Lei Yang

So I want to guarantee, I want to kind of govern the, I want to sandbox the behavior of the agent.

So something we're thinking about is what if the blockchain can, as a built-in primitive, let's user describe either in natural language or in very simple like drag and drop user interface, describe what the transaction should achieve

in order to be successfully included on the chain, in the sense that me as a user can literally just type in whatever interface I use, saying that, okay, this transaction should at most deduct $1,000 from USDC, for example, from my account, and should at least give me back 0.5 ETH.

As long as this case is hit, I don't care what transaction it is, right?

Just agent go wild.

Then the agent can literally go wild.

And I have my peace of mind knowing that whatever the agent do, I don't really care.

I at least got 0.5 Ether back and I spent at most $1,000.

I think this is something that can be very useful.

And we're just very actively brainstorming

similar features we can help agents like this.

I think there's a great parallel to draw between software programming agents like the agents that you run on your desktop and the agents that run on a chain or uses the chain because I think sandboxing and just establishing this trust boundary is always the hardest part.

And I think as an, I would say, infrastructure company, I think this is what we should focus on.

Of course, we should also try to build our own agents, what have, but I think there are a lot of interesting stuff that we can actually meaningfully do to make agents' life easier, right?

So, yeah.