I noticed an interesting initiative from Ant Digital Technologies at the Web3 festival in Hong Kong. They introduced the 4R architecture, which essentially rethinks how AI agents should operate within the economy.

The point is that there are many problems in this area now — agents can't interact properly with each other, payments go through complicated channels, and no one knows who is responsible for what. Yan Yin from the company explained that this creates barriers to transactions and serious risks when collaborating.

Their solution is built on four levels. First is the Agentic Runtime with integration of DTClaw and the CARLI security model — this ensures control over AI operations. Then there are Payment Rails, which operate directly on the blockchain and enable secure payments. The third level is the Agent Registry using DID and ERC-8004 standards for identification. And finally, the Root Infrastructure, which supports microtransactions through Jovay Layer2 and ZKVM.

What I like about this approach is that it’s not just another framework. The 4R architecture actually addresses the trust problem in a system where AI agents need to interact autonomously. The company relies on its experience in financial security and blockchain technologies, so it looks serious.

This is a direction worth following if you're interested in the development of agent economies and how technology can scale securely.