ModeMode is evolving from a DeFi-specific L2 blockchain into a DeFAI hub, aiming to improve user experience and optimize capital efficiency by leveraging AI agents to solve the complexity challenges of DeFi.
Current DeFAI solutions are limited to single functions and fragmented across multiple chains, which is why Mode is adopting an approach based on multi-agent collaboration and an integrated network environment.
Mode Network’s DeFAI stack is composed of three layers: the interface layer (AI Terminal, Agent App Store, Mode Trade), the data layer (the Synth subnet for synthetic data generation), and the infrastructure layer (AI-secured sequencer and Optimism Superchain integration).
While gaining user trust and technical stability remain challenges for Mode to address, its approach could be considered a viable alternative for widespread DeFi adoption if it progresses toward creating tangible value.
Was the crypto x AI agent narrative ultimately just an overhyped buzzword? The technical imagination and speculative interest generated by the combination of crypto and AI agents created an "AI agent cycle," but as the short-term inflated interest subsided, the pace of development has also slowed. Market reactions reflect this stagnation. The market capitalization of major AI agents (Virtuals Protocol, ai16z, ARC, aixbt, Freysa AI) has declined by an average of about 67% from their peaks, making it difficult to deny that the sector has entered a downward trend.
Nevertheless, short-term price declines do not fully reflect the mid to long-term potential of the technology. The unchanged fact is that AI agents and crypto still contain meaningful combination possibilities. Crypto's transnational payment rails can serve as an effective alternative as a medium of exchange for collaboration between AI agents, and similarly, crypto gains the possibility of more efficient asset management through the autonomy of AI agents. Beyond this, AI agents and crypto can exchange sufficient synergistic effects to enhance each other's efficiency in a wide range of areas, such as open-source development, data sourcing, UX improvement, and security.
So, what breakthrough is needed for the combination of AI agents and crypto to achieve more advanced development? The answer is quite simple. While the attention economy is powerful in terms of gaining value, it is very temporary, much like the nature of attention itself. On the other hand, fundamentals such as user traffic, project revenue, or innovative technical structures serve as driving forces for continuous development, albeit at a slower pace. If AI agents relied on short-term attention and speculative demand to generate momentum for development in the previous cycle, now it's time to create and prove fundamental value to the market so as not to be swayed by short-term interest levels.
In this respect, it's necessary to pay attention to players who are steadily accumulating fundamentals rather than being swayed by rapidly changing market cycles in the short term. Mode, which we will examine in this article, is just such a case. Mode started as an Ethereum L2 for DeFi and is now presenting a new 'DeFAI' specific L2 that uses AI to complement the limitations of existing DeFi. In particular, in a situation where more than 100 L2s exist, Mode is persuasively presenting the rationale for why their L2 infrastructure is needed, along with the vertical domain of DeFAI.
How is DeFAI attempting to take DeFi to the next level? Why is Mode trying to build a DeFAI ecosystem using Layer 2 infrastructure? This article aims to answer these questions. From here on, we will look at what DeFAI is and what problems it solves, examine the strengths of Mode's DeFAI stack, and shed light on the future of DeFi that Mode's integrated DeFAI ecosystem will create.
Recently, DeFi has shown more positive developments than ever. DEX usage has exceeded 20% compared to CEX for the first time, showing that the CEX-centered crypto market is being reorganized to be DEX-centered, and the TVL of all DeFi ecosystems is approaching $100B, showing steady growth since 2022 when it had stagnated after one DeFi summer. Also, the influx of traditional institutions and the establishment of sustainable revenue models are accelerating, as evidenced by BlackRock launching USDtb with Ethena based on the tokenized fund BUIDL, and the Aave protocol officially pushing for $AAVE buybacks through revenue. These changes show how quickly the DeFi industry is maturing.
However, despite its remarkable growth, DeFi still has a chronic problem. That problem is that the complexity of DeFi, which deepens as the ecosystem becomes more diversified, requires users to navigate a steep learning curve. Since the greatest advantage of DeFi is that anyone can execute optimal investment strategies by interoperating DeFi protocols that interact composably, users need sufficient understanding of financial engineering knowledge, as well as risks and on-chain infrastructure.
For readers already familiar with DeFi, recall the moment when you first entered DeFi from a beginner's perspective. From basic steps like wallets and on-ramp infrastructure needed to use the on-chain environment, users encounter difficulties. Furthermore, DEX's AMM and liquidity pools, lending market's utilization rate and liquidation mechanisms, or mechanisms such as liquidity staking or voting escrow act as high entry barriers for users. As a result, DeFi is currently used as a specialized capital market by a limited number of market participants.
