Background
Twenty years building digital products, across a lot of different industries. The last ten on productivity software, mostly the kind you use every day without thinking about it.
I am CTO of Umane, an athletic health club in Berlin where members train and recover the way professional athletes do, with coaches, physiotherapists and sports scientists around them. My side of it is the platform: a digital twin of each member built from their diagnostics, training and biomarkers, and the AI that turns all of it into what they should do next. The point is to give one person the kind of attention a pro athlete gets from a full staff.
urbanvention
My first company was urbanvention, a rapid-prototyping agency in Berlin. We built somewhere between 150 and 200 prototypes, MVPs and web apps, mostly for startups trying to get off the ground. One was The European, then the German answer to the Huffington Post.
Some of it was work almost nobody was doing in a browser yet. Carhartt WIP, David Lynch, Hering Berlin: highly interactive sites with a real design language behind them, built before the tools made that easy. The detail was the job. Some came through Edenspiekermann, Erik Spiekermann’s agency, where we built what they designed, Kearney and Otto Bock among them. Others we won ourselves.
Wunderlist and Microsoft
I was lead engineer on Wunderlist, the Berlin to-do app. Microsoft acquired the company in June 2015, citing more than 13 million customers who had created over a billion to-dos between them. My work was mostly on the backend, and specifically on sync.
After the acquisition I stayed at Microsoft as a Principal Product Manager, working on product and strategy for task and memory experiences across the company. Wunderlist’s team went on to build Microsoft To Do.
informed
In January 2021 I co-founded informed with Martin Kaelble, a business journalist from Capital, and Axel Bard Bringéus, who had spent six years at Spotify as global head of markets. 468 Capital led our pre-seed, and HV Capital led a €5M seed in April 2022.
The premise was that good journalism was becoming unaffordable to exactly the people who wanted it most. We bundled paywalled reporting from The New York Times, the Financial Times, The Economist, The Washington Post, The Guardian and Reuters into human-curated reading lists under one subscription, curated by people rather than an algorithm. The point was to work against filter bubbles instead of building another one.
informed has since been sold and closed. I learned more from that than from the acquisition. The lesson was distribution: we built something people liked and could not get enough of them to find it.
Now
Alongside Umane I run AI workshops and hands-on sprints through Supercharged, the practice I co-founded with Ole Tillmann. He works with leadership on strategy and adoption. I build the technical side, the systems that have to keep running after everyone goes home. And I write What Actually Works every week.
The work is mostly unglamorous. Wiring an agent into a real workflow and finding out it breaks on the third run. Working out what a task costs once you run it a hundred times instead of once. That is what I publish.
Why I write
The gap between a demo and something you can rely on Monday morning is the whole subject. Most of what gets published about AI is opinion from people who have not shipped anything with it. The failures are the part I want, and the part nobody writes down.
I try to explain complicated things simply. That is easier when you built the thing, because you know which details you are allowed to drop.
Elsewhere
LinkedIn · GitHub · X · Email · RSS
Outside work I write short stories and DJ badly. Neither is on this site.