Zum Inhalt springen
AI Learning

AI-Powered Programming

Knowledge

Writing Code with AI Support

AI is changing the way software is developed -- and this affects not only professional developers. Whether you occasionally write scripts, build automations, or are programming for the first time: AI assistants make coding more accessible and productive.

This module examines AI-powered programming from the user's perspective: How do you use AI coding assistants productively? What workflows and patterns exist? And where are the limits?

iUser vs. Expert Perspective

This module shows how to use AI coding assistants productively in everyday work. If you're interested in the technical details -- Agent-Driven Development, Vibe Coding, architecture decisions -- you'll find that in the Intermediate Path: AI Coding Tools.

What AI Coding Assistants Can Do

AI coding assistants provide support across various phases of programming:

  • Generate code: Create working code from natural language descriptions
  • Complete code: Make intelligent suggestions while you type
  • Explain code: Break down existing code in an understandable way
  • Find errors: Identify bugs and suggest corrections
  • Write tests: Automatically generate test cases for existing code
  • Refactoring: Improve code without changing its functionality

Different Approaches at a Glance

There are fundamentally different ways to program with AI:

Approaches to AI-Powered Programming

(As of: May 2026)

Understanding

When to Use Which Approach?

The choice of approach depends on your experience and your goal:

Chat-based is suitable when you want to solve a specific problem -- for example, a Python script for renaming files or an Excel formula. You don't need an IDE or setup.

Inline assistance is ideal for more experienced users who work in a development environment. The AI speeds up repetitive tasks and suggests code patterns you might not have memorized.

Agent mode is suited for larger changes to existing projects. The AI understands the context of the entire project and can coordinate changes across multiple files.

Prompt-to-App is the fastest path from zero to prototype -- but the results often require refinement, especially with more complex requirements.

*The Most Important Skill: Giving Good Instructions

Regardless of the approach: the more precisely you describe what the code should do, the better results you'll get. Treat the AI like a competent but context-free colleague -- describe the goal, the constraints, and the desired behavior.

AI Code Is Not a Final Product

Generated code is a starting point, not a finished result. Good practice with AI coding means:

  • Understanding what the code does -- don't adopt it blindly
  • Testing -- does the code work for all relevant cases?
  • Checking security -- does the code contain vulnerabilities or hardcoded values?
  • Considering maintainability -- is the code readable and well-structured?

!Trust, but Verify

AI-generated code can contain errors, use outdated libraries, or have security vulnerabilities. Especially for code that handles sensitive data or goes into production, careful review is essential.

Application

Start with a concrete, small task: ask an AI to write a script that reads a CSV file and sorts the rows by a specific column. Compare the results of chat-based generation (e.g., ChatGPT or Claude) with what an inline assistant suggests in your editor. Notice how different the solution approaches can be.

Reflection

AI-powered programming changes the developer's role: instead of typing every line yourself, you become the commissioner, reviewer, and architect. The following sections dive deeper into three central aspects: AI assistants and their patterns, code review with AI, and debugging with AI support.