Written by: GO2MARS

Before officially starting the analysis, it is necessary to clarify a core concept: DeFAI.

DeFAI is a combination abbreviation of DeFi (Decentralized Finance) and AI (Artificial Intelligence), referring to the introduction of AI Agents into on-chain financial scenarios, enabling them to perceive market conditions, autonomously formulate strategies, and directly execute on-chain operations—thus completing a series of financial behaviors that traditionally require professionals, such as asset allocation, risk management, and protocol interactions, without relying on real-time human intervention.

In short, DeFAI is not a simple AI upgrade of DeFi tools, but an attempt to build a self-operating financial execution layer on-chain.

This field has rapidly gained momentum since Q4 of 2024, with three landmark events worth noting, each corresponding to different levels of AI Agent entering Web3: narrative breaking out of the circle, infrastructure building for assetization, and the real landing of execution capabilities.

The first event occurred in July 2024. The Twitter bot Truth Terminal, built by developer Andy Ayrey, quickly gained attention after receiving a $50,000 BTC donation from a16z co-founder Marc Andreessen, sparking viral dissemination of the GOAT coin. This marked the first time AI Agents meaningfully entered the public view as participants in the on-chain economy.

The second event occurred in October of the same year. Virtuals Protocol exploded on the Base network, tokenizing the AI Agent itself, with its ecosystem market cap reaching a peak of over $3.5 billion, becoming a typical representative of the assetization infrastructure building phase in the DeFAI field.

The third event involved projects like Giza, HeyAnon, and Almanak successively landing on the on-chain execution layer, pushing the industry from narrative-driven to productization stage—AI Agents began to truly "get to work" executing on-chain operations, rather than just staying on the information exchange level.

From a global market scale perspective, several research institutions have highly consistent growth expectations for the AI Agent field:

Chart 1: Global AI Agent Market Size Forecast Comparison, Data Source: MarketsandMarkets (2025), Grand View Research (2025), BCC Research (2026.01)

However, there is still a significant gap between capital heat and industrial implementation. According to a report titled "The State of AI in 2025" released by McKinsey in November 2025 (based on 1,993 respondents from 105 countries), although 88% of organizations are using AI in at least one business function, nearly two-thirds remain at the experimental or pilot stage. Specifically in the AI Agent field: 62% of organizations have begun experimentation, and 23% are advancing scaling in at least one function, but the proportion of achieving scaled deployment in any single function is less than 10%.

This data suggests that the narrative enthusiasm in the DeFAI track currently surpasses the actual implementation progress. Understanding this gap is a prerequisite for an objective assessment of the value of this track.

The Technical Foundation of DeFAI: How AI Agents Interact with the On-chain World

To understand how DeFAI operates, we first need to answer a key question: How does AI intervene in on-chain financial operations?

The core execution unit of the DeFAI system is the AI Agent built on a large language model. According to the academic review by Wang et al. (2023), its core capabilities can be summarized into a three-layer architecture, with each layer having corresponding specific functions in on-chain scenarios:

• Planning layer, responsible for goal decomposition and path optimization, corresponding to strategy generation and risk assessment in the on-chain scenario;
• Memory layer, accumulating cross-cycle information through external storage such as vector databases, carrying historical market data and protocol status;
• Tool layer, expanding model capabilities to enable calls to DeFi protocols, price oracles, and cross-chain bridging as external systems.

But one thing needs to be clarified: AI models cannot interact directly with blockchains. Almost all current DeFAI systems adopt a separated architecture of off-chain inference and on-chain execution—AI Agents carry out strategy calculations off-chain, then convert results into on-chain transaction signals, submitted by the execution module. This design choice reflects the realities of current technological conditions, and introduces a series of security issues such as private key authorization and permission management.

AI Agents are essentially autonomous decision-making systems based on large language models, achieving closed-loop execution through task decomposition, memory management, and tool invocation. Currently, AI Agents have already begun to shape interactions with on-chain asset endpoints.

Chart 2: Three-layer architecture of AI Agents

The Evolution of DeFAI: From Information Interaction to Execution Closure

Having clarified the technical foundation of DeFAI, a natural question arises: How has this system progressed to where it is today?

According to research by The Block, DeFAI's evolution has not been instantaneous, but has gone through two distinct phases—from early information-processing interactive Agents to the current execution systems capable of truly intervening in on-chain operations.

The two differ fundamentally in goal positioning, technical means, and risk levels.

Chart 3: Comparison of Two Waves of DeFAI Evolution Paths

The evolutionary context of these two phases can be understood as follows:

The first wave is the interactive Agent, focusing on building a conversational and analytical intelligent agent framework. Representative projects include ElizaOS (originally ai16z’s Eliza framework), Virtuals' G.A.M.E., etc. The essence of this stage remains an informational tool—Agents can read, speak, and analyze, but their functional boundaries stop at the information layer, not touching any asset execution operations.