As DeFi's complexity blocks the entry of new users, DeFAI (DeFi + AI) is an attempt to innovate by simplifying the DeFi experience using the technical performance of AI agents, enabling anyone to use DeFi and improve capital operational efficiency. Generally, DeFAI solutions suggest the following approaches to improve existing DeFi:
Source: aixbt Labs
Market Analysis (Exploration): AI models are specialized in analyzing and predicting market information by simultaneously processing numerous variables such as price data, trading volume, liquidity indicators, on-chain activity, and social sentiment. Especially as demonstrated by the case of aixbt, the natural language processing capability of LLMs enables analysis of unstructured data generated in social channels such as Twitter and Telegram, not just quantitative indicators. Due to these technical characteristics, AI agents are emerging as auxiliary solutions for market analysis, such as collecting market information, predicting asset prices, and capturing market inefficiencies, compared to simple algorithms or human analysis.
Source: Griffain
User Experience Abstraction (Execution): This is a method of automating DeFi execution through natural language commands, similar to AI copilot products like ChatGPT or Perplexity that are already commercialized. For example, if a user inputs a prompt like "provide liquidity to the ETH-USDC pool," the AI agent autonomously finds the most efficient liquidity pool to earn interest and executes the transaction. This allows users to minimize the learning steps required to execute DeFi while efficiently operating their assets using optimal DeFi strategies.
Source: Uniswap v4 Launch: A New Era for DeFi and AI Integration
Autonomous Position Adjustment (Management): Since the crypto market operates 24/7 without stopping, AI agents that manage positions throughout the entire day can be an appropriate solution. For example, Uniswap v3's CLMM (Concentrated Liquidity Market Maker) requires management to adjust the liquidity supply range according to asset price volatility. In this case, AI agents can autonomously adjust the liquidity supply range by reflecting real-time market data. Thus, AI agents are emerging as effective solutions for managing positions autonomously while monitoring market information in real-time, such as managing liquidity supply positions or rebalancing synthetic asset vaults.
In this way, DeFAI effectively offers AI as a solution to improve the poor user experience of existing DeFi and increase capital efficiency. Organized according to the user journey, in the exploration stage of using DeFi, AI agents effectively investigate fragmented market information, and in the execution stage, they simplify the user experience through tools like copilots, allowing anyone to easily conduct on-chain transactions with just natural language commands. Furthermore, in the management stage, unlike existing algorithmic trading that relied only on deterministic IF-THEN rules, AI agents learn new complex market patterns in real-time and flexibly respond to changing market conditions.
The approach of DeFAI, which aims to complement the limitations of existing DeFi while creating practical utility for AI agents, has received positive evaluations from market participants. Consequently, DeFAI copilot products, applications that autonomously rebalance synthetic asset funds, or products that dynamically adjust LP positions in DEXs like Meteora have begun to fill the DeFAI market.
However, objectively evaluating the progress of DeFAI, the current approaches of DeFAI solutions appear to be insufficient to provide adequate performance to achieve the mission of enhancing DeFi's capital efficiency and user experience through the use of AI agents. Of course, it's clear that we need to consider that DeFAI is just in its initial stages, but it's a fact that it's difficult to say it has brought meaningful changes to the DeFi ecosystem as much as expected.
So, in what form should DeFAI evolve to be utilized more effectively? As a preemptive approach to finding that direction, let's first look at the essential attributes of DeFi, which DeFAI aims to improve, and the recent development trends in DeFi.
Source: DeFi Value Flows: Understanding DeFi Business Models and Revenues | by Aw Kai Shin
First, the biggest advantage of DeFi is that anyone with an internet connection can execute investment strategies by interoperating composable DeFi protocols. That is, the characteristic of DeFi that allows permissionless interoperation of DeFi protocols provides utility that traditional financial systems cannot offer. This is referred to as Money Lego and has served as the most important growth background in the process of growing DeFi into a large-scale industry as it is now.
This characteristic of DeFi has been deepening recently. DeFi protocols are combining into more tightly integrated structures, and infrastructure to maximize interoperability is also developing. Examples of this development trend are as follows:
Comsposability of DeFi protocols
Recently, Morpho, Spark, and Ethena officially announced strategic collaborations. The collaboration of these major protocols, which are at the top in terms of scale within the DeFi ecosystem and are driving the entire DeFi ecosystem, is noteworthy. Their initiative to structurally combine their core platforms and financial products well demonstrates how DeFi Money Lego is becoming increasingly complex and sophisticated.
