What I'm doing now.

Technical Excellence, one essay at a time

The newest work on the bench is Technical Excellence: the craft layer underneath the architecture, walked across all nine phases of the lifecycle. Eight new essays are in so far: The Review That Earns Its Hour in Design, 404 Is a Security Answer in Develop, The Stack Aged. The Categories Didn't. in the Change phase, Not Run Is Not Passing in Test, Every Error Has an Address in Monitor, Who Pays When It Varies, on what a team should decide for itself, Money Has Rules the Framework Does Not Know, on the rules money needs in code, and The Capacity Model Fits on One Page, on the arithmetic behind how much a system can carry. The rest of the guide fills in the same way, a piece at a time.

Before it came The Curiosity Cabinet, a wing of classic problems from computer science, mathematics, and physics, rebuilt as playable exhibits, with mystery case files that train the reasoning this craft actually runs on. The room opened with the primes, and six exhibits are open now.

Alongside it, the Transitions build continues – field guides for the forks a software career actually takes – Engineer to Architect, to AI Engineer, to Manager, and to Product Manager. Each shares a shape: the mindset and values that travel with you, a research-grounded skillset map, a role map, and a short quiz to test the fit. The Transitions hub ties them together, and Engineer to Architect is the furthest along.

The AI in Practice guide is now complete – an honest, no-hype map of AI for working engineers, eighteen chapters from first principles to staying current. It stays live and maintained; the focus has moved to the paths above.

Alongside both, The Journal archive keeps growing – essays on craft, leadership, and the long game, now also delivered as a newsletter.

Learn. Excel. Lead.

A media brand and community for engineers – from first job to leadership – built around three phases of a career. Currently under construction: brand, content pipeline, and the first episodes. The chevrons are ready; the cursor is blinking.

This site itself is also a workshop project: hand-built, no framework, and lately growing an interactive self-assessment, dark mode, and a reading-experience layer.

Cloud architecture, by day

Designing compliant, cost-modelled AWS platforms for payments and FinTech – the kind that answer to GDPR, PCI/DSS, and DORA, and to the harder regulator: production. Munich is home base; the work spans borders.

The newest mandate: carrying that architecture discipline into applied GenAI – AI-assisted testing and AI-driven cloud operations on AWS, working through the practical questions of model selection, RAG, and agent pipelines. Early, hands-on, and honest about what works.

New grammars: Go and Python

After twenty-five years of thinking in Java, I am deliberately learning to think in two new languages – Go for its ruthless simplicity, Python for its reach into data. Not to chase a stack; to stay a student.

AI-assisted engineering remains the daily practicum – using agents in real workflows, testing where they genuinely help and where they quietly fail. The findings become essays.

On the reading desk: Pandas Cookbook (putting the Python to work), Co-Intelligence (on thinking with AI), and Thinking in Bets (on deciding well when you cannot be certain).

Claude helped me build this site

Not only the writing about it – the build itself. The structure, the design system, the article catalogue, even the memory behind the scenes were shaped hand in hand with Claude. A few real runs:

The totals keep climbing: as of September 2026, 363 million tokens across all my work and 101 active days. Opus 4.8 still carries the largest share, with Opus 5 close behind. The running tally – sessions, messages, models – lives on Uses.

Three sons, one garden, many books

Raising three boys in Bavaria with Hafsa – which is to say: living inside a five-language household where the children have become the teachers. Reading slowly, in a cosy corner, preferably while it rains.