Stanford CS Professor: AI Can Code. That’s Why You Should Learn | Chris Piech
In a Nutshell
Chris Piech argues that AI's coding capabilities shouldn't discourage learning to program—instead, students must maintain self-awareness to avoid outsourcing too much of their own problem-solving growth. The most critical skill is becoming a high-contributor engineer through "time on task" and creating prototypes while using AI to teach concepts, not just generate code. Human teachers remain irreplaceable for motivation and inspiration that AI cannot provide, and the real opportunity for junior engineers lies in understanding valuable human problems rather than just technical implementation.
These notes were generated by AI and may contain inaccuracies.
Chris Piech is a professor at Stanford University who teaches large introductory computer science classes and introductory math for AI. He runs Code in Place, an online programming class with approximately 17,000 students and more than 1,000 teachers, making it the class with the most teachers in the world. The program has been running for 6 years, operating both before and after tools like Cursor and Claude Code became available.
Since implementing these AI coding tools, enrollment in Code in Place has essentially doubled, indicating increased interest in learning to code despite AI capabilities.
Piech argues that AI's ability to code, perform probability calculations, and write should not discourage people from learning these skills. The wrong answer to "Should I learn to program?" is "No." Students should still learn to formalize arguments, understand probabilistic reasoning, and program. The reasoning is that AI abilities may magnify human capabilities, and being smart in these spaces will remain important.
"We're not giving up on the next generation being smart."
Piech observes more students experiencing motivational crises than in the past. This stems from uncertainty about future contributions and job markets. Students starting 4-year programs must consider what jobs will exist in 2030 when AI is 4 years more advanced, creating significant uncertainty.
Sign in to read the full notes
Get access to AI-generated notes, topic timestamps, and more.