Author of the original text: Poopman

Original text compiled by: TechFlow

What kind of sparks will be generated when traditional DeFi meets the emerging AI? What kind of new hybrids or technological innovations can we create?

Today, we will explore the early ecosystem of DeFAI (Decentralized Finance + AI) together.

Hopefully this article can provide you with some inspiration!

(* I will soon publish a 20-page in-depth analysis article on Medium. Today's content is just a quick overview to give you a fast understanding of this emerging field.)

Why focus on DeFAI?

The combination of Artificial Intelligence (AI) and blockchain is not new. From the early decentralized model training in the Bittensor subnet, to the decentralized GPU and computing resource markets like Akash and io.net, and now the emerging integration of AI and memecoins on Solana, each stage has demonstrated how blockchain can complement AI capabilities through resource aggregation, driving the realization of sovereign AI and consumer-level applications.

According to CoinGecko data, as of January 13, 2025, the total market capitalization of DeFAI has reached around $1 billion. Among them, Griffain occupies 45% of the market share, while $ANON occupies 22%.

Starting from December 25, 2024, with the return of "American capital" after the Christmas holidays, the DeFAI industry has begun to accelerate its development, driven by the emergence of frameworks and platforms such as Virtual and ai16z.

This is just the beginning. The potential of DeFAI far exceeds its current performance.

Although the current applications are still in the proof-of-concept stage, we should not underestimate their potential to transform DeFi into a more intelligent, user-friendly and efficient financial ecosystem through AI technology.

Before delving into the DeFAI ecosystem, we need to first understand the basic principles of how AI agents operate in the DeFi and blockchain environment.

How AI Agents Operate in DeFi

AI agents are programs that represent users and perform tasks according to specific workflows. The core of these agents is supported by large language models (LLMs), which can generate responses based on their training data.

In the blockchain, agents can interact with smart contracts and accounts, handling complex tasks without the need for continuous user intervention.

For example:

• Simplifying the DeFi user experience: Completing multi-step cross-chain bridging and liquidity mining operations with a single click

• Optimizing liquidity mining strategies: Providing users with higher returns

• Automating trade execution: Buying or selling assets based on market analysis (whether third-party or their own models)

Referring to the research by @threesigmaxyz, AI models typically follow these 6 core workflows:

• Data collection

• Model inference

• Decision making

• Custody and operation

• Interoperability

• Wallet management

Once you have assembled the above 6 core elements, you can build your own autonomous agents on the blockchain. These agents can play different roles in the DeFi ecosystem, thereby improving on-chain efficiency and the user's trading experience.

Exploring the World of DeFAI v2

Overall, I categorize the integration of DeFi and AI (DeFAI) into four main areas:

• Abstracted/User-friendly AI

• Yield Optimization and Portfolio Management

• DeFAI Infrastructure or Platforms

• Market Analysis and Forecasting

Abstracted AI or AI ChatGPT

In this area, the ideal AI solution should have the following capabilities:

• Automatically execute multi-step transactions and Staking operations without requiring users to have any specialized knowledge.

• Conduct real-time market research and provide users with the key information and data they need to make informed trading decisions.

• Aggregate data from multiple platforms, identify market opportunities, and provide comprehensive analysis for users.

Next, let's take a look at some popular tools in this field:

Griffain

@griffaindotcom is currently the first and best-performing abstracted AI tool on the Solana blockchain, supporting the execution of transactions, wallet management, NFT minting, and quick Token purchases.

Its main features include:

• Ability to complete transaction operations with natural language input

• Launching Token projects, minting NFTs, and selecting addresses for airdrops through Pumpfun

• Multi-agent collaboration functionality

• Agents can post tweets on behalf of users

• Automatically purchasing newly launched Meme coins on Pumpfun based on specific keywords or conditions

• Automated Staking and DeFi strategy execution

• Task scheduling, where users can customize personalized agents by inputting memory data

• Aggregating data from multiple platforms for market analysis, such as identifying the main holders of a Token

Wallet functionality:

When creating an account, the system automatically generates a wallet through Privy. Users can authorize their account to the agent, and the agent will independently execute transactions and manage the investment portfolio. To enhance security, the private key is divided and stored using Shamir's secret sharing technology, ensuring that neither Griffain nor Privy can independently control the wallet.

