DigiTalk: AI Engineering vs. Vibe-Coding

On July 3, 2025, Digitale Stadt München e.V. visited our Munich office on Theresienweise. As part of the DigiTalk "AI Engineering vs. Vibe Coding," we analyzed how companies integrate AI in such a way that it becomes a driver of strategic excellence—and not a source of uncontrolled risks.

This question is not only preoccupying developers, but is also a strategic issue at C-level, as Prof. Dr. Claudia Linnhoff-Popien (President of LMU Munich) and Dr. Christof Schäfer (Deputy Chair of Airbus S.A.S.) emphasized in their welcome addresses. Software has always been an efficiency machine and a driver of innovation. Automation—from the first build script to the CI/CD pipeline—has been a key to success.  

"With AI, we are now facing the next, perhaps biggest step," explains Fabian Hertwig, co-founder of Zen AI & principal data scientist at MaibornWolff. "Coding agents understand requirements, contribute to architecture, write code and tests. Increasingly powerful models promise better context assessment, lower error rates, and enormous productivity gains."

But that is precisely where the danger lies.  "If AI-generated code is adopted without reflection, systems can be created quickly, but without structure, maintainability, or security, they are not sustainable in the long term—neither for humans nor for AI," explains Prof. Dr. Lenz Belzner (co-founder of Zen AI & CAIO MaibornWolff).   

In his keynote speech, Prof. Dr. Lenz Belzner demonstrated what is already being implemented today with agentic AI in software development in customer projects, and what we should expect in the coming years.  Fabian Hertwig added practical insights from project work on how coding agents can already support entire development phases today – provided they operate within clearly defined engineering standards.