ETHPrague 2026

TheDAO needs no introduction, its 2016 hack split Ethereum in two and defined an era of smart contract security. But the contract itself never went away. It's still there, on mainnet, exactly where it was deployed. Old contracts never die.

Ten years later, 46 ETH remained locked inside the original DAO contract. Recovering them was not as simple as replaying the 2016 attack. DAO tokens were required to interact with the contract, and the internal accounting had been left in a broken state, uint underflows made direct extraction impossible. Before we could drain anything, we had to apply precise tricks to undo those underflows and stabilize the contract's state into something we could work with.

Once the state was clean, we re-exploited the same reentrancy vulnerability, but we optimized it far beyond what the original attacker did. Our approach chained multiple reentrant calls, moved tokens between two contracts mid-execution, re-entered again, moved again, repeating until the gas per tx said stop. It was, by a wide margin, the most gas-efficient TheDAO exploit ever written.

The operation itself was as much about opsec as it was about code. We couldn't let other actors notice and front-run us, so we first deployed what appeared to be a voting contract. Only when we confirmed we were alone did we upgrade it to the attacker contract. We targeted a window of cheap gas during the night. When the first run turned out too slow, we had to improvise and optimize the attack further on the spot. Eight hours later, before gas fees climbed back up with the morning, the recovery was complete.

This talk will walk through TheDAO's contract architecture, the exact mechanics of the reentrancy, the state manipulation required to make the recovery possible, the optimized exploit code, and the operational decisions we made executing it live. Alongside Jordi Baylina and Griff Green, two of the people who were there recovering TheDAO funds in 2016, we closed a ten-year loop in Ethereum's history.