Maximilian Marowsky

Psychology taught me how people learn. Founding a startup taught me how to ship. Game-based learning taught me that the medium IS the message. And building with Claude taught me what AI-native actually means. I’m not a PM who read about education — I’ve published research, built products, and iterated prompts until 89% of AI assessments matched human judgment.

Three things founding taught me: speed beats perfection, your first idea is almost certainly wrong, and hiring is harder than product. I learned to love iteration and to let go of ego. Building something from zero — and then letting someone else buy it — teaches you what really matters.

I want to work at the company that takes AI ethics seriously — not as marketing, but as a constraint on revenue. Turning down military contracts when the money is right there? That’s integrity. And the culture of autonomy, where people build prototypes because they’re curious? That’s how I work best.

My M.Sc. thesis studied motivation in computer science education. The finding that stuck with me: learners need to feel autonomous, competent, and connected (Self-Determination Theory). Every product I’ve built since tries to hit all three.

In Q1 2026 I led “Make Quality Visible” at eduki: an AI assessor that evaluates teaching materials across 12 criteria. I wrote the prompt myself, iterated it through 10 versions with Claude, and validated it with human reviewers. Version 10 achieves 89% agreement. The framework was co-developed with Prof. John Hattie.

eduki — Germany’s largest marketplace for teaching materials, ~150 people — acquired us in 2022. I led the product integration with a team of seven. Then I spent three years finding product-market fit as an intrapreneur, before moving to own the core commerce experience: product page, cart, checkout.

I co-founded pearprogramming — a game where students learn to code by running a virtual startup. We won a federal grant, grew to 10 people, and in 2022 eduki acquired us. I was 28 when we started. It was the hardest and best thing I’ve done.

Students didn’t just “learn to code.” They founded a virtual startup, made business decisions, and solved real programming challenges to grow it. We started with Google Blockly for visual programming, then progressed to real code. Completion rates were far above traditional e-learning — because intrinsic motivation beats curriculum every time.

A learning app for paramedic trainees with spaced repetition. A German language app for refugees with audio-first interactions for low-literacy users. A vocabulary app I built in days, not months. All built with Claude Code. All in education. I can’t stop building things that help people learn.

Schools optimize for the wrong metric. They test whether you can recall facts — but in a world with Claude, facts are free. What’s scarce is the ability to ask the right question, evaluate an answer critically, and know what to do with it.

Agency. The confidence to say “I don’t know this yet, but I know how to figure it out.” Anthropic calls it AI fluency — not just knowing AI exists, but being able to direct it, evaluate it, and adapt as it evolves. That’s the skill of the century.

Anthropic’s Education Labs isn’t building another EdTech tool. It’s studying how AI transforms human capability — part research lab, part product studio. Claude’s Learning Mode is Socratic on purpose: it refuses to hand over answers. The outcome metric isn’t engagement. It’s whether learners become more independent.

I’d want to build learning experiences where AI doesn’t replace the teacher but amplifies the learner. Systems that identify what you don’t understand yet — not just what you got wrong. Products where the measure of success is: did this person become more capable and more curious?

I co-authored a Springer book chapter on game-based learning with my wife Anna (M.Sc. Neuroscience). I worked with Prof. John Hattie — whose “Visible Learning” is the largest meta-analysis in education — on two published studies validating a quality framework for teaching materials with 2,000+ teachers.

I use Claude Code every day to build full applications. I don’t write code — I describe what I want, test the result, and iterate. Right now I’m building three learning apps. The best way to understand AI is to ship with it.

Building with Claude changed my understanding of product management. The bottleneck isn’t implementation anymore — it’s taste, judgment, and knowing what’s worth building. PMs who can’t prototype with AI will be like PMs who can’t read a spreadsheet.

I’m hypothesis-driven. I don’t ask “what should we build?” but “what do we believe, and how do we test it?” Deep discoveries before building. A/B tests to validate. And the discipline to kill ideas that don’t work — including my own.

I’m a 34-year-old new dad, specialty coffee nerd, road cyclist, and ambitious home cook. I live in Cologne with my wife Anna and our daughter Frieda. I once dreamed of becoming a chef. Now I dream of opening a cafe someday — and building things that help the next generation learn better.

Frieda was born in August 2025. Becoming a father changed how I think about education. It’s not abstract anymore — what kind of learning do I want for her? Not memorization. Not standardized tests. I want her to stay curious.

Pour-over, hand-brewed, single origin. I was a barista once. I love geeking out about extraction ratios and water temperature. Filter coffee is my love language.

I recently discovered road cycling and fell in love. Long rides clear my head. I’m outdoors as much as I can be — hiking, cycling, just moving.