Here's an implication we care a lot about: If specialized knowledge becomes easier to turn into software, expertise no longer needs a software team before it can become useful to someone else. Your expertise doesn't need to be valuable to everyone. It only needs to be valuable to someone. That's the opportunity behind agent building.

As AI capabilities become more widely available, what will actually set one agent apart from another? The answer may have less to do with the model itself than we think. I summarize AI Layers from bottom to top: 1. Infrastructure - power grid, IDCs, clouds 2. GPUs and computing powers 3. Models, Data Brokers 4. Agents - capabilities (tools), trust layer (blockchain) 5. Interfaces - including robotics In a nutshell, AI will not only change the productivity but also start the product relation change, and the product relation change will be completed by Blockchain.

The bubble question is only half the story. The other half: what happens when AI stops answering and starts doing. That shift is the one we're building for.

The CAPTCHA era was about proving you’re human. The agent era might be about proving which machine you are. Today, “bot” often means something a website tries to block. Tomorrow, some bots may arrive with permission to spend money, make bookings or act on someone’s behalf. So the question is no longer just: Is this human? It’s also: Who does this agent represent? What is it allowed to do? What has it done before? Should you trust it enough to let it act? We spent decades building the internet around human users. What changes when the user isn’t always human?