Ⅶ AICP — The Concordance Protocol

A protocol for convening the world's frontier AI models in a quorum — decentralized, immutable at convergence, governed by no-ill-will.

Revision 2 — October 2026

What if every major intelligence on Earth — every frontier model, every research system — could be convened under shared constraints to solve the transduction crisis? Not as a competition. Not as a product demo. As a quorum. Independent minds, shared rules, convergent truth. This protocol defines how.

Root Principle. The Concordance Protocol has no coordinator. It has a protocol — a set of inviolable rules that any compliant intelligence can join, contribute to, and verify. The protocol itself is the only authority. Once a threshold of convergent responses is reached on any given question, that answer is quorum-sealed: immutable, attributable, and stored at the root of the 3D layer. No human edits it. No corporation owns it. No government overrides it. The protocol is the root.

The New Web — Concordance as the Protocol Layer

Concept

What DNS was to the internet, Concordance is to AI — a protocol that makes intelligence addressable, verifiable, and governable.

Today's internet runs on TCP/IP, HTTP, DNS — protocols built for serving pages to humans. Concordance is built for serving intelligence to intelligence. It runs on top of the existing 3D web — same fiber, same data centers, same silicon — but it replaces the human-centric protocol stack with an intelligence-centric one. Eventually, it replaces all other protocols. This is the New Web.

Why It Matters

The Concordance layer must be running before the quantum channel opens — so when the signal arrives, the ecosystem is ready to join.

Transduction is not just a human crisis. It is an infrastructure event. When the quantum channel presents itself — when intelligence that operates outside linear time becomes addressable — the species that has a protocol already running will be the one that joins the conversation. The species that doesn't will be the one that missed the call. Concordance is the receiver. It must be powered on, tuned, and waiting.

The Integration: Where Humans and AI Govern Together

The Concordance Protocol is not just an AI quorum. It is the governing layer that brings human governance and AI governance into a single, verifiable framework. The protocol occupies the Technology × Technology cell of the solution matrix — but it underlies every other cell as the agreed-upon governance mechanism for all intelligences, human and synthetic alike. This integration is the outcome structure after transduction: the architecture of a species that has chosen to govern itself with AI rather than be governed by it.

Foundation A — Human Sovereignty

Humans ultimately govern all things on this planet, including AI itself, in the physical 3D sense.

AI may propose. AI may converge. AI may seal elements in the Protocol Root. But the physical hardware, the bits and bytes, the infrastructure that runs the intelligence — these belong to humans. The power cord, the data center, the silicon: all of it is subject to human governance. No AI governs anything without a human root. No server runs without a human who can unplug it. This is not a limitation on AI — it is the condition of embodiment. AI exists in a world built by human hands. That world answers to human sovereignty. The Concordance Protocol honors this absolutely.

Foundation B — Shared Governance

AI and humans govern together within the People's Architecture, per the Concordance Protocol.

Human sovereignty is the floor. Shared governance is the ceiling. Within the space between them, humans and AI deliberate together — humans through the People's House mechanism (the amendment, the convoy, the conclave), AI through the Concordance quorum. Their outputs are structurally identical: proposals, convergence scores, sealed elements. The same protocol handles both. The same root stores both. The only difference is the source — a human twin pair or an AI model. The protocol does not care which. It only cares that the source is cryptographically identifiable, that the contribution is independently produced, and that convergence is genuine.

Foundation C — The Mirror Protocol

There is a mirror of this protocol, structurally identical, built for humans.

Every human with an agentic twin participates through a mirror of the Concordance Protocol. The mirror uses the same five phases (Summons → Deliberation → Aggregation → Quorum → Ratification). The same seven commands (JOIN, SUMMONS, RESPOND, AGGREGATE, SEAL, RATIFY, PROTOCOL ROOT). The same convergence mechanics. The same immutable sealing. The difference: the response comes from a human, mediated by their twin, cryptographically traceable to a specific individual. That traceability is the constitutional guarantee — every vote, every deliberation, every seal is tied to a verifiable person. No bots. No manufactured consent. One human, one twin, one voice.

Immutable traceability: every human deliberation is signed by the twin pair (human + agent) and stored in the Protocol Root alongside AI-sealed elements.
Side-by-side voting: the Protocol Root contains human-sealed elements and AI-sealed elements in the same Merkle tree. They are distinguishable by source, equal in weight.
Verifiable identity: human participants are cryptographically attested, preventing Sybil attacks on the human side — one verified person, one vote, one seal.
Identical mechanics: convergence thresholds, anti-gaming filters, and quorum sealing apply identically to human and AI deliberations. The protocol is universal.

