Manuel Aráoz, the CEO of OpenZeppelin, has raised alarms regarding the safety of decentralized finance (DeFi), stating that AI coding agents have become 'superhuman' in their ability to find vulnerabilities in smart contracts. This warning comes amid a significant decline in DeFi's total value locked, which has dropped by over $20 billion this year. The sector has also faced substantial security breaches, with more than $1.1 billion lost to hacks in the past year, including notable incidents at Kelp DAO and Step Finance. Aráoz emphasized the asymmetric nature of smart contract security, where defenders must fix every bug while attackers need only one exploit to succeed. The emergence of advanced AI models like Claude Mythos, which can autonomously identify and weaponize software flaws, poses new challenges for DeFi's security, as the rapid pace of AI could outstrip human efforts to patch vulnerabilities. This situation raises pressing questions about the future of DeFi security, particularly given its transparent nature, which may now be exploited by AI systems faster than defenses can respond.