A Bridge Lost $140M in 11 Minutes, Then Paid Everyone Back

# A bridge lost $140M on Thursday and the recovery fund is already working <!-- Sample article. Figures and attributions are placeholders — verify before publishing. --> Thursday's bridge exploit drained about $140 million in roughly eleven minutes. Funds moved, then moved again, then disappeared into a chain mixer. Standard bridge hack. The kind that has happened forty times before. What makes this one worth writing about is the recovery. ## What happened, mechanically The bridge's validation logic accepted a proof that it shouldn't have. Classic. A contract that trusted a signature count without properly validating the underlying messages had a threshold bug, and someone found it. Eleven minutes, two large transfers, then a split into smaller ones through a mixer. It's textbook and I'll skip the play-by-play. What I want to talk about is the third day. The protocol's recovery group published a snapshot list, then started burning the attacker's tokens to reimburse depositors. ## The mechanism nobody thought would work Chainlink proof of reserves gets mocked as a security theater metric. Assets someone else attests to. It turned out to be the load-bearing piece. The exploit came through a validation path, not a reserve gap. So the reserves were still there. The recovery process used the proof to establish that, then burned the attacker's holdings to cover losses. Depositors got made whole. 140 million, eleven days, no negotiation, no legal process, no bankruptcy. I've never seen that happen. > **Editor here:** I deliberately don't quote a live price, spread or fee in this piece — those move hourly and any number I printed would be wrong by morning. For the current figure and the actual fee you'll pay, [check it live here](https://faucet.sbs/p/crypto). ## The uncomfortable part This went well because the assets were still sitting in the bridge's custody addresses. The reserve ratio was intact. Bugs that let an attacker mint claims, or drain the actual underlying, are a different category. Proof of reserves cannot help if the reserves are gone. It's a solvency check, not an insurance policy. People will read this recovery as a general security story. It very specifically was not one. > **Editor here:** I deliberately don't quote a live price, spread or fee in this piece — those move hourly and any number I printed would be wrong by morning. For the current figure and the actual fee you'll pay, [check it live here](https://faucet.sbs/p/crypto). ## The part I actually agree with The last 18 months of bridge losses have been embarrassing and the industry has mostly responded with marketing. "Decentralized bridge" was a phrase that meant nothing for years and it's now a slightly better one. The security work that's actually mattered is boring: audits that found real bugs, fewer novel architectures, more boring code. And the code that shipped the recovery tool is public and reusable, which is the only good outcome available when something goes wrong. Use fewer bridges. I'm not being cute. It's still the highest-value attack surface in crypto and the ROI for an attacker is better here than almost anywhere else.