its intelligent characteristics. Currently, contextual information is mostly stored in local databases, and if the database unexpectedly shuts down or is tampered with, it may lead to abnormal AI Agent operations. While this may not matter for knowledge information text dialogue processing, any tampering with the database regarding "transactions" could lead to irreversible losses for the AI Agent.

focEliza aims to extend the original on-chain Data Availability (DA) characteristics by constructing an on-chain DA layer framework, providing a decentralized full-chain interaction solution for the memory data storage, invocation, and verification of AI Agents.

Imagine that in the future, all content stored on the chain becomes necessary contextual information for the Agent, allowing the Agent to achieve "immortality" on the chain. The potential behind this is immense.

In summary, the emergence of focEliza will accelerate the further "chainification" upgrade of the ELIZA ecosystem, practically driving the integration of AI Agents and Crypto. However, it remains unclear whether there are opportunities for integration between focEliza and the original ELIZA framework in areas like Tokenomics. Perhaps it will become a new member of the ai16z family? Or will there be a "The DAO"-level hard fork of the ELIZA community? Everything is uncertain, and all is worth looking forward to!

Anyway, let's focus on innovative opportunities in the direction of framework standards and explore the value capture logic behind interesting innovations. Perhaps the next big opportunity is brewing within!

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。