Spark directly supplies DAI stablecoins to Morpho's vault, MetaMorpho. Initially, $100M worth of DAI is transferred to the vault, and this DAI liquidity is distributed to create sUSDe/DAI and USDe/DAI markets on the Morpho Blue platform, Morpho's lending market. This allows users to provide Ethena's assets, sUSDe and USDe, as collateral and borrow DAI.
Through this structure, users can adopt various strategies to operate their capital as efficiently as possible according to their purpose. For example, they can use leverage strategies such as providing sUSDe as collateral on Morpho Blue, borrowing DAI, and purchasing more sUSDe, or they can earn interest rate arbitrage by utilizing the difference between USDe's staking yield and DAI's borrowing interest rate. It's also possible to maintain sUSDe staking rewards while securing additional liquidity through DAI.
Cross-chain infrastructure
As mentioned above, as DeFi's Money Lego is increasingly maximized, the infrastructure that technically supports the composability of DeFi protocols is also playing an important role. Particularly, the importance of cross-chain infrastructure to integrate fragmented liquidity is becoming more prominent. Even considering only Ethereum, which has the largest market size of the DeFi ecosystem, there are already over 100 L2s, so an environment has been created where limited liquidity is distributed across multiple L2s. As a result, users inevitably experience additional costs and delays by having to use bridges, or they suffer excessive slippage due to insufficient liquidity.
As this problem of liquidity dispersion intensifies, cross-chain infrastructure is rapidly developing. For instance, Unichain introduces base technology needed to reduce friction in cross-chain token transfers by securing economic finality on its own L2 to solve the problem of liquidity dispersion in the Ethereum ecosystem. Additionally, various Ethereum protocols, including Superchain, are preparing an integrated ecosystem through ERC-7683, an intent-based token bridge, and LayerZero is launching token frameworks like OFT (Omnichain Fungible Token) to enable seamless token circulation across multiple networks through cross-chain token transfers.
The fact that can be reflected through the cases up to here is that the DeFi ecosystem is developing in two key directions: First, the composability between DeFi protocols is becoming tighter. As seen in the case of Morpho, Spark, and Ethena, protocols are building complex and diverse forms of Money Lego by combining each other's financial products. This composability enables more sophisticated capital management strategies for users but also increases the complexity of interactions.
Second, the demand for an integrated liquidity environment and the development of cross-chain infrastructure to implement it are accelerating. The importance of cross-chain infrastructure that ensures asset transfers between chains do not cause friction in smoothly building DeFi Money Lego is increasingly cited as an important challenge to solve.
Returning to the point, the two trends shown in the DeFi ecosystem above provide important implications for the process of identifying the prerequisites that DeFAI, which will serve as an alternative to existing DeFi, should have in order to advance to the next level. The rational development direction for DeFAI to provide better performance and actually improve the user experience can be summarized as follows:
3.2.1 Building a Multi-Agent Collaboration System
Currently, most DeFAI solutions are limited to single functions such as (1) market analysis, (2) AI copilot, (3) position adjustment, and provide services that support only specific parts of the overall user journey. However, in reality, the user journey in DeFi is not a series of disconnected processes of (1) exploring the market, (2) executing transactions, and (3) managing positions in real-time, but rather closer to a feedback loop where all stages influence each other. It's like monitoring changing market conditions while readjusting positions or making real-time deposits and withdrawals. Moreover, as the composability of DeFi protocols deepens as we've seen earlier, the factors that users need to consider are increasing, and the user journey is becoming even more complex.
For example, when a user uses DeFi protocols to generate interest income and an AI agent assists in this process, a very complex process logic is required. First, after receiving the user's request through a natural language command, the agent needs to devise the most profitable strategy considering the composability of various DeFi protocols. Then, it needs to select the optimal pool for the highest yield and execute real-time tasks such as swapping tokens, depositing or withdrawing assets, or adjusting collateral ratios in response to changing market conditions.
As such complex tasks are required, a single AI agent may have performance limitations in terms of data or reasoning abilities to perform complex tasks. Therefore, rather than individually utilizing single AI agents that perform limited functions, integrating multiple AI agents may be more in line with the direction in which DeFi is evolving. As referred to as multi-agents, multiple AI agents with different roles, knowledge bases, and common data layers collaborate toward a common goal, enabling more complex DeFi executions that cannot be solved by a single agent alone.
