I've been pondering something over the past few weeks: if AI is no longer just your chat partner, no longer just giving advice, but actually starts doing things for you—transferring money, placing orders, calling resources, interacting with other systems, is the world really ready for that?

This is not just about writing a copy or a trivial task. I'm talking about something much more hardcore—real money circulation and cooperation calls between systems. The more I think about it, the more it feels a bit spine-chilling. We spend all day praising how smart the models are and how fast the reasoning is, but very few people have properly dissected: when AI truly starts to take action, what insurance is relied upon at the base level?

It was also because of this question that I began to pay attention to the Kite project.

To be honest, it is particularly "low-key". There isn't any overwhelming promotion, and the concepts aren't piled up much. But upon closer inspection, you'll find that these people are a bit "unusually honest". It feels less like they are chasing trends, and more like a group of people sitting down to thoroughly dissect the potential real issues that may arise in the future.

Most AI projects talk about capabilities. Kite instead talks about boundaries first. This approach touches a pain point.

It breaks down the identity aspect into three dimensions: who a person is, who the agent is, and who is responsible for this operation. On the surface, it sounds a bit technical, but the essence is quite simple— you cannot let a system handle money without being clear on "who is doing what and to what extent." Once this logic is established, everything becomes stable: users gain a sense of security, agents have room for activity, and the system will not go out of control.

What impresses me the most about Kite is its deep understanding of the "payment" aspect. I have always believed that in the future, the frequency of AI transactions will definitely far exceed that of humans. I am not talking about one or two transactions, but rather large-scale, high-frequency automated trading. Once this scenario arises, the traditional payment model will need to be redesigned.