The second wave is the executive DeFAI Agent, which truly enters the decision execution closed loop. Representative projects include HeyAnon, Wayfinder, Giza (ARMA Agent), and Almanak, among others. The common feature of these systems is that AI runs off-chain, outputs structured strategy signals, and completes transactions through on-chain execution modules—it does not replace existing DeFi protocols but introduces a layer of AI decision-making above them, transforming the operation chain from "human giving orders" to "Agent executing independently."

The fundamental difference between the two waves of evolution lies not in technical complexity, but in whether they genuinely touch assets. This also determines that the second wave systems face challenges in trust mechanisms, permission design, and security architectures that are far more complex than those of the first wave—this will be the focal point of the next chapter.

Landing Scenarios of DeFAI: Four Major Mainstream Application Scenarios

From technological architecture to evolutionary paths, what DeFAI "can do" is becoming increasingly clear. So in terms of actual products, what real problems is it solving?

Overall, current explorations of DeFAI applications have formed a relatively mature landing pattern around four core directions, corresponding to "yield efficiency, strategy execution, interaction thresholds, and risk control," four core pain points in on-chain operations.

Yield Optimization: Automated Cross-Protocol Rebalancing

Yield optimization is currently the most mature application scenario of DeFAI. Its core logic is: continuously scanning the annual deposit yields of mainstream DeFi protocols such as Aave, Compound, Fluid, etc., determining whether rebalancing is necessary based on preset risk parameters, and performing transaction cost analysis before each operation—only when the yield increase can cover all gas and transaction fees will the funds be genuinely shifted, thereby achieving automated optimal allocation across protocols.

Taking Giza as an example, its ARMA Agent launched a stablecoin yield strategy on the Base network in February 2025, continuously monitoring interest rate changes of protocols like Aave, Morpho, Compound, Moonwell, etc., intelligently allocating user funds to maximize yield after considering protocol APYs, fee costs, and liquidity. According to public data, ARMA currently has around 60,000 independent holders, over 36,000 deployed Agents, and manages assets worth more than $20 million.

In a market environment where DeFi protocol yields are continuously fluctuating, the efficiency and timeliness of manual monitoring and manual rebalancing are far inferior to automated systems, which is the core value of this scenario.

Chart 4: Giza Platform ARMA Agent Example Image

Quantitative Strategy Automation: Democratizing Institutional-level Capabilities

In the scenario of quantitative strategy automation, DeFAI platforms aim to modularize and automate the entire process of traditional quantitative teams' operations, allowing individual users to access institutional-level strategy execution capabilities.

For example, Almanak, supported by Delphi Digital, has launched the AI Swarm system, which breaks down the quantitative process into four steps:

• Strategy modules support writing investment logic through a Python SDK and completing backtesting;
• Execution engines automatically run the audited strategy code and trigger DeFi calls after obtaining user authorization;
• Safe-based secure wallets implement a multi-signature system through Zodiac Roles, granting strategy execution rights to AI Agents while ensuring funds remain within the user's control;
• The strategy treasury packages the strategy as a tradable treasury based on the ERC-7540 standard, allowing investors to participate in strategy profit distribution like fund shares.

The significance of this architecture lies in that the AI agent takes on data analysis, strategy iteration, and risk management functions, allowing users to only conduct final reviews of system output results without needing to build specialized quantitative teams—achieving so-called "equality of institutional-level strategies" (as claimed by the project).

Chart 5: Almanak Platform Homepage Display Image

Natural Language Command Execution: Making DeFi Operations as Simple as Sending Messages

The core of this scenario is intent-based DeFi operations: using natural language processing technology, users issue trading commands in everyday language, which the AI parses and converts into multi-step on-chain operations, significantly lowering the operation threshold for average users.

HeyAnon has created a DeFAI chat platform where users input commands through a dialog box, and the AI can execute token swaps, cross-chain bridging, borrowing, staking, and other on-chain operations, integrating LayerZero cross-chain bridge and protocols like Aave v3, supporting deployment across multiple chains including Ethereum, Base, Solana, etc.

Chart 6: HeyAnon Platform Homepage Display Image

Wayfinder, backed by Paradigm, offers further comprehensive cross-chain trading services. Its AI Agent (called Shells) automatically navigates the optimal trading pathways between different chains, executing cross-chain transfers, token swaps, or NFT interactions, allowing users to avoid concerns about underlying gas fees, cross-chain compatibility, and other technical details.

Chart 7: Wayfinder Platform Homepage Display Image

Overall, natural language interfaces significantly reduce the operational threshold of DeFi, but they also raise higher demands for the accuracy of underlying intent interpretation—once the AI's understanding of commands deviates, the operation results may be far from user expectations.

