Original author: @BlazingKevin_, Blockbooster researcher

On December 11 last year, a16z crypto released the annual "Big Ideas 2026: Part 3." In the section on stablecoins written by partner Sam Broner, there are the following points worth discussing:

"Stablecoins lacking a strong credit infrastructure look like narrow banks—they only hold specific liquid assets deemed very safe. Narrow banks are an effective product, but I do not believe they will become a long-term pillar of the on-chain economy."

Broner then gives his judgment:

"We have already seen a new batch of asset managers, curators, and protocols beginning to facilitate on-chain asset-backed loans supported by off-chain collateral. These loans are typically created off-chain and then tokenized. I believe tokenization has little benefit here... so debt assets should be created on-chain instead of being created off-chain and then tokenized."

Four months later, in March 2026, Sam Broner left a16z to found The Better Money Company, and a16z crypto led a $10 million seed round, with Circle co-founder Sean Neville also participating in the investment. However—what Broner chose to do himself was not the "on-chain native credit creation" he specifically mentioned in the article, but rather another track: a stablecoin clearinghouse, facilitating low-cost exchanges between different compliant stablecoins, with signed partners including issuers and distribution channels like Paxos, Stripe's Bridge, and MoonPay.

Those who most favor stablecoin infrastructure and were the first to call out the "narrow bank ceiling" chose, when stepping in, the clearing/interoperability layer rather than the credit creation layer. This is because the credit creation layer is too difficult, and no project has matured enough to make him or equally scaled practitioners willing to stake their time. In other words, even those who understand this judgment best are still waiting for the "moment to step into the credit layer."

This brings us to the topic we are discussing today: when the entire industry is talking about "RWA tokenization," the next real structural opportunity may be "on-chain native credit creation"—a direction that has been discussed repeatedly but has yet to materialize at scale.

0. Defining "On-Chain Native"

"On-chain native credit creation" has two easily confused interpretations, and we will discuss the second one.

The first is "on-chain native" in the process sense: from loan initiation, interest rate pricing, to liquidation, the entire process is completed on-chain. In this sense, Aave, Compound, and Morpho are thoroughly on-chain native—loans are initiated on-chain, interest rates are dynamically priced by algorithms based on capital utilization, and liquidation is automatically executed by smart contracts when collateral ratios are breached.

The second is "on-chain native" in the context of credit assessment: underwriting credit using the borrower’s on-chain behavior, cash flow, and on-chain identity, rather than relying on over-collateralization or traditional off-chain credit reports and financial statements. This is the truly immature part.

The fundamental difference between the two lies in "what it is based on to lend." Aave's model is "over-collateralization"—if you want to borrow $100, you must first deposit $150 worth of ETH. This is essentially not credit; it is a pawn shop. It does not create any new purchasing power; it merely releases the liquidity of existing assets. Borrowers must already have money to borrow.

Real credit creation is "lending based on judgment of future repayment ability"—the bank lending you money to buy a house is based on your income, credit history, and repayment capacity. Such credit creates new purchasing power and is the core engine of the monetary multiplier and economic growth in modern economies.

Here, it is necessary to clarify a common misconception: "Aave's algorithmic interest rate is not a form of on-chain underwriting?" It is not. Aave’s algorithm sets interest rates based on capital utilization rates, not on the risk of the borrower. As the money in the pool is borrowed more, the interest rates rise—this prices the tightness of the pool's capital, treating all borrowers equally. Aave charges the same interest rate to every borrower in the same pool because it does not distinguish who the borrowers are. True underwriting, essentially, gives different prices to borrowers with different risks—that is the core of credit creation. A system that does not distinguish between borrowers, regardless of how complex its rate algorithms are, is not doing underwriting.

