Lei Yang

I don't think I check the diff AI works out anymore.

I just care if it works, like if it kind of adheres to some kind of spec, I give the AI.

So I think the same can be said for on-chain transactions.

We envision that AI agents

will help users make sense of contracts, hidden and public ones.

And we expect that AI agents will help users to transact and to actually achieve their, say, financial goals or whatever they want to achieve, for example, swapping, right?

So what we're working on is trying to put the necessary boundaries

to AI agents, it is almost similar to intent protocols.

We want a way for users to specify the end goal they have for their AI agents, and also what the AI agents can do or cannot do.

I'll just make an example.

If I want to swap from, say, $1,000 of USDC to, I don't know, maybe 0.5, maybe 0.25 Ether,

i can set i can state that this ai agent can at most withdraw this amount of usdc from me and if they withdraw the money they have to at least repay at least for example 0.5 ether to the same address and then the ai agents can go all crazy and do whatever they want as long as they do not break my my my constraints so we're thinking about

how we can optimize the chain for this kind of use models, kind of compartmentalization and adding restrictions to AI agents.

And also we think MegaEth is a great place for AI agents to use as a playground, because if you let AI agents go crazy, they burn a lot of tokens.

They also have many failed attempts, right?

So if you try to use AI agents to automatically send transactions on your behalf on an expensive chain, maybe the transaction cost they can add up.

But on a chain like MegaEth, this is not an issue.

So this is kind of a snippet of what we think about AI agents.

We do have something that is very interesting that we're trying to announce next month.

So yeah, stay tuned.