Risk Management and Liquidation Monitoring: Mechanisms Embedded in On-chain Protocols

In DeFi lending and leverage scenarios, the most common application of AI Agents is to continuously monitor on-chain position health, executing protective operations automatically before the liquidation threshold approaches; this heavy application is gradually being integrated into mainstream DeFi protocols, becoming a native function of DeFi platforms.

• Aave uses "Health Factor" to measure position safety; when the health factor falls below 1.0, the borrower's position triggers eligibility for liquidation;
• Compound adopts the "Liquidation Collateral Factor" mechanism, triggering liquidation when an account's borrowing balance exceeds the upper limit set by this factor, with specific parameters for collateral assets set through on-chain governance.

Manual monitoring struggles to maintain consistent response efficiency in a 24/7 high-volatility on-chain market, and AI Agents can achieve continuous tracking, intelligent assessment, and automatic intervention, improving risk control efficiency to levels that manual or rule-based automated systems find hard to reach.

Chart 8: Four Major Mainstream Application Scenarios of Agent × DeFi

In summary, the above four scenarios are not mutually independent but form a complement around the same main line: yield optimization and quantitative strategy automation target advanced users with a certain asset scale, with core advantages in execution efficiency and strategy accuracy; natural language interaction aims to lower the operational thresholds for average users; risk management is the underlying security guarantee that runs through all scenarios. The synergy of the three constitutes the basic landing pattern of the current DeFAI ecosystem and lays the foundation for more complex on-chain Agent applications in the future.

The Security Bottom Line of DeFAI: Private Key Management and Permission Control

The aforementioned four application scenarios, whether yield optimization or quantitative strategy automation, have only one prerequisite for realization: AI Agents must possess some form of signature authority, i.e., access to private keys. This is the most critical technological challenge of the entire DeFAI track and is easily overshadowed by narrative enthusiasm—once the signing mechanism fails, all upper-layer strategy capabilities will lose meaning.

Currently, the mainstream private key security management solutions in the industry are divided into two categories: MPC (Multi-Party Computation) and TEE (Trusted Execution Environment). Each has its emphasis on security models, levels of automation, and engineering complexity.

Chart 9: Comparison of Two Mainstream Solutions for Private Key Security Management

• The core idea of MPC (Multi-Party Computation) is to eliminate single points of failure through key splitting. For example, in the common 2-of-3 threshold signature, even if one key is leaked, the attacker cannot independently complete the signing, ensuring the safety of funds. Vultisig is a representative product in this direction, an open-source multi-chain self-custody wallet built on MPC/TSS technology, using a non-single mnemonic architecture to combine key security with user self-custody.
• TEE (Trusted Execution Environment) takes another route: it seals private keys and agent code within an isolated area (enclave) protected by hardware, allowing the AI agent to complete strategy calculations and signing within the enclave, outputting only the signature results to the on-chain, rendering the private keys completely invisible to the external environment. Mainstream chips such as Intel SGX, AMD SEV, and ARM CCA provide hardware-level isolation and encryption support. Chainlink has introduced TEE into the oracle network for processing sensitive data, proving the integrity of the execution environment to the outside through remote attestation mechanisms.

However, key security is just the first line of defense. In actual deployment, regardless of which key management solution is adopted, permission control mechanisms must also be layered on top to prevent Agents from overstepping their bounds. Almanak's practice provides a more complete reference framework: the platform uses TEE to protect strategy logic and confidential parameters while inserting a Zodiac Roles Modifier permissions layer between the deployment engine and the user's Safe smart account—every transaction initiated by AI must be compared one by one with predefined contract addresses, functions, and parameter whitelists, and transactions that do not meet the authorized scope will be automatically rejected.

This implementation of the principle of least privilege has become an important reference for the security design of DeFAI systems. It reveals a deeper logic: the security issues of DeFAI are fundamentally not just about single technology choices but a system engineering formed by the synergy of key management, permission boundaries, and execution auditing—any gap may become the weakest node in the entire chain. This is also the starting point for risk analysis in the next chapter.

The Gap Between Reality and Narrative: Core Risk Analysis of DeFAI

The above analysis reveals a core conclusion:

VCX did not obtain a premium due to outstanding asset selection or higher return expectations but because it sells the channel itself. Thus, we need to answer a question: What kind of product is VCX?

From a legal perspective, it is a closed-end fund registered with the SEC, with transparent holdings and compliant structure, fundamentally no different from any ordinary equity ETF on the market. However, from the actual function perspective, what it sells is not the "expected investment returns" in a traditional sense, but a qualification for access to asset endpoints—previously accessible only to top VC institutions and qualified investors—packaged into unit shares that can be traded on the NYSE.

Therefore, the market is willing to pay a 16 to 30 times NAV premium, which essentially prices this access right rather than assessing the future earnings of the underlying assets.