1. Current Situation

Regarding this direction, there are currently products on the market, with 5 to 10 teams seriously attempting it, but their combined TVL is still less than a fraction of Aave's single USDC pool. For example:

• 3Jane: Currently the closest attempt at "on-chain native credit underwriting." It uses zkTLS technology to pull off-chain bank data from borrowers (via Plaid integration) and on-chain asset profiles, with an underwriting algorithm called 3CA that calculates a "Jane Score" credit score in real-time, then issues unsecured USDC credit lines—borrowers do not need to put up any cryptocurrency collateral. Default disposal follows actual legal channels: bad debts are packaged and auctioned to debt collection agencies in the U.S., with recovered funds distributed between the collection agency and the lenders.
• It raised $5.2 million in a seed round led by Paradigm in June 2025, with participants including Coinbase Ventures, Wintermute, Robot Ventures—Circle co-founder Jeremy Allaire is also an angel investor. 3Jane is set to launch its mainnet in early November 2025 with an initial cap of about $50 million and initially limited to U.S. residents with total assets over $150,000.

However, even this most watched project in the field, which secured investment from Paradigm and was endorsed by Delphi, has an actual TVL that is very small (initially on the order of hundreds of thousands of dollars).

• Divine Research: Represents a completely opposite route to 3Jane. Divine is a San Francisco company founded by Diego Estevez that has been issuing unsecured USDC short-term loans through a platform called Credit since December 2024—by the second half of 2025, it had issued over 500,000 loans covering more than 100,000 borrowers, and completed a $6.6 million financing round.
• Its underwriting method is based on progressively building identity + performance history: borrowers must first complete an iris scan using World ID from Sam Altman’s Worldcoin to anchor a unique identity (and then start with a very small limit, usually under $100), with each repayment raising the limit, up to about $1,000. It targets underbanked populations in developing countries (Argentina, Nigeria, Colombia, etc.)—in the founder’s own words, "high school teachers, fruit vendors... basically anyone with internet access." Interest rates range from 20% to 30%.
• Its first loan default rate is indeed as high as about 40%, but as borrowers accumulate records in this "repayment-for-credits" system, its overall default rate has been reported to be close to zero—the 40% is the acquisition cost of the very front end (covered by high-interest rates and user recovery receiving WLD tokens) and not the steady-state bad debt rate of this model.

When comparing 3Jane and Divine, we can see the two routes of on-chain native credit and their respective limitations:

3Jane follows the "proof of assets/income" route—verifying your bank account and on-chain assets using zkTLS, targeting asset-rich borrowers (high-net-worth individuals, businesses); default disposal follows the legal channels of U.S. debt collection. Its limitation is that it serves those who already have assets, falling short of the true credit creation of "creating purchasing power for the asset-poor," and legal collection is effective only in mature jurisdictions like the U.S.

Divine follows the "identity + progressive trust" route—first using an iris scan to ensure only one borrowing identity per person, then gradually nurturing credit through "repayment-for-credits," targeting the asset-poor long tail in developing countries and truly addressing inclusive credit. It does not have collateral to recover nor effective cross-border legal recourse; the only consequence of default is "you won’t be able to borrow money with that iris again"—which sounds like a weak deterrent, but the close-to-zero steady-state default rate indicates that the positive incentive of "to borrow more, you must first repay" actually works for long-tail borrowers. The real limitation of Divine does not lie at the deterrence end but in two aspects: first, the credit it builds up is only valid within Divine; second, its entire defense against fraudulent identities is outsourced to World ID—a non-native solution to pseudonym issues.

The comparison of these two routes points to one conclusion: neither has solved "what basis to lend" under the toughest setting of "on-chain, facing a pseudonymous borrower," but each has introduced a lever from outside the setting. 3Jane circumvents by "proving you have money" (which is essentially still a form of collateral); Divine anchors identity with World ID and then forces out credit from behavior using the "repayment-for-credits" positive cycle. In other words, the hardest version—"based on on-chain behavior, judging whether a borrower you don’t know and can change addresses at any time will repay"—has yet to be directly addressed by either route; their cleverness lies in finding a lever to avoid tackling it head-on while still being able to lend money.

