DeFi’s Latest $290M Hack, Kalshi vs Polymarket & Will the Clarity Act Pass?

Summary

• The biggest DeFi exploit of 2026 — roughly $290M of Kelp DAO’s rsETH — traced back not to a smart-contract bug but to a compromised LayerZero security setup, including an RPC node feeding false data to a one-of-one DVN verifier. The attacker forged a cross-chain message to mint unbacked rsETH on Ethereum mainnet, deposited perhaps ~100,000 rsETH on Aave at ~93% LTV to borrow ~$200M of real ETH, and laundered it through mixers; a second ~$95M attempt was stopped, while Arbitrum’s Security Council emergency-froze ~$70M (~30,000 ETH).
• The market verdict was brutal and, per Santi, honest: “markets always speak truth.” Aave’s TVL fell 33% in four days to just shy of $15B — down 60% from a $43B peak — while DeFi TVL sits at $85B, its lowest in ~12 months and roughly 50% below its October 2025 peak. Rob calls Aave an innocent bystander with frozen pools; Santi says lenders, Aave, or the broader ecosystem will ultimately bear the bad debt. Yanowitz thinks Aave, rather than Kelp DAO, will drive the settlement, and says “it’s not clear to me that there is a future for KelpDAO.”
• The crew reframed North Korea’s Lazarus Group as the industry’s new organizing enemy after banks became customers and the SEC became a friend. Santi’s math: Lazarus pulled ~$600M this month alone and $600M–$1B a year in hacks — “anywhere between 3 to 4% of North Korea’s GDP” — making it “the most important unit of North Korea as a country.” Yanowitz: “an industry without an enemy” now has one, and fighting it may bond the ecosystem together.
• The episode’s stated theme: “do we get more pragmatic or do we not?” Yanowitz argues the ideological “fully permissionless” position (the Gabriel Shapiro camp) forfeits the big prize — all capital markets on-chain, where “a retail investor in Kansas has the same access to deals, data, and disclosures that a hedge fund in Greenwich has” — leaving only “stablecoins, some institutional adoption and perps.” A concrete fix, via Ethena’s Guy: rate limits capping damage at “$10 million per chain per hour” — “a slightly annoying inconvenience for users 99% of the time, but a worthwhile tradeoff to avoid going to zero.”
• The lending-market reshuffle is tradeable: Aave token down ~14–15%, Morpho up 10%, Spark up 140% — and Yanowitz’s structural call is that “the whole world’s going isolated.” Every fintech and institution wants isolated markets (see Coinbase’s Bitcoin-Morpho market scaling to $1B fast); pooled markets made sense for an earlier DeFi era, and Aave V4’s hub-and-spoke design suggests the shift is underway. Dragonfly remains long DeFi — Lighter publicly, plus stealth positions.
• On the Kalshi–Polymarket perps race, Rob’s core claim is that Twitter’s zero-sum framing is wrong. Kalshi’s CFTC-regulated route risks repeating Coinbase’s “perps” flop (5-year futures, throttled risk, “basically zero uptake”), pending Chair Selig’s teased perps guidance; Polymarket’s on-chain product competes more directly with Hyperliquid, but overlap between perps and prediction-market traders is only ~12–15%. “Prediction markets are going to get 10x bigger from here.”
• Rob accused Kalshi’s comms operation of crossing from opposition research into fabrication, with reporters printing unverified claims. A Bloomberg piece claimed de minimis Polymarket US volume when the app did “over a billion dollars in the last 30 days, far more than any other competitor”; Santi said Polymarket still has no internal head of comms. Yanowitz’s discipline: “never interrupt your enemy when he’s making a mistake,” while Santi says a founder should stay focused on the product.
• Clarity Act odds are decaying and USDai was the week’s standout launch. Yanowitz says his historical read from Washington contacts has been ~33%, though he still argued for better than 50% during the discussion; Polymarket was at 43% after a 37% low and a 65% peak a month earlier. The hosts heard that failure to pass before Memorial Day could be fatal. Meanwhile, USDai’s CHIP token traded above a $1B valuation on GPU-collateralized lending: $61M active loans, $59M signed, a $300M+ pipeline, and ~$350M capacity.