From this perspective, the comparison between VCX and MicroStrategy (MSTR) is quite illustrative. On the surface, both do similar things: packaging scarce assets (Bitcoin / top Pre-IPO equity) that are hard to obtain directly into securities that can be traded in the secondary market, showing a premium far exceeding the value of the underlying assets. However, there are fundamental differences in their capital operation logic:

• MSTR raises funds through continuous issuance of convertible bonds and preferred stocks, then adds the funds to buy more Bitcoin. This mechanism gives it the ability to dynamically expand the balance sheet and sustain its holdings, making its share price premium have an inherent basis for maintenance.
• VCX, on the other hand, is constrained by the structural limitations of closed-end funds: the asset scale is essentially locked after the issuance is completed, it cannot continuously buy new assets through refinancing, and its holding liquidity is highly dependent on the underlying company's IPO or M&A exit. Once retail sentiment declines, or the six-month lock-up period ends and circulating chips increase, the pressure for its premium to narrow will far exceed that of MSTR.

Comparison of VCX and MSTR (Strategy) Models

In other words, MSTR's premium is supported by a continuously operating capital mechanism, while VCX's premium mainly derives from chip scarcity + sentiment drive. This product logic itself carries no inherent right or wrong, but the risks it embodies are far harder for the market to price correctly than those of ordinary closed-end funds:

If retail investors buy at prices far exceeding NAV, they are essentially paying not for the value of the asset itself, but for the premium of this access qualification—and this premium will face rapid pressure to vanish after the underlying company completes its IPO and a direct trading channel forms in the public market.

Trend Assessment

Based on the preceding analysis, we can make a phased judgment on the evolution path of DeFAI. Overall, this track is at a critical node transitioning from concept validation to productization, and its evolution is expected to experience three progressive stages:

Chart 11: Forecast of DeFAI Development Stages

Note: The table above is based on a comprehensive assessment of industry public reports, project progress, and technology maturity; it is not a definitive timetable.

At the current node, DeFAI is overall transitioning from a decision-support phase to a semi-autonomous phase—some projects have begun to undertake limited autonomous execution capabilities, but human review and backstop mechanisms remain the mainstream deployment form. In this context, combined with the current technology maturity and market conditions, three key judgment points are worthy of emphasis.

Firstly, the essence of most current DeFAI projects is still automation tools rather than truly autonomous Agents. Products currently labeled as "DeFAI" primarily excel in translating human commands into preset DeFi operation sequences; they are essentially more akin to efficient execution interfaces rather than autonomous systems with independent reasoning and decision-making abilities. According to McKinsey's 2025 report, less than 10% of organizations have achieved scaled deployment of AI Agents in any single function, even in general enterprise contexts. The trust threshold and operational complexity in on-chain scenarios are higher, and there is still quite a distance from technological demonstrations to real commercial closed loops.

Secondly, the most mature and easiest place to gain institutional trust for AI Agents is not high-risk autonomous trading, but on-chain monitoring, alerts, and governance assistance. 24/7 position monitoring, liquidation alerts, governance proposal analysis, etc., offer a higher tolerance for LLM hallucinations—output errors do not directly trigger fund losses; additionally, they can effectively compensate for humans' inherent lack of sustained attention. These scenarios provide a more realistic path for DeFAI from "technological showcase" to "institutional adoption."

Thirdly, the integration of AI Agents with RWA is a cross-direction worthy of key attention in this track. According to data from RWA.xyz, as of early April 2026, the total value of on-chain tokenized RWA assets has exceeded $27 billion (excluding stablecoins), covering multiple categories including US Treasury bonds, private credit, commodities, and corporate bonds. If AI Agents can intervene in managing asset portfolios that include government bond RWAs and stablecoins—for instance, by automatically adjusting the allocation ratio of both according to market conditions—their accessible asset scale will far surpass the current realm dominated by native DeFi assets, and they may truly connect on-chain and off-chain asset endpoints, achieving the linkage of Web3 + AI + TradFi, significantly expanding market imagination.

Conclusion

AI Agents and on-chain asset management are at a critical transition period from concept validation to productization. The technological feasibility has been preliminarily validated, but challenges facing the industry, from LLM hallucination risks and on-chain data heterogeneity to the lack of trust infrastructure, cannot be resolved solely by technological iterations; they require systematic advancement in project architecture design, compliance path planning, security system construction, and business model validation.

This also precisely means that this track is still in the early construction phase, and the true competitive pattern has yet to form. For teams capable of mastering both Web3 and AI dimensions, now is the window period for intervention—whether in building a more reliable on-chain Agent system at the execution layer or in connecting data, permissions, and trust in critical links at the infrastructure layer, there are significant gaps waiting to be filled.

The competitive barrier of DeFAI will ultimately not lie in a single model capability or the depth of protocol integration but in whether it can build a truly coherent closed loop among technology, compliance, and security.