Students and professionals said they are increasingly using artificial intelligence assistants to code — with increasingly efficient results.
Marco Papa, a senior lecturer of computer science, said “vibe coding” is a term coined earlier this year by Andrej Karpathy, the former director of AI at Tesla and a co-founder of OpenAI.
“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,” Karpathy wrote in a February 2025 post on X, adding, “I’m building a project or webapp, but it’s not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.”
At USC, Papa teaches Principles of Software Development and Web Technologies, which are offered to undergraduate and graduate students, respectively. He said he embraces vibe coding in those classes and feels he would be doing students a disservice to deny them the preparation of using AI tools in their work.
“We will have to learn it,” he said. “And in fact, they’ll probably have to learn not just one of these configurations, but most of them.”
In his courses, Papa provides students with guides on how to use AI to assist their coding. He tells students that many AI coding assistants are free for them with verification of their academic email.
He said the use of AI tools has also helped him adjust to the loss of dozens of grading and teaching assistants following significant university layoffs. Students’ questions on Piazza, a discussion board tool typically used by computer science classes, have been halved, Papa said, and office hours appointments have become sparse.
Further, he added he has taken to creating quizzes for his class by uploading lecture materials to an AI assistant, saving him time.
“In terms of score with or without the tools, the average is the same, which means they probably did less work and they spent less time querying the instructor or the TAs during office hours,” Papa said. “Instead, they interacted with the LLM to ask the same questions that they would have asked the instructor.”
While the university leaves the decision up to professors of whether to prohibit, limit or encourage AI in classes, Papa said he has found policing its use difficult and impractical. Furthermore, knowledge of AI tools in coding has begun to weave its way into job interviews, as Microsoft itself has revealed that as much as 30% of their code is now AI-generated, he added.
“If you’re one of the lucky ones that gets an interview, they may give you what is called the impossible project,” he said. “It’s a description of a programming project that normally would take three weeks, and during the interview, they say, you have three hours. The idea is that if you are knowledgeable about using the AI tools, you actually may make it.”
Students and faculty to dive deeper into AI
USC is also offering programming on the use of AI at the university, including through vibe coding. On Nov. 12 and 13, the Viterbi Office of Technology Innovation and Entrepreneurship will host a “Vibe Coding Workshop” from 5 to 7 p.m. The workshop is open to all students and requires no coding experience. The workshop plans to offer instruction on use of the AI coding assistants Lovable and Github Copilot.
Additionally, on Nov. 13, USC will host a Fall Faculty Showcase on AI in Teaching, where Papa and other university staff and instructors are scheduled to speak on topics ranging from AI literacy to the use of AI in design.
In surveys of his fall semester students, Papa found that about half of students felt they had to make significant changes to the code their AI assistants generated, while about half felt they did not have to make significant changes. He added that the surveys showed the vast majority also preferred to use AI and indicated they felt they still had a good understanding of the code produced by the AI tools.
Eshaan Kapoor, a graduate student of computer science, said he had signed up for the workshop to see what methods could help him save time and energy, although he emphasized the importance of people maintaining an adequate understanding of the coding languages they would be using.
Although the use of AI assistants in coding has become common among students and professionals, it has been uncommon to come across direct academic instruction on the subject, he said.
“This will be the first time that I’m gaining a third person’s perspective into vibe coding and their own set of rules to follow,” Kapoor added. “This was the first time that someone will be teaching me or telling me how a certain vibe is to be followed — so that is new, and I like to see the dynamic shift here.”
Yashaswi Sharma, a graduate student of computer science, said he typically uses the AI coding assistant Cursor but hoped to attend the workshop to learn about using other assistants. He felt the use of AI had already dramatically improved his productivity.
“I just bootstrapped an end-to-end project in like four days that would typically take something like 40 days,” Sharma said. “So if you really understand what’s going on under the hood, you can scale up yourself as an engineer on the technical front, 10 to 100 times more than what you could do effectively before.”
Papa said that while job losses may also result from the continuing integration of AI into professional fields, he believed there will also be changes to the industry that will create new, AI-powered jobs that had not previously existed.
Kapoor said nevertheless, he felt nothing can replace human skill and understanding in editing and diagnosing issues with code, and added AI can not be trusted to be fully accurate.
He said AI-assisted coding appears to be an inevitable future, with room for optimism in the changes it would bring.
“There is no going back, I’ll be honest — once you start to vibe code, once you see that your time is getting saved, and you can invest that time somewhere else,” Kapoor said. “Once you see that happening, I don’t think there’s any time any person could go back.”