The mission
Through an open, machine-executable standard, we envision a world where machines execute exactly the authority people grant, while humans retain the power to authorize it, decide the cases it reserves, and remain accountable for the outcomes.
Why
For centuries, institutions have written their authority down: policy, regulation, standards, contracts. That prose was written for people to interpret, and it quietly relied on what people bring to it: time to read, judgement to fill the gaps, and someone to ask.
Machines now act at a speed and scale no institution was designed for, and they do not stop to ask. Guidelines and after-the-fact filters cannot bound what a machine does at the moment it acts. When authority stays in prose while execution moves to code, the rules stop governing the decisions they were written for.
The discipline asks
What has to be true of that writing for a machine to run it, and for a stranger to check the run?
Answering it replaces soft promises with boundaries that hold: authority that is sealed, versioned, evaluated against proven facts at the moment of action, and checkable by anyone, without trusting the machine that acted.
What the mission means
A · the granted authority
Capability is not authority. A system acts within a sealed, versioned grant, and nothing outside it.
The Issuer
People write and approve the grant. Every version names who approved it, and no machine can issue one to itself.
The Escalation Target
Where the rules say a decision is not the machine’s to make, it returns to a named person, for that exact action.
The Sovereign
A machine cannot be held to account. Every action traces to the authority that permitted it and the people who granted it, and that authority can be withdrawn.
As machines grow more capable, the authority stays human.
The full argument: What remains human (R-18) →The initiative
This site exists so that someone can spend ten minutes here and leave understanding a concept they did not have a name for when they arrived.
Institutions are about to delegate an enormous amount of decision-making to machines. The rules those machines will operate under are currently held in documents that no machine can read, interpreted by people whose reasoning is never recorded, and revised in ways that cannot be traced.
That gap has a name now: the authority layer. Building it is a discipline in its own right — part legal drafting, part systems engineering, part verification. This initiative documents that discipline as it emerges: the vocabulary, the failure modes, the open problems, and the worked examples.
This is not a product site. NOMOS Protocol is one implementation of the framework described here, and it appears where implementations belong — at the end, after the ideas. If the framework is right, other implementations should follow, and the discipline should outlive all of them.
This initiative is maintained as a public field guide for Computable Authority. When referencing it, cite the project name, page title, retrieval date, and URL so definitions and examples can be traced as the discipline evolves.
Suggested citation formats — plain text and BibTeX, for the discipline as a whole and for individual pages — are on the citation page. A machine-readable citation is also published as CITATION.cff.
The written content on this site — the Manifesto, the Laws, the syllabus, the research library, the glossary, and the illustrative examples — is licensed CC BY 4.0: reuse and adapt freely, with attribution.
Instruments and internal policies worth decomposing in public.
See current examples →