I believe many people feel like they are picking up treasures when they open DEX, as there are plenty of popular projects with decent tech ranging from 10M to 300M. However, when opening CEX, one finds only old narrative projects ranging from 300M to 3B, leaving mixed feelings. But in fact, CEX also has a logic for picking treasures along the trend; let me briefly explain:

1) The current wave of AI Agent narrative is rapidly evolving, from AI MEME -> single AI application -> AI Agent Launchpad -> AI Agent framework standard -> AI Agent chainization.

Pure MEME, no more to say, has always been a logic of rapid passing momentum; Single AI, due to the need to compete for experience, can dominate the short-term AIXBT; The Launchpad platform will definitely be consumed by Virtual, taking advantage of first-mover advantage; Frameworks and standards, backed by complex technical logic, have great imaginative space, but are currently still in a chaotic battle stage; Only the direction of AI Agent chainization is relatively 'traceable'.

2) The logic of 'chainization' is definitely not that a public chain suddenly says, I can support all AI Agents running on the chain. Such public chains must exist, for example,

@NEARProtocol, ICP, BNBChain all have such potential. But the problem is that at the current stage, AI Agents are still in the primary stage of MEME 'asset issuance', they can run AI Agents but can't stimulate developer enthusiasm, which is futile.

Therefore, the first wave of 'chainization' narratives must come from the ai16z ecosystem series. For example, @focEliza.

3) Moreover, the core point of 'chainization' lies in building infrastructure that better accommodates and serves large language models (LLM) for AI Agents, for example:

1. For AI Agents to achieve autonomous generation of private keys and asset management, it is necessary to rely on TEE infrastructure. Therefore, projects that can provide mature TEE solutions are potential Alphas in the narrative of middle-aged CEX. Because TEE has remained inconspicuous under competitive pressure from traditional ZK, MPC, FHE, and other cryptographic algorithms, but has suddenly become the core infrastructure of the new wave of AI narratives.

Imagine some small market cap, low presence, technically solid projects that ride the new trend; they must be potential treasures among many CEX targets (the performance logic of $PHA is precisely like this; if you agree with the value of TEE infrastructure for AI Agents, the momentum in the TEE track is definitely still there).

2. For AI Agents to achieve layered optimization and matching of Memory, the goal must be to build Data Availability (DA) capabilities suitable for AI Agents. Traditional EVM public chains have extended the DA sector with many exciting narratives such as Blob plug-in space and third-party DA War.

How to create DA capabilities specifically for AI Agents will also be a focal topic: including how DA block space accurately records effective semantics in LLM context, how Character role-setting plugins interact in real-time with DA blocks, how data in multi-block spaces effectively matches multi-modal interactions, and the cost issues of layered data storage in multi-block spaces (short-term, long-term, working memory, etc.) are all challenges that new DA builders need to tackle. This is also why focEliza takes DA capabilities as its core breakthrough point.

3. For AI Agents to achieve verifiability in single modalities and trustworthy interactions in multi-modalities, they must rely on public-chain-level trusted verification processing. In the TEE infrastructure, while solving the generation and application of private keys for large models, TEE alone clearly cannot solve the problem of physical cracking of single-point TEE hardware, as well as the consensus verifiability issue of TEE execution programs.

In the long run, these must rely on the blockchain's inherent decentralized node verification consensus and smart contract collaborative invocation environment to achieve.

In other words, a blockchain distributed system with strong consensus will make the intelligence of AI Agents become 'reliable' and guaranteed. Therefore, following this logic, projects that are currently committed to providing zkVM underlying frameworks, ZK Oracle solutions, ZK Bridge cross-chain solutions, and 'chain abstraction' public-chain-level application solutions are also part of this wave of 'chainization' narrative trend.

Above

Once a series of solutions providing 'chainization' capabilities for AI Agents mature, such as TEE + DA + Oracle + zkVM + chain abstraction, at that time, the decentralized distributed computing power, decentralized fine-tuning inference environment, decentralized data sources, decentralized IP communication and incentives that AI Agents need will truly become 'essential'.

To some extent, based on this effort towards AI Agent chainization, large AI platform projects like io, Aethir, Vana, SaharaAI will also have their place.

Clarifying this logic, how to obtain true insights amidst the chaotic era above DEX, and how to find treasures amidst silence in CEX; without further mention, clear judgment will naturally arise.