The core advantage of DeFi is actively destroying itself.

Author: Thejaswini, Source: Buttercup Network, Translation: Shaw Golden Finance

“Extinction” is a very heavy word. As I have always believed, I am optimistic about DeFi itself. But on the other hand, every few months, DeFi repeats the same risk drama, only to act surprised by the outcome afterward. Over time, I have to admire this persistence.

On April 18, an attacker exploited a vulnerability in Kelp DAO’s LayerZero cross-chain bridge, stealing 116,500 rsETH tokens, worth about $292 million, accounting for roughly 18% of the token’s circulating supply. The attacker did not sell immediately—today’s hackers are more cautious in their actions.

The hacker deposited 89,567 of the stolen tokens into Aave V3 as collateral, borrowed native ETH against it, and then withdrew completely.

Thus, Aave was left with an uncollateralized bad debt of about $196 million, while the collateralized tokens, which shouldn’t have existed just hours ago, suddenly appeared out of nowhere.

Within 48 hours, Aave’s total value locked (TVL) plummeted from $26.4 billion to nearly $20 billion, and has now fallen to $14 billion; its native token AAVE also dropped by 16% in price.

SparkLend and Fluid urgently froze rsETH-related lending markets, Lido paused the earnETH product, and Ethena, even without direct exposure, quickly shut down all its LayerZero cross-chain bridges.

When the neighbor’s house catches fire, their uncle, fearing spillover, simply throws all the furniture out the window. In today’s industry environment, how can I blame such self-preservation behavior?

Aave founder Stani Kulechov clarified on Twitter: The Aave contract itself was not hacked, the emergency pause mechanism worked normally, and the entire system operated exactly as designed.

If this system truly functions as designed, then that very design is a disaster.

Kelp’s cross-chain bridge uses a single-node decentralized verification network (1-of-1 DVN) architecture. Here’s what that means:

• Single Validator: Only one entity is responsible for transaction verification.

• Single Signature Authorization: Only one digital “confirmation signature” is needed to approve asset transfers.

• Unlimited Authority: With this one signature, the entity has the highest authority to mint billions of dollars in new tokens.

Even a traditional bank, for a $10k wire transfer, usually requires approval from two or three managers.

But Kelp’s protocol, handling cross-chain assets worth billions, prioritizes low cost and “simple” architecture, completely lacking effective checks and risk controls.

Now, let’s talk about Aave. Previously, Aave accepted rsETH as collateral with a loan-to-value (LTV) ratio as high as 93%. Simply put: For every $1 of rsETH deposited, Aave loans out $0.93 of real assets. Only with extremely high confidence in the collateral would such aggressive parameters be set. But Aave’s trusted token can be inflated out of thin air by a single transaction, increasing its total supply by 18%.

The deposit cap for this token is set very loosely—one wallet can deposit up to 90k rsETH at once, without triggering any risk warnings. How to describe such risk parameters? Perhaps the answer is obvious.

Every post-mortem concludes the same: Smart contracts executed correctly, oracle data reported normally, liquidation mechanisms triggered as per rules, emergency pause functions activated within minutes. These are facts, but they miss the core issue.

Aave’s protocol design is fundamentally flawed: as long as the risk isn’t publicly exposed, the system allows anonymous wallets to deposit billions of dollars’ worth of unbacked collateral assets and issue real loans based on them.

This design stems from a core philosophy: as long as assets have oracle prices and deposit limits are sufficiently loose, they can be unconditionally used for lending and collateralization.

Groups that believe in “code is law” always take this phrase as a reassurance. But code being law doesn’t mean the rules are reasonable; it only means they are mechanically enforced without buffer. If the rules state “any token with an oracle price can be highly collateralized,” then this unconditional enforcement is the most terrifying aspect.

Over the past six years, DeFi has been touting composability. Everyone talks about “financial LEGO”: protocols nested layer upon layer, funds flowing freely, mechanisms efficient and intricate.

In contrast, traditional banks would never casually treat interbank promissory notes as collateral; behind the scenes, there are dedicated risk control teams conducting multiple layers of review. This is a lesson from the 2008 financial crisis: highly interconnected financial positions only share risks, not profits. Globally, all financial regulators saw through the fatal flaws of DeFi architecture as early as 2019.

Layered counterparty risks, zero-discount risk controls, no credit limits, no last-resort lenders—yet the entire industry blindly believes that on-chain transparency can replace all risk management bottom lines.

Aave’s core liquidity pool utilization once hit 100%, making it impossible for depositors to withdraw their ETH. In traditional finance, this is a classic bank run.

When users deposit assets into Aave, they are simultaneously exposed to four layers of risk: Aave’s risk parameters, rsETH oracle, Kelp cross-chain bridge, and Kelp’s decentralized verification network configuration. And most Aave users have never even heard of the last.

Cost of the bill

• Kelp incident loss: $292 million;

• Three weeks ago, Drift incident loss: $285 million;

• In just 21 days, two industry blue-chip protocols lost nearly $500 million in total.

In Q1 2026, crypto venture capital deals plummeted 48.9% year-over-year, with funding rounds shrinking from 358 to 183. Meanwhile, AI companies raised as much as $242 billion, accounting for about 80% of global VC funding.

Capital used to improve DeFi’s underlying infrastructure is now flowing into other more speculative sectors. The real innovations DeFi desperately needs—conservative loan-to-value ratios, multi-validated node cross-chain bridges, effective insurance mechanisms, routine audits of core configurations—are ignored.

No one has ever become rich by pondering “what disaster might happen if a single validation node (DVN) becomes a single point of failure.” Of course, the hacker behind the April 18 attack is an exception.

Let’s do a simple calculation: on April 18, Aave’s safety reserve pool was $390 million, while the platform’s outstanding loans totaled $17.82 billion.

The implied capital adequacy ratio was only 2.2%.

Basel III requires banks to maintain a minimum capital adequacy ratio of 8%, with JPMorgan Chase actually meeting 15.5%. Aave’s maximum penalty for collateral liquidation of AAVE is 30%, meaning the actual available safety reserve is only about $180 million, and the effective capital adequacy ratio drops further to around 1.0%.

Simple and feasible solutions have always existed: adopt multi-validator (Multi-DVN) architecture, reduce loan-to-value ratios for questionable assets, set truly effective token supply caps. The most costly solution, however, is to accept a harsh reality: composability is never an advantage; it’s just a sophisticated risk transmission mechanism.

I don’t believe this is merely a security issue. For six years, DeFi has been trapped in a misconception: on-chain transparency equals asset safety. And every 18 months, larger-scale incidents repeatedly prove this fallacy.

“Financial LEGO” sounds wonderful. But it also perfectly describes the current reality: stacking plastic blocks on a burning tabletop. This lesson was taught to me by my elders—sharp and to the point.