Advisory
AI produces outputs; humans make every decision; the substrate captures the lineage.
Decision Lineage Protocol (DLP)
Open governance infrastructure: the structural layer that turns model prediction into governed, traceable, reversible decisions. Every consequential move carries its lineage.
AI participates in governance in graduated stages — but it never modifies the authoritative record. Every AI output carries its truth type, source, and confidence; the substrate captures the lineage at each stage.
AI produces outputs; humans make every decision; the substrate captures the lineage.
AI proposes actions; humans review and promote; both the proposal and the decision are recorded.
AI routes within delegation bounds; humans review by exception.
The Decision Lineage Protocol names the connective tissue between raw prediction and consequential organizational decisions — 19 structural primitives, 10 behavioral invariants, a five-tier architecture, and a four-type truth system — so decisions are traceable, constrained, and epistemically rigorous, not merely statistically coherent. It is runtime-agnostic: it governs the decision, whatever produces it.
Always required — cannot be removed, merged, or derived.
Required for any substrate — the core can't be instantiated without it.
Required once governance maturity exceeds basic capture.
Required for AI participation; unnecessary in purely human systems.
Required for temporal governance legibility across substrates.
A human assertion of best current understanding; versioned.
AI-generated, in staging — never canonical without human promotion.
Behind an inspection boundary — impenetrable to governance.
The complete Decision Lineage Protocol specification — thirty-plus sections, from irreducible primitives through governance activation and the SDK workbench — is maintained in the GrytLabs governance substrate. A browsable spec reader lands with the next revision.
Read the Specification →