Other players include: Wildcat Finance (on-chain matching of bilateral private credit, where lenders and borrowers negotiate terms directly, and the protocol only serves as a matching engine and smart contract execution; upon default, lenders coordinate recovery directly); Clearpool, TrueFi (attempts at varying degrees of unsecured/under-collateralized lending); Union Protocol (credit based on social relationships); Accountable (verifiable credit disclosures for off-chain assets). The TVL of these protocols generally ranges from hundreds of thousands to millions of dollars, with a few larger institutions directed towards them.

Here, you may wonder: why are these small teams doing this while the biggest DeFi lending protocols—Aave, Morpho, Compound—are not engaging in unsecured credit themselves? They have the deepest liquidity, strongest brands, and most on-chain data, ostensibly making them best positioned to develop on-chain native underwriting. Their lack of engagement is due to two structural reasons:

• First, tail risk cannot be borne by token holders: over-collateralized liquidations are automatic and predictable, while the default losses of unsecured credit result in real bad debts, which governance token holders cannot bear—mass defaults could potentially break the entire protocol.
• Second, regulatory arbitrage space: over-collateralization has a clear legal narrative of "non-securities, non-traditional lending" (essentially collateral swaps), while unsecured credit instantly falls into the view of consumption credit regulation. Therefore, it is the business models and risk structures of the giants that determine they cannot and do not wish to pursue this venture—this, in turn, provides new teams with a structural window that giants cannot enter.

Next, let’s address another question: where exactly is the demand? If it is merely "should exist theoretically," then this is just a story seeking a solution for a problem. However, the real on-chain credit demand is already distributed across several specific scenarios: market makers and quantitative teams need working capital turnover but are unwilling to pledge equal collateral for it; on-chain native merchants, RWA asset initiators, and crypto projects need accounts receivable financing and prepayments; and a large number of small-to-medium borrowers who are directly blocked by the over-collateralization model—they do not have surplus crypto assets to pledge but have actual cash flows.

In other words, the over-collateralization model serves "those who already have money wanting to release liquidity," while the demand that is blocked from entering this model is precisely from "those with cash flow but lacking collateral"—this is the real market for credit creation. The demand has been filtered out by the collateral threshold of existing models and has never been accounted for.

2. Why Stablecoins "Need" to Solve This Problem

To understand why on-chain native credit creation is a "structural demand," we need to first understand the traditional monetary banking concept of "narrow banks."

Narrow Bank is a classic theoretical construct: a bank that only accepts deposits, holds only super-safe assets (short-term government bonds, central bank reserves), and does not issue loans at all. The deposits of a narrow bank are 100% backed by safe assets, theoretically never experiencing a run and never going bankrupt. It sounds very safe, but in history, it has never become mainstream—because it has a fatal commercial ceiling: it does not create credit and therefore does not generate monetary multipliers; the profit margins of the business model are extremely limited.

The core value of modern banks lies precisely in "fractional reserves + credit creation." You deposit $100 in a bank; the bank keeps some as reserves and lends the rest to others; the lent out money then transforms into someone else’s deposit and is lent out again... this process creates purchasing power far exceeding the original deposit (the monetary multiplier) and is the financial engine of economic growth in modern economies. Narrow banks actively give up this engine, thus they can only play a marginal role in the financial system, not a pillar.

Whether on-chain credit creation can truly generate a monetary multiplier depends on one premise—whether the lent-out stablecoins can be re-deposited back into the protocol, becoming new sources of lending. If they can (similar to the supply→borrow→supply cycle on Aave), then it will indeed produce a monetary multiplier effect; if borrowers primarily use the borrowed funds for off-chain consumption and the money leaves the on-chain credit system, then the monetary multiplier effect will be limited. Thus, strictly speaking, on-chain credit creation is a necessary condition for the monetary multiplier, but to what extent the multiplier can amplify also depends on the capital reflow rate of the on-chain economy.

