When established DeFi protocols begin to "decentralize": an experiment on speed and control

A Noteworthy Shift

Recently, while observing the DeFi lending sector, I noticed an interesting phenomenon. Some established lending protocols that have been operating for several years are making a bold decision—gradually opening up the market creation power, which was originally strictly controlled by DAOs, to external curators. The logic behind this shift actually reflects some deep changes that the entire DeFi industry is undergoing. From the data, this adjustment seems to have made the overall DeFi lending sector more prosperous. As of now, a total of $46.583 billion is flowing through lending protocols.

Let’s start with the background. Traditional DeFi lending protocols have always been typical DAO full-stack governance models, where every step from asset listing to risk parameter setting requires DAO voting. This model was indeed meaningful in the early days of DeFi, as the industry was still in an exploratory phase, and being cautious was always the right approach. However, the problem is that the environment surrounding DeFi has changed too rapidly in recent years. The types of on-chain assets have exploded, and in the early days, it was just the familiar ones—stablecoins, mainstream coins, LSTs; now everything is moving on-chain. RWA, tokenized stocks, basis yield products, private credit—these were once unimaginable assets being moved on-chain, and now they are growing at a rate of several billion dollars each year. But there’s a problem: the traditional approach is to tokenize these assets and then throw them onto DEXs, hoping the market will generate liquidity on its own. The reality is that many yield-bearing assets are simply not suitable for this model; they are designed to be held, not traded. Thus, we see more and more issuers starting to bypass traditional paths, creating treasury-style mint-and-redeem models, or like Convex, simply not tokenizing and directly doing staking. These new play styles cannot be supported by traditional lending protocols, as their architecture is designed around standardized tokens.

At the same time, the group of Risk Curators has grown more than 20 times in the past year. These individuals are neither developers nor DAO members; they are professionals focused on risk management for specific types of assets. Their understanding of specific markets may be more precise than any parameters voted on by a DAO, but under the traditional model, they have no room to operate because all decisions must go through the DAO process. Changes at the network level are also worth mentioning. The fastest-growing networks over the past year—Plasma grew by $8.6 billion, BSC by $6 billion, and Base by $4.9 billion—share a common trait: they are all EVM-compatible. The entire EVM ecosystem added over $20 billion in TVL last year, more than the combined total of the top five non-EVM networks. For protocols, if they cannot quickly deploy to these emerging EVMs, it means missing out on growth opportunities.

Next is the unavoidable topic of compliance. Assets like Ethena's USDtb, Tether Gold, and Superstate, which have KYC requirements, contributed over $10 billion in TVL growth last year. When institutional funds come in, compliance tools must keep up, but they cannot sacrifice the composability and openness of DeFi. Finding this balance is quite a test for protocol design. Speaking of institutions, take a look at the recent survey conducted by EY and Coinbase, which suggests that the number of on-chain institutions may triple in the next two years. More importantly, half of the eight core use cases these institutions are focused on are related to lending, yield, and strategies. Traditional financial giants like Superstate, Securitize, and VanEck are starting to take on-chain business seriously, no longer just testing the waters. Retail participation is declining, while institutional funds have become the main driver of growth; this trend is already quite evident.

Various Paths for DeFi Protocol Transformation

In the face of these changes, different protocols have chosen different ways to respond. GearBox DAO has launched a new model called "Permissionless," which is quite straightforward in name, aiming to make market creation a permissionless process. How is this achieved? Simply put, it allows curators and institutions to directly launch their own lending markets on GearBox's infrastructure without coding.

The technical implementation is quite clever. Curators can deploy multiple markets, each with its own risk parameters and interest rates, but all markets draw liquidity from a single unified pool. The benefit of this design is that curators do not touch the funds deposited by users; they can only set rules and cannot transfer assets. This is safer for lenders and more compliant for curators, as they do not have to deal with the hassle of fund custody. Interestingly, this system supports scenarios that traditional lending protocols cannot handle, such as non-tokenized yield systems or illiquid semi-liquid assets, allowing credit markets to be established under GearBox's architecture. It also integrates over 20 mainstream DeFi protocols—Curve, Pendle, Convex, Uniswap, and others—so curators can directly call upon them without needing to integrate themselves.