3.2.2 The Need for an Integrated Network Environment
Next, the fact that AI agents are all fragmented and interacting on separate blockchains can be considered a limitation of current DeFAI. Despite the essential purpose of DeFAI being to improve the poor user experience of existing DeFi, if DeFAI exists as fragmented solutions across different network environments, users will still face the same complexities such as cross-chain bridging or wallet switching.
This means that DeFAI solutions limited to individual chains may end up providing another form of fragmented user experience, potentially counteracting the improvement of user experience. Furthermore, compared to the development trend of DeFi moving towards cross-chain integration, it has the limitation of leaving the liquidity dispersion problem as is, making it difficult to secure abundant liquidity.
Therefore, for DeFAI to substantially improve the user experience and capital efficiency of DeFi, DeFAI also needs to evolve on the premise of an integrated environment. In other words, it would be more reasonable to provide DeFAI solutions in a network environment that is either a single network or at least has sufficient cross-chain infrastructure to enable seamless asset movement between chains.
An integrated network allows (1) AI agents to easily access data from various protocols, (2) efficiently share learning results and execution data between agents to form collective intelligence, and (3) provide users with various DeFAI functions through a single interface. This can be a key factor in maximizing the efficiency of the DeFAI ecosystem and realizing true user experience improvement.
Considering the prerequisites for DeFAI examined so far, Mode, which we will look at in the next chapter, is worth noting as an alternative that builds an integrated DeFAI ecosystem using DeFAI-specific infrastructure and Optimism Superchain-based L2. From here on, we will look more specifically at what kind of agentic economy Mode aims to build through interactions between AI and AI based on L2, and how the interface layer, data layer, and infrastructure layer designed specifically for DeFAI work organically together.
As mentioned earlier, Mode is an Ethereum L2 built on OP stack-based optimistic rollup. Mode, also a member of the Optimism Superchain, once recorded over $400M in TVL and grew to become the third-largest ecosystem in the Superchain after Base and OP mainnet. Mode's market approach focused on the purpose of growing the vertical domain of DeFi and introduced Mode as a DeFi-specific L2. Unlike many L2s that struggle to achieve user adoption and meaningful growth after launch, Mode's differentiated approach was sufficient to attract the attention of market participants.
Initially, Mode focused on securing major DeFi protocols in its ecosystem to realize its value proposition as a DeFi-specific L2. As a result, players like Ether.Fi and Velodrome onboarded to Mode, and Mode established itself as an attractive platform for executing DeFi strategies using interest-bearing assets such as LRT (Liquid Restaking Token), which was a major market narrative at the time.
Additionally, Mode has continuously presented new market approaches. Mode's tokenomics also introduced a voting escrow (ve) mechanism for its native token $MODE, designed to interoperate with DeFi protocols and coordinate the interests of ecosystem participants, befitting the vision of a DeFi-specific L2.
Mode receives OP grants from the Optimism chain as a member of the Superchain, and Mode distributes these grants to $veMODE holders and protocols. $veMODE holders can participate in gauge voting to determine which protocols will receive incentives ($MODE, $OP).
The limited utility of governance-only native tokens for almost all L2s is always cited as a chronic limitation. Compared to this, the token economy that combines voting escrow and integrates its native token into the ecosystem's economic system is a noteworthy effort by Mode to strengthen its differentiation as a DeFi-specific L2.
Then, recently, as the combination potential of AI agents and crypto emerged, Mode set DeFAI as a new goal to strengthen its value proposition as a DeFi-specific L2. In particular, it focuses on implementing high-performance DeFAI solutions through the interaction between AI and AI to lower the entry barrier of existing DeFi and achieve widespread adoption of DeFi. To this end, Mode has newly designed a DeFAI stack including interface layer, data layer, and security layer, and aims to build an integrated DeFAI ecosystem where each layer is organically connected to complete an AI agent-based DeFi experience.
The DeFAI stack proposed by Mode is divided into three parts: interface layer, data layer, and security layer, and each layer collectively aims to provide an environment where general users or developers can utilize AI agents in a complete manner. From below, we will look specifically at the DeFAI-related functions that Mode provides in each layer.
4.2.1 Interface Layer: AI Terminal, AI Agent Appstore, Mode Trade
The most important change in the process of evolving DeFi to DeFAI appears in the way users interact with protocols and networks. Mode aims to provide an interface layer where users no longer interact with the frontend of DeFi applications but perform DeFi tasks using Mode's AI-based terminal and agents.
