Vibe coding, a relatively new approach in software development, involves using large language models (LLMs)—advanced AI systems capable of generating human-like text—to write some or all of the code for a project. The term was popularized by Andrej Karpathy, a well-known AI researcher and former leader of Tesla’s Autopilot Vision team. He described vibe coding as a method where developers "fully give in to the vibes, embrace exponentials, and forget that the code even exists," highlighting the almost intuitive, less structured nature of the process. As LLMs have become more sophisticated, the practice has gained traction, especially among those looking to streamline or simplify software development. Supporters of vibe coding argue that it makes software development more accessible, allowing people without formal training in coding to build applications. For example, a neighbor of one reporter used vibe coding to create an app that helps track insulin shots for her senior cat. This approach can reduce the learning curve and potentially lower the cost of development. However, concerns have emerged about the security risks associated with AI-generated code. A study by the School of Cybersecurity and Privacy at Georgia Tech identified 74 vulnerabilities in 43,000 security advisories that could be linked to AI-generated code, with 14 of those classified as critical. Experts believe the actual number of such vulnerabilities could be much higher, due to limited transparency about the use of AI in coding. Despite the rise in AI-assisted tools, many developers remain cautious about fully relying on AI for coding. While a survey of 1,100 professional programmers found that 72 percent use AI coding tools daily and 42 percent of their code is either generated or assisted by AI, others are more hesitant. A Stack Overflow 2025 survey revealed that 72 percent of respondents did not use vibe coding as part of their development workflow, with 5 percent stating it was "emphatically" not part of their process. These figures may have shifted since mid-2025, as the field continues to evolve. A key issue in the debate is the difference between code that is entirely generated by an LLM and code where AI is used to assist with tasks like debugging or cleanup. While AI can handle some tasks previously done by junior developers, this shift raises questions about the future of coding careers. As AI becomes more integrated into the development process, the role of human coders may evolve, prompting discussions about the long-term impact on the profession and the need for new skills in the industry.