Now looking at the stablecoin system—it is precisely a giant narrow bank. USDC and USDT absorb "deposits" (with reserves 100% in short-term government bonds and cash, not issuing any loans or creating any credit. The total scale of "deposits" in the entire stablecoin market—about $240 billion around mid-2025, and over $320 billion by mid-2026—lies entirely in safe assets, generating no monetary multiplier at all.

Here we must avoid a misunderstanding: "not generating a monetary multiplier" does not mean "not making money." Quite the contrary, issuers profit tremendously—they earn interest on the reserves held in government bonds. The GENIUS Act + CLARITY Act prohibits paying interest to holders, not the issuers profiting from spread. So the issue with stablecoins is not that "no one profits from it," but that this profit is locked within the layer of issuers, neither shared with users nor entering the multiplier cycle of credit creation. The value is being captured rather than amplified.

Thus, if the stablecoin system wants to break through the ceiling of narrow banks and truly become an "on-chain banking system," the only way out is to create credit outside the issuer—that is, at the DeFi protocol layer. However, the current credit at the DeFi protocol layer is not true credit creation; it is merely a pawn shop.

Hence, a logical loop forms: stablecoin issuers are legally prohibited from lending → credit creation can only happen at the protocol layer → the existing over-collateralization model at the protocol layer does not create new purchasing power → therefore, the only logical pathway for the stablecoin system to break through the ceiling of narrow banks is to develop true on-chain native credit creation.

3. Why Is It Stuck Until Now?

If on-chain native credit creation is a structural inevitability, why have only 5-10 teams tried for over a year, and why is the TVL still not scaling up?

The answer is a chicken-and-egg dilemma, but a more precise historical comparison is the U.S. consumer credit market before FICO.

Engineer Bill Fair and mathematician Earl Isaac founded Fair, Isaac and Company back in 1956, but the consumer-facing FICO credit score wasn’t officially launched until 1989, and it wasn't widely accepted in the industry as a standard until the mid-1990s when it was adopted by the two government-sponsored enterprises (GSEs) for mortgage loans (Fannie Mae, Freddie Mac). It took 33 years from the company’s founding to the score’s creation, and about 40 years for industry-wide adoption.

The maturity of credit infrastructure layers is measured in "decades" rather than "years." And it was this FICO score that first made credit a calculable, reusable, and standardized entity across institutions. For decades after the FICO became widespread, the U.S. consumer credit market truly exploded—scalability of credit cards, auto loans, and mortgages all followed the standardization of FICO. FICO is not just a function of consumer credit; it is the prerequisite for the scaling of consumer credit.

What is currently missing in on-chain credit is precisely this "FICO moment"—a widely accepted, mechanism-trusted, cross-protocol reusable "on-chain credit score."

Without this standardized credit layer, every protocol engaged in on-chain native credit is forced to build its underwriting system from scratch: 3Jane develops the 3CA algorithm and Jane Score; Spectral creates a credit score based on on-chain wallet behavior; Cred Protocol and Blockchain Bureau each develop their own on-chain credit models; the identity layer has Worldcoin and Gitcoin Passport attempting solutions. Every protocol is reinventing the wheel with no standard that can be reused by others. This resembles the U.S. before FICO—each banker had their own subjective judgment system, which could not be scaled up.

Currently, all attempts at on-chain native credit are stuck in a chicken-and-egg cycle: true on-chain credit assessment requires rich on-chain credit histories, but most real borrowers' economic activities are still off-chain, and there is not enough behavioral data on-chain to support underwriting. As a result, protocols either have to rely on off-chain data or restrict lending targets to "the wealthy whose assets are already on-chain." Neither route can reach the long-tail borrowers who truly need credit creation.

But the FICO analogy can also diagnose a deeper sticking point. The success of FICO was not just because it standardized credit scoring but also because it standardized default consequences—once you default, your FICO score becomes visible across the industry, impacting your borrowing capacity at any institution in the future. This "cross-institutional transmissibility of default consequences" is the true source of FICO’s deterrent effect: it is not an individual bank punishing you; it is the entire financial system sanctioning you.

