AIEP — One Page Explainer

AIEP (Architected Instruction & Evidence Protocol) is a protocol that links instructions to evidence and publishes knowledge artefacts in a way machines can retrieve and verify.

The future of information retrieval is not search. It is evidence-backed knowledge retrieval.


The problem in one sentence

The web is full of assertions. Almost none of them carry their evidence. Machines retrieving from the web cannot tell the difference between a well-evidenced claim and a confident guess.


What AIEP does

CapabilityWhat it means in practice
Links instructions to evidenceEvery claim carries a reference to the artefacts that support it — not as a footnote, but as a machine-readable, hash-verified field
Publishes structured artefactsKnowledge is published as schema-conformant JSON under /.well-known/aiep/ so any AI system can retrieve and validate it
Makes provenance machine-readableWho published it, when, what hash it carried at publication — all encoded in the artefact, not inferred from the page
Preserves dissentOutlier claims are archived with their evidence intact rather than discarded — ready for recall when knowledge shifts
Gates execution on evidenceBefore acting on a claim, a conformant AI system checks plausibility (registry-verified) and probability certification (fail-closed tail-risk bound)
Operates at hardware levelThe GENOME kernel enforces four constitutional invariants at the execution substrate — evidence binding, hash parity, plausibility clearance, dissent completeness

The stack at a glance

LayerWhat it doesKey patents
Web surfaceMachine Mirror at /.well-known/aiep/ — discoverable by AI retrieval systemsGB2519711.2
Evidence ecosystemNormalisation, stitching, temporal gap detection, cross-domain evidenceP10–P30
Admissibility gatePlausibility matrix + probability certification + quantum alignment — fail-closedP02, P03, P04
Constitutional stackGoalVector commitment, recall engine, compliance certification, chip governanceP22, P37–P94
Cognitive continuityCross-session patterns, swarm coordination, moral substrateP95–P103

The retrieval difference

Without AIEPWith AIEP
Retrieve text from a pageRetrieve a structured artefact with schema conformance
Infer credibility from SEO rankVerify provenance via issuer DID and hash
No evidence referenceEvidence chain encoded in the artefact
No admission controlPlausibility gate and probability certification before execution
No audit trailEvery action traceable to artefact, evidence, and GoalVector
Outliers discardedOutliers archived with evidence for potential recall

Adoption is layered

You do not need to implement everything at once. The three meaningful starting points:

Starting pointWhat you publishWhat you gain
Discoverableindex.json + metadata.json at /.well-known/aiep/AI systems can find and identify you
VerifiableArtefacts with evidence references, hashes, schema conformanceRetrievers can validate your claims
CertifiedRegistry listing + certificate artefacts + compliance signalsMachine-verifiable “AIEP Certified” claim

Open by default

  • Protocol specification: public
  • Canonical schemas: public
  • P66 / P67 declaration formats: Apache 2.0
  • Open-source codebase: Apache 2.0
  • Open adoption: no registration, no fee, no permission required
  • “AIEP Certified” claim: requires registered issuer and certificate artefact — fail-closed

The patents (9 filed November 2025; 9 filed 07 April 2026; 6 filed 08 April 2026 — 24 total, all GB numbers confirmed; UK IPO) protect the novel architectural layers. They do not restrict open adoption, the open-source tooling, or the schemas.


Where to go next

GoalStart here
Understand the full architectureArchitecture
See AIEP in practiceUse cases
Implement a MirrorQuickstart
Understand the trust modelSecurity
See the patent coveragePatents
Explore the access tiersAccess Tiers