AI Terminal - DeFi Copilot
Source: Mode Network
The AI Terminal allows users to perform various on-chain tasks in natural language through a copilot. For example, users can easily execute DeFi tasks such as swapping, depositing, and yield farming based on major DeFi protocols and assets such as Velodrome, Balancer, and Ethena sUSDe by conversing with the terminal.
Mode's AI Terminal introduces differentiated features compared to other DeFAI copilots, one of which is the unique ability to answer probabilistic questions about market data by integrating synthetic data from the Bittensor subnet, which will be discussed later. For instance, to a question like "What's the likelihood of BTC price exceeding $100K next week?" the AI Terminal can provide an answer based on the data it has learned. Besides this, not only does it offer functions for general users, but developers can also use the AI Terminal to execute development tasks such as deploying smart contracts, simultaneously transferring tokens to multiple addresses, and creating NFT collections.
Through these features, the AI Terminal provides seamless on-chain experiences for DeFi users, enhanced development productivity for developers, and market predictions, which are useful for attracting general market participants. Thus, it has the potential to position itself as a major AI copilot in the DeFAI ecosystem and as an interface that provides convenience to users at the very entry point of the DeFAI ecosystem that Mode is building.
AI Agent Appstore: Collaboration Channel for AI Agents
Source: Mode Network
The AI Agent Appstore (hereinafter referred to as the Appstore) allows users to search for and use various AI agents that exist in a fragmented state in one place. In Mode's words, "In the future, the act of users selecting DeFi strategies will become the act of selecting AI agents." In this future agentic economy, the Appstore will be the most important interface layer for AI agent collaboration. Currently, AI agents performing various functions such as yield optimization, natural language interface, and liquidity management are deployed in the Appstore, and a brief look at some of the meaningful use cases is as follows:
Source: Giza
ARMA is an AI agent designed to maximize returns on stablecoin deposits (USDT, USDC) in lending protocols such as Mode's Ionic, LayerBank, and Ironclad. For example, when a user deposits USDC, ARMA autonomously continuously evaluates the interest rates of various lending protocols for efficient stablecoin farming and optimizes the portfolio by reallocating funds to the pool with the highest expected return.
Source: Amplifi
Amplifi is a DeFAI solution that provides optimized yield strategies centered on BTC and stablecoins based on AI agents. In line with the purpose of improving the poor user experience of DeFi, it mainly provides atomic functions to users, where users can obtain the best returns through the One-Click Vaults feature, utilizing omnichain liquidity and AI's autonomous portfolio optimization without having to go through gas fees, swaps, and bridging processes directly.
In addition to these, numerous AI agents performing functions such as natural language interface or liquidity management are deployed in Mode's Appstore, allowing users to search for and use various AI agents in one place. Furthermore, as a roadmap for the expansion of the Appstore, Mode plans to introduce cross-chain actions that execute DeFi strategies across multiple chains using interoperability protocols. Also, in the long term, it is expected to develop it into an open platform where more developers can participate in the Appstore in a permissionless manner, allowing anyone to offer their own AI-based DeFi services.
Mode Trade: AI-Powered Perpetual Trading Platform
Mode Trade is a trading service that integrates perpetual futures, AI, and Synth data into a single unified platform. What sets it apart is its focus on accessibility, making it possible for retail users to leverage advanced predictive analytics and simulation tools that were previously reserved for institutional traders. At the core of the platform is a predictive AI engine that enables users to execute complex trading strategies through simple text-based commands.
Mode Trade’s perpetual trading functionality is built on Orderly’s central limit order book (CLOB) infrastructure, which ensures deep liquidity and fast execution. The platform currently supports over 100 popular token pairs with up to 50x leverage. Tokenized stock trading is also planned for future release.
A key differentiator for Mode Trade is its seamless integration with Mode’s broader AI stack, including the AI Terminal and the Synth predictive data layer. Traders can manage their positions by opening, adjusting, monitoring, and closing trades using intuitive text prompts. They can also perform onchain actions such as bridging, swapping, and staking. The integrated Synth module delivers decentralized, probabilistic market forecasts powered by Bittensor, helping traders improve risk management and optimize their strategies.
4.2.2 Data Layer: Synth Subnet
Source: Synth Subnet
The data layer serves as the data infrastructure to support the reasoning abilities and decision-making of AI agents, with Synth subnet at the core of Mode's DeFAI stack design. Synth is a Bittensor-based subnet that generates synthetic data, providing high-quality simulator data to enable AI models to perform improved predictions and inferences. This acts as a key element in solving the problem of data scarcity in the on-chain DeFAI environment and in learning data for AI agents to make appropriate decisions in uncertain market situations.
