I build with AI, study how it behaves, and question what it should be trusted to do.
I'm based in Toronto. I use AI every day to turn ideas into working tools, examine the decisions behind those tools, and develop practical ways to think about governance, risk, and human control. Everything here is personal and independent.
Curiosity first. Structure close behind.
I like questions that are technical enough to require careful investigation and practical enough to end in a decision.
I grew up in Germany and was the teenager who built his own computer piece by piece. I wanted to know what sat underneath the interface, how the parts fit together, and what changed when one part was replaced. That instinct still shapes how I approach technology.
I later studied business law and spent seven years in Frankfurt before moving to Toronto. Law trained me to ask a different set of questions: who made the decision, what authority supported it, what evidence remains, and who can challenge the result?
Generative AI brought those interests together. It gave me a way to turn more ideas into working products, while raising harder questions about behaviour, reliability, responsibility, and control. A morning briefing, a communication coach, and a governance assessment tool all started as useful personal problems. Building them became a way to investigate the systems themselves.
The contribution I try to make
I am most useful when a question sits between product ambition, technical possibility, and responsible execution.
Frame the real decision
Turn a broad AI question into a clear problem, a set of tradeoffs, and a decision someone can actually make.
See the Oversight Threshold →Build enough to learn
Move from an idea to a working prototype, then use the result to find the product questions that a concept alone cannot reveal.
Review the case studies →Make controls inspectable
Name where a control sits, who owns it, what it should prevent, and which evidence would show that it worked.
Open Agents that Spend →Explain without flattening
Make a difficult idea understandable while preserving the uncertainty, limits, and source trail that give it meaning.
Explore the research →Three ways into my thinking
These pieces show how I move between research, product decisions, control design, and working software.
The rabbit holes I am in now
Calm, clarity, and depth in a noisy field
AI can make everything feel urgent: another model, another breakthrough, another claim that changes everything. I made this space to slow the conversation down enough to understand an idea properly and leave with a useful next step.
Publishing is part of how I learn. A build is stronger when I have to explain the product decisions behind it. A framework is stronger when the sources, assumptions, limits, and open questions remain visible. The goal is not to sound certain. It is to make the reasoning clear enough that someone else can inspect it, challenge it, and improve it.
Travel changes the frame
I travel whenever I can. New York, Moscow, Cairo, and London stayed with me less for the landmarks than for the people and the different assumptions each city makes about how life should work. Seeing those assumptions up close is a useful reminder that the defaults around us are choices, even when they feel inevitable.
Difficult AI questions are usually the interesting ones.
I am especially interested in thoughtful work where AI, product decisions, governance, and risk meet. If that overlaps with what you are building, if something here raised a useful disagreement, or if you see a better way to test one of these ideas, I would like to hear from you.