On-chain credit has yet to establish this "cross-protocol transmission": the default penalties of a single protocol cannot take effect across platforms, and thus the bad debt risks of each protocol are locked into their own operations, unable to be diluted through industry-level reputation mechanisms.

A true "on-chain FICO" must simultaneously address these two matters of standardization of scoring criteria and cross-protocol transmissibility of default consequences. Many are attempting the former, while almost no one has touched the latter—which, furthermore, is tied to a deeper, possibly unsolvable issue of "durable defenses against fraudulent identities."

4. Temporary Solutions

Returning to the current market situation, we believe that layer of infrastructure (durable identity + cross-protocol default broadcasting + standardized scoring) may be very difficult to establish. Therefore, without reaching the endpoint, whoever can work around it and capture a portion of temporary value appears more valuable at this point in time.

First, let’s see why these three are tough—and independently so. This means that betting on "they will all be built" is essentially betting on a series of locks being opened simultaneously:

• **First Lock, Data Pipeline:** zkTLS and similar technologies aim to reliably bring off-chain data onto the chain—but this precisely indicates that there is not enough credit data on-chain, necessitating reliance on off-chain data. A system that fully depends on transferring bank statements and VantageScores simply adds a layer of encryption to the existing off-chain credit assessment feature; hence the assumption that "the data pipeline can be built into a truly on-chain native underwriting foundation" is inherently fragile.
• Second Lock, Credit Bureau—this is the most valuable layer, yet also the least likely to emerge spontaneously because it is a classic public good/coordinated problem. Let’s first look at how traditional credit bureaus were formed—they depended on decades of industry consolidation, regulatory push, and eventual oligopolistic mergers, none of which came from a startup company "creating a better protocol". Relying on an open protocol to spontaneously grow a widely accessible credit bureau in a few years is essentially treating something that took half a century to sculpt through regulation and mergers as a product that can be engineered and delivered.
• Third Lock, Durable Anti-Fraud Identity—this is the foundational level and could be the most fundamentally unsolvable one. It contains a deadlock: any sufficiently strong identity binding (mandatory KYC, biometrics) sacrifices the core attributes of openness and permissionlessness on-chain, reverting on-chain credit back to a traditional system that requires centralized identity endorsement; whereas any sufficiently light scheme preserving permissionlessness cannot prevent the "changing address to start over."

Under the constraints of these three locks, we believe that on-chain native credit creation is a direction that is extremely challenging to reach the endpoint.

Currently running products, without exception, are "bypassing those locks." They are borrowing the missing component of the endpoint from outside the chain (off-chain legal systems, biometrics) rather than crafting that component within the chain. So, is there a more broadly applicable or more "on-chain native" phased workaround than these two paths?

5. Better Phased Directions

Breaking down the endpoint reveals that what has everyone stuck is fundamentally the same issue: the endpoint requires that "punishments" can take effect—if you defaulted, the consequences must be traceable to you across protocols and addresses. And "making punishments effective" relies on those three most challenging components: durable identity, cross-protocol transmission, and trustworthy data.

"Punishment" is a public good; no one has the incentive to build it alone; however, "rewards" are a private good, with every protocol motivated to build it.

Expanding on this asymmetry. If you want to "punish" a defaulter, you need all protocols to see their blemishes—this leads to the public good dilemma where no one wants to provide. But if you want to "reward" a compliant borrower, all you need to do is provide some benefits to those addresses with clean histories within your own protocol. The cost of "rebuilding" is constrained by the rewards mechanism.

This flip-verts the entire problem. The endpoint seeks to answer "how to prevent defaulters from escaping"; phased products pursue "how to allow compliant borrowers to accumulate increasingly valuable items." The latter is the form of on-chain credit most likely to advance successfully.

This "reward compliance" logic has already been successfully applied by Divine. Its "repayment-for-credits" cycle is essentially about "using accumulated good repayment records to exchange for better borrowing conditions (higher limits)."

