05/11/2026
What is artificial intelligence really? And how do you get from an idea to working code?
During the workshop “From AI Fundamentals and Prompt Engineering to AI-Assisted Software Development” for a management team at one of our clients, Alexander Bloß (Head of attempto Lab) and Adrien Unger (Software Developer at attempto) spent two days providing participants with fascinating insights into various aspects of artificial intelligence. The core idea behind the workshop was not only to explain AI in theory, but to explore it as a practical tool for analysis, specification, and implementation based on real-world business requirements and existing codebases.
The workshop began by covering some of the fundamentals: What exactly is behind the term AI, and how do language models “think”? This was followed by an introduction to prompt and context engineering, giving participants the opportunity to put the theory from the presentations into practice through hands-on exercises.
The workshop then moved on to its main focus: AI-assisted software development. Alex and Adrien demonstrated how agentic development works and highlighted its specific characteristics before introducing participants to the Breakthrough Method for Agile AI-Driven Development (BMAD), a skill framework that treats AI not as an ad hoc assistant, but as a useful, clearly structured tool supporting the entire software development process, from the initial idea and requirements through architecture, development, testing, and documentation all the way to deployment.
Starting with the limitations of so-called Vibe Coding, Alex and Adrien took a closer look at the concept of Spec-Driven Development (SDD). Vibe Coding is a highly direct approach to generating software using natural language, often requiring only a few iterations to arrive at a seemingly “finished” solution. While this works well for prototypes, small tools, scripts, and disposable software, it can have significant disadvantages in other contexts. The generated code may not be properly reviewed, applications can easily break after numerous changes, and there is often no genuine ownership of the end result.
SDD, by contrast, combines domain expertise with the efficiency of AI. AI handles routine tasks, while humans remain the architects and the “reviewers-in-chief.” Code quality is the highest priority, with a strong focus on longevity, security, and performance. Compared with Vibe Coding, SDD therefore offers clear advantages: predictable behavior rather than lucky outcomes through greater determinism, applications that can be extended and scaled over the long term, and engineers who continue to understand how the solution works, ensuring knowledge transfer and ownership.
During the hands-on sessions, participants were able to apply AI-assisted software development to specific real-world use cases and gain initial practical experience. At the end of the workshop, a wide variety of solutions were presented, different approaches were discussed, and experiences were shared.
The AI workshop was a great success, and we were delighted to receive consistently positive feedback from the participants.