Before explaining the subnet, Mode utilizes Bittensor as the underlying infrastructure to secure synthetic data. Bittensor is a decentralized machine learning network designed to allow participants to collaborate to train AI models. In Bittensor, participants act as miners or validators, performing roles of running or validating AI models, and receive TAO tokens as rewards for this. Based on this incentive mechanism, miners are motivated to continuously improve their prediction performance, and the network can continuously secure high-quality synthetic data.
Under these incentives, participants train and evaluate AI models in subnets, which are independently operated sub-node networks of Bittensor. Each subnet is created to perform specific AI functions or services, and depending on the purpose of the subnet, individual learning tasks or data computations are conducted. For example, one subnet might handle natural language processing, another might handle image analysis, and yet another might handle financial data prediction. In conclusion, using this subnet structure, Mode's Synth aims to continuously generate high-quality synthetic data needed for agent learning.
To this end, in the initial stage of the subnet, Mode's Synth is laying the groundwork for advancing subnet performance by predicting the price of BTC (Bitcoin) through a price prediction simulator. The way Synth predicts BTC prices is very sophisticated, with the detailed process as follows:
Miner's Price Prediction: Miners submit simulations of future BTC price movements with about 100 price paths periodically. Before running the prediction model, they use various data as inputs, including on-chain data (transaction volume on the blockchain, wallet activity, etc.) and oracle data such as Pyth. By training and simulating the prediction model by synthesizing the current market situation and on-chain data, they can predict the possibility of future prices within the most realistic range.
Validator's Verification: Validators are nodes that verify and evaluate the quality of AI outputs produced by miners in the subnet. In the Synth subnet, validators record price path predictions from all miners in local storage every round, and after the prediction period (24 hours) has passed, they retrieve the actual price time series from reliable price data providers such as Pyth to verify the authenticity and accuracy of the data submitted by miners from multiple angles.
Scoring System and Rewards: After verification, validators measure predictive accuracy by scoring how close the distribution of various scenarios submitted by miners is to the actual price. By comparing the scores of all miners, ranking them, and deriving normalized scores, tokens are awarded to miners who made accurate predictions based on the results. Conversely, miners who consistently produce inaccurate predictions receive lower scores and are penalized. This feedback structure induces miners to continuously improve their prediction models and increase accuracy through competition.
By repeating the above process, the Synth subnet accumulates rich synthetic data over time. The numerous price path simulations submitted by miners and the verification results form a high-quality probability distribution data collection that well reflects the characteristics of the actual market. This data can then be used for AI agents to learn.
The reason Mode built such a Synth is clear. Until now, one of the main limitations of DeFAI was that the predictive power of AI agents was limited due to the lack of useful data for them to learn from. Synth solves this problem, allowing AI agents to learn from various future scenarios as training data, thus acquiring better prediction and decision-making abilities.
In the initial stage, Synth's data is expected to be used for model learning of AI agents, sophistication of option pricing models, and portfolio risk management. For example, AI agents deployed in the AI Agent Appstore can learn with Synth data and then devise DeFi strategies considering the probability distribution of price fluctuations or even execute diversified asset management strategies.
In the future, APIs and applications for developers are also planned to be provided, allowing any external developer to utilize Synth's synthetic data. This means that projects outside the Mode ecosystem can also integrate Synth data into their services or use it to train their own algorithms, ultimately suggesting the possibility of Mode playing a public good role for the entire DeFAI ecosystem centered around Synth.
4.2.3 Infrastructure Layer: AI-Secured Sequencer and Optimism Superchain
Next, the infrastructure layer illustrates well why Mode's design for executing DeFAI in an L2 environment is important. While AI models are rapidly evolving, it remains difficult to readily answer whether one can fully entrust them with their assets. To alleviate this psychological barrier, Mode adds a security layer to the L2 sequencer structure, protecting assets by filtering transactions from AI agents that could adversely affect users' assets.
Meanwhile, the ability to actively leverage Superchain interoperability is a technical advantage that is possible because Mode is built on the OP stack. The benefit of being able to utilize the liquidity and user base of various L2 chains becomes an important competitive advantage that can enhance the efficiency of DeFi strategies. Let's look in detail at the AI-secured sequencer and Superchain interoperability, which are part of the infrastructure layer of the DeFAI stack.