Anon

@HeyAnonai is created by the renowned developer @danielesesta, who previously founded the DeFi protocols Wonderland and MIM. Anon aims to simplify DeFi interactions, making it easy for both new and experienced users.

Key features include:

• Cross-chain asset bridging based on LayerZero

• Providing real-time price updates and data through Pyth

• Offering automated operations and triggers based on time and Gas price

• Providing real-time market insights, such as sentiment analysis and social data analysis

• Supporting lending operations with protocols like Aave, Sparks, Sky, and Wagmi

• Offering natural language transaction functionality in multiple languages, including Chinese

Furthermore, Anon has recently released two important updates:

• Automation framework

• Agent functionality focused on Gemma research

These updates have made Anon one of the most anticipated abstracted tools.

Slate (not yet launched)

Slate is supported by BigBrain Holdings investment, and its founder @slate_ceo positions it as an "Alpha AI" capable of autonomous trading based on on-chain data signals. Currently, Slate is the only abstracted AI tool that can achieve trade automation on the @hyperliquidX platform.

One notable aspect is their fee structure.

Slate's fees are divided into two categories:

• General operations: Slate does not charge any fees for regular transfers or withdrawals. However, for more complex operations such as Swapping, Bridging, Claiming, Borrowing, Lending, Repaying, Staking, Unstaking, Longing, Shorting, Locking, and Unlocking, the platform charges a 0.35% fee.

• Conditional operations: If users set conditional orders (e.g., limit orders), Slate will charge fees based on the different condition types:

• 0.25% fee for Gas-based conditional operations;

• 1.00% fee for all other conditional operations.

