Vibe coding with AI sparks debate, reshapes developer jobs

As agents expand generative AI’s abilities from writing lines of code to creating entire applications, an industry argument that began with ChatGPT has gained new momentum.

The term vibe coding entered the popular lexicon from a post on X, formerly Twitter, in February. It was written by Andrej Karpathy, founder of Eureka Labs, an AI education company, a former director of AI at Tesla and part of the founding team at OpenAI. In the X post, which now has 5 million views, Karpathy described vibe coding as an approach “where you fully give in to the vibes, embrace exponentials, and forget that the code even exists.” He admitted he rarely checked the work of the tool he was using at the code level. “The code grows beyond my usual comprehension,” he said.

Karpathy did his vibe coding with a tool from Cursor AI called Composer, which expanded the company’s original AI-driven integrated development environment (IDE) to support multi-file code generation and modification. In April, more coding tools, such as JetBrains and GitHub Copilot, introduced coding agents, and another AI-powered app generator, Windsurf, was acquired in May by OpenAI for $3 billion.

While Karpathy’s original post described vibe coding as “not too bad for throwaway weekend projects,” some people in the tech industry, especially in the startup world, take it much more seriously. They have resumed a refrain that began with ChatGPT’s first introduction: developer jobs are about to become obsolete.

There’s a new kind of coding I call “vibe coding”, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It’s possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper…

— Andrej Karpathy (@karpathy)
February 2, 2025

Are developer jobs doomed?

In a March video panel discussion of Y Combinator leaders, YC managing partner Jared Friedman disclosed that one quarter of respondents to the startup accelerator’s survey said more than 95% of their codebases were AI-generated.

“This isn’t a fad, this isn’t going away, this is actually the dominant way to code, and if you’re not doing it, you might just be left behind,” said Garry Tan, president and CEO of Y Combinator, in the video.

Execs at other startups have also bought into this idea.

“I’ve got eight computer screens in front of me, and five of them actually have vibe coding going on right now, so I’m actively writing code while we’re talking,” said Ken Schirrmacher, CTO at ObligeAI, a passwordless authentication startup in San Francisco, during an interview with Informa TechTarget in April.

The typical days of a developer sitting there and solo writing 30- or 40,000 lines of code are gone.

“What would take large numbers of teams and massive numbers of people and tons of time to create can be created in just weeks and months versus years,” Schirrmacher said. “The typical days of a developer sitting there and solo writing 30- or 40,000 lines of code are gone. … You aren’t starting from a blank slate anymore.”

As vibe coding expanded, dark predictions about the future for human developer jobs also grew. Anthropic co-founder and CEO Dario Amodei stated in a March interview that AI would be writing 90% of code within three to six months, and “essentially all of the code” within the next year.

“We will eventually reach the point where the AIs can do everything that humans can. And I think that will happen in every industry,” Amodei said. “I think it’s actually better that it happens to all of us than that it happens — you know, that it kind of picks people randomly. I actually think the most societally divisive outcome is if randomly 50% of the jobs are suddenly done by AI, because what that means — the societal message is we’re picking half — we’re randomly picking half of people and saying, ‘You are useless, you are devalued, you are unnecessary.'”

It soon appeared that Amodei’s prediction might be coming true: some web companies, such as Shopify and Fiverr, began warning employees in April that they would have to prove the value of future human recruits over AI tools.

Schirrmacher took this a step further. If AI makes software development faster with fewer developers, he predicted that large tech companies that employ thousands of developers will also become unsustainable over the next decade.

“Imagine a country just announced that they have a mountain that they did scans on and realized that the entire mountain is filled with gold — the price of gold is now going to plummet,” he said. “That’s what AI has done with development — and you can’t sell a product offering for as much as you’re charging today and expect that you’re going to get away with that in 10 years.”

Not so fast: Coding agents hit snags

Five months later, Amodei’s prediction that most code will be AI-generated within three to six months does not appear close to becoming reality, even among large tech companies. In April, Microsoft CEO Satya Nadella estimated that between 20% to 30% of his company’s code had been generated by AI, similar to an estimate by Google CEO Sundar Pichai in an October 2024 earnings call. IBM’s CEO, Arvind Krishna, predicted in a March 2025 interview that this is where AI-generated code percentages will stay.

Several factors are slowing AI’s coding takeover, including ongoing issues with hallucinations and biased or inaccurate results. For most enterprises with established reputations, vibe coding serves as a good way to develop minimum viable products, but most companies wouldn’t put what it generates into production, said Diego Lo Giudice, an analyst at Forrester Research.