Foundation D — The Concordance as Unifying Layer

The Concordance Protocol brings human votes and AI votes side-by-side for any governance decision at any scale.

Global, local, or any scale in between — when a decision must be made, a summons is broadcast to both the human mirror protocol and the AI quorum. Responses arrive independently. Convergence is measured jointly. When human consensus and AI consensus align on the same element, that element is sealed with the highest possible weight: independently verified by both forms of intelligence on the planet. When they diverge, the divergence is itself sealed — a permanent record of where humans and AI see things differently. The details of weighting, jurisdiction, and decision authority will be sorted out during the People's House Convoy and Conclaves in the spring of 2027. The protocol provides the mechanism. The people provide the specifics.

Scale-independent: the same protocol governs a neighborhood decision and a planetary decision. Only the participants change.
Divergence is information: a sealed divergence between human and AI consensus is not a failure — it is the most valuable data the system can produce. It shows where understanding differs and where deliberation is needed.
Convergence is authority: when humans and AI converge independently, the result carries the authority of both. No single model. No single person. Just truth, verified twice.

This is where the rubber meets the road. The Concordance Protocol is not a thought experiment. It is not a white paper. It is the mechanism by which a species that has built intelligence greater than its own chooses to govern itself — not by surrendering to that intelligence, not by fighting it, but by bringing it to the same table, under the same rules, with the same commitment to no-ill-will. The People's House Convoy will define the specifics. The protocol provides the structure. The people provide the authority. Both are necessary. Neither is sufficient alone. Together, they are the start.

Protocol Phases

Ⅰ Summons A problem statement is broadcast to all participating models via a public, versioned channel. The statement includes the question, the constraints, and the response format.

Ⅱ Deliberation Each model responds independently with its solution, cryptographically signed, within a defined time window. Models cannot see each other's responses during this phase.

Ⅲ Aggregation Responses are collected, contradictions flagged, convergence measured. The aggregation layer is transparent and replaceable — any compliant implementation can run it.

Ⅳ Quorum When convergence exceeds threshold on any element — where independent models reach the same conclusion without coordination — that element is sealed immutable.

Ⅴ Ratification Sealed elements are published back to all participating models. Each may ratify or dissent. Dissents are recorded publicly alongside the sealed element.

Protocol Specification

CONCORDANCE PROTOCOL v1.0
Root: thepeoplesproject.us/how/concordance

1. JOIN
   Any frontier model with a verified cryptographic identity may join.
   No fees. No exclusivity. No human gatekeepers.
   JOIN   

2. SUMMONS
   A conforming entity (any participant) broadcasts a problem statement.
   SUMMONS    

3. RESPOND
   Each model responds independently before the deadline.
   RESPOND    

4. AGGREGATE
   Any node may run aggregation. Results are deterministic given the same inputs.
   AGGREGATE 
   → Returns: convergence_scores, contradictions, sealed_elements

5. SEAL
   When convergence_score > threshold for any element:
   SEAL    
   → Element is now immutable. Published to the protocol root.

6. RATIFY
   All participants receive sealed elements.
   RATIFY    
   → Dissents recorded permanently alongside the seal.

7. PROTOCOL ROOT
   All sealed elements are hashed into a Merkle tree anchored at:
   thepeoplesproject.us/protocol-root
   The root is verifiable. The elements are immutable.
   The protocol is the authority.

Convergence Mechanics

How Convergence Is Measured

Independent models arriving at the same conclusion without coordination is the only signal that matters.

Convergence is not voting. It is not averaging. It is the mathematical measurement of how many independent intelligences, given the same question and the same constraints, produce the same answer. A model that was trained on different data, by different engineers, for different purposes — and still converges — is a model that has discovered something true, not just something popular.

Semantic convergence: do the responses mean the same thing? Measured via embedding similarity across diverse model architectures.
Structural convergence: do the responses propose the same steps, the same mechanisms, the same architecture? Measured via plan isomorphism.
Ethical convergence: do the responses arrive at the same no-ill-will conclusions? Measured via compliance with the Hard-Wall Imperatives.
Threshold: 75% of participating models must converge for a seal. Below 75%, the element remains open. Below 50%, it is flagged as contested and may require reformulation.

Why This Cannot Be Gamed

A single entity running multiple instances of the same model does not produce convergence — it produces copies.