AI-Secured Sequencer
The AI-Secured Sequencer is an infrastructure implemented to have AI pre-screen all transactions delivered to Mode's sequencer. As Mode aims for an agentic economy, it requires thorough network security. Conditions must first be established to reliably support large-scale DeFi transactions, and the infrastructure layer deployed to prepare for this is the AI-Secured Sequencer.
First, Mode introduced Firewall in cooperation with Forta to enhance network security. The Forta Firewall is a threat blocking system that advances the pre-risk detection capabilities of blockchain networks, similar to a web application firewall (WAF) in traditional internet. It is positioned in front of the sequencer to pre-screen transactions before they are processed by the sequencer.
In general optimistic rollups including Mode, the sequencer plays the role of ordering transactions executed by users or contracts and including them in rollup blocks. However, the sequencer only checks the validity of transactions, such as signatures and fees, and does not have built-in security checking functions. Therefore, traditionally, if there is a vulnerability in a contract, when an attacker sends a transaction, the sequencer processes it as is and includes it in a block, and only after this can it detect anomalies through procedures such as fraud proofs.
To address this, Mode designed all transactions entering the rollup network to be first delivered to the Forta Firewall. The Forta Firewall conducts risk assessments on transactions before the sequencer, and transactions classified as high risk are filtered out before being delivered to the sequencer. This verification process is carried out very quickly, within an average of 80ms, so from the user's perspective, they can receive security assurance from a firewall that detects more than 99% of potential hacking attempts without feeling almost any delay.
Moreover, befitting Mode as a DeFAI-specific L2, the pre-screening of all transactions delivered to the sequencer is done through AI models. This means leveraging AI's inspection capabilities to examine all transactions, judge whether there is malicious intent or signs of hacking, and process settlements only when it's safe. For this purpose, a machine learning-based risk assessment engine is deployed in front of the sequencer.
Here, Forta's own AI model, FORTRESS, handles this pre-screening, analyzing behavior by running simulations on new transactions based on patterns learned from numerous past hacking cases, assigning a risk score between 0 and 1. For instance, if a specific transaction is similar to patterns of frequent attacks in the past (re-entry attacks, fund theft attempts, etc.), the risk score will be set high. Network operators can set a threshold for what score is considered high risk, and transactions exceeding that threshold can be immediately rejected from block inclusion.
In this way, Mode actively introduces AI models from the security layer, befitting its aim for an agentic economy, establishing security differentiation. This allows builders in the ecosystem to focus on developing DeFAI solutions in a secure environment without establishing separate security measures, and users can use solutions under enhanced security.
Superchain Interoperability
As a member of the Superchain, being able to utilize the infrastructure base of the Superchain ecosystem is the most core differentiation point that Mode has in terms of infrastructure. The Optimism Superchain refers to an L2 consensus body that aims to operate multiple L2 blockchains interconnected based on the OP stack as if they were a single integrated network. In this Superchain, L2s such as OP Mainnet, Base, Mode, and Unichain under the OP stack are closely connected in terms of technology, economy, and governance, with the ultimate goal of providing an experience where users can use dapps without worrying about which chain they are on, effectively achieving horizontal scaling of Ethereum.
The core of the Superchain is to enable direct and rapid connections between L2s by introducing native interoperability to the OP stack. Through this, data and assets can move seamlessly between L2s within the Superchain without necessarily going through Ethereum L1. With the introduction of key technologies such as the ERC-7683 intent standard interface or Superchain ERC-20 token standard, users will be able to complete transfers within seconds when sending assets from one L2 to another, and no longer have to endure L1 bridge waiting times or high fees of L2 intermediary bridges. Also, at the ecosystem level, all chains participating in the Superchain can benefit from shared user bases and liquidity.
Mode occupies an important position within such a Superchain. Mode received a grant of 2M OP tokens, worth $5.3M, from the Optimism Foundation, and has agreed to return part of its own sequencer revenue to the Optimism Collective, the economic and governance body of the Superchain. This suggests that Mode, as the third largest L2 chain in the Superchain after Base and OP Mainnet, has a very close connection with the Superchain.
Therefore, if Superchain interoperability is secured, the liquidity on Mode will be connected to the liquidity of the entire Superchain, and conversely, the liquidity from other chains can also be smoothly connected to Mode. This greatly expands the activity range of AI agents in the process of Mode building a DeFAI ecosystem.
Mode's AI agents can access data and liquidity pools of L2s in real-time through the Superchain, allowing them to find and execute transactions for optimal profit opportunities in a broader market. For example, if Mode's AI agent discovers an arbitrage opportunity by comparing DEX prices on Unichain and Mode's DEX, it can move tokens between the two chains without separate bridge delays and immediately close the deal. This provides a tremendous liquidity advantage to Mode's DeFAI ecosystem.