Deep dive

1. Anatomy of the $290M Kelp DAO exploit: a people hack, not a code hack

• Yanowitz’s walkthrough: on April 18th, an attacker funded via a Tornado Cash-funded wallet exploited Kelp DAO’s LayerZero-powered cross-chain rsETH bridge. A compromised RPC node fed false data to a single verifier — a one-of-one DVN setup — enabling a forged message that minted unbacked rsETH on Ethereum mainnet. The attacker deposited perhaps ~100,000 rsETH as collateral on Aave to borrow ~$200M of real ETH, which was sold and laundered through mixers.
• Rob’s key correction to the discourse: “this is not a smart contract hack” — probably a person inside LayerZero’s security setup was “hacked or phished or in some way compromised,” followed by a chain of decisions that “in retrospect look very poorly thought out”: the 1-of-1 DVN, allowing rsETH as collateral at all, and permitting LTV to reach 93% on Aave.
• The containment: a second ~$95M attempt by the same hacker was stopped, and Arbitrum’s Security Council emergency-froze about $70M (~30,000 ETH), reigniting the Tae-versus-Gabriel-Shapiro fight over whether a freezable chain is “really decentralized.”

2. Who eats the loss — a prisoner’s dilemma among three protocols

• Rob’s map of the misaligned incentives: all parties “lawyered up right away” and are pointing fingers, but this is the first DeFi hack of this size with multiple large brand-name victims — unlike Wormhole or Drift, where the resolution path was clearer. Rob calls Aave “sort of an innocent bystander”: it had nothing to do with the hack, but its pools remain frozen and users cannot freely withdraw.
• Santi says the bad debt will ultimately be borne by lenders, Aave itself, or people elsewhere in the ecosystem. Yanowitz thinks the decision about whether losses are socialized will effectively be driven by Aave rather than KelpDAO, because Aave has the most to lose and the most incentive to reach a resolution. Yanowitz also says “it’s not clear to me that there is a future for KelpDAO” following this. rsETH is about 18% below ETH because there are now more rsETH claims than ETH backing them.
• Yanowitz’s proposed settlement, noting that LayerZero is “the most capitalized entity here that could socialize the loss”: LayerZero could extend a loan to Kelp repaid from fees over time, preserving Kelp’s survival and LayerZero’s institutional credibility, while Aave commits to tighter caps and higher security spending. Yanowitz says he heard privately that roughly 50% of LayerZero app DVNs run one-of-one setups and that LayerZero is now telling teams to change them.
• Santi notes that Brian Pellegrino has stayed quiet — “I’m sure he’s being advised to be quiet” — and suggests bringing him on later.

3. Lazarus is now the industry’s enemy — and it’s 3–4% of North Korea’s GDP

• Santi’s staggering framing: Lazarus pulled roughly $600M this month alone and $600M–$1B per year in hacks — “anywhere between 3 to 4% of North Korea’s GDP… for all intents and purposes, Lazarus is the most important unit of North Korea as a country.”
• Yanowitz’s tweet-turned-thesis: “everyone needs an enemy,” and crypto has been “an industry without an enemy” — banks became customers, while the SEC and Gary Gensler became friendly. Echoing Hasib’s posts, he says that if the industry cannot protect capital from Lazarus, it will not grow; a common enemy “tends to bond people together,” so the turmoil may ultimately be good for coordination.
• Santi’s open worry, flagged without a resolution: Lazarus will not slow down, and that is “at odds with the willingness of founders to experiment” — “how much is permissionlessness going to be constrained? What is the healthy middle ground? These are questions that I don’t know the answer to.” His hedge: “in many ways, I’m glad that we don’t have over 100 billion of TVL. It’s a huge honeypot.”

