The guy who coined ‘vibe coding’ predicts it will ‘terraform software and alter job descriptions’

• Andrej Karpathy led AI at Tesla and cofounded OpenAI. He also also coined the term “vibe coding.”
• Reflecting on the last year, Karpathy wrote that vibe-coding has produced a new type of code that is “free” and “discardable.”
• He predicted that vibe coding will “terraform software and alter job descriptions.”

He coined “vibe coding” earlier this year. Now, he has something to say about it.

Andrej Karpathy led AI at Tesla for five years, steering the company’s Autopilot effort and briefly working on its humanoid robot Optimus. He sandwiched his Tesla job with two stints at OpenAI, making Karpathy a cofounder of the AI pioneer.

As 2025 comes to a close, Karpathy published his year-in-review for large language models on X. He reflected on the famous term he originated in February, a term that has since shaken up the software engineering industry.

“With vibe coding, programming is not strictly reserved for highly trained professionals,” Karpathy wrote. He called it an example of how “regular people benefit a lot more from LLMs compared to professionals, corporations and governments.”

Vibe coding has likely benefited businesses, too. Tech companies have equipped their engineers with tools like Cursor, Claude Code, and OpenAI’s Codex, aiming for productivity gains.

Karpathy wrote that vibe coding “empowers trained professionals to write a lot more (vibe coded) software that would otherwise never be written.”

It may also change the makeup — or the use case — of the code itself. Karpathy threw out a slew of adjectives to describe this new body of code: It is “free, ephemeral, malleable, discardable after single use.”

“Vibe coding will terraform software and alter job descriptions,” he wrote.

How does Karpathy feel about being the term’s origin?

“Amusingly, I coined the term “vibe coding” in this shower of thoughts tweet totally oblivious to how far it would go,” he wrote.

It’s not yet clear how efficient vibe coding is making engineers. In a METR study published in July, AI coding assistants were found to decrease the productivity of participating experienced software developers by 19%. The developers in that study were also overconfident in the tools, its authors said, expecting a 20% productivity boost even after using them.

What is clear, though, is that the practice is unlocking a whole new form of tech products. Twitter founder Jack Dorsey vibe-coded a new messaging app this year. Non-technical workers are easily building, shipping, and, in some cases, even selling apps they build in hours, if not minutes.

Karpathy gave some other reflections. He praised Google Gemini’s Nano Banana image model, and wrote that Claude Code was the “first convincing demonstration of what an LLM Agent looks like.”

Overall, Karpathy wrote that 2025 was an “exciting and mildly surprising year of LLMs.”