A new direction for AI Agent, a new round of DeFAI internal competition has begun.

Original author: Haotian (X: @tme l0 211 )

“Just as a spring breeze comes overnight, even the iron tree can blossom,” how did so many DeFai projects emerge like magic in such a short time? We haven’t even fully understood the standards and frameworks, and a new round of DeFai competition has begun? Well, next, I will share from an educational perspective, what’s the deal with the various categories of DeFai projects?

1. These days, @poopmandefi has shared a DeFai ecosystem project distribution chart, which has been widely circulated within the community. There are still many related projects that have not been included in the chart according to the comments. Many people may feel anxious and worry about missing out on wealth opportunities, but there’s really no need to.

First of all, it needs to be clarified that while there are indeed some AI ‘new’ projects with good Mindshare at present, such as: $AIXBT, $BUZZ, #NEUR 、#GRIFT, #Cod3x, etc., most of them are entirely emitting the ‘old’ flavor with new faces.

The core reason is that most of them are old projects that have been given new expectations through the new narrative of AI Agent, some old projects that have done a lot of optimization and experience work in the DeFi field but have been ignored, and some old projects that are difficult to discover under the background of the last VC traction and scattered attention of retail investors.

2. There are a total of four major categories in the figure below, and I will try to dissect my understanding one by one:

1. AI Abstraction: As the name suggests, it is a front-end product experience that encapsulates the information processing capabilities of AI large models through abstraction ability, allowing users to directly interact semantically. Users can input prompts and directly call the transaction interface in the dialog box front-end, and the AI backend will automatically complete the transaction.

As a result, such products are often scorned at first glance because of the significant friction in the early product interaction experience. For example, the ‘fuzziness’ of user input prompts and the ‘precision’ required for AIGC backend processing information and executing requests need ‘fault tolerance’ mechanisms. Either users find the instructions they can input and execute too simple to compete with the current DeFi experience, or they input too many high-expected instructions and discover that the program backend cannot execute the Solver accurately and cannot handle it.

But such products can also gain the trust of a large number of users with novel interactive modes and solutions to some basic issues such as Swap and Staking. The reason is that they have strong potential. Because the way users input prompts can be text, audio, etc., which is convenient for their habits, it will greatly reduce the threshold for use. At the same time, the processing capabilities of the AIGC backend 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 back then, and was also initially complained about the large slippage friction. In the short term, the AI Abstraction sub-track does seem somewhat mediocre, but the long-term opportunity for nurturing a major paradigm shift is worth attention.

2. Autonomous Portfolio Management & Yield Optimization: This type of product is the result of the last round of DeFi market frenzy. A large number of projects aiming to get a share in the DeFi race have been continuously working from the perspectives of personalization, customization, vertical segmentation, and specialized experience. However, before they could reap the fruits of victory, the DeFi industry has almost become desolate.

Most of these DeFi yield optimization strategies come from the team’s ability to monitor and analyze on-chain data, such as trading depth, fund flows, APY fluctuations, estimated slippage, price divergence, arbitrage opportunities, risk alerts, etc. Based on real-time on-chain data analysis, a set of execution strategies is formulated, such as position capital allocation, arbitrage opportunity capture execution, Yield income estimation, single pool or combination strategy, impermanent loss management, liquidation risk control, etc.

Simply put, the core of such products is the real-time on-chain data + transaction opportunity capture capability, plus a complete set of automated analysis and execution operation experience optimization based on smart contract. At first glance, what does it have to do with AI? The combination lies in the fact that data analysis and strategy formulation can be used in traders’ strategy training and fine-tuning to produce a set of investment opportunities that may be more efficient than manual work.

And, with the combination of AI Agent, the imagination space becomes even larger. Everyone can use their own strategy to fine-tune a personal trading preference AI Agent, to automatically help themselves find opportunities on the chain and execute trades automatically. Making AI Agent a sophisticated trading assistant for humans is a long-term, attractive, and online narrative.

3. Market Analysis or Prediction: This type of product, as a powerful standalone AI, has captured the majority of users’ mindshare. For example, @aixbt_agent has indeed become a key information acquisition platform for many traders as a top KOL. However, there is a lack of recognition for the actual application capabilities of AI agents that only provide trading strategy analysis, but less long-term imagination. For example, can my AI agent monitor AIXBT’s news, automatically help me decide to buy low and arbitrage, and so on.

In theory, it is feasible. In practice, AI agents like AIXBT could autonomously manage user assets and assist users in making trading decisions based on their own information. However, this step has not been taken yet. Currently, these types of products have gained popularity rapidly among users, and considering the commercial potential driven by traffic, the imagination space is not small.

4. DeFai infrastructure or Platform: This type of protocol covers a wide range of areas. In addition to emerging platforms such as Virtaual AI Agent Native, other project business frameworks related to AI computing power, data, fine-tuning, and other business areas, such as Bittensor, io, Atheir, @hyperbolic_labs, Vana, SaharaAI, etc., can be extended to this category.

After all, AI Agent needs to operate normally, data is oil, computing power is the power grid, reasoning is the transformer, AI Agent is the terminal, and they are all suppliers that provide services upstream.

So, there’s not much to say, the latter part of the AI Agent needs to gather strength, this type of DeFi platform will definitely exert great effort. Originally, AI Agent’s narrative is just the earliest part of AI Narrative, and the framework, standards, DeFai, Gamfai, MetAiverse, and other focal narratives are all inseparable from these AI infra platforms.

Above.

Although I have clarified a clear DeFai understanding perspective for everyone, it does not mean that I am not optimistic. Compared to the current chaotic and difficult-to-judge narrative framework and standard ‘chaotic era’ market, DeFai at least incorporates a more AI Agent-oriented application, and can see progress and expectations step by step through experience, PMF product landing, and so on.

This is also a manifestation of the current AI Agent market’s transition from virtual to real heat. Moreover, so many old species could not find opportunities in the old DeFi era. Isn’t it the opportunity for them to unleash their potential in the face of the new trend?

Original Link