A thought-provoking analysis from a Chinese tech blog examines why traditional software systems felt more secure than modern AI-driven platforms. The key insight: humans were always in the middle, providing oversight and judgment. As AI systems increasingly automate execution, the human role diminishes, raising new safety concerns. The author, writing under the name Havenlon, explores how execution control has evolved and what lessons we can carry forward. For engineering leaders building AI products, this perspective highlights the importance of maintaining human oversight even as automation advances. The piece does not prescribe solutions but frames a critical conversation about trust and control in AI systems.
This article argues that traditional systems appeared safer because humans remained in the control loop, whereas AI systems automate execution with less transparency.