Introduction
Jonas Berg failed a maths exam and then failed the feeling that he could pretend it had not happened. The result sheet stayed on his desk for days. What finally moved him was not motivation, but irritation: why had nobody explained the steps in a way that matched his mistakes?
Jonas lives in Freiburg and now works as a junior developer on learning AI for vocational education. His route includes an aborted engineering degree, side jobs and a second attempt at study.
Story of the Path into AI
An open-source tutor helped Jonas understand where he had gone wrong. It did not make him brilliant overnight, but it asked questions without rolling its eyes. He began to imagine tools for people who do not move through education in a straight line.
He learned programming through tutorials, coding meetups and small daily functions rather than heroic all-nighters. The first learning bot he built dealt with fractions. It recognized common mistakes and asked counterquestions. It also made absurd errors: once it praised a learner for “simplifying” a fraction into a larger number. Jonas kept the screenshot.
Over time he understood that patience has to be designed. The bot needed to wait, rephrase and avoid making the learner feel watched or graded every second.
Current Work
Today Jonas works as a student assistant on adaptive materials for vocational schools. In tests with apprentices, he found that short explanations worked better than perfect theory. The bot now asks for the next step rather than the final answer.
Several learners use the prototype voluntarily before exams because it does not create shame. Jonas is careful with that feedback. A tool can support practice, but it cannot replace teachers who notice fatigue, frustration or a learner pretending to understand.
Personal Advice
“Build for the person you used to be,” Jonas says. He means the confused version, not the polished success story. The weaker moment often knows more about the product than the confident presentation.
Key Facts
Age and place: 21, Freiburg.
Background: study interruption, working-class family, second attempt.
Entry into AI: fraction tutor that detects common thinking errors.
Focus today: vocational education.
Typical tools: Python, language models, learning analytics.
Werkstattnotiz
Jonas keeps a folder called “kind but wrong.” It contains feedback messages that sound supportive while reinforcing an error. His current test is whether the tutor can correct without sounding disappointed. That tone, he has learned, is not decoration; it is part of the learning system.