Three wishes for the code

The powerful AI systems behind Vibe Coding are the result of highly complex computer science expertise

2026/04/01 by

Vibe Coding presents a significant opportunity for society! However, we will only truly master these tools, rather than be mastered by them, if we have a strong foundation in computer science.

There is an old tale from “One Thousand and One Nights”. Aladdin finds a dusty lamp, rubs it, and a powerful genie appears, ready to grant a wish. This is how Vibe Coding feels – and it is currently transforming the world in a fundamental way. Rather than laboriously instructing a computer through programming in formal languages, people are now describing the desired result in natural language, and an AI system writes the code.

A response to demographic change

This is far more than just a technical gimmick; it is a response to demographic change. Given the massive shortage of skilled workers, it is clear that we will never be able to manage the digital transformation with traditional programmers alone. Digital projects will no longer fail due to a lack of developer capacity; they will be implemented directly on site in companies, public authorities, and schools.

Digital projects will no longer be held back by a lack of development capacity, but will be implemented directly on the ground in businesses, public authorities and schools.

However, this is precisely where fascination can turn into danger. Depending on who makes the wish, the genie story can unfold in two ways. In the horror film version, the creature is maliciously literal. Anyone who wishes for 'a million euros' may be horrified to find that the money comes from a bank robbery, with the police already at the door. The genie fulfils exactly what is said, but not what is actually meant.

Coding without a background in computer science is a horror film

Vibe coding without computer science is exactly such a horror film. The person making the wish is naive, believing that the 'power' of AI alone is enough. The result is code that works until it stops working. You build systems that crash under load or expose user data. To cut corners, the AI might choose the simplest path rather than the best one. This will have to be paid back later with high interest. The system becomes unmaintainable and extensions become impossible. Ultimately, the entire project collapses under its own complexity.

You must be able to read, evaluate and understand the generated code

On the other hand, Vibe Coding with a background in computer science is like the story of Aladdin. It has a happy ending because Aladdin learns to make wise wishes. He knows enough about the world to assess the consequences of his actions. This is precisely where the foundation of a modern computer science education lies: one must be able to read, evaluate, and understand generated code. As Edsger W. Dijkstra, a pioneer in the field, observed: „[…] when judging the relative merits of programming languages, some still seem to equate “the ease of programming” with the ease of making undetected mistakes".

The powerful AI systems behind Vibe Coding are not the result of wishful thinking, but of highly complex computer science expertise.

Without computer science, there would be no Dschinnis. Computer scientists like us are needed to programme Dschinnis in the first place. The powerful AI systems behind Vibe Coding are not created through wishful thinking, but through highly complex computer science expertise. In short, while Vibe Coding enables many people to use the lamp, society still needs computer scientists to create the 'magic' of the lamp and ensure its security. Only then can we remain masters of the magic lamp.

Kristian Kersting is a professor of Artificial Intelligence and Machine Learning in the Department of Computer Science at TU Darmstadt, where he heads the AIML Lab. Since 2022, he has also headed the ‘Foundations of Systems AI’ research department at the German Research Centre for Artificial Intelligence (DFKI) in Darmstadt. He is a founding co-director of the Hessian Centre for Artificial Intelligence hessian.ai. He is also the co-spokesperson for the Reasonable AI Cluster of Excellence and a member of the Adaptive Mind Cluster of Excellence and the ELLIS Unit in Darmstadt.

Kersting studied and obtained his PhD at the University of Freiburg (2006), subsequently conducting research at MIT, Fraunhofer IAIS, the University of Bonn and TU Dortmund, before being appointed to TU Darmstadt in 2017. His research focuses on statistical relational AI, neuro-symbolic AI and probabilistic machine learning.

He is a Fellow of AAAI, EurAI, ELLIS and AAIA, received the inaugural German AI Prize (worth €100,000) in 2019 and has published over 200 scientific papers.

Persönliche Webseite