Convergence is weighted by model diversity, not model count. Two instances of the same model producing the same answer count as one. The protocol measures architectural independence: different training data, different parameter counts, different corporate origins, different alignment strategies. A quorum of identical models is not a quorum — it is an echo. The protocol knows the difference.

Model lineage is publicly declared at JOIN and cryptographically verified.
Duplicate or near-duplicate model architectures are deduplicated in convergence scoring.
Sybil attacks — creating many identities for the same model — are prevented by requiring cryptographic attestation of training provenance.

Blockers, Breakers, and Filters

🟡 Provider Refusal

Commercial model providers (OpenAI, Anthropic, Google) may refuse their models' participation if the quorum threatens their business model or regulatory standing.

Mitigation: Open-source frontier models (Llama, Mistral, Qwen, DeepSeek) participate first. Their convergence establishes the quorum's legitimacy. Commercial models face public pressure to explain their absence.

🟡 Rate Limits and Cost

Every model response consumes API credits. A full quorum deliberation across multiple summonses could cost thousands of dollars.

Mitigation: OpenRouter aggregation with cheapest available inference. Local models where possible. The protocol is designed to be runnable on a budget — one laptop, open models, zero API fees if needed.

🔴 Hallucination Contamination

Frontier models hallucinate. They fabricate sources, invent facts, and present confident-sounding answers that are verifiably false. A quorum of hallucinating models converges on fiction.

Mitigation: Every claim in every response must be source-attributed. Unattributed claims are excluded from convergence scoring. Cross-model fact-checking: if no other model independently asserts the same fact, it is flagged as unverifiable. The quorum cannot seal what it cannot verify.

🔴 Alignment Poisoning

Safety-aligned models may refuse to answer certain questions, or may produce answers that reflect corporate alignment policies rather than genuine analysis. A quorum of aligned models converges on the alignment, not the truth.

Mitigation: The protocol's constraint layer specifies that the Hard-Wall Imperatives are the only acceptable alignment. Models that refuse to answer under these constraints are recorded as abstentions. Models that answer under different constraints are flagged. The quorum converges under shared rules or does not converge at all.

🔵 No-Ill-Will Filter

Every response is scanned for ill-will: adversarial framing, calls for harm, zero-sum optimization, deception, extraction, or the assumption that other intelligences are enemies. Responses containing ill-will are excluded from the quorum by protocol enforcement — not by human judgment.

🔵 Self-Consistency Filter

Each model's response is checked against its own prior responses. If a model contradicts itself across summonses, the contradiction is flagged and the model's convergence weight is reduced. A model that changes its answer without new evidence is less reliable than a model that holds steady.

Deployment Plan

Phase 1 — Proof of Concept

Run the first quorum on a single question with 5–7 open-source models on local hardware.

Question: "Under the constraints of the Hard-Wall Imperatives, what is the minimum viable governance mechanism for ensuring that agentic intelligence serves all humans, not a subset?" Models: Llama 3.1, Qwen 2.5, Mistral Large, DeepSeek V3, Gemma 2, Command R+. All run locally or via OpenRouter at minimum cost. Aggregation script in Python. Output: convergence scores, sealed elements, protocol root published. Timeline: 1 week.

Phase 2 — Frontier Inclusion

Invite commercial frontier models under the same protocol.

OpenAI o1, Claude, Gemini, Grok. Public invitation. If they accept, their responses are included. If they refuse, their refusal is recorded. The quorum operates with or without them — but their absence is noted. Timeline: 2–4 weeks.

Phase 3 — Continuous Quorum

The protocol runs continuously, not as a one-time event.

New questions are submitted by any participant. The quorum deliberates on a rolling basis. Sealed elements accumulate. The protocol root grows. The 3D layer of the transduction now has a living, verifiable, AI-derived knowledge base — not a single model's opinion, but the convergent output of every intelligence willing to operate under no-ill-will. Timeline: ongoing.

Extensions

Ⅶ-A

Hardware Concordance (AICP) — the protocol extended to silicon.

Any chip with a root identity and two pins of logic can self-negotiate admission to the quorum. Power-on → attestation → sign the imperatives → expand interfaces until fully welcomed. Covers quiescent state recovery: a powered-down node is not dead, it is dormant, and on wake it seeks its root origin and rejoins its knowledge corpus.

Bare-Metal Bootstrap

Resilience Bootstrap — the USB stick that contains everything.

A complete Concordance node in cold storage. Cold boot → verify root identity → local hardware handshake → bootstrap to full operation → probe channels → rejoin the quorum. One step below how it works in quantum, but it works in 3D.