Ethan Mollick introduces "vibecoding" for programming automation

A new blog post by Ethan Mollick discusses the rise of AI tools that can create software based on simple prompts. He highlights a recent conversation by Andrej Karpathy, who emphasized that using English may become the main way to program computers. Mollick experimented with an AI called Claude Code. He shared his experience creating a 3D game, where he could design buildings and navigate through them. Within four minutes, he obtained a working game after giving the AI a straightforward prompt. While he pointed out that he still needed some technical know-how to improve the game, he praised the convenience of using AI for coding tasks. He coined the term "vibecoding" to describe this process, stating it is not about removing the need for expertise, but rather about shifting how expertise is applied. Familiarity with systems is essential for successful collaboration with AI. Mollick also reviewed a Chinese AI agent named Manus, which he instructed to create a course on elevator pitching. Although the output was impressive, he noted it lacked interactive components. This raises questions about human roles in AI-driven projects and how to adapt in an evolving workspace. He believes the relationship between human skills and AI capabilities is changing and not fully established. Current AI tools are still not entirely reliable but can significantly enhance productivity. As technology advances, society must reconsider traditional work roles and how to effectively integrate AI into workflows.