Specification Architecture

The AIEP ecosystem contains 700+ specifications organised across eight coordinated architectural layers. Of these, 137 specifications are published as open-source prior art disclosures under Apache 2.0. The remainder are either patent pending (● FILE), proprietary trade secrets (◆ TS), or under review (HOLD).

This page describes how those layers relate to each other, and where each set of OS specifications sits in the full-stack architecture.

Full-Stack Layer Diagram

                    APPLICATIONS
                         │
              Forecast, Enterprise Apps
                         │
                         ▼
                    AGENT LAYER
                         │
                     PIEA Agent
                         │
                         ▼
                  REASONING LAYER
     ┌──────────────────────────────────────────┐
     │ Replayable Reasoning Chains              │
     │ Semantic Branch Detection                │
     │ Dissent Signal Generation                │
     │ Hypothesis Forking                       │
     │ Multi-Model Reasoning                    │
     └──────────────────────────────────────────┘
                         │
                         ▼
                   EVIDENCE LAYER
     ┌──────────────────────────────────────────┐
     │ Evidence Discovery                       │
     │ Evidence Hashing                         │
     │ Evidence Mirrors                         │
     │ Evidence Index                           │
     │ Trust Scoring                            │
     └──────────────────────────────────────────┘
                         │
                         ▼
                   KNOWLEDGE LAYER
     ┌──────────────────────────────────────────┐
     │ Knowledge Graphs                         │
     │ Temporal Knowledge Reassembly            │
     │ Jurisdictional Knowledge Segmentation    │
     │ Persistent World Models                  │
     └──────────────────────────────────────────┘
                         │
                         ▼
                   GOVERNANCE LAYER
     ┌──────────────────────────────────────────┐
     │ Confidence Tier Assignment               │
     │ Evidence Sufficiency Evaluation          │
     │ Compliance Certification                 │
     │ Meta-Governance Reasoning                │
     └──────────────────────────────────────────┘
                         │
                         ▼
                INFRASTRUCTURE LAYER
     ┌──────────────────────────────────────────┐
     │ Distributed Evidence Index               │
     │ Global Mirror Network                    │
     │ Sovereign Knowledge Nodes                │
     │ Protocol Registries                      │
     └──────────────────────────────────────────┘
                         │
                         ▼
                   HARDWARE LAYER
     ┌──────────────────────────────────────────┐
     │ Trusted Node Hardware                    │
     │ Secure Evidence Storage                  │
     │ Attestation Infrastructure               │
     └──────────────────────────────────────────┘

Layer Summary — OS Specifications

Layer	Spec Range	OS Specs	Key Protocols
Layer 0 — Filed GB	P01–P09	P01–P09	Constitutional substrate, canonical primitives, plausibility, probability, quantum alignment
Layer 1 — Evidence Foundation	P10–P24	P10–P25	Normalisation, divergence detection, temporal stitching, source verification
Layer 2 — Machine Mirror	P60–P79	P60–P72	Well-known manifests, mirror generation, evidence site maps, innovation ledger
Layer 3 — Determinism Kernel	P30–P59	P30–P44, P46–P57	GoalVector, dual-ledger memory, recall, divergence attestation, compliance
Layer 4 — Network & Market	P133–P145	P133, P134, P140, P142	Evidence index, mirror generation, evidence market, global mirror network
Layer 5 — Swarm & Cognitive Pattern	P90–P98	P90–P98	Swarm consensus, cognitive pattern, reasoning fingerprint
Layer 6 — AGI & Protocol Extensions	P80–P132	P102–P132	Swarm consensus, retraction, challenge, dissent, PIEA surface, compliance
Layer 7 — AGI Cognitive Architecture	P200–P259	P201–P207, P211–P217, P220–P252, P255–P258 (OS) · P200, P209, P210, P264, P265, P269 filed (FILED-GB)	World model, reasoning memory, multi-agent, federated knowledge, compute governance

Layer 0 — Filed GB (P01–P09)

These specifications underpin the core AIEP protocol. All nine are filed with the UK Intellectual Property Office (priority date 20 November 2025, GB2519711.2 et al.).

Spec	Description
P01	AIEP Constitutional Substrate
P02	Quantum Alignment Layer
P03	Plausibility Matrix
P04	Probability Certification Engine
P05	Canonical Record Protocol
P06	Instruction–Evidence Binding
P07	Schema Governance Protocol
P08	Evidence Presentation Layer
P09	Hardware Canonicalisation

Layer 1 — Evidence Foundation (P10–P24)

Governs how evidence is acquired, normalised, and verified.

