At Snowflake’s Build conference this week, Google Brain founder and Stanford professor Andrew Ng described how the way people learn programming is changing fast.
Speaking before an audience of developers and tech executives, he urged everyone to keep learning to code but not by relying on old manual habits. His argument was simple, that is, artificial intelligence has made the barrier to software creation far lower, and this shift is rewriting who can build and how fast they can do it.
Ng, who also leads Landing AI and DeepLearning.AI, said the tools now available can turn almost anyone into a creator. He pointed to the arrival of AI-assisted code generation (often called vibe coding) as a major turning point for both professional engineers and people in other roles. Rather than spending hours writing every line from scratch, users can now guide the machine to write, test, and refine code for them. He called this a practical step for managers, analysts, and even marketers who want to automate tasks or build small applications without waiting on technical teams.
During the session, Ng explained that AI-powered development has already started to erode one of software’s long-standing advantages - the difficulty of replication. In the past, complex systems demanded large engineering teams working for years. Now small groups or even individuals can reproduce that work with AI help. The result, he said, is a faster cycle of creation and a much broader class of people who can participate in it.
Drawing on his experience running AI-focused startups, Ng noted that costs once associated with coding are dropping as consumption-based tools replace traditional subscription software. Developers now pay mainly for what they use, and the efficiency gains make the expense worthwhile. The transcript also showed his concern about how universities are responding. Computer-science graduates, he observed, are seeing higher unemployment because many courses still teach pre-AI methods. He said curricula should adapt quickly to include AI-assisted development if students are to stay competitive in the job market.
He also touched on a deeper issue within the industry, the growing divide between open and closed AI models. Ng highlighted how China has become a key player in releasing open-weight systems while the United States remains more guarded. He believes that encouraging open models will prevent innovation bottlenecks and help startups experiment freely without depending on a few large companies for access.
Ng’s remarks connected with broader conversations at the event about how AI is altering the economics of software. As cloud infrastructure grows more elastic, pricing is shifting toward pay-as-you-go models. Companies such as Amazon and Snowflake are already experimenting with platforms that automatically select the most efficient AI model for a given task. This trend, Ng said, could make future coding tools both cheaper and smarter over time.
He closed on a practical message. The next generation of workers, whether in engineering or marketing, should learn to build with AI rather than compete against it. Those who use these tools will get more done, he said, and discover that programming no longer belongs only to the technically trained. It is quickly becoming a universal skill for anyone willing to learn the new way.
Notes: This post was edited/created using GenAI tools. Image: Snowflake Inc. / YT
Read next: Researchers Discover AI Systems Lose Fairness When They Know Who Spoke, With China Becoming the Main Target of Bias
Speaking before an audience of developers and tech executives, he urged everyone to keep learning to code but not by relying on old manual habits. His argument was simple, that is, artificial intelligence has made the barrier to software creation far lower, and this shift is rewriting who can build and how fast they can do it.
Ng, who also leads Landing AI and DeepLearning.AI, said the tools now available can turn almost anyone into a creator. He pointed to the arrival of AI-assisted code generation (often called vibe coding) as a major turning point for both professional engineers and people in other roles. Rather than spending hours writing every line from scratch, users can now guide the machine to write, test, and refine code for them. He called this a practical step for managers, analysts, and even marketers who want to automate tasks or build small applications without waiting on technical teams.
During the session, Ng explained that AI-powered development has already started to erode one of software’s long-standing advantages - the difficulty of replication. In the past, complex systems demanded large engineering teams working for years. Now small groups or even individuals can reproduce that work with AI help. The result, he said, is a faster cycle of creation and a much broader class of people who can participate in it.
Drawing on his experience running AI-focused startups, Ng noted that costs once associated with coding are dropping as consumption-based tools replace traditional subscription software. Developers now pay mainly for what they use, and the efficiency gains make the expense worthwhile. The transcript also showed his concern about how universities are responding. Computer-science graduates, he observed, are seeing higher unemployment because many courses still teach pre-AI methods. He said curricula should adapt quickly to include AI-assisted development if students are to stay competitive in the job market.
He also touched on a deeper issue within the industry, the growing divide between open and closed AI models. Ng highlighted how China has become a key player in releasing open-weight systems while the United States remains more guarded. He believes that encouraging open models will prevent innovation bottlenecks and help startups experiment freely without depending on a few large companies for access.
Ng’s remarks connected with broader conversations at the event about how AI is altering the economics of software. As cloud infrastructure grows more elastic, pricing is shifting toward pay-as-you-go models. Companies such as Amazon and Snowflake are already experimenting with platforms that automatically select the most efficient AI model for a given task. This trend, Ng said, could make future coding tools both cheaper and smarter over time.
He closed on a practical message. The next generation of workers, whether in engineering or marketing, should learn to build with AI rather than compete against it. Those who use these tools will get more done, he said, and discover that programming no longer belongs only to the technically trained. It is quickly becoming a universal skill for anyone willing to learn the new way.
Notes: This post was edited/created using GenAI tools. Image: Snowflake Inc. / YT
Read next: Researchers Discover AI Systems Lose Fairness When They Know Who Spoke, With China Becoming the Main Target of Bias
