“Dylan Field asserts that allowing AI agents to build software systems without human engineering oversight currently leads to security vulnerabilities and systems that break at scale.”