Vibe coding expands as software makers weigh AI’s risks
The use of large language models in software development is growing, but concerns over security, maintenance and junior developer opportunities are intensifying.
Vibe coding, the use of large language models to generate or assist with software code, is gaining popularity while prompting debate over how much control developers should retain over the code they produce.
The term was coined by AI researcher Andrej Karpathy in February 2025. Engadget describes the practice as a spectrum, ranging from software generated almost entirely by an AI model to more limited uses such as debugging, code clean-up and working through development problems.
Supporters argue the tools broaden access to app development. Engadget cited the example of a former veterinary technician who used vibe coding to create an app for tracking her senior cat’s insulin injections. Critics say inexperienced users may struggle to identify security flaws or maintain a codebase when an AI system cannot resolve an error.
Preliminary research cited by Engadget points to potential security risks. Researchers at Georgia Tech’s School of Cybersecurity and Privacy reportedly traced 74 vulnerabilities, including 14 classified as critical, to AI-generated code in a review of 43,000 security advisories covering three months. The researchers estimated the actual number could be five to 10 times higher because attribution was often unavailable.
Professional use also appears to be increasing, although the available figures do not show that all AI-assisted coding is vibe coding. Engadget reported that a survey of 1,100 professional programmers who had tried AI tools found 72 per cent used coding tools daily and estimated that 42 per cent of their codebase was AI-generated or AI-assisted. The survey’s methodology and publication details were not provided.
Stack Overflow’s 2025 developer survey found 47.1 per cent of respondents used AI tools daily, while 72 per cent said vibe coding was not part of their workflow. Engadget said many professionals were using AI for tasks including autocomplete, code review and navigating technical roadblocks. The wider adoption of those tools has also raised concerns that work once delegated to junior developers could increasingly be automated, potentially narrowing an established entry path into the profession.
