Vibe Coding: Build Websites & Apps with AI

Vibe coding is a loose name for building software by describing what you want in natural language and working with an AI coding system to create it. At its best it dramatically lowers the barrier between an idea and a working prototype. At its worst it becomes “I kept asking the AI to fix things until I had no idea what any of it did.”

We teach a more useful version: enough product thinking, interface design, technical structure and debugging to stay in control of what you are building.

• Problem before solution, who the product is for, user needs and jobs
• MVP thinking and feature prioritisation
• User stories, user flows, pages and application states
• Basic data requirements, turned into a build brief

Exercise: take a broad idea and cut it down to a first version that can actually be completed.

• Browser-based builders versus local coding tools
• Projects, files and repositories; version control in plain English
• Giving an AI coding assistant project context, rules and instructions
• Planning before generation and asking the AI to explain its approach
• Building one change at a time, and recovering when a change breaks something

Exercise: create the project, define the first technical plan and generate the basic application structure.

• Layout, navigation, components and forms
• Responsive behaviour and design systems
• Interaction, empty, loading and error states
• Accessibility basics, and reviewing generated UX instead of accepting it

Exercise: build the main user journey and test it across screen sizes.

• What a database does: tables, records and basic relationships
• Authentication, user accounts and permissions
• Saving and retrieving information
• Environment variables and secrets; APIs in plain English
• What should never be exposed in frontend code

Exercise: connect the application to stored data and add at least one real user-specific feature.

• Reading an error, reproducing a bug and giving AI useful debugging context
• Logs and browser developer tools
• Testing complete user journeys and edge cases
• Basic security checks and AI-generated technical debt
• Knowing when to stop patching and rethink

Exercise: run a structured test on your product, document problems and fix the highest-priority issues.

• Deployment, domains and production environment variables
• Analytics, feedback, basic monitoring and backups
• Updating a live application and planning version two
• When a prototype needs professional engineering

Final project: a working web application with a clear user problem, a focused first version, a designed user flow, responsive UI, stored data or another real backend feature, error and empty states, a live URL and a short README.

This is a build-along course, but everybody does not have to build the same product. We teach a common method, demonstrate the important steps, then help you apply them to your own project. Expect things to break: seeing how an instructor diagnoses a messy AI-generated problem is often more valuable than a perfect demo.

AI tools change quickly. We review the teaching stack before each cohort; it may include Replit, Cursor, major AI coding assistants, GitHub and Supabase or Firebase.

AI does not turn every beginner into a senior software engineer. You will be able to build useful prototypes and small products with far more independence than before, and you will understand where your limits are.

For products involving high security, complex financial transactions, sensitive medical information, unusual scale or difficult infrastructure, experienced engineers still matter.