As artificial intelligence continues to reshape the software industry, a growing number of developers are embracing a trend known as “vibe coding” — a style of programming where engineers describe what they want and let AI tools generate most, if not all, of the code. While the approach promises speed and convenience, Cursor co-founder and CEO Michael Truell believes it comes with serious long-term risks.

Speaking at the Fortune Brainstorm AI conference earlier this month, Truell cautioned that excessive dependence on AI-generated code can quietly erode the structural integrity of software systems. According to him, as projects scale and become more complex, these hidden weaknesses may eventually cause everything to “crumble”.

Over the past decade, Truell noted, the nature of programming has shifted dramatically. Developers once spent hours inside text editors, carefully writing, reviewing, and refining code line by line. Today, generative AI tools allow engineers to step back and delegate entire tasks to machines — from writing functions to refactoring large sections of applications. This change has undoubtedly accelerated development and lowered the barrier to entry, especially for newcomers.

However, Truell warned that stepping back too far can be dangerous. “If you were vibe coding, you would close your eyes and just ask for a house to be built. You wouldn’t examine the foundations, you wouldn’t look under the floorboards, and you wouldn’t look at the wiring,” he said.

Using a construction analogy, he explained that vibe coding is similar to building a house without ever inspecting what holds it together. “If you close your eyes and you don’t look at the code and you have AIs build things with shaky foundations as you add another floor, and another floor, and another floor, and another floor, things start to kind of crumble,” Truell explained.

According to him, this approach may work for small experiments, personal projects, or rapid prototypes, where speed matters more than long-term stability. But applying the same mindset to production-grade or enterprise software can be risky. When developers add new features on top of code they do not fully understand, problems compound over time. Bugs become harder to trace, systems grow fragile, and maintenance turns into a challenge.

What makes Truell’s warning particularly notable is Cursor’s own success in the AI coding space. Founded in 2022, Cursor integrates AI directly into the developer workflow, helping engineers write code faster, debug issues, and automate repetitive tasks. The platform has been widely adopted by professional teams working on complex, high-stakes software, including enterprise and infrastructure systems.

Cursor’s rapid growth — now boasting over a million daily users and a valuation in the billion-dollar range — underscores the widespread appeal of AI-assisted development. Yet Truell’s message is clear: AI should augment human judgment, not replace it. Developers, he argues, must remain actively involved in reviewing and understanding their code to ensure systems are built on solid foundations.