Cross-chain deployment is also done flexibly. GearBox is currently activated on 28 EVMs, but only 9 networks are actually operational. This design is smart; the activation cost is low, and if a market needs to be established on a particular network, it can be formally operated then, saving operational costs while maintaining expansion flexibility. Imagine if every time a new chain had to go through a complete DAO governance process; the speed would simply not keep up with market rhythms.

In contrast to GearBox's "fully open" path, Morpho has chosen a different model. Morpho is also promoting a "curator-driven" transformation, but its core logic is to break down the lending market into independent "vaults." Each vault is operated by a professional risk management team, setting its own collateral, risk parameters, and interest rate strategies, but the entire system is connected through a unified liquidity layer. This architecture allows Morpho to maintain flexibility while isolating risks more thoroughly—problems in one vault do not affect others. From the data, Morpho's TVL grew from less than $1 billion in 2024 to over $3 billion, attracting top curators like Steakhouse Financial and Block Analitica. Its advantage lies in giving curators greater autonomy; each vault is almost an independent lending protocol, but the downside is that users need to choose between different vaults, increasing decision costs.

Aave has chosen a more conservative path. As a leader in DeFi lending, Aave launched Aave V3 in 2023, introducing "isolation mode" and "efficiency mode," but the core market creation authority remains in the hands of the DAO. New assets must still go through a strict proposal, audit, and voting process to be listed on Aave, which often takes weeks or even months. The benefit is that Aave has maintained a solid safety record with a very low bad debt rate, which is a huge attraction for risk-averse institutional users. However, the cost is speed; when new yield opportunities arise in the market, Aave often reacts slowly. This is why we see some emerging asset classes, such as certain RWA tokens or LST derivatives, launching on Morpho or GearBox first, and appearing on Aave months later.

These three models represent different trade-offs between "speed and control" in current lending protocols: GearBox pursues maximum speed and openness, completely delegating decision-making power to curators; Morpho finds a middle ground between speed and control, balancing flexibility and safety through vault isolation; Aave adheres to centralized review, trading speed for safety and reputation. Interestingly, all three models are experiencing growth, indicating that there is demand in the market for products with different risk preferences.

While GearBox opens up market creation authority, it also places a high emphasis on security, establishing an "optimistic liquidation" system that forks the Ethereum mainnet every three hours for stress testing, and provides curators with testing modules and automatic liquidation bots to ensure system stability and low risk. Since the launch of the Permissionless model in March this year, TVL has grown by 209%, adding $224 million, and EVM coverage has increased from 4 to 28, attracting top curators managing over $1.5 billion in assets. During the extreme market fluctuations on October 10, all markets remained stable with no bad debts, further validating the robustness of the system. With these achievements, GearBox DAO has officially transitioned to the Permissionless model through a proposal (GIP-264), with a TVL growth rate exceeding 220%, reaching $240 million. Currently, this model accounts for over 70% of the total TVL.

Where Are the Boundaries of Decentralization?

This decision may make some purists uncomfortable, as it seems like the DAO is "giving up control." However, the reality is much more nuanced. The core control at the protocol level—smart contract security, emergency parameters, systemic risk thresholds—remains firmly in the hands of the DAO. What is being opened up is only the decision-making power at the market level: what assets can be used as collateral, how to price risks, and what kind of liquidation lines to set. Delegating these decision-making powers to professional curators can actually allow the entire system to respond more quickly to market demands. This brings to mind a more fundamental question: is decentralization a goal or a means? Early DeFi treated decentralization as an almost religious belief, where all decisions had to go through the DAO process, even if it meant sacrificing efficiency. But now the environment has changed; institutions want speed and certainty, new asset classes are emerging endlessly, and regulatory requirements are becoming more specific. If we still cling to the "everything must be voted on" model, the result may be watching opportunities being seized by more agile competitors.

