Subscribe to Bankless or sign in
gm Bankless Nation,
Sunday's exploit struck at the heart of DeFi.
The entire industry is going to have to move forward from this exploit with a new mindset – and a new design. In today's essay, David Hoffman lays out the path.
Thanks for being a Premium subscriber,
luma 🫡
Sponsor: MetaMask - Trade prediction markets on mobile, powered by Polymarket.
Although not even in the top 10 crypto hacks in terms of dollar-size, the LayerZero<>KelpDAO<>
Aave exploit will go down in history as the most consequential DeFi hack of all time.
The entire onchain industry is now on a completely different path than we were before.
The implications to crypto come in two buckets:
- DeFi needs to be Rearchitected
- Validators/Security Councils will need to standardize recovery operations (or give up control)
DeFi Needs to be Rearchitected
The LayerZero<>KelpDAO<>Aave exploit occurred because each component in their composable DeFi stack trusted each other.
KelpDAO trusted that LayerZero’s DVN wouldn’t ever be compromised (it got compromised).
LayerZero trusted KelpDAO to choose the appropriate level of security (they didn’t)- Aave trusted KelpDAO’s rsETH collateral to always be fully backed (it wasn’t)
Trust lowers the costs of transactions and improves the welfare of everyone involved. So, it's enticing to assume these protocols are operating inside of trustworthy environments.
But they are not.
Permissionless systems are inherently adversarial environments. Building under the premise that 99.99% of users are good doesn’t change the fact that it only takes one bad actor to attack the system.
The LayerZero<>KelpDAO<>Aave structure forgot this paradigm.
DeFi must never forget it lives and operates inside of an adversarial environment.
Building composable systems means that each component must assume that every component it is connected to will fail one day. It can never trust that an external component will work as it's intended to. With AI, all possible attack vectors will one day be discovered, and any vulnerable trust assumption will be exploitable.
As Odysseus from Phylax Systems wrote in his incredible article: “crypto needs an aerospace mindset.” The aerospace industry builds aircraft in a way where maximum safety and redundancy are non-negotiable. All commercial aircraft can experience a total engine failure at their worst possible moment – during takeoff – and still be able to climb using their other engine.
Take the Airbus A320 for example. It runs five flight control computers from two different manufacturers. Two different software implementations. Written by two independent programming teams. The architecture assumes the hardware will fail. Safety comes from the system being designed to survive when its components aren't. (Fun fact, my freshman year of high school was at Aviation High School in Seattle, WA – everything was aviation-themed.)
For Aave – and lending protocols broadly – this means rate limiters and circuit breakers.
Rate limiters constrain the size of the exits, so that an attacker can’t depart with its entire stolen haul all at once. For example, if Aave had a $10M/day rate-limiter on withdrawals, then its exposure to the KelpDAO<>LZ exploit would have gone from $280M to $10M because people quickly took notice of the attack.
Having completely unconstrained withdrawal rates from Aave allows the attackers to withdraw all the exploited funds in a single block. Surely the 99.99% of good actors using Aave can wait a few hours to withdraw tens of millions of dollars so that Aave can protect itself and its collateral from thieves.
This is just one example of a way in which a component in a composable system can be hardened and protect itself from its connection to other components that might fail.
All of DeFi will need to be rebuilt in this way.
Validators/Security Councils Have a Decision
Around $70M of the stolen funds from Aave happened on Arbitrum. In an unprecedented act, the 
Arbitrum Security Council (9 of 12 multisig) froze and recovered the ~$70M of stolen ETH on its chain.
It was a complete violation of the state of Arbitrum: the Security Council voted to change the state of Arbitrum and yoinked funds from the attacker to the custody of Arbitrum, who is holding it awaiting clarity from the Aave/KelpDAO/LZ recovery effort.
I think this was the right choice, and I believe that’s the consensus take too.
People taking the opposite stance will argue that this one ‘correct’ choice will lead to tens, then hundreds, then thousands of similar choices being made, which eventually takes us to a place not dissimilar from TradFi.
The argument is that the center cannot hold – all chain operators are now faced with a decision: come up with rules and standards for when the decision to execute top-down asset recovery is used, or completely burn the ability to recover the assets at all.
The Arbitrum recovery has set a precedent for the industry.
Now that this event happened, I feel that it was inevitable that our industry arrived here. The security council was able to seize assets. It was given an opportunity to do so, without any collateral damage to anyone or anything else on Arbitrum. It took the opportunity. It was surgical, with an unequivocally good outcome.
Future instances of lost or stolen funds might not be so clean, and will likely create much messier, contentious decisions. How will the security council make decisions like this in the future? Will they be ready to make the same choice again, if the fog of war is a bit thicker or the outcomes less certain? Will they want that liability? What if someone shows up with lawyers and makes demands?
