A dimly lit basement classroom at Stanford University is filled with anxious computer science students, and their instructor, Mikhail Erik, tells the class that he’s going to teach them how to write code without writing a single line of code.
Eric’s class modern software developerhas quickly become one of the most popular Stanford CS courses this semester, and is touted as the first at a major university to employ coding tools such as Cursor and Claude.
It’s an anxious time to major in computer science, even at a prestigious university like Stanford University, knowing you’ll be graduating into a world where AI is getting better at programming every day.
“It can be scary to think that your job security is at risk and that you might be fired,” said Brent Ju, one of the dozens of students in the class. Joo will graduate this spring, but so far she has not received any job offers. “The market is a little tough. We’re still interviewing.”
“It gives me more power to go through this entire class without writing a single line of code,” said Eric, a Stanford graduate. Eric intentionally designed this course to be an antidote to the majority of classes that still prohibit the use of AI.
Some of the biggest names in AI coding, including Claude Code creator Boris Charney and Vercel’s head of AI research, Gaspar Garcia, stopped by the idyllic Palo Alto campus to give guest lectures. Andreessen Horowitz general partner Martin Casado will speak at the final class next week.
Inside a classroom on a recent morning, Cyrus Alberti, Cognition’s director of research, gave a talk titled “Your Unique Guide to AI Coding in 2025.”
Cyrus Alberti, Head of Research at Cognition, gave a talk titled “Your Unique Guide to AI Coding in 2025.” Ben Bergman/BI
“I think we’re always a little bit behind in what we learn in school, so I’m glad that this course exists to teach us what’s new,” Alberti said after his talk, surrounded by students who lined up to greet him like celebrities. “Learning the same way yesterday won’t make you super competitive, but if you really use the tools, you’ll become a super engineer.”
excitement and fear
The mood of the students in the class reflects Silicon Valley’s current zeitgeist, excited about what many consider one of the most important technological advances of our lifetimes. But there are many concerns that AI will make expensive Stanford degrees obsolete.
When Eric graduated in 2016, a CS degree from Stanford University was his golden ticket.
“People thought, ‘I’m going to go to an elite university and then I’m ready for life and I’m just going to work a six-figure job at a FAANG company for as long as I want,'” he says.
The number of CS students has skyrocketed as technology companies embark on large-scale recruitment drives.
“On the other hand, many companies that made mass hiring during the coronavirus pandemic found themselves overemployed,” Eric said. “Now we not only have a surplus of young people, but we also have a surplus of experienced people who have been newly laid off.”
To make matters worse, AI is already proficient at coding and continues to improve rapidly. Microsoft CEO Satya Nadella said up to 30% of the company’s code is written by AI, while Anthropic CEO Dario Amodei predicted in March that AI could write “essentially all” of the company’s code within a year.
Ju, whose dream job is to work at Anthropic, said she is trying to stay positive.
“It’s really interesting if the tool acts as an assistant rather than a replacement for the user, because it can significantly increase productivity and make you a more effective developer,” says Ju. “I’m more of an optimist and lean in that direction.”
Zach Lloyd, founder and CEO of Warp, an agent workflow developer tool, gave a guest lecture last month and asserted that he remains very interested in hiring CS students.
“The idea that someone with a CS education from a university like Stanford can’t get a job as an engineer is a bit of a stretch,” he said, adding that knowing the basics of programming is still important to using Warp and Claude effectively. “These tools are accelerators, but they are not replacements yet. The people who actually use these tools best are those who have a solid foundation.”
Eric plans to teach the course again next year, but he says the class will likely look very different because AI is advancing so quickly.
“People asked me if I was worried that what I said in week one would be outdated by week seven,” he said. “Yes, that is a concern. So far it hasn’t happened yet.”