There is also a deeper issue: the logic of ecological niche competition will change. Previously, lending protocols competed on who had more efficient DAO governance and more active communities. Now, if everyone shifts to a curator model, the competition will be about who can attract the best curators. This means that the protocols themselves must possess sufficient appeal—perhaps a better tech stack, lower fees, or stronger brand and network effects. From a broader perspective, the transformations of GearBox and Morpho are actually a microcosm of the overall maturation of DeFi. We are witnessing a shift from "ideology-driven" to "pragmatism-driven." It’s not that ideology isn’t important, but when ideals and reality conflict, more and more projects are choosing to solve real-world problems first rather than sticking rigidly to ideals. What kind of long-term impact will this shift bring? It’s hard to say. Optimistically, specialization can improve the efficiency of the entire industry, encouraging more traditional financial institutions to participate and accelerating the mainstreaming of DeFi. The level of risk management will also improve overall due to the involvement of professional curators, as these individuals rely on this for their livelihood and are certainly more committed than amateur DAO members.

On the pessimistic side, this may lead to DeFi gradually aligning with traditional finance, and the qualities that initially made DeFi unique—complete openness, censorship resistance, community autonomy—may be diluted bit by bit. As more markets require KYC and more decisions are made by professional institutions, the boundaries between DeFi and CeFi will become increasingly blurred. By then, can we still say DeFi is "different"? However, this evolution may be inevitable. Any industry goes through a process of fading idealism and the rise of pragmatism from its inception to maturity. The early internet was also filled with utopian colors, and now it is dominated by a few giants, right? But this does not mean the value of the internet has disappeared; it has just been realized in another form. For GearBox, this transformation seems to be successful at least in the short term. TVL has doubled, market expansion has accelerated, top institutions are willing to participate, and it operates stably under extreme conditions—these are all tangible results. They have leveraged nearly four years of accumulated safety records and tech stacks, which have now become their core competitive advantages in attracting curators. From over $3 million in security investments, more than ten audits, and over $200,000 in bug bounties, to a history of never being attacked, these are not defenses that can be built overnight.

Interestingly, the transformations of GearBox and Morpho may inspire other protocols. It’s not that all projects should copy this model, but rather that when facing industry changes, they may need to think more flexibly about "what to persist in and what to adjust." Some protocols may continue to adhere to pure DAO governance, taking a small but beautiful route; others may embrace specialization like GearBox, pursuing scale and efficiency. Ultimately, which path is more correct may take years to clarify. But at least for now, we have multiple samples to observe and learn from. This is actually a good thing for the entire industry. After all, one of the charms of DeFi is that it allows various models to experiment and compete, right?

As I write this, I suddenly recall a reflective article written by Ivan, an early contributor to the GearBox team, discussing the pitfalls they have encountered and the lessons learned over the years. The title of that article is quite interesting: "HUMBLE: Hurdles Unlock Meaning, Bringing Lessons & Evolution." Perhaps this is the most core aspect of the whole matter—whether it’s the DAO or curator model, they are just tools. What’s important is to maintain the ability to learn and evolve, rather than dogmatically adhering to a certain form. Of course, as observers, we must also remain cautious. Data can be manipulated, narratives can be packaged, and the real test often comes from unexpected moments. GearBox's Permissionless model has only been running for a little over six months, and Morpho's vault model is still in its early stages, both needing longer periods of verification. But at least for now, these are experiments worth paying attention to. If successful, they may redefine how lending protocols operate; if they fail, they can still leave valuable lessons for the industry.

Regardless, it is certain that DeFi is entering a new phase. The era purely driven by ideology, rejecting any compromise, may indeed be over. Next, we will see more attempts to find a balance between ideals and reality; some will succeed, some will fail, but the exploration process itself may be more valuable than any single outcome.