AI-powered coding tools are fundamentally transforming the software engineering profession, with a new survey revealing how developers are adapting to what Andrej Karpathy dubbed “vibe-coding” - the creation of code using artificial intelligence. The term has gained such widespread adoption that Collins Dictionary named it Word of the Year for 2025.
In a survey of 167 software engineers, nearly half (46.9% or 75 engineers) reported they were “keeping up” with AI coding tools, while 30 felt ahead of the curve and 27 felt behind. Notably, 28 engineers (17.5%) opted out entirely, citing that the tools weren’t advanced enough or took too long to learn. None of these holdouts agreed to speak on the record.
The survey reveals a profession in flux, with engineers using tools like Cursor, Google’s Antigravity, and Claude Opus 4.5 to varying degrees. Ryan Shah, a 23-year-old AI consultant, uses these tools extensively and believes the future lies in learning to “read” code rather than write it from scratch. He considers AI models like Claude Opus 4.5 to be at “midlevel engineer status.”
The productivity debate remains contentious. Ed Gaile, a 55-year-old principal solutions architect, claims AI tools have doubled or tripled his productivity by reducing context switching. However, a July METR study found that AI-assisted developers were actually less productive overall - they spent 10% less time coding but 20% more time reviewing AI outputs, prompting AI, waiting on AI, or being idle.
Job security concerns vary by experience level. Ryan Clinton, a 46-year-old developer, believes senior engineers focusing on architecture and design are safe, while junior coding positions may be more vulnerable. Barry Fruitman, 56, doesn’t expect major job market impacts for 5-10 years - conveniently around his retirement. Meanwhile, Javanie Campbell warns that engineers who treat AI as “the God or the expert” will be replaced.
The definition of “vibe-coding” itself remains contested. While it’s come to encompass most AI-assisted coding, Karpathy originally defined it as when developers “fully give in to the vibes, embrace exponentials, and forget that the code even exists.” Data analyst Lara Fraser argues true vibe-coders are those who can create apps but can’t maintain or fix them when something breaks.
Key Quotes
Did I really need to learn how to write code?
Ryan Shah, a 23-year-old AI consultant who recently graduated with a computer information technology degree, questions the value of traditional coding education in an AI-powered world. Despite his doubts, he believes learning to ‘read’ code keeps him employable and prevents him from being ’the first one laid off.’
For people who turn to the LLM as the God or the expert, they will be replaced
Javanie Campbell, 35-year-old CEO of DevDaysAtWork based in Jamaica, warns against over-reliance on AI coding tools. His statement highlights the ongoing debate about whether AI tools are job replacements or merely assistive technologies that still require human expertise and oversight.
I wish I had this 15 years ago
Ed Gaile, a 55-year-old principal solutions architect, expresses enthusiasm about AI coding tools after experiencing doubled or tripled productivity gains. His comment reflects the perspective of experienced engineers who see AI as an enhancement rather than a threat to their careers.
Inevitably, something’s going to break. Can you fix it? If you can’t, you’re a vibe-coder
Lara Fraser, a data analyst and epidemiologist from Florida, provides a practical definition that distinguishes between AI-assisted professionals and those overly dependent on AI. Her perspective emphasizes that true engineering competence requires the ability to maintain and troubleshoot code, not just generate it.
Our Take
This survey captures a pivotal moment in software engineering’s evolution. The fact that nearly half of engineers report “keeping up” suggests the industry is adapting faster than many predicted, but the 17.5% opting out entirely reveals significant resistance that could create a skills gap.
What’s most striking is the generational divide in perspectives: younger engineers like Shah question whether traditional coding skills were necessary, while veterans like Gaile wish they’d had these tools decades ago. This suggests AI coding tools may be more disruptive to career entry points than to established professionals.
The METR study’s productivity findings deserve serious attention - they suggest we’re still in the early stages of understanding how to effectively integrate AI into development workflows. The 20% time spent reviewing AI outputs indicates these tools require new skills around prompt engineering, code review, and quality assurance rather than replacing traditional engineering entirely. The real winners will be those who master both traditional fundamentals and AI augmentation.
Why This Matters
This survey provides crucial insight into how AI is reshaping one of technology’s core professions at an unprecedented pace. Software engineering has long been considered a stable, high-paying career path, but AI coding tools are forcing a fundamental reassessment of what skills matter and which jobs are secure.
The findings reveal a profession divided: nearly half are adapting successfully, but significant minorities feel either left behind or are rejecting the tools entirely. This fragmentation suggests the industry is in a transitional period where competitive advantage may increasingly depend on AI tool proficiency rather than traditional coding skills.
The productivity paradox highlighted by the METR study is particularly significant - it challenges the assumption that AI tools automatically increase efficiency and suggests implementation requires careful consideration of workflow changes. For businesses investing heavily in AI coding tools, this raises important questions about ROI and training requirements.
Most critically, this shift affects not just current engineers but the entire pipeline of computer science education. If reading and reviewing code becomes more valuable than writing it, curriculum and training programs must evolve rapidly to prepare the next generation for an AI-augmented development environment.
Related Stories
- Andrej Karpathy Reflects on ‘Vibe Coding’ Revolution in AI
- The Future of Work in an AI World
- Microsoft AI CEO’s Career Advice for Young People in the AI Era
- Tailwind CEO Blames AI for 75% Engineering Layoffs, 80% Revenue Drop
- AI’s Role in Tech Hiring Freeze: White-Collar Job Market Slump
Source: https://www.businessinsider.com/software-engineers-on-vibe-coding-ai-tools-2026-1