Many people feel like they are picking up treasures when they open a DEX, as there are many popular projects with 10-300M in technical assets. However, when they open a CEX, they see projects with 300M-3B, which can be confusing. But in fact, there is a logic to picking up treasures on the CEX by following the trend:
1) The narrative of AI Agent is evolving rapidly, from AI MEME to single AI applications, AI Agent Launchpad, AI Agent framework standards, and AI Agent chaining.
Pure MEME has always been a logic of rapid hype; single AI applications can dominate a field in the short term due to the need to compete on experience; Launchpad platforms will be completely eaten up by Virtual's first-mover advantage; frameworks and standards have a lot of imagination space due to the complex technical logic, but are currently in a state of chaos; only the direction of AI Agent chaining is relatively "traceable".
2) The "chaining" logic must not be a public chain that says it can support all AI Agents running on the chain. There are public chains that have this potential, such as @NEARProtocol, ICP, and BNBChain. But the problem is that AI Agents are currently in the initial stage of "issuing assets" as MEME, and can only run AI Agents without being able to mobilize the enthusiasm of developers, which is useless.
Therefore, the first wave of "chaining" narratives will definitely come from the ai16z ecosystem. For example, @focEliza.
3) The core of "chaining" is to build infrastructure to better accommodate and serve AI Agents for large language models (LLMs), such as:
1. AI Agents need to realize autonomous generation of private keys and asset management, which will inevitably rely on the TEE infrastructure. Projects that can provide mature TEE solutions are potential Alphas in the old narratives of CEX.
2. AI Agents need to realize layered optimization and matching of Memory, with the goal of building Data Availability (DA) capabilities suitable for AI Agents. Traditional EVM public chains have extended the DA field with Blob add-ons, third-party DA Wars, and other interesting narratives.
3. AI Agents need to realize verifiable single-modal and trustworthy multi-modal interactions, which must rely on the trusted verification processing at the public chain level.
After the solutions for AI Agent "chaining" such as TEE, DA, Oracle, zkVM, and "chain abstraction" are mature, the things that AI Platform projects like io, Aethir, Vana, and SaharaAI are doing, such as decentralized distributed computing power, decentralized fine-tuning inference environment, decentralized data sources, and decentralized IP communication and incentives, will truly become "indispensable".
Clarifying this logic, it will be clear how to pick up the true treasures on DEX in the chaotic era, and how to pick up the treasures on CEX in the silence.
