About

I work on Cognos Analytics, open-source test automation, and fast operational dashboards — systems that must stay clear under load. I speak for Cognos User Group DACH, AIDA, and IBM TechXchange.

Right now I run parallel AI agent loops on real product repos — Karpathy-style autoresearch with keep/discard gates, overnight Cursor improve ticks, and local Ollama strategy search. The active set includes mac-stats, ai-stock-checker, npp-rust, and Satisfecho/POS, with watchdog and GitHub-watch loops beside them. That fleet burns more than five billion tokens a month. The volume is unusual; the point is continuous, measured shipping — not chat demos.

The path started before the public web: FidoNet sysop of Over_the_Edge in Bremen (nodelist 1992–1994), then Bremen’s early ISP years as co-founder of Internationale Stadt Bremen / ISB (is-bremen.de, isb.net). That era also includes EMAF and a 1997 Prix Ars Electronica honorary mention for programming on *Energized Gaming Culture*, printed in the Cyberarts 97 catalog.

Expertise

AI agent loops and autoresearch harnesses, Cognos Analytics and REST APIs, test automation (COMETA), dashboard engineering, system architecture, early internet / ISP operations

Philosophy

Prefer tools you can own, host, and understand. Keep the stack simple. The open internet — from BBS echoes to self-hosted services — is still the point. Agents earn their keep only when they verify, discard, and ship.