This paper presents the integration of Large Language Models (LLMs) and Physics-Informed Neural Networks (PINNs) into a Computational Fluid Dy-namics course based on deep learning and constructive alignment. The course included a lecture and a computer laboratory, conducted via Jupyter-Hub.NRW, enabling students to access remote Python environments and hardware. A survey of 14 students revealed that LLMs, primarily ChatGPT, were used mostly for programming assistance and debugging, while support for mathematical and physical understanding or generating examples was moderate, and planning activities was rare. Exam results indicate that per-formance differences were driven by cognitive levels as defined by the SOLO taxonomy rather than by AI versus traditional content, suggesting lim-ited intrinsic motivation toward AI. Students actively engaged with three AI Literacy Framework domains—Engaging with AI, Creating with AI, and Managing AI—while the fourth, Designing AI, was addressed through hands-on PINN tasks in lectures and laboratories. These activities exceed secondary education expectations of the AI Framework and represent a tertiary-level deepening of AI Literacy competencies. The findings highlight how AI can support numerics education, enhance learning outcomes, and provide a framework for developing advanced competencies in designing and critically evaluating AI systems in higher education.