AI350: Vibe Coding for Developers

Agenda:

• Getting Started & Mindset
  • What “Vibe Coding” is – and what it is not (speed vs. diligence)
  • Strengths/weaknesses of LLMs in day-to-day development (reality instead of hype)
  • Quality definition: “Done” means running, tested, reviewed, secure
    
    
• Prompt Patterns Developers Actually Need
  • Spec First: formulate requirements, acceptance criteria, edge cases cleanly
  • Plan First: have the model propose architecture/steps first
  • Constraints & Guardrails: lock down tech stack, style, performance, security, compatibility
  • Diff/patch requests instead of “redo everything”: controlled changes to existing code
  • “Explain then implement” vs. “Implement then explain”: when each pattern fits
    
    
• From Zero to Feature: Scaffolding & Implementation with AI
  • Generate structure (modules, classes, files, interfaces)
  • API design & data models: fast proposals, then hard reviews
  • Iteration loop: small increments, frequent compile/run/check cycles
    
    
• Testing with AI
  • Test strategy: unit/integration, “happy path” and edge cases
  • AI as a test idea generator: better coverage, targeted negative tests
  • Debug loop: provide error messages/logs with proper context, validate hypotheses
  • Anti-pattern: tests that the model “prompts green” instead of verifying real quality
    
    
• Refactoring, Code Review & Maintainability
  • Refactoring prompts (readability, decoupling, reuse, naming)
  • Review prompts (bugs, race conditions, error cases, API misuse, performance)
  • Documentation & comments
  • Team conventions: making AI-driven changes traceable (commits, PR descriptions)
    
    
• Security, Data Protection, IP – Practical Guardrails
  • What can go into prompts (and what cannot): secrets, customer data, proprietary code
  • Handling dependencies: new libraries only with justification/review
  • “Trust but verify”: security checks and linting
  • Prompt checklist: spotting risks (injection patterns, insecure defaults, input validation)

Objectives:

• Use AI as a productive pair programmer – in a controlled and reproducible way
• Formulate requirements so that correct code can be produced (including edge cases)
• Apply reusable prompt and workflow patterns
• Accelerate testing and debugging with AI without losing quality
• Use AI for refactoring/review to increase maintainability
• Identify and avoid security and data protection risks in AI workflows

Target audience:

The training AI350 Vibe Coding for Developers is targeted at:

• Developers (Backend, Frontend, Full Stack), Tech Leads, Senior Engineers
• Development teams that want to standardize AI assistance in daily operations

Prerequisites:

For participation in the course AI350 Vibe Coding for Developers the following prerequisites are required:

• Practical experience in software development
• Familiarity with Git and basic test/build workflows
• No AI prior knowledge necessary (will be built up in the course)

Description:

The workshop AI350 Vibe Coding for Developers is a practice-oriented developer workshop that shows how to use generative AI (LLMs) as a “coding copilot” in a way that increases speed without compromising code quality, security, and maintainability. By “Vibe Coding” we mean an iterative way of working: quickly turning ideas into running code, refining it together with the model, consistently following up with tests/checks, and hardening the result as in a real review.

The course AI350 Vibe Coding for Developers provides a robust working model for AI-assisted programming: when AI is particularly strong—and where it systematically falls short. Participants learn reusable prompt and workflow patterns that turn “Generate code for me” into a controlled engineering process: Spec-First, Plan-First, Diff-Requests, Constraint-Prompts, Test-First loops, as well as Review and Hardening prompts.

At the core is an end-to-end exercise case: the group builds or extends a small component/service/feature and uses AI throughout the entire cycle—from requirements and design through implementation to tests, documentation, code review, and hardening. By the end, participants have a personal “Vibe Coding” toolkit that can be used directly in day-to-day team work.