Now
A snapshot of what I'm actually doing this month. If any of it overlaps with your world, the fastest thing is a coffee in Zürich.
last updated 2026-11-18
Building
- Automated evaluation harness for LLM and agentic systems — deterministic runs, regression suites and scored output, so agent quality is a number that gates a release instead of an opinion in a review meeting.
- Hands-on with agent runtimes: making agentic workflows fast, stable and production-ready with the same operational discipline (rollback, observability, eval gates) I used to apply to Kubernetes.
- Mapping the AI-native ecosystem in Zürich and Zug — who is building what, on which stack, and which of them will need serious compute and platform support in the next twelve months.
Thinking about
- How to tell an AI-native company from an AI-enabled one in a single conversation — and why that distinction predicts almost everything about how a partnership will go.
- Where the Kubernetes adoption curve of 2015 repeats itself for agents, and where it genuinely doesn't.
- Why inference economics, not model quality, is what quietly kills AI-native startups in year two.
Where to catch me
- 19 Nov 2026AI Agents for Science Zurich — Zürich
- 20 Nov 2026AI Breakfast #56 — Lucerne
- 26 Nov 2026open-i 2026 — Zürich
Communities
- DevOpsDays Zürich — co-founder and organiser, 2016–2019
- KubeCon + CloudNativeCon — regular attendee
- GDG Cloud Zürich @ Google
- Zürich AI Meetup · Zürich Agentic Coding · Mindstone Zürich
- Zug Tech Circle
- Founders Running Club Zürich
- Agent Economy Association