Practitioner-friendly terms mapped to the discipline's canonical terms.
The discipline's canonical terms are in Source-Engineering/source-engineering-foundations/terminology.md. This glossary maps practitioner-friendly terms to the canonical terms; the canonical terms are normative.
Audit -> Audit trail. The persistent record of who saw what response, when, with what evidence, and what they did with it. Part of Layer 4 of the discipline's four-layer stack.
Audit history -> Audit trail (per-decision). The append-only log of changes to a single DecisionRecord.
Appeal -> Feedback surface (type: appeal). A COTS's validation.feedback_surfaces entry that lets the user file an appeal against a decision.
Blueprint -> Architecture blueprint. A versioned architectural specification that floors one or more layers of the stack.
Bundle -> COTS (Chain of Orchestrable and Traceable Sources). The structured, machine-readable bundle that every Source Engineering response carries.
Chain of evidence -> Evidence chain. The chain that links a decision back to its supporting and contradicting evidence.
COTS -> COTS (Chain of Orchestrable and Traceable Sources). The discipline's headline artefact. Five parts: Inputs, Context, Decisions, Evidence, Validation.
Compatibility -> Backwards compatibility. The discipline's guarantee that a v1.x COTS producer is acceptable to a v1 COTS consumer.
Conformance -> Conformance. The property of an implementation that satisfies the discipline's specifications. Verified by the conformance suite.
Counter-evidence -> Counter-evidence. The subset of evidence items that contradicts a decision or contradicts a claim the decision relies on. Mandatory.
Decision -> DecisionRecord. The persistent artefact that captures a decision. Part of the COTS.
Decision basis -> DecisionBasis. The subset of the COTS that links a decision to its supporting evidence, counterevidence, context, constraints, and known gaps.
Domain -> High-stakes domain. A vertical where a confidently wrong decision carries a cost the user is not willing to absorb through latency.
Evidence -> Evidence chain + EvidenceItem. The chain that links a decision back to its sources, and the items in the chain.
Explanation -> DecisionBasis + Evidence chain + Context. The discipline's surface for explaining a decision. The explanation is auditable.
Feedback surface -> Validation surface. An endpoint in the COTS's validation.feedback_surfaces that lets the user interact with the decision. Three types: appeal, rerun, correction.
Five pillars -> The five pillars (P1-P5). Data Provenance, Contextual Signature Mapping, Chain of Evidence, Explainability Anchoring, Human Validation Points. The 2025 essay's intellectual shape.
Four layers -> The four-layer stack. Typed Model, Runtime Validation, Business Guardrails, Audit Trail. The 2026 essay's architectural shape.
Hesitation factor -> Hesitation factor. The reluctance to act on a model's answer because the user cannot verify its reasoning in time. The discipline exists to close this.
Judgment gap -> Judgment gap. The gap between a structurally valid answer and the right answer. The discipline's second-order problem.
Layer -> Layer. One of the four named layers of the discipline's stack: Typed Model, Runtime Validation, Business Guardrails, Audit Trail.
Log -> Audit trail (operator-facing). A COTS's audit trail is more than a log; a log is operator-facing, an audit trail is user-facing.
Manifesto -> Manifesto. The discipline's canonical home, in Source-Engineering/source-engineering-foundations/docs/00-manifesto.md.
Model card -> Model card. A COTS-aligned model card. The model card is generated by the model-card-gen tool in Source-Engineering/source-engineering-tools/.
Pillar -> Pillar. One of the five named claims introduced in the 2025 essay. The pillars remain the language of the foundations; the layers are the language of the architecture.
Provenance -> Provenance. The chain that links an evidence item back to its source.
Reasoning trace -> Trace + COTS. A COTS is more than a reasoning trace; a trace is the model's internal log, a COTS is the discipline's auditable response.
Re-run -> Feedback surface (type: rerun). A COTS's validation.feedback_surfaces entry that re-runs the decision with the same evidence under a different model or version.
Source -> SourceArtifact. A discrete piece of content that can be cited as an input to a decision.
Spec -> Specification. A normative technical requirement, in Source-Engineering/source-engineering-specifications/.
Stack -> Four-layer stack. The composition of the discipline into four named layers.
Trace -> Trace + COTS. A COTS is the discipline's trace; the operator-facing trace is the model's internal log.
Typed question -> Typed question (per a registered decision-type). A question that declares its answer shape in advance.
Validation -> Validation. The subset of a COTS that captures the feedback surfaces exposed back to the user. Not the same as the runtime-validation layer (Layer 2).
Version -> Version. A specific state of a source artifact, addressed via content hash.