Charles Esther
Andrew Ng, the former Google Brain scientist and Stanford professor, has spoken out about the growing trend of AI-assisted coding, challenging the casual notion behind the term “vibe coding.”
While many may think vibe coding sounds relaxed and effortless, Ng describes the term as “unfortunate” because it misleads people about the real demands of coding with AI.
Ng explained at the LangChain Interrupt conference in May that vibe coding suggests developers simply “go with the vibes” and accept or reject AI-generated code without much thought.
In reality, Ng says, coding with AI remains “a deeply intellectual exercise” that requires intense mental effort. He shared that after a full day of coding with AI assistance, he often feels “frankly exhausted.”
Despite this, Ng is enthusiastic about AI’s role in software development. He calls AI-assisted coding “fantastic” because it enables programmers to write code faster and sometimes with barely glancing at the underlying code.
The term “vibe coding” was coined by OpenAI cofounder Andrej Karpathy in February to describe how developers can lean heavily on AI prompts and “forget the code even exists.”
This approach has significantly changed software development. It has sparked concerns among engineers about job security and raised debates on whether startup founders still need strong technical skills.
Vibe coding has also empowered nontechnical individuals to build apps. For example, a Block product designer with no formal coding background used vibe coding to create a dog ID app in just two months.
Ng believes that more companies should embrace AI-assisted coding instead of resisting it. He pointed out that many CIOs and CTOs still have restrictive policies that prevent engineers from using these tools.
Ng urged, “We have to get past that,” adding that his own teams dislike coding without AI support.
Everyone Should Learn to Code
Ng is concerned about a misleading trend advising people to skip learning programming because AI will automate all coding.
He called this advice “some of the worst career advice ever given.”
Ng strongly believes “everyone should learn to code,” at least one programming language. He sees coding as a critical future skill for both developers and non-developers alike.
Learning to code helps people communicate precisely with computers and understand how to identify and fix errors. This leads to writing better prompts and instructions for AI tools.
At his startup AI Fund, Ng revealed that all employees, including the CFO, general counsel, and receptionist, know how to code at some level.
“It’s not that I want them to be software engineers. They’re not,” Ng clarified. “But by learning a little bit about how to code, they’re better able to tell a computer what they want it to do.”
This increased clarity has resulted in “meaningful productivity improvements” across the company, Ng said.
In summary, while AI-assisted coding may appear simple on the surface, Andrew Ng stresses it remains a challenging intellectual task.
He encourages widespread coding education and embracing AI tools to unlock new levels of productivity.
The rise of AI coding tools signals an evolving future for programming, where human insight and machine assistance work hand in hand.
