Here is the English translation: "As if a spring breeze came overnight, even an iron tree can blossom with pear blossoms." How come so many DeFai projects have emerged in such a short time like magic? The standards and frameworks are not yet clear, and a new round of DeFai internal struggle has begun? Okay, next, I will share from a popular science perspective, what's the deal with the major categories of DeFai projects? 1) In the past two days, @poopmandefi shared a DeFai ecosystem project distribution map, which has been widely circulated in the community, and the comment area shows that there are still too many related projects that have not been included. Many people will be very anxious, worrying about missing out on one financial code after another, but there is really no need to worry. First, it needs to be pointed out that among them, there are indeed some "new" AI projects with good Mindshare at the moment, such as $AIXBT, $BUZZ, #NEUR, #GRIFT, #Cod3x, etc., but most of them are old faces with an "old" smell. The core reason is that most of them are old projects that have been given new expectations by rewriting the narrative with AI Agent, some old projects that have done a lot of optimization work in the DeFi field but have been ignored, and some old projects that were difficult to discover under the background of the previous VC-driven retail attention. 2) The figure below has a total of four major categories, and I'll try to analyze my understanding one by one: 1. AI Abstraction: In simple terms, these are products that encapsulate the information processing capabilities of large AI models into user-friendly front-end product experiences that allow users to directly call transaction interfaces through semantic interaction, and the AI backend automatically completes the transactions. Due to this, these products are often looked down upon at first glance, because the early product interaction experience has a lot of friction, such as the "ambiguity" of user input prompts and the "precision" required for the AIGC backend to process information and execute requirements, which requires a "fault tolerance" mechanism. Either the user feels that the instructions that can be input and executed are too simple to match the current DeFi experience, or the user inputs too many high-expectation instructions and finds that the backend program does not have precise execution Solver to handle them. However, this type of product can also gain the trust of a large number of users with its novel interaction mode and the solution of some basic Swap, Staking and other problems. The reason is that its future potential is predictable. Because the user's input of Prompt in the form of text, audio, etc. can be a convenient way that fits the usage habits, which will greatly reduce the usage threshold, and the AIGC backend processing capabilities will gradually encapsulate more new Solver execution solutions to improve the user experience. Anyway, this is an attempt to explore a new trading paradigm, just like Uniswap's introduction of the AMM Swap trading pool paradigm to the market in the early days, which was also criticized for large slippage and friction. The AI Abstraction subdivision track may indeed be quite weak in the short term, but the long-term opportunity for a major paradigm shift is worth watching. 2. Autonomous Portfolio Management & Yield Optimization: These products are the result of the internal struggle in the DeFi market in the previous round. A large number of projects that want to get a share of the DeFi track have been working hard from the perspectives of personalization, customization, vertical subdivision, and specialized experience, but before they could reap the fruits of victory, the DeFi industry was almost desolate. The yield optimization strategies of these DeFi products are mostly derived from the team's ability to monitor and analyze on-chain data, such as trading depth, fund flows, APY fluctuations, slippage estimates, price deviations, arbitrage opportunities, risk warnings, etc. Based on these real-time on-chain data analysis, they formulate a set of execution strategies, such as position fund allocation, arbitrage opportunity capture and execution, yield income estimation, single pool or portfolio strategy, impermanent loss management, liquidation risk control, etc. In simple terms, the core of these products is real-time on-chain data + trading opportunity capture capability, plus a whole set of automated analysis and execution operation experience upgrade capabilities based on smart contracts. At first glance, what does this have to do with AI? The connection point is that data analysis and strategy formulation can be improved in trading strategy training and fine-tuning, with the possibility of generating a more efficient investment opportunity than manual. Moreover, when combined with AI Agent, the imagination space becomes even greater. Everyone can use their own strategies to fine-tune a personalized AI Agent to automatically help them find opportunities on the chain and automatically execute trades. Allowing AI Agent to become a high-level trading assistant for people is a long-term, sexy, and online narrative. 3. Market Analysis or Prediction: This type of product has already captured the majority of user Mindshare as a super-powerful single-body AI, such as @aixbt_agent, which has indeed become a key information acquisition platform for many traders. However, people recognize the actual application scenario capabilities of AI Agents that only provide trading strategy analysis, but lack a long-term imagination space. For example, can my AI Agent monitor AIXBT news, automatically help me decide to buy the dips and arbitrage? Theoretically, it is naturally feasible. In fact, AIXBT and other AI Agents may very well autonomously manage and custody user assets, and based on their own information decision-making, help users with trading operations. It's just that this move hasn't been made yet. For now, the user mindshare occupation speed of this type of product is so fast, and the commercial monetization capability driven by traffic is actually not small. 4. DeFai Infrastructure or Platform: This type of protocol covers a wide range, including not only the emerging platforms native to AI Agents such as #ai16z and #Virtaual, but also projects related to AI computing power, data, fine-tuning, and other businesses, such as Bittensor, io, Atheir, @hyperbolic_labs, Vana, SaharaAI, etc. After all, for AI Agents to operate normally, data is the oil, computing power is the power grid, reasoning is the transformer, and AI Agents are the terminals, and they are all upstream service providers. So there's not much to say, the second half of the AI Agent narrative needs to be accumulated, and these DeFi platforms will definitely play a big role. Originally, the AI Agent narrative is just the earliest part of the AI Narrative, and the frameworks and standards, DeFai, Gamfai, MetAiverse and other focus narratives are inseparable from these AI infra platforms. That's it. Although I have clarified a clear understanding perspective of DeFai for everyone, it does not mean that I am not optimistic. Compared to the current chaotic and narrative-difficult "Disorder Era" market of frameworks and standards, DeFai at least incorporates an AI Agent application, and can see progress and expectations step by step through experience and PMF product landing. This is also the performance of the current AI Agent market's transition from virtuality to reality. Moreover, so many old species that could not find opportunities in the old DeFi era, in the face of new trends, is not this an opportunity for them to release their potential?