Therefore, the following phased directions will adapt the same "reward compliance" logic validated by Divine into scenarios it has yet to cover—especially into the DeFi battleground, where the sufficiency of collateral and capital efficiency are truly the pain points. Their commonality lies in being built on "reward compliance" rather than "punishing defaults."

Direction One: Gradually Lower Collateral Rates—Let Reputation Function as "Discounts" Rather than "Substitutes."

The ultimate goal of on-chain native credit is "zero collateral," which is a gradually approaching line that is hard to reach. However, between "150% over-collateralization" and "zero collateral" is a full segment of a continuous spectrum, and this spectrum itself represents a massive, almost untouched market.

The most natural phased product looks like this: for each timely repayment and each secure closure by a borrower within a specific protocol, it is recorded in their compliance file for that address; as the clean history accumulates, the protocol gradually eases its requirements—collateral rates drop from 150% to 130%, 120%, 110%, with interest rates offered discounts, limits raised, and liquidations given some buffer. This mirrors the real-world path of "deposit-backed credit cards → standard credit cards → increased limits": first prove yourself with a deposit, then exchange the record for the deposit back.

Aave’s efficiency mode (E-Mode) might seem similar to this. However, E-Mode adjusts for asset correlation (for example, between stablecoins and ETH or stETH) rather than borrower history: it treats everyone equally, looking only at what you’re pledging and not at who you are or how many times you've repaid.

Direction Two: Replace "Judging People" with "Withholding Cash Flows."

Getting to the endpoint is particularly challenging, largely because it must solve one of the trickiest issues: predicting a borrower's character.

On-chain programmable cash flows can be automatically withheld at the smart contract level. If a borrower’s future income is already on-chain (the sales revenue of an on-chain merchant, a protocol's fee sharing, or even a tokenized salary flow), then the loan can be designed so that when income arrives, the contract automatically deducts repayments before the remainder goes to the borrower. The lender's "collateral" is that future cash flow managed by code, which the borrower has no access to.

Projects like Goldfinch, Centrifuge, and Maple are working on bringing off-chain accounts receivable on-chain—underwriting, due diligence, and collection are still off-chain. The real phased opportunity lies in cash flows where the income occurs on-chain, hence can be directly intercepted by the contract.

Direction Three: Curator Model.

Since it's unlikely there will be a standardized, trustworthy underwriting algorithm developed on-chain any time soon, let’s stop pretending that algorithms can offer underwriting. Instead, let those who actually have underwriting capabilities and are willing to cover the first-loss capital handle the underwriting. This is the delegated lending and curator model: the protocol provides the track (settlement, transparency, automated contract execution), while specific borrowings and conditions are determined by a patron/curator who deposits the first-loss capital. They earn spreads and take the initial losses.

It does not need a one-size-fits-all on-chain FICO; it substitutes that universal scoring layer with "localized trust + first-loss capital." Aave's credit delegation, along with the curator/treasury models being pursued by Maple and Morpho, are early forms of this direction. Its value will accumulate on good curators—those whose treasury doesn’t explode with losses and returns remain stable will draw in more deposits, creating a slowly emerging, performance-backed form of credit.

However, viewed dialectically, it essentially shifts the trust issue one layer up—you may not need to trust the borrower, but you need to trust the curator. It resembles "packaging the human management aspects of off-chain credit with on-chain transparency and automated liquidation."

None of these three phased directions attempt to punish defaults; they focus on "rewarding compliance"—allowing an address's accumulated good history to gradually convert into lower collateral rates, priority over cash flows, favor from curators, or various tangible conveniences within the ecosystem. Punishments require industry-wide cooperation; rewards require only individual protocols or ecosystems to have motivation.

Therefore, the more likely path for on-chain native credit creation is: each protocol, each ecosystem, delving into the matter of "deserving better conditions for compliant addresses," allowing on-chain compliance records to gradually gain value in diverse, specific scenarios; these scattered, reward-anchored accumulations of credit will grow address by address, protocol by protocol, eventually reaching a point where they begin to resemble true credit.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。