Pandora's box has been opened.
Maybe the center can hold… and the fact that ARB token holders have the ability to remove individual security council members, or the entire existence of the security council, will keep things in check. But if individual security council members are forced to be liable for decisions… who would possibly want that? Maybe the security council itself decides to self-destruct.
The center where the security council maintains itself and also continues to recover obviously-stolen funds is possible… but it seems in the fullness of time, Arbitrum and chains like it will have to pick a strategy – either throwing the ring into Mordor or creating legally-defensible structures and processes for asset recovery (which is expensive and might not be economically rational!)
Closing Thoughts
If chains choose the latter path – the path in which they burn their control over their chain, this only increases the need for DeFi to implement an aerospace mindset to its safety architecture. The chain is the last layer of defense for saving DeFi assets, and it's a responsibility that no chain operator wants or signed up for anyway.
As an industry, we should be building for a world in which the chain cannot recover assets – because that seems like the more logical, sustainable equilibrium.
I think there will also be chains that choose the other path - the more fintechy path of committing to do their best to protect their user assets from exploits and attacks. While I think this is better for users and certainly a selling point that people would consider when choosing a chain, I fear that the costs associated with this are simply too high.
A chain committed to this path would need to hire a team to manage this whole process. The legal costs would be large. The liability risk could ultimately be extraordinary. It doesn’t seem like it’s in the best interest of the chain operators.
For now, my eyes are on 
Coinbase and their own decision on burning their keys. Base's Azul upgrade is ready for implementation, and would move Base from a Stage-1 to a Stage-2 rollup, removing Coinbase's ability to do any sort of asset recovery or change the state of the chain. Will Coinbase take the plunge?
My money is that Base will eventually burn the keys and go fully decentralized. Not only for the reasons stated above, but because this opens the door for further products and services they can bring to Base that they otherwise couldn’t – think tokenized securities and securities marketplaces for example.
While this exploit was a dark moment for DeFi, it seems that the hole is being filled via Arbitrum's recovery effort and voluntary donations from generous teams – 
Mantle donating 30K ETH, EtherFi, Stani, and 
Ethena all donating 5K ETH apiece. I’m not even an rsETH holder and I am incredibly grateful for their commitment to protect the brand and value of DeFi.
There's a lot to observe from this exploit and plenty to ruminate on for this industry's future. I'll be here, and hope you will too. Onwards.
Trade the outcomes of real-world events across sports, politics, crypto, finance, all on mobile—with prediction markets on MetaMask, powered by Polymarket. No KYC, total self-custody, and simple two-tap trades.
📈 The Asset
- Bitmine bought 10,000 ETH OTC from the EF and is now staking +70% of its total holdings
- Etherealize published the Productive Money thesis
- The EF swapped 21,269 aWETH to wstETH
- Grayscale staked another 102K ETH for its Ethereum Mini Trust
🏛️ The Protocol
📱 The Apps
- Aave froze its rsETH markets in the wake of the Kelp exploit
- Shutter unveiled Perpetual Endowment Networks
- Alchemix postponed lifting its V3 deposit caps
- Catalysis introduced Covered Vaults
- EtherFi updated its LayerZero DVN threshold to 4/4
- Eth.limo added support for local ENS gateways
- Fileverse launched Comments v2
- Fluid created the aWETH Redemption Protocol
- Lido outlined its EarnETH Vault plans re: the rsETH shortfall
- Lighter added support for ETH collateral
Polymarket is officially migrating to a new chain- Safe published its Q1 2026 report
🤫 The Privacy Stack
- Privacy Cash launched on Base
- Starknet rolled out its privacy-centric Shinobi upgrade
- Sunnyside Labs launched Privacy Boost on OP Mainnet
- Tom Lehman proposed baking private transfers into the Ethereum L1
🐸 The Culture
- DeFi United has nearly raised 100% of the funds needed to cover Kelp’s rsETH hack
- MegaETH is holding its TGE on April 30
- Giveth kicked off its Ethereum Security QF round
💽 The Tech
- Base introduced Azul, the L2’s first independent network upgrade
- L2BEAT launched its Interop page and added “Quantum Resistant” tags to its ZK Catalog
- Mantle transformed from a Validium into a ZK Rollup
- Ronin is migrating to the OP Stack
- ZKsync Lite is shutting down on May 4
Markets are hitting new highs, but crypto just took a major hit.
Ryan and David break down the $300M KelpDAO exploit and why it exposed deeper flaws across DeFi and Layer 2s, including Arbitrum’s controversial decision to freeze funds.
They also explore whether AI will drive deflation or inequality, unpack a new bullish ETH thesis, and debate why the biggest risks in crypto may be building beneath the surface.
Tune into this week’s Rollup! 👇
