Author: Haotian

"Suddenly, a spring breeze comes, and even an iron tree can blossom with pear blossoms." How come so many DeFai projects have emerged like magic in such a short period of time? The standards and frameworks are not yet clear, and a new round of DeFai internal competition has begun? Okay, next, I will share from a popular science perspective, what's the deal with the major categories of DeFai projects?

1) These days, Poopman shared a DeFai ecosystem project distribution map, which has been widely circulated in the community, and the comment area shows that there are still many related projects that have not been included. Many people will be anxious, worrying about missing out on a financial code, but there is no need to worry.

First, it needs to be clarified that among them, there are indeed some AI "new" 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" flavor.

The core reason is that most of them are old projects that have been given new expectations through the new narrative of AI Agents, 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, it is some 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 will automatically complete the transactions.

Because of 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 the user's input prompt and the "precision" required for the AIGC backend to process the information and execute the request, 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 cannot accurately execute the Solver and cannot handle it.

But these products can also gain the trust of a large number of users with their novel interaction models and the solution of some basic Swap, Staking and other problems. The reason is that their 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, it will greatly reduce the usage threshold, and the AIGC backend's 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 brought the AMM Swap trading pool paradigm to the market years ago, which was also complained about the large slippage and friction at the beginning. The AI Abstraction subdivision track may be a bit weak in the short term, but the long-term opportunity for a major paradigm shift is worth watching.

2. Autonomous Portfolio Management & Yield Optimization: This type of product is the result of the internal competition 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 DeFi yield optimization strategies of these 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 capital allocation, arbitrage opportunity capture, yield 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 optimization 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 further optimized through the training and fine-tuning of trading strategies by AI agents, with the possibility of achieving higher efficiency investment opportunities than manual.

Moreover, when combined with AI agents, 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 agents to become high-level trading assistants 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-strong single-body AI, such as aixbt, which has indeed become a key information acquisition platform for many traders. However, users 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 the messages of AIXBT and automatically help me decide to buy the dips and arbitrage?

Theoretically, it is naturally feasible, but in fact, AIXBT and other AI agents may very well autonomously manage 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 these products is so fast, and with the commercial monetization capabilities driven by traffic, the imagination space is actually not small.

4. DeFai Infrastructure or Platform: This category of protocols 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 so on, such as Bittensor, io, Atheir, Hyperbolic, 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, 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 frameworks and standards, DeFai, Gamfai, MetAiverse and other focus narratives are all inseparable from these AI infra platforms.

That's it.

Although I have clarified a clear understanding 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 scenario, and can see progress and expectations step by step through experience, PMF product landing, etc.

This is also a manifestation of the current AI agent market's transition from virtuality to reality. Moreover, so many old species that couldn't find opportunities in the old DeFi era, isn't this an opportunity for them to unleash their potential in the new trend?