Teaching

2026-05-12 · Source: Kyunghyun Cho · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, long

Summary

An NYU computer science course for seniors was fully revamped to integrate modern software engineering practices and AI coding assistants. Taught post-ChatGPT, the course aimed to teach fundamental machine learning concepts by having students build complete software packages from scratch for each topic. Initial weeks focused on mathematical and statistical foundations, while teaching assistants ensured students had access to coding assistants like GitHub Copilot and OpenAI Codex with free credits. The core methodology involved weekly "vibe coding" sessions where the instructor demonstrated building a full web application for a new ML concept, emphasizing prompt setup and debugging with AI agents. Students then replicated this process, developing applications for concepts such as vector space retrieval, dimensionality reduction, clustering, and reinforcement learning. Lab sessions provided deeper dives into algorithmic mechanisms using PyTorch and Transformers. Final projects, built without constraints, showcased impressive student creativity and functional applications, highlighting a new era of education where students learn by doing and making mistakes with advanced tools.

Key takeaway

For AI educators designing modern curricula, consider integrating AI coding assistants to foster practical, project-based learning. You should structure courses to allow students to build complete applications from scratch, embracing iterative development and debugging with AI tools. This approach cultivates essential problem-solving skills and prepares students for an industrialized software engineering landscape, moving beyond theoretical exercises to hands-on creation.

Key insights

Integrating AI coding assistants enables a project-based, "learn-by-doing" approach in machine learning education.

Principles

• Embrace uncertainty in machine learning.
• Systematize convex optimization techniques.
• Allow students to learn from mistakes.

Method

Instructors "vibe-code" full-stack applications for ML concepts, demonstrating prompt engineering and debugging with AI, then students build their own from scratch weekly.

In practice

• Provide students with coding assistants.
• Encourage full-stack app building.
• Demonstrate prompt engineering.

Topics

• Machine Learning Education
• AI Coding Assistants
• Project-Based Learning
• Software Engineering Curriculum
• Prompt Engineering
• Full-Stack Development

Code references

• kyunghyuncho/nyu-course-search
• kyunghyuncho/nyc-taxi-trip-visualization
• kyunghyuncho/clustering-countries
• kyunghyuncho/checker-bot
• kyunghyuncho/federal-payroll-prediction-antigravity

Best for: AI Student, AI Scientist, Research Scientist

Related on AIssential

• Learn by Building with Coding Agents — A workflow leveraging AI coding agents offers a custom, effective way to learn programming by building.
• 5 Projects To Get Started With AI Engineering — Building small, practical projects is the most effective way to learn applied AI engineering concepts.
• Are Engineers Forgetting How to Think? The Hidden Cost of AI-Assisted Coding — AI-assisted coding, while boosting speed, risks eroding engineers' fundamental problem-solving and critical thinking skills.
• Presentation: State of Play: AI Coding Assistants — AI coding assistants are evolving towards greater autonomy, necessitating advanced context engineering and robust safety harnesses.
• Agents That Teach: Towards Designing Incidental Learning Back into AI-Assisted Software Development — AI coding agents reduce incidental learning, necessitating deliberate design to reintroduce teaching moments and prevent "Knowledge Debt."
• The Hitchhiker's Guide to Monoculture — AI coding assistants standardize code syntax but do not homogenize underlying problem-solving strategies or semantic approaches.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Kyunghyun Cho.