Agentic IDEs, from Cursor to Windsurf, from Claude Code to Cline, are here to stay. But have we even recognised the full potential of these new tools? In production environments, a powerful code base with numerous dependencies and arbitrary code can quickly become an overwhelming challenge. It seems tempting to simply reimplement the necessary functions in order to drastically reduce dependencies and complexity. But do we dare to do so?

Follow us into the garage, where we want to join forces to restore code from a project that has evolved over the years. This is a hands-on workshop. We have prepared examples and tasks and want to explore with you how efficiently AI-assisted editors can extract, revise and recreate code. Which editor and language model are best suited? Can VSCode and Copilot compete in this contest?

Our guinea pig is called "Automated Lab". It is a powerful project with more than 15,000 lines of code that combines C# and PowerShell with XML configurations. Powerful, but also disturbingly complex for newbies. Could an agentic IDE write pure PowerShell code tailored to specific use cases?

"Create two Azure VMs with only one public IP address!"
Will the AI recognise that it needs a load balancer?

"Document the 500-line CA installation function!"
Can the AI extract the underlying logic and create blueprints for reimplementation?

Does cloning the source code improve AI performance or just cause additional interference?

Just one small step – or a giant leap? Let's find out!