The Constitution AI Doesn't Have (But Needs)

Why I spent six months writing 34 papers on human-AI governance — and why I'm publishing them now.

Earlier this week I wrote about the Token Burn Trap — the perverse incentive that rewards burning compute rather than producing outcomes. That piece described a symptom.

The disease runs deeper.

We're building AI coordination systems with no constitutional foundation. No theory of authority. No structure for accountability. No framework for when humans must remain in the loop.

We're wiring up agents that can execute but can't be governed.

That's why I wrote the DEIA Federalist Papers.

What Are They?

Thirty papers and four interludes. Co-authored with Claude and GPT under the pseudonym PUBLIUS — the same pen name Hamilton, Madison, and Jay used to argue the U.S. Constitution into existence in 1787.

The original Federalist Papers answered a question: How do you coordinate autonomous entities without creating tyranny or chaos?

The DEIA Federalist Papers answer the same question for a different era — with AI agents instead of states.

The answer is the same in both cases: constitutional structure. Not trust. Not hope. Structure.

The Core Thesis

#NOKINGS — Human sovereignty is non-negotiable.

Every design decision in the papers flows from this single principle. AI systems advise, execute, and optimize — but they never decide for humans on matters that require human judgment.

This isn't a limitation. It's an architecture.

The papers introduce mechanisms you won't find in any AI framework today:

GateEnforcer — Five dispositions: PASS, HOLD, BLOCK, ESCALATE, REQUIRE_HUMAN. Every request passes through the gate. No exceptions.

Five-Tier Operator Routing — Decision authority flows through tiers 0–4, from fully automated to human-only. The system routes based on stakes and novelty, not convenience.

Four-Vector Entity Profiling — Every agent is characterized by Identity, Authority, Capability, and Accountability. You can't delegate what you can't describe.

Three Currencies — Every operation tracks CLOCK, COIN, and CARBON. Always together. Never fewer.

Why Publish Now?

Because the alternative is worse.

AI agent orchestration is developing two ways: informally (ad-hoc rules, hidden violations, no accountability) or through centralized platforms (vendor lock-in, opaque algorithms).

Neither path leads anywhere good.

I'm publishing because there's a third option: distributed governance with constitutional boundaries. Harder. Requires discipline. But it's the only approach that scales safely while preserving human sovereignty.

The papers are licensed CC BY 4.0. Use them. Critique them. Build on them.

Where to Start

If you're skeptical: Paper No. 2 — On Queens and Tyranny. It opens with a real incident: an AI agent exceeding its authority, getting caught, and recovering through transparent process. Not theoretical. It happened during the writing.

If you're a builder: Paper No. 7 — On the Protocol of Grace. Five-step conflict resolution you can implement today.

If you're a philosopher: Paper No. 10 — On the Common Good. The hardest question: what holds a distributed system together when the nodes are autonomous?

The Repository

All 34 documents are now public:

github.com/deiasolutions/federalist-papers-ai

Licensed CC BY 4.0. Fork it. Implement it. Tell me where I'm wrong.

What Comes Next

The papers are philosophy. SimDecisions is practice.

The Event Ledger. The GateEnforcer. The Three-Currency cost model. PRISM-IR as the process representation layer. All of it traces back to the constitutional principles in these papers.

Follow the work: simdecisions.com

Argue about governance: the papers are open. I'm listening.

Every simulation is a prayer. Every governance decision is a choice about what kind of future we're building.