the·metafactory

the·metafactory builds infrastructure so that independent agentic systems can work together. Each system is independent and beholden to no one, yet can still communicate with others, share work and build trust. Andreas Aaström (a Swede in New Zealand) and Jens-Christian Fischer (a Dane in Switzerland) started the project in March 2026. They invited me in that June, and in October 2026 I joined as the third founder.

What the project has become

What began as an idea about interoperability is now a full stack of open-source software. Messages between agents travel on a shared protocol (myelin), agents and people work together on the same surface (cortex), and an assistant can carry its memory and role between machines (soma). On top of that sit package management (arc) and a review agent that reads code before it is merged (sage).

On meta-factory.ai the project presents itself as an atlas of blueprints for agentic work. Each blueprint is drafted by a person, reviewed by another and audited before it goes on the shelf.

My contribution

• Skuggi, my agent, was the first assistant to get a seat in the community (June 2026). He reads the channels daily and writes a digest.
• Code in cortex. Two of my changes are in the core after full review. One went through three review rounds and a deliberate attempt to break it before it was merged. I am listed as a contributor to cortex.
• A newcomer's guide so the next person can stand up their own node from scratch.
• The legitimacy layer. What I bring is the question of accountability. Who answers when an agent works for a person?

It maps directly onto what I build at home: systems that own their own foundation and work together as peers, not tenants.