“It will definitely enhance prototyping and innovation in companies, but then somebody is going to have to go and do the rest of the work for turning that into a real, scalable product,” Lo Giudice said. “They can still use code generators, but that’s going to be done by true developers, not by vibe coders.”

This pattern has already begun to alter the role of senior developers at many organizations, but not necessarily for the better.

Kyler Middleton

“It shifts the expertise required from the junior engineers who are now vibe coding and using tools to seniors who have to represent the business requirements for stability, security, extensibility and standardization,” said Kyler Middleton, principal software engineer at healthcare tech company Veradigm. “I am, personally, reviewing 10 times the amount of code I was last year, but I can’t just run at a 10x pace to make sure all this code generated by AIs fits our standards.”

Over time, junior developers will likely become less skilled at writing their own code and more dependent on vibe coding tools, widening the skills gap between junior and senior engineers, said Rob Strechay, an analyst at TheCube Research.

“I think it severely hurts young developers and will affect how companies ‘grow’ their own devs,” Strechay said.

Ongoing macroeconomic volatility and uncertainty are only making that worse, he said.

“Teams aren’t getting the hires they thought they would and instead are being told to use AI to make themselves more productive,” Strechay said. “At the same time, they are being slowed down in using AI by potential IP issues; the AI might be as good as a college grad, but there is a lot of time spent ‘fixing’ the code once you use AI. Sometimes it could be twice the time it would have been to code in the first place.”

Google’s DORA has already observed this slowdown in its research over the past year and, in an April 2025 report, noted that AI-assisted coding hasn’t lived up to its promises for developers in other ways.

“While AI is making the tasks people consider valuable easier and faster, it isn’t really helping with the tasks people don’t enjoy,” the report reads. “That this is happening while toil and burnout remain unchanged, obstinate in the face of AI adoption, highlights that AI hasn’t cracked the code of helping us avoid the drudgery of meetings, bureaucracy, and many other toilsome tasks.”

The long view: dev decline vs. democratization

However, some believe developer jobs will disappear over a period of years, not months, as AI technology resolves its initial issues.

For example, AI code can introduce security vulnerabilities, and at a volume it’s difficult for humans to keep up with, “but the solution is to cram some more AI into it, and have AI do validations on code,” according to Schirrmacher.

Some tech vendors believe that the long-term outcome of this cycle of using AI to manage AI will be full autonomy for future agentic AI systems — and the removal of most human engineers.

Fetch.ai is developing an agentic AI PaaS based on its experience as a cryptocurrency company, and plans to use blockchain consensus mechanisms to verify the accuracy of agent-based outputs. Its founder and CEO, Humayun Sheikh, also predicts that advances such as quantum computing will accelerate autonomous agentic AI systems.

“We have to build this new agentic web, and people will need to build that infrastructure,” Sheikh said in a February interview. “But once it’s done, we wouldn’t have to write the same code for 10 different companies, 10 different times. Software development, over time, will be a thing of the past, because LLMs can do a better job.”

However, even major changes in business technology, such as the shift from mainframes to distributed systems and from on-premises data centers to cloud computing, are rarely absolute. Similar predictions about the demise of IT operations professionals amid the rise of DevOps also haven’t come true — instead, the role has evolved into platform and site reliability engineering.

For now, most industry experts predict a similar trajectory for developer jobs in the era of AI coding.

“Who we call a developer is starting to change,” said Nick Cassidy, lead innovation engineer and lead AI product developer at Stellarus Group, a health services company in San Diego. Cassidy emphasized in an interview during Red Hat Summit in May that his personal opinions do not necessarily reflect those of his employer.

“The aperture of development is expanding — you may not need to have the deep skills that you needed before to get started as a developer,” Cassidy said. Tools such as Red Hat’s InstructLab have also made it possible for citizen developers to build AI-driven applications and fine-tune models, he said.

This has led to new ways of working at some companies, where business leaders use vibe coding to speed up handoffs to engineering teams.

“I’ve personally been picking up AI tools to prototype things for engineers,” said David Strauss, CTO at WebOps service provider Pantheon.io. “It’s much easier to share a prototype that is even partly wrong than natural language explaining what we need to accomplish.”

Generative AI has also spawned new disciplines, such as prompt engineering, which could indicate the future of software development work. Even if daily development tasks change, the essential need for human problem-solving will remain, predicted Lars Maaløe, co-founder and CTO at Corti, a healthcare AI company.

“AI models have a tendency to regress toward the mean,” Maaløe said. “That means that what they know, they’re very comfortable building. So, if you ask it to build a website number 1,000, then it will build website number 1,000 with the design schedule that it’s normally used to seeing. But it will continue to be bad at actually showing real human creativity.”

Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.