Model
Anything that governs a system carries a picture of it. When the picture is wrong, it pushes the wrong way — hard.
A book about metasystems engineering
Why the systems that run our work fail — and how to build ones that don't.
Every organisation builds a second system on top of its work: the dashboards, the platforms, the reviews, the compliance process — and now the AI agents that write the code and the pipelines that check it. When that layer goes wrong, it rarely breaks loudly. It keeps reporting green while the forest dies.
This book asks one question, again and again: what does the system that runs your system actually know — and who is allowed to tell it it's wrong?
Anything that governs a system carries a picture of it. When the picture is wrong, it pushes the wrong way — hard.
To govern something you have to see it, and to see it you simplify it. What the picture leaves out gets managed away.
Automation takes the easy part of a job and leaves people the hardest part — with the least practice.
The most important question about any governing layer: who may halt it, and what happens to them when they do.
Layers stack: product, factory, platform, agents. Each jump creates a layer somebody has to govern.