Constitutional Proposal

A Constitutional Orientation for the Agent Economy

6 min read

The Twelve Articles

Canonical text, version 2026-07-16. Where any other statement of the Articles differs, this text governs.

Each names a right attaching to any person subject to coercive computational authority: any entity that is a non-substitutable chokepoint for access to money, work, housing, communication, identity, or legal standing at population scale. Coverage follows unavoidable passage and practical dependency, not an operator's label or size alone. A smaller service may qualify where affected people lack a reasonable alternative; a large but readily substitutable service may not. Together the articles constitute the constitutional floor below which computational governance lacks legitimacy.

1. Right to a receipt for any coercive computational act, at the time of the act, with five fields evaluable by a literate person or standardized verification software.

2. Right to contest any act through independent review, on a timeline proportionate to the severity of the harm.

3. Right to human review by an independent arbiter for material adverse actions above defined severity thresholds (those restricting access to funds, employment, housing, medical care, communication, or legal standing). Other adverse actions require expedited algorithmic reconsideration plus sampled audit.

4. Right to export data, identity credentials, reputation attestations, and relational context in interoperable, machine-readable formats.

5. Right to present portable credentials to any compatible system without the issuing platform's permission.

6. Right against cross-domain composition of personal records (directly or by proxy, including inferred joins) without explicit, informed, revocable consent.

7. Right to time-bounded personal records. The right to become someone new is constitutionally protected. Records of coercive authority are exempt and persist indefinitely.

8. Right to know the specific rule invoked in any adverse act, stated in evaluable terms. Disclosure is tiered: full rule and threshold to the affected party and their advocate under confidentiality; aggregate rule-invocation data published publicly; narrowly defined, time-bounded red-team exemptions subject to independent audit.

9. Right against data collection beyond the minimum necessary for the contracted service.

10. No contractual waiver extinguishes Articles 1–9. Terms of service that purport to waive receipt, appeal, exit, or portability rights are void against the constitutional floor.

11. Right to an independent record. No coercive computational authority may control the sole authoritative record of its own acts. The record must be independently verifiable and resistant to unilateral revision, suppression, or censorship. Its integrity must rest on a constraint external to the authority's control, whose defeat is publicly detectable and costly in proportion to the power the record constrains.

12. Mandatory independent audit for coercive computational authorities above the threshold of systemic significance.

Canonical construction of Article 7. After the applicable threshold, renewed adverse use requires fresh, purpose-specific justification and independent review.


The Boundary Condition

Delegation without maintained understanding produces domination by default.

The orientation is not anti-delegation. The boundary condition insists that the delegating party (the surviving principal) must retain enough understanding to evaluate whether the constitutional constraints are being honored.

A radiologist who signs an AI-generated report without the capacity to catch the system's errors has not delegated. She has abdicated. A municipal council that adopts an algorithmic sentencing tool without any member who can evaluate its classification logic has not automated justice. It has outsourced judgment to a vendor and kept only the liability. In both cases the signature assures the world that a qualified human reviewed the output. If the assurance is hollow, the signature is not oversight. It is ceremony, and ceremony is a species of fraud when it substitutes for the judgment it claims to represent.

The danger is not that delegation happens. The danger is that delegation erodes the capacity to evaluate what was delegated, and the erosion is gradual, invisible, and self-reinforcing. Each month of unchallenged confirmation makes the next challenge less likely. The competence that would detect failure atrophies precisely because the system works well enough, often enough, to make independent evaluation feel redundant. By the time the failure arrives, no one remains who can recognize it as a failure rather than an output.

The architecture works only if the beings for whom it was built remain competent to operate it, and maintaining that competence is itself a constitutional obligation, not a personal virtue.


The Wager

The orientation bets that four claims are true, and structures the bet so the world can collect if the claims are wrong.

First, that domination without a dominator is a structural condition, not a metaphor. If systems composed of individually compliant components produce no emergent interference, the claim is wrong.

Second, that verification cost is the primitive variable determining the structure of trust and therefore of power. If plausibility without proof suffices for accountability in a world of cheap fabrication, the claim is wrong.

Third, that power may not be the sole custodian of the evidence against itself. If a record controlled solely by the authority it documents remains durable, independently verifiable, and resistant to capture at civilizational scale, the claim is wrong.

Fourth, that freedom requires mercy — that perfect verification without designed forgetting produces permanent caste. If perfect verification produces no pathology of permanent exclusion, the claim is wrong.

Each claim names its falsification condition. A constitutional orientation that does not subject itself to its own critique expects belief, not evaluation, and belief is precisely the disposition computational republicanism is designed to replace.


The Choice

Two equilibria are technically achievable with tools that exist today.

In the first, coordination systems are property: owned, operated, and optimized for the builders' objectives. Exit is impractical. Voice is meaningless. The governed discover the terms of their subjection by waking one morning to find the world has already decided against them. This is the Quiet Foreclosure.

In the second, coordination is infrastructure: governed by rules that apply to builders and users alike, producing receipts that make power inspectable, preserving exit that makes voice credible, grounding the record in a constraint the receipted authority cannot issue, waive, or silently alter. Conflict persists. Politics persists. What does not persist is unanswerable coercion.

Every day without the constitutional floor in place, the first equilibrium hardens through drift and convenience.

The architecture can be specified. It cannot build itself.


The question is whether we will pay as citizens or have it extracted as subjects.


The companion document, "The Constitutional Machine: A Design Brief," specifies the definitions, institutional architecture, implementation pathway, coalition logic, and opponent model that operationalize the commitments described here. This Orientation states the rights; the Design Brief operationalizes them. Where the two conflict, this Orientation governs and the Design Brief is amended.