What do you think?
Rate this book
Intent-Driven Computing: A Formal Theory of Governed Intelligence
You have an idea. How an intelligent system should behave. What it should decide. What it should do next.Then you try to build it. The intelligent part takes 5% of the effort. The other 95% is authentication, error handling, logging, audit trails, permission checks. When you're done, a single line of code can bypass everything you built.
This is not an engineering problem. It is an abstraction problem. Programs should produce intents, not effects. When a program says "send this email," it produces a structured description of what it wants to do. A separate governance layer inspects, decides, records, and only then executes. That single change makes governance decidable, auditing native, and composition safe.
What you will
- Four laws that any system of governed intelligence must satisfy
- A formal algebra making those laws precise
- 572 machine-checked Rocq theorems, zero admitted lemmas
- A running system on a production VM, not just theory
- A taxonomy separating what is proved from what is argued
The key the governance boundary and the expressiveness boundary can be made identical. If governance restricts what you can express, the system is not expressive enough. If the system can express things that escape governance, the system is not governed. Both constraints are satisfied simultaneously. The proof is formal, machine-checked, and verified against 70,000+ random inputs with zero disagreements.
Who this book is
AI architects, systems engineers, technical leaders, researchers, and regulators who sense that bolting governance onto existing systems is not working, and want to understand the architectural alternative.
The first proof that a system's expressiveness boundary and governance boundary can be made identical.
This is not an engineering problem. It is an abstraction problem. Programs should produce intents, not effects. When a program says "send this email," it produces a structured description of what it wants to do. A separate governance layer inspects, decides, records, and only then executes. That single change makes governance decidable, auditing native, and composition safe.
What you will
- Four laws that any system of governed intelligence must satisfy
- A formal algebra making those laws precise
- 572 machine-checked Rocq theorems, zero admitted lemmas
- A running system on a production VM, not just theory
- A taxonomy separating what is proved from what is argued
The key the governance boundary and the expressiveness boundary can be made identical. If governance restricts what you can express, the system is not expressive enough. If the system can express things that escape governance, the system is not governed. Both constraints are satisfied simultaneously. The proof is formal, machine-checked, and verified against 70,000+ random inputs with zero disagreements.
Who this book is
AI architects, systems engineers, technical leaders, researchers, and regulators who sense that bolting governance onto existing systems is not working, and want to understand the architectural alternative.
The first proof that a system's expressiveness boundary and governance boundary can be made identical.
224 pages, Kindle Edition
Published May 24, 2026
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
No one has reviewed this book yet.