The companion platform for CSCI 455/555 at William & Mary. Interactive teaching modules, student-built projects, and open resources — designed for students to learn, experiment, and showcase their work with modern AI tools.
CodeLab is the open, browser-based companion platform for Generative AI for Software Development (CSCI 455/555) at William & Mary.
It provides self-paced interactive learning modules, student project showcases, and hands-on demos — all free, no install required. CodeLab bridges the gap between academic teaching and real-world AI-assisted development by giving students a place to learn, experiment, and publish their work.
Every module features interactive slides with embedded runnable code, live demos, and exercises. Student projects showcase full applications built through “vibe coding” — building software primarily through natural language and AI pair-programming.
Free for anyone — no login, no install. Open your browser and start learning.
Interactive slides with live code, runnable demos, and exercises — all in-browser.
Real projects built by undergraduates using AI pair-programming and vibe coding.
Developed by an active researcher in AI for Software Engineering at W&M.
Students develop foundational skills in deep learning and generative AI applied to software engineering — from mining repositories and training language models to prompting LLMs and detecting hallucinations in generated code.
This course covers the full pipeline of AI-enabled software development: collecting and processing source-code data, building probabilistic and neural language models, evaluating generated outputs, and applying LLMs to real-world coding tasks.
Go from zero to shipping real applications using LLMs as a coding partner. Learn what works, what doesn't, and why — through hands-on vibe coding projects.
Explore how language models learn from code: tokenization, probability, attention, and the training pipeline behind tools like GitHub Copilot and ChatGPT.
Evaluate AI-generated code rigorously. Measure quality with established metrics, detect hallucinations, and understand when to trust model outputs — and when not to.
Seven self-paced, browser-based modules covering the full pipeline — from mining repositories to detecting hallucinations. Each module includes interactive demos and hands-on exercises. No install required.
CodeLab was born from first-hand experience with vibe coding. As a researcher and practitioner, Antonio started building software through natural language prompts and AI pair-programming — and the impact on the development chain of command was striking. That hands-on experimentation sparked a deeper reflection on how far this novel approach to writing code can actually take us.
This website is itself part of those original experiments: an entirely vibe-coded platform, built and iterated through AI-assisted development. It now serves as the open companion to Generative AI for Software Development (CSCI 455/555) at William & Mary, where students explore the same workflow — from prompt engineering and model evaluation to shipping real applications.
All teaching modules are self-paced, browser-based, and free to use. Student projects demonstrate what undergrads can build when armed with modern LLMs and a bit of ambition.