Failure rates in foundational computer science courses at the University of California have seen a staggering increase, with introductory class CS 10 reaching a record 35.3% failure rate in the first semester of 2026. This is a dramatic shift compared to previous years, where failure rates for both introductory and gateway programming courses consistently stayed below 10%.
Faculty members point to a growing dependency on AI tools as the primary culprit for this academic decline. While students may be completing assignments using generative models, they are arriving at exams without a fundamental understanding of the logic required to solve problems manually. This disconnect has led to several critical issues in the classroom:
- Students are frequently caught cheating on take-home assignments using AI-generated code.
- A lack of computational thinking skills is becoming evident during proctored, hands-on examinations.
- The CS 61A gateway course saw its failure rate climb to 10.6%, signaling that even students deeper in the major are struggling to move past automated assistance.
The transition from AI-assisted homework to individual performance in testing environments is proving to be a significant barrier. As educators grapple with these statistics, the trend suggests that the convenience of large language models may be eroding the rigorous problem-solving abilities traditionally expected of computer science graduates.
