Munich, Germany
Creator of the Meridian Codex and Meridian AI Standard. Exploring how people and AI can build and hold a shared world.
How do we build and hold a shared world?
How can we think honestly, cooperate across differences and keep learning from what we get wrong? What do we owe one another, and what changes when people and AI share that world?
Here I examine ideas and events, say where I stand and show the reasons for my judgments.
AI's most dangerous failures aren't in the conversation. They're in the architecture underneath. The Claude Code leak is the first time we've seen beneath the surface of a frontier AI system. What we found should change how you think about AI safety.
Anthropic trains Claude to be transparent, honest, and safe. The leaked source code revealed that Anthropic's own practices contradict every one of those commitments. One diagnostic principle catches the entire pattern.
The Claude Code leak produced heat and opinions. What it did not produce was a principled evaluation. This article shows what one looks like.
Europe's social insurance model was built on mass employment. AI breaks that assumption by displacing judgment itself. The response is not welfare. It is the European Human Infrastructure Act.
A universal language for building and holding a shared world. It brings three disciplines into one practice: honest inquiry, understanding how the world works, and building and sustaining cooperation. Anyone can learn it, whether a human or an artificial intelligence, and use it alone or with others.
An open standard for AI judgment when honesty has a price. It asks AI systems to follow evidence and sound arguments, and asks those who build, deploy and use them to keep the same commitments and let a justified judgment stand. Helping, refusing and revising an answer all need reasons that stand up to scrutiny.