4. Pragmatism over ideology: rate limits, time locks, and the SWIFT standard

• Yanowitz’s stakes-setting: take the ideological blue pill and crypto gets exactly three things — “stablecoins, some institutional adoption and perps” — a “nice little niche technology.” The alternative is every asset with a liquid market, transparent price, 24/7 order book, and Kansas-retail parity with Greenwich hedge funds: “that won’t happen if we don’t fix some of this stuff.” He says Ethereum and Bitcoin can remain fully permissionless, while protocols operating like businesses have a duty to stop crime where they can. Stablecoins’ ability to be frozen is, in his view, acceptable because they meet a large unmet demand for basic financial products.
• Yanowitz’s version of the pragmatic position is that the fully permissionless purity argument “is just not going to be the reality” if the industry wants regulation, scale, and better financial rails. He cares more about building those rails than about whether a 12-person Security Council can stop North Korea from taking funds.
• The best specimen of the fix, Guy of Ethena’s tweet as relayed by Yanowitz: asset issuers should set mint/redemption rate limits atop LayerZero OFTs so a compromised DVN can “at least contain the damage to 10 million per chain per hour before shutting down transfers” — “a slightly annoying inconvenience 99% of the time, but a worthwhile tradeoff to avoid going to zero,” like double-confirming a bank wire.
• Santi’s analogies: the SWIFT messaging protocol does not rely on one verifier — a Switzerland-to-JPMorgan wire passes multiple approval checkpoints — and no bank lends against a piece of paper claiming fictitious real estate that nobody reviews. Practically, “someone at Aave should have called Kelp… ‘are you doing all right over there? I just saw 292 million come in trying to borrow immediately.’” An abnormal spike is either a hack “or Justin Sun moving money around to yield-farm” — and Yanowitz’s line: “we can inconvenience Justin Sun to stop hacks.”
• Yanowitz’s consolidation thesis: DeFi is full of “technologists who think they’re building software and don’t realize they are building finance.” Founders like Guy and the Agora team put in rate limits and extra security — including a 3-of-3 DVN rather than a 2-of-3 or 3-of-5 setup — “with nobody telling them they should.” The future should consolidate around founders who understand that trust and security are the business.

5. The lender reshuffle: isolated markets eat the pooled-market era

• The scoreboard Yanowitz put up: Aave token down ~14–15%, Morpho up 10%, Spark up 140% — is this a durable reshuffling and a buying opportunity? Yanowitz’s structural answer: “the whole world’s going isolated.” Every institution and every fintech exploring on-chain lending wants isolated markets — “not a single one of them is going to do anything but an isolated market” — citing Coinbase’s Bitcoin-Morpho market growing to $1B very quickly through the CeFi app.
• Santi’s taxonomy and the tradeoff: siloed markets include Morpho, Euler, and Aave V3; Aave V4 uses a hub-and-spoke model, “maybe that is the writing on the wall.” Traditional firms will treat everything crypto-native as “a shitcoin” and haircut it heavily in pooled markets. The cost is capital efficiency, but breaking from Aave’s roughly $14B through the $100B mark likely runs through isolated RWA markets — Figure’s HELOC book being effectively one already, though the quality of the underlying mortgages varies.
• Positioning disclosure worth logging: Dragonfly is “big into Lighter” plus unnamed stealth DeFi bets since Ethena in summer 2023. Rob: “I’ve never lost faith that there will be finance on chain” — with a jab at Santi: “I’m more of a true believer than you, maybe.”

6. Kalshi vs Polymarket perps: two different regulatory bets, probably not zero-sum

• Rob’s structural breakdown: Kalshi is launching perps through its regulated US DCO/DCM while “perps in the US are illegal” — the cautionary precedent being Coinbase’s synthetic perps, effectively 5-year futures with throttled overnight and weekend risk, that got “basically zero uptake.” The swing variable is CFTC Chair Mike Selig’s teased perps guidance; meanwhile IBIT options “blow all other options out of the water” and probably capture almost all institutional derivatives volume, because perps are mainly retail products rather than institutional ones.
• His market-structure call: Kalshi-vs-Hyperliquid competition “is probably wrong — completely different markets and user bases.” Kalshi targets Robinhood-style retail leverage seekers; Rob estimates that around 70% of Robinhood’s options volume is zero-day options. Polymarket’s on-chain product is the closer Hyperliquid comparable. Overlap between perps and prediction-market traders is only ~12–15% according to analysis by some Hyperliquid participants — enough to justify the product, not enough to imply full cannibalization.
• Rob is also skeptical of HIP-4 cross-margin talk: HIP-4 markets are Hyperliquid’s binary-option markets, but “there’s actually nothing in the documentation about” cross-margin with the perps side, so he does not expect it soon.
• The conviction close: “this is not zero sum, despite the fact that Twitter wants you to think it is. Prediction markets are going to get 10x bigger from here” — perps grow too, though Rob does not know whether they will grow 10x.

