Year-end is here, and various capital sources are predicting the trends for next year. But I’ve noticed a phenomenon—those beautiful trend reports are easy to spread, while the developers who actually get things done often scratch their heads when comparing these predictions.

For example, a leading VC’s crypto research team recently proposed three directions: iteration of AI research methodologies, the definition of proxy identities, and the cost structure of open networks. Sounds impressive, but when broken down, these are three urgent technical challenges we need to solve.

Instead of repeating others’ conclusions, it’s better to treat these as a technical requirements list. If you also agree that the future Web3 ecosystem will be driven by complex collaboration among AI agents, then now is the time to design the underlying protocols, system architecture, and value cycle mechanisms. Today, we’ll discuss practical technical solutions for these three major challenges.

**Building a "Nested" AI Collaboration Framework**

Existing AI agent frameworks solve the problem of "how multiple agents communicate," but frankly, they are still linear or tree-structured execution processes. The real breakthrough lies in "agent nesting"—an ecosystem model closer to a real research team: agents observe each other, evaluate each other, constrain each other, and optimize each other.

This requires a completely different system thinking approach. The key is to establish a "meta-evaluation layer"—a specialized review agent. Their prompt engineering is not centered around the task itself, but around the rigor of methodology, logical flaws, and innovation dimensions. These agents do not output the final answer but produce structured assessments of the process.