Andrej Karpathy's term "vibe coding" describes a carefree approach to programming where developers prompt AI, accept its outputs without scrutiny, and iterate based on error messages. While useful for rapid prototyping, personal projects, and creative brainstorming, this method lacks rigor. Engineers often find themselves in trouble when they try to modify or scale what they’ve created, revealing that vibe coding leads to poor understanding of the codebase. The article argues that this casual approach can lead to significant issues in real-world applications. In contrast, "agentic engineering" represents a more disciplined use of AI in software development. Experienced engineers are achieving productivity boosts—sometimes doubling or tripling output—while maintaining high code quality. Unlike vibe coding, agentic engineering involves planning, writing specifications, and thorough code reviews. Developers treat AI like a junior colleague who requires oversight, leveraging AI for tasks while focusing on architecture and testing. The key difference lies in the rigorous testing and documentation that accompany this method, ensuring reliability and long-term maintainability. However, agentic engineering has a downside: it tends to benefit seasoned engineers more than juniors. Those with strong fundamentals can leverage AI effectively, while newcomers risk skill atrophy by relying too heavily on AI without a solid grasp of coding principles. The article emphasizes the need for clearer terminology to distinguish between vibe coding and agentic engineering, as well as better evaluation frameworks to assess how AI-assisted workflows affect software reliability. It underscores the importance of investing in foundational skills as AI takes on more implementation tasks, raising the bar for what constitutes good software engineering.