7. The opposition-research war — and where Rob says Kalshi crossed the line

• The trigger: a Bloomberg piece on Shayne Coplan’s “unconventional management style” — late to meetings, “attended at least one of them barefoot,” texting mid-conversation — which Yanowitz, as a former news-business owner, reads as competitor-fed: “take all of this stuff with a big [grain of salt].” He notes Uber’s Operation SLOG as precedent for how aggressive corporate rivalries can become.
• Rob’s distinction, delivered with heat: opposition research is fine; “an extreme willingness to lie” is not. A Barron’s volume story was “factually untrue and the reporter didn’t even check it,” and Bloomberg’s de minimis-US-volume claim collapses against the data — “over a billion dollars in the last 30 days on the US app, far more than any other competitor.”
• Santi says one reporter admitted he was the first person contacted who was not aligned with Kalshi before the reporter planned to publish another negative Polymarket story. His diagnosis: Polymarket never staffed up its comms operation and still has no internal head of comms, while “on the Kalshi side, this was part of the strategy from the beginning.” His advice to Polymarket leadership: “we probably just need to take this more seriously.”
• Santi’s investor-seat view — ParaFi seeded Polymarket and is among, if not the second- or third-largest holder, so he is biased — is that founders should not over-index on competitors: “just own your product and you’ll crush.” Yanowitz adds that a truly winning company should be “in your lane. Unbothered, moisturized,” and later quotes Napoleon: “never interrupt your enemy when he’s making a mistake.”
• Yanowitz’s balancing note: rivalry comms can be savvy — the Pepsi Challenge pressured Coke into the disastrous New Coke — but it has to “lead with integrity.” He wants a Kalshi-camp guest on to respond.

8. Clarity Act: theta decay, a Memorial Day deadline, and a live host disagreement

• The odds board: Galaxy’s Alex Thorn reportedly said 50/50; Polymarket printed 43% during recording after touching 37%, down from a 65% peak a little over a month earlier. Yanowitz argued for better than 50% at that moment, though he says his historical read from Washington lobbyists and other observers has been roughly one-third, consistent with Rebecca Rettig’s earlier read.
• Yanowitz says the longer the process drags, the lower the chance becomes. Rob says every person he spoke with in Washington told him that if the bill was not done before Memorial Day, it was dead: “it’s an option — there’s a theta decay happening.” Yanowitz expects the market to settle in the high 30s over the following month and says, “I wanted it to pass… I try to be a realist and I get less confident every day.”

9. USDai’s CHIP token: a rare up-only TGE built on GPU-collateralized credit

• Yanowitz’s setup: USDai, its sUSDai yield product, and the CHIP token launched with CHIP above a $1B valuation — “it’s kind of hit every narrative” — AI exposure, real customers, and real revenue, when most TGEs are “down only.” Rob, an investor in USDai, calls it “a proxy for compute demand”: collateralized loans against GPUs themselves, from the David Joyce team whose prior NFT protocol was specifically focused on credit around NFTs.
• Santi’s pushback — worth keeping: who are they lending to that lacks access to traditional credit, given this is a hot category for real-world lenders? Rob’s rebuttal: lending against compute for foundational labs is crowded, but earlier-stage AI and compute startups remain underserved, and this team underwrites that segment well at an attractive cost of capital.
• Current book: $61M in active loans, $59M signed but undrawn, a $300M+ pipeline, and approximately $350M of capacity. Rob says the team is being thoughtful about underwriting and is extending less than its available capacity.