Latest Posts

EXCESSIVE AGENCY: WHY HUMAN-IN-THE-LOOP IS NOT ENOUGH

Tools such as Claude Code and Claude CoWork are increasingly embedded in systems where they are granted a degree of agency. They are no longer limited to generating text, but can call functions, interact with external tools, modify code, and execute multi-step workflows in response to a prompt.

WHY THREAT MODELS MATTER IN SECURITY RESEARCH ?

Over time, while reviewing and reading cybersecurity and privacy papers, I notice something that bothers me more and more: some papers claim security or privacy guarantees without clearly defining a threat model.

WHEN THE MAP NO LONGER MATCHES THE TERRITORY

In the context of my new role within the BATTwin project, I have been thinking about how we observe systems and how far observation alone can take us. Working on reliability and industrial adoption quickly reveals a recurring pattern: we see problems, but often too late.

WHY I CREATED THIS WEBSITE ?

I created this website for a simple reason: to have a place to think in public about systems I build, operate, or study.