Protocol Specification
The Decision Lineage Protocol (DLP) creates governed semantic infrastructure that enables machines and humans to share context at scale—while preserving clear authority boundaries and full audit trails.
Core Innovation
"AI should interpret declared reality—it should not invent it. AI should assist orientation—it should not silently operate."
ProtoLex introduces a new category: AI-Native Accountability Substrates—governed semantic infrastructure that enables machines and humans to share context at scale while preserving clear authority boundaries and full audit trails.
AI-Native
Systems designed from inception to support shared machine-human context. The substrate holds meaning; interfaces project it.
vs. AI-Augmented
Traditional systems with AI features added. The architecture wasn't designed for AI context requirements.
The Three-Layer Separation Model
A novel separation of concerns between substrate, operating systems, and advisory systems.
Layer 1: Substrate
Governed semantic infrastructure. NOT a product, NOT a workflow engine. Defines MEANING, not behavior. Stewarded, versioned, auditable.
Layer 2: Command Operating Systems
Authority: ACTS. Owns organizational intent. Assigns accountability. Executes state changes. Records operational reality. System of record.
Layer 3: Advisory Operating Systems
Authority: ADVISES. Interprets public rules. Surfaces risk & uncertainty. Produces bounded artifacts. Claims are scoped & evidenced. Non-binding, non-certifying.
Key Innovation
The explicit separation of "meaning governance" (substrate) from "execution governance" (command OS) from "interpretation governance" (advisory OS) prevents the dangerous conflation of AI recommendations with authoritative decisions.
The Three Truth Types Model
A fundamental architectural innovation: distinguishing three categories of truth with different governance rules.
| Truth Type | Definition | Mutability | Examples |
|---|---|---|---|
| Authoritative | What happened | Append-only, immutable | Events, receipts, transactions, decisions |
| Declared | What conditions apply | Versioned, time-bounded | Context declarations, constraints, policies |
| Derived | What we conclude | Ephemeral, recomputable | Gap registers, projections, summaries |
Critical Rule: Derived outputs may be stored for convenience but are NEVER treated as canonical. They are projections, not truth. This separation prevents semantic drift and regulatory fragility.
The Four Primitives Model
A novel simplification revealing the universal structure underlying all accountability substrates.
Decision
"What was chosen?"
The core unit of accountability. A choice made at a point in time, with alternatives considered and one selected.
Authority
"Under what power?"
The source of legitimacy. Who or what grants the right to make this decision? What bounds exist?
Context
"Given what conditions?"
The state of the world at decision time. Immutable snapshot of what was known, believed, and constrained.
Evidence
"Based on what proof?"
The supporting artifacts. Documents, data, claims, and sources that informed the decision.
Domain Projections
The primitives are constant. The projections vary by domain:
Authority Doctrine
Substrate defines meaning.
Command OS acts.
Advisory OS advises.
This separation is non-negotiable and represents a key architectural innovation.