AI has radically changed what coding looks like. We asked software engineers how they felt about it.
Andrej Karpathy coined the term “vibe-coding,” or the creation of code using AI. The term has since gained traction among developers worldwide and was named Collins Dictionary’s Word of the Year for 2025.
Less than a year after his post, Karpathy wrote that he had “never felt this much behind as a programmer.”
We asked developers: When it comes to vibe-coding, do you feel ahead, behind, or like you’re keeping pace?
167 software engineers responded to our survey. The biggest cohort — 75 engineers, or 46.9% — said that they were “keeping up.” 30 engineers said they felt ahead of the curve, while 27 felt behind.
28 engineers (or 17.5% of respondents) said that they were opting out of using AI code editing tools entirely. These engineers wrote that the tools weren’t advanced enough, or that they took too long to learn how to use. None of the 28 agreed to speak on the record after Business Insider reached out.
While the survey isn’t scientific, the results offer insight into how software engineers are feeling about their rapidly changing industry.
In follow-up conversations, eight engineers told Business Insider how they feel about AI code editors. All found them helpful in some form, though their usages ranged from one-off tools to lifesavers.
Industry disruptor — or just another tool?
Ryan Shah sometimes wonders: “Did I really need to learn how to write code?”
The 23-year-old AI consultant from Atlanta recently graduated with a degree in computer information technology. Now he uses Cursor and Google’s Antigravity, paired with Claude Opus 4.5, which he said was at “midlevel engineer status.”
Shah said he doesn’t regret his software engineering courses, though. They taught him to “read” code, he said, a skill that, coupled with his vibe-coding proficiencies, keeps him from being “the first one laid off.”
Ryan Shah
Javanie Campbell swung the other way: He warned that over-reliance on vibe-coding tools will put your career in danger.
“For people who turn to the LLM as the God or the expert, they will be replaced,” said the 35-year-old CEO of DevDaysAtWork, who is based in Jamaica.
Among software engineers, there’s a debate brewing: Just how bad will the effects of AI code editors be on jobs? Some say they will shrink the industry’s workforce; others call them tools, not replacements for engineers.
The first time Ryan Clinton tried vibe-coding, he got scared for his job. He’s not fearful anymore, he said.
Clinton’s engineering level won’t be affected, said the 46-year-old software developer from Nashville. More experienced engineers work on “architecture and design,” he said, while more junior staffers code. At this point of AI coding, human intervention is also still routinely necessary.
“You want to make sure it makes sense,” he said. “Only an idiot would randomly click ‘yes’ and commit it.”
Barry Fruitman is more worried — but not for himself. At 56, the Android developer from Toronto doesn’t think the job market will feel the effect until five to 10 years out.
“Today, I think the threat is overstated, and hopefully it will stay that way until I retire,” he said.
Are there real productivity gains?
Ed Gaile said AI tools have doubled, if not tripled, his productivity.
The 55-year-old Appfire principal solutions architect from Atlanta was impressed by the decrease in context switching that vibe-coding tools brought.
“I wish I had this 15 years ago,” he said.
For AI code editors, the word “productivity” still looms large. Many people feel that they’re saving time by using these tools. Others cite the additional time spent reviewing and correcting lines of code.
Ed Gaile
A July METR study added fuel to the fire.
The study asked experienced developers to complete a series of tasks. Study participants working without AI’s help spent 10% more time coding — but those with AI assistance spent 20% more time reviewing AI outputs, prompting AI, waiting on AI, or being idle. Ultimately, the study found that the AI-assisted developers were less productive.
Shawn Gay, a 54-year-old R&D manager from El Paso, Texas, spends time keeping up with the industry’s changes. He said he felt behind the curve.
“I have decades of experience, so I feel like it’s a huge effort to try to change the way my brain thinks about software,” Gay told Business Insider.
Gus De Souza said that he saved time on coding, but spent more time reviewing the AI-generated code. The real productivity gains were in troubleshooting, said the 48-year-old software architect from Kitchener, Ontario.
What even is a vibe-coder? While the term has grown to encompass most forms of AI-assisted coding, Karpathy’s X post first defined it as when developers “fully give in to the vibes, embrace exponentials, and forget that the code even exists.”
Lara Fraser, a data analyst and epidemiologist from Sarasota, Florida, doesn’t consider herself a vibe-coder.
Fraser codes in R and uses tools like ChatGPT and Claude to assist. She’s tried other tools, but found high rates of hallucination. The model generation also matters, Fraser said: GPT 5.1 was great, but 5.2 was a “disaster.”
For Fraser, vibe-coding depends on the programmer’s skill. Anyone can create an app, but not everyone can maintain it.
“Inevitably, something’s going to break,” she said. “Can you fix it? If you can’t, you’re a vibe-coder.”
