Some cool AI work coming out of @AnthropicAI: While it's no surprise that AI will have a large impact on software security (good and bad), this is another data point showing that blockchains are uniquely vulnerable: 1. Smart contracts are open-source (many are even static and can't be modified/upgraded on-chain). 2. Vulnerable assets are public (e.g., token balances are visible). This makes ROI quantifiable ahead of time. 3. Smart contract languages are still young/naive. Many (most?) are not designed to be security-first. 4. The barrier to entry for smart contract development is dropping. Vibe-coding will keep making it easier to deploy code that puts real user funds at risk. 5. Attackers have a higher change of avoiding detection and legal action when stealing crypto vs. fiat (e.g., due to the nature of decentralization). AI will keep getting better and cheaper at finding vulnerabilities (graph below). While Move on Aptos was already designed with safety and security in-mind, nothing is impenetrable. I'm thankful that the Move and security teams at @AptosLabs are always paying close attention to smart contract security. Blockchains have a lot of security work to do in the coming years! At least until the robots take over, in which case, we're all done. 😄