Author:Haotian

 

I saw @shawmakesmagic retweet a project introduction of focEliza based on ELIZA for full-chain environment adaptation. My intuition tells me that the#ELIZAecosystem is going to make another bull run. So, what exactly does#focElizawant to do? Can its newly proposed TEE+ELIZA and DA+ELIZA become a necessity for the "chainization" of the ELIZA framework?

focEliza (Fully-on chain ELIZA) is a collection of ELIZA plug-ins designed for full-chain AI Agents. It is fully compatible with ELIZA and its core goal is to solve the problems of AI Agent trusted communication and subsequent permanent storage of AI Agent data in Memory.

Because the current ELIZA standard framework is just a "connector" that connects the LLM big model and social platforms such as Discord and Twitter, it only solves the developer's need to quickly deploy AI Agents, and does not have a "chain" design. What if we want AI Agents to have the ability to independently keep private keys, sign transactions and interact with chains? You will find that the existing ELIZA framework functions are far from supporting this.

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

1) TEE + ELIZA. The original AI Agent had to interact on the chain, but there was actually someone behind it controlling it. The Agent could not manage its private key independently, and it was difficult for it to independently determine when to sign a transaction. How could we make sure that there was no human control behind the AI ​​Agent's signature? The solution is TEE (Trusted Execution Environment);

I have introduced in many previous articles that TEE is a secure enclave environment that is isolated and extended at the hardware level. It can process data in a closed manner to ensure privacy. Its application logic is "available but invisible", which means that the computer can process the data in the application TEE, but cannot see the data in it. What if we let the AI ​​Agent generate the private key and put it in the TEE environment? This can solve the problem that the AI ​​Agent can independently generate, host and apply the private key.

In the early stage, if the AI ​​Agent publisher is not confident about the decision-making ability of AI, it can also use TEE for multi-signature processing, allowing humans and AI Agent to jointly manage private keys. In this way, some major behavioral triggers can be ensured to be included in the AI's own judgment, which is much more reassuring than being completely controlled by humans behind the scenes.

Imagine if focEliza can solve the problem of trustworthy custody on the asset side, it means that when interacting with the chain, the chain can directly verify that the signed transaction issued by the AI ​​Agent is from the Agent itself. The Verifiable capability at the chain level can record the usage log of each TEE, thus ensuring that the focEliza framework has the ability to process assets in a decentralized manner when interacting with the chain, which is a great benefit to the on-top extended AI fund management, AI lottery pool games, AI custody intention transactions and other landing scenarios.

2) DA +ELIZA. The inclusion of TEE mentioned above can solve the problem of verifiable transactions on the chain. focEliza also needs to solve the problem of AI Agent Memory storage reliability.

The logic is actually very simple. Every input and output of the AI ​​Agent must have context information. Only in this way can it avoid repeated information push and reflect its intelligent characteristics. Now the context information is more stored in the local database. If the database is abnormally closed or tampered with, it may cause the AI ​​Agent to work abnormally. If it is a text dialogue of knowledge information, it does not matter, but once it involves "transaction" issues, any database tampering may cause irreversible losses to the AI ​​Agent.

focEliza aims to extend the original on-chain Data Avaliablity (DA) features and build an on-chain DA layer framework to provide a decentralized full-chain on-chain interaction solution for AI Agent's memory data storage, call, and verification.

Just imagine that when all the content stored on the chain in the future becomes the necessary contextual information of the Agent, the Agent will achieve "immortality" on the chain. It is conceivable how much imagination space there is behind this.

In short, the emergence of focEliza will accelerate the further "chainization" upgrade of the ELIZA ecosystem, and in practical terms drive the integration of AI Agent + Crypto. However, it is still unclear whether focEliza and the original ELIZA framework have the opportunity to integrate in the direction of Tokenomics. Maybe it will become a new member of the ai16z family bucket? Or will there be a "The DAO"-level ELIZA community hard fork? Everything is unknown, and everything is worth looking forward to!

Anyway, lock in the innovation opportunities in the direction of framework standards and explore the value capture logic behind interesting innovations. Maybe the next big opportunity will be nurtured there.