University overhauls Intro AI course to beat LLM cheating

The University of Washington Bothell’s Computer Science program faced a growing crisis: students were using large language models (LLMs) like GPT-4o to complete nearly all assignments in its introductory AI course, CSS~382. Rather than ban AI tools, Yusuf Pisan and his team redesigned the curriculum to embrace them—while ensuring students understood the technology inside and out.

The new approach kept the course’s classical core (search algorithms, reinforcement learning, etc.) but added a hands-on strand where students built an LLM from scratch. Assessment shifted to AI-resistant tasks like in-class exercises, reflective writing, and a defended team project, eliminating traditional exams. The most radical change? The course policy flipped from ignoring AI in 2023 to *requiring* it in 2026. Students even drafted a 'Student Bill of AI Rights'—a participatory ethics framework that forced the instructor to complete any AI-generated assignment before using it in class.