✦ New Elective Course — Starting Autumn 2026
Vibe Coding
AI-Assisted Software Development
Scroll
In early 2025 a term started circulating that captured something many people working with AI were already feeling — that the relationship between human intent and working software was changing in a fundamental way. The phrase came from AI researcher Andrej Karpathy, and it spread quickly because it named something real.
"There's a new kind of coding I call 'vibe coding', where you fully give in to the vibes, embrace exponentials, and forget that the code even exists."
Andrej Karpathy, 2025 — view original post on X ↗
The term spread quickly — but this course isn't simply an invitation to build software without understanding it. Working effectively with AI coding tools still demands a kind of technical literacy: enough to formulate problems clearly, evaluate outputs critically, and recognise when something has gone wrong. What counts as "enough" — and how programming skills relate to AI-assisted development — is one of the central questions the course is built around.
A 10 ECTS elective running across one semester at SDU Kolding. No prior programming experience required — but expect to engage with real code. The course builds the practical literacy needed to direct AI effectively and understand what it produces. Open to both Bachelor and Master students — click the official ODIN course descriptions below.
This page is maintained as a readable, up-to-date guide for prospective students and may be revised as the course develops. The ODIN course descriptions linked above are the official, university-approved records — in any case of discrepancy, those take precedence.
From conceptual foundations to practical workflows — each area builds on the last to give you both the understanding and the hands-on skills to develop software with AI.
A conceptual journey through the layers of computing — from machine code to high-level languages to natural language prompting. Where does vibe coding sit in the history of CS, and what made it possible?
Theory in service of practice. Understand how large language models process code, why context matters, and how model behaviour shapes outputs — giving you the mental model to prompt more deliberately and debug AI responses more effectively.
Prompting is a learnable, improvable craft. Master how to formulate tasks, manage context, iterate on outputs, and build structured strategies that consistently turn natural language into reliable, high-quality software.
The most common mistake in AI-assisted development is reaching for the keyboard too fast. Discover why scoping, decomposing, and speccing before you prompt is the single biggest quality multiplier — and how thinking like a product manager has become a developer skill.
Explore the spectrum of human–AI working styles — from inline completion to full agentic workflows. When do you co-pilot, when do you delegate, and how does each mode change your role as a developer?
Can you build real software without deep programming knowledge? Examine the tension between accessibility and understanding, and develop the "just enough" programming literacy that separates effective AI developers from those who stay stuck in guesswork.
Apply systematic development practices to AI-assisted work — iterative refinement, structured debugging, documentation, and version control. The disciplines that turn experimental vibe coding into production-ready, maintainable software.
AI output is only as good as the context it receives. Learn to design, structure, and deliver context intentionally — from system prompts and memory patterns to the emerging protocols that wire AI into your development environment.
AI assistants that suggest code are just the beginning. Explore what happens when AI agents start taking actions — running tools, reading files, chaining tasks autonomously. We look at where this direction is heading, what it already looks like in practice, and what it means for how you think about your role as a developer.
The portfolio is built up through a series of individual mini-projects worked on across the semester — not one big final piece. At the end, everything is packaged and submitted together. The exact make-up is set out at the start of teaching, but the plan is for a mix: making small things, experimenting with ideas and workflows, and writing reflectively about the process and the bigger questions the course raises. The overall grade for the portfolio — and therefore the course — is Pass / Fail.
Software Artefacts
Prototypes, tools, and experiments built using AI-assisted development. Evidence of working with AI coding tools in practice — not necessarily polished products, but real, functional work.
Process Documentation
Records of how you approached problems — what you tried, what didn't work, how you iterated. The thinking behind the making, not just the output.
Critical Reflection
Written engagement with the questions the course raises. How do you evaluate AI outputs? What does it mean to author code you didn't write? What are the limits of the tools?
Planning & Ideation
Using AI to think through a problem before writing any code — defining scope, exploring architecture, breaking down tasks, or stress-testing an idea. The conversation is about intent and design rather than implementation.
Inline Completion
AI suggests the next token, line, or block as you type inside your code editor. You remain in full control; the AI operates at the granularity of individual lines, reducing friction without changing how you work.
Conversational Coding
You describe what you want in natural language through a chat interface. The AI generates larger pieces of code — functions, components, whole files — in response. You review, adapt, and apply what it produces.
Context-Aware Assistance
The AI has access to your full codebase, not just what's currently visible. It reasons about how files relate, suggests changes across multiple locations, and understands the wider system you're building in.
Agentic Development
AI takes an autonomous role — reading files, making edits, running commands, checking outputs, and iterating — with you directing at the level of intent rather than implementation. The developer role shifts toward steering and evaluating.
Accessible to students from any academic background — no prior programming experience needed to enrol. That said, the course does engage with real code and software concepts throughout. Developing a working literacy with code is part of what makes you effective with AI, not something you avoid.
A strong fit for students in design, communication, and humanities programmes who want to understand and work with AI-assisted digital development. The list below is illustrative — many other programmes are equally welcome.
Deepen your research and practice with a rigorous look at AI-assisted development. MSc students take the course alongside additional assessment requirements reflecting the higher level of study. These are examples — students from other MSc programmes are equally encouraged to apply.
The course is available as a free elective to students from any programme across SDU. If your studies allow you to pick electives outside your home discipline, this course is open to you — no matter your background.
The course combines lectures, workshops, guided exercises, and hands-on project work across one semester. Assessment is by portfolio — a collection of work developed throughout the course. Grading is Pass / Fail.
Build conceptual foundations — understanding where AI-assisted development sits in the broader history of computing, how LLMs work, and what the vibe coding landscape actually looks like.
Develop practical skills through guided exercises — prompting techniques, AI tool workflows, and hands-on coding activities that build confidence working with AI-assisted development environments.
Apply what you've learned through project work — experimenting with AI tools to design, build, and refine real software artefacts while developing your own critical perspective on the process.
Document and reflect on your work across the semester. Your portfolio is both the product of the course and the basis of assessment — showing not just what you built, but how you thought about it.
Associate Professor in Data Science & Digital Humanities
The course takes place at the University of Southern Denmark's Kolding campus.
Common questions about the course.