We asked over 150 software engineers about vibe-coding. Here's what they said.

• 46.9% of 167 engineers feel they're keeping up with AI coding tools
• 17.5% opt out entirely, citing tool limitations and learning curves
• Productivity gains disputed: some report 2-3x improvements, others cite review overhead

The great coding identity crisis is here, and it's messier than a junior developer's first pull request. When nearly half of software engineers say they're just "keeping up" with AI tools that can supposedly code at midlevel engineer status, we're witnessing a profession grapple with its own potential obsolescence in real time. The fact that 17.5% are completely opting out suggests this isn't just about learning new tools—it's about fundamental questions of what programming even means anymore.

The productivity debate reveals the classic tech industry paradox: tools designed to save time often create new forms of work. While some engineers report doubling their output, others find themselves spending more time reviewing AI-generated code than they would have spent writing it themselves. This mirrors every productivity revolution in tech history, where the promised efficiency gains get eaten up by new complexities and quality control measures. The real question isn't whether AI makes coding faster, but whether it makes coding better.

Perhaps most telling is the generational divide emerging in responses. Younger engineers who grew up with AI assistance worry they never truly learned to code, while veterans fear their decades of experience might become irrelevant overnight. This tension between foundational skills and AI augmentation will likely define the next decade of software development, as the industry figures out whether we're training the next generation of engineers or the last generation of human coders.