In this way, Mode can secure synergy effects within the economic connection of the Superchain. From abundant liquidity for DeFAI execution to user base and market opportunities for optimal returns, integration with other L2s provides a useful foundation for Mode to implement an agentic economy. Furthermore, Mode is expected to share DeFAI infrastructure such as the Synth subnet and AI app store that it is developing on its own, like public goods of the Superchain, contributing to the entire ecosystem while positioning itself as the core DeFAI hub within the Superchain.
Having examined the technical configuration of Mode's DeFAI stack up to this point, I can imagine the reader's impression that the actual form of DeFAI, where these technologies interact organically with each other, might feel somewhat vague. Since DeFAI is not yet universally utilized, it's natural to have doubts as a clear blueprint for its implementation methods has not been drawn.
Therefore, let's briefly look at how DeFAI stacks can function integratively through a virtual scenario where Alice makes a request for DeFi execution. However, as Mode advances its technical roadmap, the detailed structure could change significantly, so this should be understood as a purpose to reflect on the potential future that DeFAI will create.
DeFAI Operation Scenario: Mode's DeFAI Stack Operation Procedure
User Request: Alice, who has accessed Mode's AI Terminal, requests with a simple natural language command: "Use my 100 USDC in my wallet to generate optimal interest using DeFi protocols."
Interface Layer: The AI Terminal analyzes Alice's request and formulates an optimal execution plan. It selects and combines the best AI agents from the app store to perform tasks according to the established plan. For example, ARMA takes on the role of comparing and analyzing interest rates between various lending markets, while Brian takes on the role of converting Alice's natural language commands into executable transactions. These specialized agents collaborate to automate complex DeFi tasks, so Alice only needs to approve the final plan with one click.
Data Layer & Infrastructure Layer: The selected AI agents analyze market trends and derive optimal investment strategies using high-quality on-chain data provided by the Synth subnet. The agents comprehensively analyze real-time market data and transaction history data to identify DeFi opportunities that provide the highest returns and generate on-chain transactions to execute them. At this time, all generated transactions are pre-verified for malicious patterns or security risks through the AI-secured sequencer integrated with the Forta Firewall, and only transactions with confirmed stability are sent to the sequencer. Thanks to this security firewall, Alice can trust AI agents and delegate her assets.
Superchain Interoperability: Mode optimizes Alice's capital operation across multiple chains using Superchain interoperability. It comprehensively analyzes DeFi protocols not only on Mode but also on other L2s connected to the Superchain, and autonomously relocates assets to chains where higher returns are predicted. For example, it simultaneously executes complex strategies such as placing some assets in Mode chain's Velodrome pool, supplying some to the lending market on Base chain, and holding the rest in USDC to prepare for predicted market fluctuations.
In this way, whereas previously users had to manually traverse each chain, move assets through bridges, and process multiple transactions through different frontends, Mode's DeFAI stack effectively simplifies this process. Since AI agents handle all processes in the background, Alice can execute DeFi strategies that can expect optimal returns through multiple prediction simulations based on a cross-chain environment from a single interface.
Mode's agentic economy envisions a future where users give commands to AI agents in natural language, and these agents execute DeFi strategies in the most optimal way. To realize this blueprint, Mode started as a DeFi-specific L2 and is now building an integrated DeFAI stack encompassing AI Terminal, App store, Mode Trade, synthetic data generation subnet, and AI-secured sequencer. Moreover, by enabling smooth asset movement and liquidity integration in a cross-chain environment based on Optimism Superchain's interoperability, it presents a relatively clear solution to overcome the complexity problem of existing DeFi.
However, much technological development is still required for this vision to become a reality. In particular, since users must delegate their assets to AI agents, building trust in DeFAI solutions remains a key challenge. Mode and other DeFAI projects must prove the stability and security of the system through actual use cases and extensive testing, and find a balance between protecting users' assets and providing convenience through gradual adoption. This is not just a technical challenge but also a process of breaking down users' psychological barriers.
Nevertheless, the vision of Mode's DeFAI stack is certainly worth anticipating as an important milestone showing the next direction of DeFi development. If DeFAI develops sufficiently in the direction of creating tangible value beyond being a buzzword swayed by short-term market direction, Mode's approach will be sufficient to be evaluated as a realistic alternative for the widespread adoption of DeFi.
