Intent: Potential Solutions to the Complexity Issues of DeFi

Before the Luna crash, I was managing a stablecoin yield strategy for a friend who wanted to understand the attractive interest rates available in January 2020. My friend is not very familiar with cryptocurrencies, and our collaboration involved him depositing funds into a hardware wallet, after which we would have regular video meetings where I guided him through the process step by step.

We are diversifying funds across multiple DeFi protocols on various chains. During each meeting, we conduct dozens of transactions involving approvals, transfers, swaps, deposits, claims, and withdrawals. Funds are allocated to customized LP trading pairs, Curve voting locks, etc., to maximize returns. We have used almost all mainstream cross-chain bridges, DEXs, and yield aggregators.

This process is quite complex for my friend. I need to explain each step in detail, and he needs to understand the interfaces of various DeFi tools. For example, exchanging USDC for FRAX/DAI LP on Polygon requires 12 transactions:

1. Exchange USDC for DAI on the DEX (2 transactions)
2. Transfer USDC and DAI across chains to Polygon (4 transactions)
3. Merge USDC and DAI on Polygon DEX (4 transactions)
4. Deposit LP into the reward pool (2 transactions)

This seemingly simple operation is actually very complicated, especially when managing a large-scale portfolio.

At a higher level, our operations have clear objectives. For example, "Converting USDC on Ethereum to FRAX/DAI LP on Polygon and depositing it into the staking pool." This is the "content" of the operation, while the 12 specific transactions represent the "how" of the operation. From the starting point to the endpoint, a series of logical steps are required, and these steps are all quantifiable.

If there is a powerful trading routing algorithm, this process can be simplified to 1-2 steps. We just need to present the goal, and the algorithm can provide the best path and even execute the trade directly. This path mapping structure is referred to as "intention" and is an important direction for the future development of Ethereum middleware.

Although there is no consensus in the industry on the definition of "intention", there are some common views. One definition is: "Intention is a set of declarative constraints that allows users to outsource the creation of transactions to third parties while maintaining full control over the transactions." Another saying is: "Transactions are imperative, while intentions are declarative. Transactions specify how to run the EVM to change the state, while intentions only focus on the desired state changes, regardless of the specific implementation process."

Both definitions emphasize the "declarative" nature of intent, which seeks external assistance through data sharing between the user and the "solver". The user states the desired outcome, and the solver provides the means to achieve it. Unlike transactions with specific parameters, intents require third-party mapping. Additionally, there are some constraints to limit the possible paths. This helps to focus the possibilities into a smaller set from which the user can choose.

For example, in the case of my friend, the intention system allows us to broadcast the final goal to a group of solvers, which calculate the optimal path. We choose the price-optimal route and execute the transaction, with the intermediate steps handled by the solvers, requiring the user to confirm only 1-2 transactions.

The basic architecture based on "intention" already exists in the EVM. Any DEX will look for the best trading route when in use. For example, in the Curve interface, after selecting the trading assets, the system will automatically find the best LP for routing. For assets without direct trading pairs, orders will go through multiple LPs to achieve the best execution path, all completed in a single transaction. The system will also provide an estimate of price impact and suggestions on how to limit slippage.

This trading intention on the DEX is just a basic example. The interface is merely a tool to help construct the trade, while the intention is to share the desired outcome (such as acquiring the most USDC) and constraints (such as only selling a specific amount of FRAX). The solver is responsible for determining the best exchange solution.

Some aggregators like 1inch or DeFiLlama have already been using similar intent systems to build swap trades. Users provide execution parameters and then receive a range of possible relay options for the trades. These options may use the same trading path, but depending on the relayer, there will be different fees and gas costs. Users ultimately choose the best price/cost option.

In addition to trading aggregation, there are several other types of "intents" on Ethereum:

1. Limit Order: Allows asset withdrawal when conditions are met.
2. CowSwap-style auctions: Third-party order execution based on non-DEX liquidity
3. Gas sponsorship: Allows the use of specific tokens to execute third-party transactions.
4. Authorization: Like a whitelist, perform a database check before executing the transaction.
5. Transaction Batch Processing: Allows for batching of gas efficiency intentions.
6. Cross-chain exchange: services provided by Socket

Despite the increasing diversification of order types, the simplest way to describe the intent may be "an upgraded version of a limit order." A limit order is executed only when someone accepts the order, aiming to buy a specific quantity of an asset at a certain price.

