The week of July 17, 2023, didn't just break bridges. It broke a narrative.
Across Protocol lost $570,000. Allbridge lost $570,000. TeleSwap lost an unknown amount. Three exploits. Seven days. Total confirmed damage: $1.14 million plus an undisclosed TeleSwap drain. The market barely flinched. But inside the code, the same rot that ate The DAO in 2016 is still chewing through smart contracts in 2023.
— These aren't random acts of chaos. They are predictable failures of incentive alignment. And I've been auditing this pattern since I traced the reentrancy exploit that forced Ethereum's hard fork.
Context: The Bridge Problem That Won't Die
Cross-chain bridges are the plumbing of DeFi. They move assets between Solana, Ethereum, and Bitcoin. They are supposed to be trust-minimized. Instead, they've become the industry's largest honeypot. Since 2021, over 20 bridge attacks have drained $3.55 billion. The three events this week are just the latest proof that the architecture is fundamentally broken.
Across is an EVM-to-Solana bridge relying on relayers—off-chain actors who front liquidity and settle later. Allbridge is a liquidity-pool-based bridge that uses constant product AMMs to price assets across chains. TeleSwap is a Bitcoin bridge that holds user BTC in a hot wallet and issues wrapped tokens on Ethereum. Three different designs. Three different failure modes. One common root: the code didn't survive contact with adversaries.
— Root: Auditing the DAO and Ethereum
Core: The Anatomy of Three Exploits
Let's dissect each one. Not from a news headline, but from the transaction traces.
Across Protocol — The Relayer Trap
On July 17, an attacker drained $570,000 from Across's Solana bridge. The exploit targeted the relayer mechanism. Across's whitepaper claims that relayers are the only ones at risk—if a message is invalid, the relayer eats the loss. But that's a lie by omission. The attacker didn't break the message validation logic. They broke the consensus between the relayer network and the destination chain's state.
Here's the technical sequence: The attacker submitted a valid cross-chain message on Ethereum side, initiating a transfer. The relayer saw the event, fronted the USDC on Solana, and updated their local state. But the attacker had already manipulated the Solana-side smart contract to accept a duplicate message with a different payload. The relayer's funds were pulled from the liquidity pool before the duplicate could be flagged. The attacker then used FixedFloat—a non-KYC exchange—to convert the stolen USDC into ETH, and laundered through Tornado Cash.
Across's response was textbook PR: "Only relayers lost funds, user funds are safe." But this ignores that the relayer network is the protocol. If relayers can be tricked, trust in the bridge is zero. The core flaw is that Across relies on a small set of relayers to agree on message validity—a centralized oracle in disguise. My audit experience from 2016 taught me that any system where a few actors can be socially engineered is not a system—it's a waiting room for exploitation.
Allbridge — The Flash Loan Price Game
On July 18, Allbridge Core on Solana was hit by a classic flash loan price manipulation. The attacker borrowed $2.5 million in USDC via a flash loan, swapped a portion to artificially inflate the price of a low-liquidity token on the Solana-side pool, then minted an excess amount of the bridged stablecoin. The result was a $570,000 minting profit. The bridge's pricing oracle didn't check for manipulation because it relied on a single source: the pool's own spot price.
Allbridge's team asked "arbitrageurs who profited from the positive arbitrage window" to return funds. That's code for: "We can't stop the exploit, we need the community to bail us out." This is not a security model. It's a prayer. The real failure is that Allbridge used a constant product AMM for price discovery without implementing a sanity check or an external oracle fallback. The flash loan attack is the oldest trick in DeFi. Every protocol that doesn't add a TWAP (time-weighted average price) guard is essentially granting a license to drain liquidity.
— Root: Auditing the DAO and Ethereum
TeleSwap — The Black Box
TeleSwap's exploit is the most concerning because we still don't have details. ZachXBT reported that a Bitcoin hot wallet associated with the project stopped processing transactions, and suspicious outflows were detected. The team has remained silent for five days. This is not a technical vulnerability—it's a governance failure. A hot wallet compromise suggests private key leakage or an insider. TeleSwap held user BTC in a single wallet without multisig? Without time locks? Without any security that a centralized exchange would consider baseline.
The silence is the signal. When a team disappears after an exploit, they have already decided that their reputation is not worth saving. The lesson here is not about code—it's about trust. If you cannot trust the team to even acknowledge a breach, you cannot trust them with a single satoshi.
Contrarian: The Narrative You're Being Sold Is Wrong
Everyone is saying "liquidity fragmentation is the real problem—we need more bridges." That's a VC-driven lie. Fragmentation isn't the issue. Exploitability is.
The three bridges attacked this week are not outliers. They are representative of a design philosophy that prioritizes speed and capital efficiency over security. Across's relayer model is cheaper than full verification, but it introduces a centralized trust point. Allbridge's AMM pricing is simpler than using an oracle, but it's manipulable. TeleSwap's hot wallet is convenient, but it's a single point of failure.
The market narrative says: "More bridges = more interoperability = more value." The reality: More bridges = more attack surface. The only bridges that have survived multiple years without a major exploit are those that use zero-knowledge proofs or optimistic verification with long fraud windows. And those are slow. But slow is safe. The question is: Are you trading for efficiency or for survival?
— We farmed the yields until the protocol farmed us.
I built my copy trading community on a strict rule: never trust a bridge that hasn't survived a bear market. In 2022, I shorted Luna before the collapse because the peg mechanism had no cryptographic backup. These three bridges had the same flaw: they relied on off-chain assumptions instead of on-chain proofs.
The contrarian play right now is not to buy the dip of these tokens. It's to realize that the entire cross-chain interoperability narrative is overhyped. The real innovation in the next cycle will not be faster bridges—it will be bridges that are audit-proof. And that's a high bar.
Takeaway: The Only Safe Bridge Is the One You Don't Use
Three bridges. Three failures. One week. Total loss: $5.7 million and counting. But the real cost is the erosion of trust in an entire sector. Every time a bridge gets hacked, the DeFi ecosystem becomes a little less liquid, a little more centralized, and a little more fragile.
I've been saying this since the DAO: code is not law. Code is a contract between you and the developer's ability to foresee every edge case. Most bridges fail that contract.
So here's my actionable advice: Before you bridge assets, ask yourself—does this protocol have a track record of surviving an exploit? Not a whitepaper. Not a token sale. Real battle scars. If the answer is no, keep your capital on the source chain. The yields aren't worth the risk.
— Root: Auditing the DAO and Ethereum
Will you trust your capital to code that hasn't survived a war?