Key mechanisms: canonical normalisation (P10), evidence stitching (P17), temporal divergence detection (P16), source quality scoring.

These specifications form the lowest buildable layer of the open-source protocol surface. They are prerequisites for every higher layer.

Layer 2 — Machine Mirror (P60–P72)

Governs how AIEP-compliant publishers expose machine-readable discovery endpoints.

The /.well-known/aiep/ endpoint tree defined in this layer allows AI agents, crawlers, and protocol validators to programmatically discover protocol-conformant sites.

Key mechanisms: site manifest (P61), metadata record (P62), mirror record (P63), innovation ledger (P64), well-known endpoint builder (P71/P72).

Three repositories implementing this layer are available now: aiep-mirror, aiep-hub-validator, aiep-well-known.

Layer 3 — Determinism Kernel (P30–P59)

The computational governance core.

Governs how AIEP systems maintain deterministic, replayable reasoning. Includes:

GoalVector (P50) — reasoning trajectory stabilisation
Dual-Ledger Memory (P80) — dual-ledger evidence and reasoning record substrate
Deterministic Recall (P83) — controlled reconstruction of archived branches
Divergence Attestation (P46) — cross-node verifiable divergence proofs
Compliance Certification (P92/P93) — governance certificate generation

The mechanisms in this layer are why AIEP reasoning is replayable and cross-node equivalent.

Layer 6 — AGI & Protocol Extensions (P80–P132)

A broad set of protocol extensions covering:

swarm consensus (P90)
cross-node retraction (challenge/dissent)
PIEA agent surface (P116–P128)
multimodal evidence ingestion (P119)
governed output generation (P120)
semantic interpretation branches (P128)

The PIEA agent implements the full P116–P128 surface in production.

Layer 7 — AGI Cognitive Architecture (P200–P259)

The Phase 2 open-source specifications. These define the cognitive substrate for governed AI systems:

Group	Specs	Function
Causal World Model	P201–P207, P220	Entity state registry, temporal event graph, counterfactual timelines, hypothesis simulation
Cognitive Toolkit	P211–P217	Tool synthesis, self-model, resource allocation, meta-reasoning, safety constraints, federation
Reasoning Protocols	P221, P227–P241	Trace schema, anomaly detection, governance policy, federated trust, epistemic budget control
Replay & Attestation	P244–P252	Hardware attestation paths, replay kernels, branch checkpointing, deterministic scheduling
Dev & Ingestion	P255–P258	Proof verification, provenance-bound cache, dev environment, multimodal ingestion

Trade secret (not published): P218, P219, P222, P224, P225, P226, P242, P243, P253, P254, P259 — operational mechanics permanently confidential.

Filed (UK IPO, 08 April 2026): P200 (GB2608060.6), P209 (GB2608061.4), P210 (GB2608062.2), P264 (GB2608063.0), P265 (GB2608064.8), P269 (GB2608066.3) — patent applications filed; specifications remain under delayed publication.

OS (Gate 1 confirmed — published): P208 (Long-Term Reasoning Memory Engine), P223 (Goal Arbitration Engine) — reclassified open-source in v2.0; published as defensive prior art (Gate 1 confirmed April 2026).

Specification Scale — OS vs Full Portfolio

Category	Count
OS (open-source prior art)	120
FILE (patent pending)	~40+ choke-point patents
TS (trade secret)	Permanent — not disclosed
THESIS (conceptual)	P600–P903 — architecture context only
Total ecosystem	700+

The OS specifications establish defensive prior art across the open protocol surface. They are the complement to the patent moat, not a substitute for it.

View All OS Specifications

Browse the full specification index →

Each specification is published at a stable canonical URL: https://aiep.dev/specs/{spec-id}.

IP and Filing Context

Priority date: 20 November 2025 (GB2519711.2 et al.)
PCT filing deadline: November 2026 (Paris Convention 12-month window)
OS specifications: published as prior art disclosures under Apache License 2.0
Phase-2 FILE specifications (P200, P209, P210, P264, P265, P269) filed 08 April 2026 — under delayed publication
Remaining FILE and TS specifications remain under embargo pending filing

See Architecture for the technical stack detail and IP/Licensing for access tiers.