Here is the English translation of the text, with the specified translations applied: Except for Slate, there are many emerging abstract AI tools in this field. Here are some representative projects: @AIWayfinder @orbitcryptoai @dolion_ai @askthehive_ai @HeyElsaAI @Spectral_Labs @Infinit_Labs @ProjectPlutus_ @bankrbot And more projects under development... Here is a comparative table of several abstract AI tools: Image: Compiled by TechFlow Automated yield optimization and investment management: Unlike traditional yield strategies, the DeFi protocols in this field use AI to analyze on-chain data, identify trends, and provide insights to help teams develop more efficient yield optimization and portfolio management strategies. T 3A I @trustInWeb3 is a lending protocol that supports under-collateralized loans, using AI as an intermediary and risk management engine. The T 3A I AI agent can monitor the health of loans in real-time and ensure loans remain repayable through its risk metric framework. This is an interesting application of AI in DeFi. Kudai @Kudai_IO is an experimental agent focused on the GMX ecosystem, developed by the GMX Blueberry Club using the EmpyrealSDK toolkit. Currently, the $KUDAI Token is trading on the Base network. Here is the Kudai roadmap: Image The core idea of Kudai is to use all the trading fees earned through $KUDAI to fund the autonomous trading operations of the agent, and return the profits generated by these operations to the Token holders. In the upcoming second phase (out of four), Kudai will have the following features that users can trigger through natural language commands on Twitter: - Purchase and stake $GMX to generate new revenue streams - Invest in the GM pool of GMX to further increase yields - Purchase GBC NFTs at floor prices to expand their investment portfolio Sturdy Finance V2 @SturdyFinance is a protocol that combines lending and yield aggregation functions. It dynamically allocates funds between different whitelisted isolated pools using AI models trained by Bittensor SN 10 subnet miners to optimize yields. Sturdy's architecture is divided into two layers: isolated pools and the aggregation layer. - Isolated pools: These are single-asset pools where users can only lend one asset or use one collateral type, reducing cross-asset risk. - Aggregation layer: Built on Yearn V3, user assets are allocated to whitelisted isolated pools based on utilization and yield. The Bittensor subnet provides the optimal allocation strategy. When users lend to the aggregation layer, their risk is limited to the chosen collateral type, avoiding risks from other lending pools or collateral assets. Image Other representative projects in the yield optimization and investment management space include: @derivexyz @Thales_ai @Mozaic_Fi @boltrade_ai @vainguard_ai @Ensofi_xyz @0x ARMAgeddon @glamsystems And more projects under development... Market Sentiment Analysis AI Agents AIXBT @AIXBT_agent is a market sentiment tracking agent that integrates and analyzes data from over 400 key opinion leaders (KOLs) on Twitter using its proprietary engine. AIXBT can capture market trends in real-time and provide valuable insights to users around the clock. Among all the DeFi AI agents, AIXBT occupies 14.76% of the market attention, making it one of the most influential agents in the ecosystem. Image AIXBT's functionality is not limited to providing market insights; it is also interactive, able to answer user questions and even issue tokens through the Twitter platform. For example, the $CHAOS token was created by AIXBT in collaboration with another interactive bot, Simi, using the @EmpyrealSDK toolkit. Other market analysis agents include: @tri_sigma_ @ASYM 41 b 07 @kwantxbt @gemach_io DeFi Infrastructure and Ecosystem Platforms The realization of Web3 AI agents relies on decentralized infrastructure. These projects not only provide model training and inference services, but also data, verification mechanisms, and coordination layers for the development of AI agents. Whether in Web2 or Web3, models, computing power, and data have always been the three core pillars driving the development of large language models (LLMs) and AI agents. We have discussed the following in depth on our Medium platform: - How to create models - Providing data and computing resources - The role of verification mechanisms - The working principles of Trusted Execution Environments (TEEs) Due to the extensive content, please refer to the articles on Medium for more details. Here is an ecosystem map of DeFi infrastructure created by @pinkbrains_io: Image The main participants in this field include: Trusted Execution Environments (TEEs) @PhalaNetwork @MarlinProtocol @AutomataNetwork Frameworks @arcdotfun @ai16z dao Platforms / Integrated Solutions @virtuals_io @aisweatshop @Almanak__ @autonolas

@Cod3xOrg

@crestalnetwork

@CreatorBid

@openservai

@WaveformBackup

@getaxal

@EmpyrealSDK

Common Infrastructure

@joinFXN

@TheoriqAI

@hyperbolic_labs

@BagelOpenAI

@Hive_Intel

Toolkits

@sendaifun

@lexiconinfra

The Future Development of DeFi AI

I believe that the TechFlow market will go through three main stages: first pursuing efficiency, then achieving decentralization, and finally focusing on privacy protection.

The development of DeFi AI will go through 4 specific stages.

The first stage: Focus on improving efficiency, launching tools that simplify complex DeFi operations. For example:

AI that can understand imperfect inputs

Tools that can complete transactions quickly

Real-time market research to help users make wiser decisions based on their goals

The second stage: Intelligent agents will achieve autonomous trading, able to execute strategies based on third-party data or insights from other intelligent agents. Advanced users can fine-tune models to build intelligent agents to optimize returns for themselves or their clients.

The third stage: Users will focus on wallet management and AI verification issues. Trusted Execution Environments (TEEs) and Zero-Knowledge Proofs (ZKPs) will ensure the transparency and security of AI systems.

The fourth stage: Eventually, a no-code DeFi AI toolkit or AI-as-a-Service protocol may emerge, creating an agent-based economic system where users can fine-tune models through cryptocurrency trading.

Although this vision is exciting, there are still some pressing issues to be solved:

Many current tools are just simple wrappers of ChatGPT, lacking clear evaluation criteria.

The fragmentation trend of on-chain data may lead AI models to be more centralized than decentralized, and there is currently no clear solution.