The man behind ‘Vibe Coding’ admits he’s never felt more behind as a programmer

The coder who coined vibe coding

Andrej Karpathy has spent years ahead of major tech shifts. He was an early founding member and research scientist at OpenAI in 2015 and later served as Tesla’s senior director of artificial intelligence, where he led the Autopilot vision team.

That history makes his latest admission striking. Karpathy says he now feels more behind as a programmer than ever before. The tools he helped inspire are moving faster than his ability to fully use them.

Why vibe coding took off

Karpathy introduced the term “vibe coding” in February 2025 to describe using natural language prompts and iterative feedback to get models to generate code.

The phrase spread quickly across tech circles. Collins Dictionary later named it Word of the Year. It captured how coding was shifting from typing lines manually to guiding intelligent systems.

A sudden shift in coding

In a recent post on X, Karpathy said software engineering is being dramatically refactored. Individual programmers are now writing fewer lines of code as AI systems handle more of the work.

He described the moment as unsettling. The pace of change has made longtime experts feel outmatched by tools that barely existed a year earlier.

Feeling left behind by tools

Karpathy said he believes he could be far more productive if he fully understood today’s tools. He suggested the gap feels like a personal failure rather than a lack of available technology.

The challenge is not access. It is learning how to combine fast-evolving tools into a reliable workflow while expectations for speed and output continue to rise.

AI tools everywhere now

Editors and agentic tools such as Cursor, Anthropic’s Claude Code, and OpenAI’s Codex now integrate with terminals and IDEs to read codebases, make coordinated edits, run tests, and propose fixes, which changes how engineers prototype and ship software.

Success is no longer measured only by writing flawless code but also by the ability to work alongside intelligent systems, continuously learning and adjusting as tools evolve at a rapid pace.

An alien tool with power

Karpathy compared modern AI coding tools to a powerful alien device. He said they were handed out without instructions, forcing everyone to learn through trial and error.

At the same time, he likened the disruption to a massive earthquake. The profession is shifting while developers are still figuring out how to hold the tools properly.

Other builders feel same way

Boris Cherny, who helped build Anthropic’s Claude Code, has described similar experiences in which teams begin solving problems manually and then realize that AI tools can complete large parts of the work much faster.

Cherny noted this happens frequently. Even those building the tools struggle to stay aligned with what the models can now accomplish.

New coders may adapt faster

Cherny suggested new graduates might thrive in this environment. Early career developers have fewer assumptions about what AI can or cannot do. This openness allows them to explore new methods and approaches without being limited by established habits or old workflows.

Without fixed habits, they may adjust more easily as models improve every few months. Constant relearning is becoming part of the job.

Tools still misfire sometimes

Karpathy cautioned that AI coding tools are far from perfect. He compared them to a weapon that sometimes misfires or shoots pellets instead of precision output.

The analogy highlights the work-in-progress nature of these systems. Reliability improves, but trust must still be earned. Developers must remain vigilant, testing and verifying results to ensure that outputs meet standards and expectations.

When everything suddenly clicks

Despite flaws, Karpathy said the tools sometimes perform remarkably. When used just right, they can solve problems with surprising speed and clarity. These moments highlight the potential of AI to significantly boost productivity and creativity in software development.

He described moments where solutions appear instantly. Those flashes hint at why developers keep pushing to adapt.

Redefining programmer identity

Vibe coding is reshaping what it means to be a programmer. The focus is moving away from purely writing lines of code to guiding and directing intelligent systems that can handle complex tasks on their own.

Even veterans like Karpathy are adjusting their self-image. Developers must understand how to guide, correct, and integrate machine intelligence into their workflows to stay relevant.

An industry still recalibrating

The pace of AI improvement means no one feels settled. Developers must constantly reassess what tools can handle versus what still needs human judgment.

Mental flexibility, adaptability, and the ability to learn quickly have become far more important than raw programming muscle. Success increasingly depends on knowing when to rely on AI and when human insight is still essential.

Want to see how AI is playing out in real time? Check out how users are already upgrading their smart homes with a simple switch.

Vibe coding future questions

Karpathy’s comments reveal how deeply AI is reshaping software work. Even the people driving the change feel pressure to keep up. The very systems he helped design are evolving so quickly that staying fully up to date has become a daunting challenge.

His comments highlight a reality many software professionals face today: the pace of AI innovation can outstrip even the most experienced developers’ ability to master it completely.

Curious how Alphabet’s broad strategy compares to a chipmaker’s big win? See how Nvidia is playing the game right now.

What do you think about vibe coding reshaping programming? Share your thoughts.

This slideshow was made with AI assistance and human editing.

Don’t forget to follow us for more exclusive content on MSN.

Read More From This Brand:

• Google brings Opal vibe coding tool to Gemini dashboard
• OpenAI Launches Codex AI Coding Agent In ChatGPT
• Google hands unreleased Pixel phones to 15 loyal testers