Similar to limit orders, an intention consists of two parts: the final state the user expects and the transaction initiated by the solver. Together, these two parts comprise everything needed to execute the trade.

The design of intent architecture carries almost no risk. Firstly, solvers have no incentive to propagate intents that contain profitable MEV. In many cases, extraction of MEV requires executing user orders on-chain, which exposes the blockchain state, allowing extractors to profit from this information. Common examples include backrunning and sandwich trades.

The core feature of intent is data exposure. Signing an intent message means you are willing to extract MEV at the cost of convenience. Since intents cannot be directly broadcast to the Ethereum mempool, they are stored in a private off-chain Interpool. These Interpools can be permissioned, permissionless, or hybrid.

Unlicensed Interpool uses a decentralized API, allowing nodes in the system to freely share intentions and granting executors unrestricted access. For example, the 0x protocol relayers and the proposed shared ERC4337 memory pool. Open memory pools are vulnerable to DDOS attacks and cannot guarantee the prevention of malicious intent propagation.

In contrast, permissioned memory pools use trusted APIs that can resist DDoS attacks and do not require intent propagation. They rely on trusted intermediaries to ensure execution quality. These intermediaries usually have a good reputation and are motivated to ensure top-notch execution. However, they still operate on strong trust assumptions, which conflicts with the core principles of open blockchains.

Hybrid solutions attempt to balance the advantages of permissioned and permissionless systems. They may combine permissioned dissemination and permissionless execution, or vice versa. For example, order flow auctions like CoW Protocol use trusted parties (off-chain order matching) to operate the auction, but participation is open.

Currently, the most popular Interpool is centralized and requires permission, and they have no incentive to share information with competitors. The risk is that if one party monopolizes most of the intent-based transactions, it may start to introduce additional fees and other rent-seeking behaviors, potentially eroding users' bargaining power.

When considering intent as a limit order, we can draw an analogy with Robinhood's order flow payment (PFOF). Robinhood offers "free" trading, which essentially involves selling users' order flow to market makers instead of sending it to traditional exchanges. Market makers profit from the bid-ask spread of the orders. Although brokerage firms have an obligation to provide the best execution for their clients, the economic incentives of PFOF may influence their decisions.

Intent can be seen as a form of PFOF arbitrage, which we refer to as MEV. The arbitrage opportunities created by long-standing unfulfilled orders may be more valuable than trades directly added to the Ethereum mempool, as solvers can determine the route rather than competing for MEV in a specific block.

If there is a lack of regulation and transparency, solvers are likely to provide the worst routes, as their profits are inversely related to execution quality. Users need to choose solvers carefully and leverage their bargaining power to force solvers to compete with each other. The solver that can deliver the highest returns for users will win the order flow.

CoWSwap adopts this design, using batch auctions to find the best settlement price for traders. CoWSwap does not execute orders immediately, but instead collects and settles them in batches. The system matches orders through public competition among solvers. After the batch ends, the solver submits a proposal for settling the orders.

Batch auctions ensure that the transaction prices within a batch are the same, avoiding the problem of miners reordering transactions. The CoW Protocol uses order flow auctions to ensure traders get the best price. However, there is still some MEV in CoW orders, as market makers need to engage in arbitrage trading elsewhere to maintain profitability.

Currently, some protocols are developing intent-based hybrid system infrastructure. Flashbot's SUAVE is building private memory pools and block construction networks to route traffic to L2 and Ethereum. Anom is dedicated to building the next generation of completely permissionless infrastructure. Several other companies have also joined this field.

Although there is no consensus in the industry on the final winner of intent systems, this is part of a middleware layer revolution happening in the cryptocurrency space aimed at improving user friendliness. Current intents are mainly used for currency swaps and order batching, but the future goal is to make them applicable to a broader range of data and arbitrary operations.

This creates opportunities for building new possibilities, such as on Fraxchain, where all wallets may default to account abstraction. A powerful intent layer can unlock new use cases for Frax products and simplify applications built on it.