Short definitions for the L4 architecture set. Normative behaviour lives in 00-kernel.md and the numbered docs. Architecture names: section 1.1.

Table: 46 rows by term
TermMeaning
AbstainTerminal: insufficient case to decide; always traced
Action footprintAssets, shared resources, crew/role, material, time window (start, end, lag) claimed by a candidate or open card
AL0–AL5Autonomy levels (direction; fast-loop stages 1–3 = AL1–AL3) — section 6
As-known-atPSM / ledger snapshot using recorded time so late corrections do not rewrite what L4 knew
AutonomyPolicySigned grants, envelopes, expiry (direction; L5 owns engine — section 5.6)
Builder readBulk/list read for the PSM builder; not an agent tool; cannot write
Case libraryEpisodic store of traces joined with L5 outcomes; authority when it disagrees with Hindsight
ClosureStateEleven code states on the live card (direction; as built: stamped_l5_domain/cards/states.py) — section 5.9
Condition keyStable id for “this plant condition”; one open card per key
DecisionCaseOne run unit: intake + snapshot + obligations + candidates + terminal
DecisionTraceAlways-on record: observed, context, action, policy, approval, outcome (and seam decisions)
DiscoveryL4-originated candidate (scanner / pattern / hypothesis), not an L3 Finding
EmitTerminal: send one Prescription draft to L5 (as built: card proposal)
EvidenceLayer contract for detector output (direction; as built: Finding finding.json 1.2.0) — section 5.2
Evidence ledgerTyped rows (measured / advisory / model partitions) frozen into the trace
Evidence tierContract tier on claims: measured / confirmed / modeled / unknown; quantity labels per D13
Exploration cardOpt-in card marked exploration=true to test soft-gate blocks
FindingAs-built L3 detector output admitted to L4 (finding.json 1.2.0)
Gate idStable id of the hard or soft gate that blocked a candidate
Grounded-hypothesis laneOwner opt-in path for LLM hypotheses that still pass L3 grounding + dual-family + kernel
Hard gateNever tunable, never backlog, never explored
HoldL4-internal portfolio outcome; staff-visible; not sent to L5
HindsightMemory product: plant bank + dialogue banks; advisory except where case library overrides outcomes
Jev seamDecision slot with LLM structured output today and a logged record so a classifier (Jev) can later beat it
KernelFrozen normative rules in 00-kernel.md
LockfileRelease pin of registry versions, kernel version, model pins
Opportunity ledgerStore of every blocked candidate with gate id and later outcome if known
Plant BoxPlant-side computer for the fast loop (direction; D4)
PortfolioDedupe, conflict, attention, supersede among candidates and open cards
PrescriptionWhat to do, why, who, check plan (direction; as built: prescription.json 1.0.0 / card-proposal) — section 5.3
Proof floorMinimum evidence/structure required before emit (asset bound, verification path, L3 condition test for discoveries)
PSMPlant Situation Model — derived per-plant cache in the L4 store
Release lockfileSee lockfile
ShadowTraced, never sent to L5 (uncertified detectors, ungrounded hypotheses, demoted patterns)
Site packVersioned, owner-reviewed plant configuration including topology
Soft gateThreshold in registry; calibrated from opportunity ledger + exploration
SupersedeTerminal: replace prior L4 proposal before owner acceptance
WithholdTerminal: do not send; reason recorded
Working ledgerPer-run ledger frozen into the trace, then discarded
WriteRequestPlant Box write path request (direction; closure/action-intent.json retired) — section 5.8
Zoom readAllowlisted typed read requested by a model; code executes and appends rows
OE knowledge corpusVersioned literature + sparse method ontology; retrieved as advisory rows only — see 23-oe-knowledge-corpus.md
Shift sweepOnce-per-shift whole-plant discovery pass — 08-discovery.md
Work queueDurable per-plant scheduler for Findings, sweeps, Ask — 25-work-queue-and-concurrency.md
Kill switchControl-plane stop of new leases + CardSink — 28-commissioning-and-controls.md
Safe-startChecklist before emit_enabled — 28-commissioning-and-controls.md

v1 slice vs later#

Glossary terms are stable for l4-kernel.v1. New terms arrive with ADRs or registry packs; this file updates in the same commit as the doc that introduces them.

Names used in the architecture

From architecture section 1.1. These names have one meaning everywhere in the pack.

NameMeansNot to be confused with
Layer 1 to Layer 6 (①–⑥ in diagrams)The conceptual layers: Plant systems, Context, Analytics, Decision, Action, ValueRepo layers
L0 to L6Repo layers (SE, connectors, L2 store, L3 intelligence, KR, L5 closure, L6 experience); mapped in section 3.5Conceptual layers
AL0 to AL5Autonomy levels (section 6)Either kind of layer
measured, confirmed, modeled, unknownContract tier of a claim (section 5.2). "Confirmed" is used only in this senseEvidence labels
Measured, Estimated, Assumed, UnknownEvidence label of one quantity, plus calibrated or uncalibrated (D13)Contract tiers
supported, refuted, untestedState of a causal hypothesis after an experimentContract tiers
verifiedOnly a counterfactual M&V result that passed its model-fit checks; never a bill claimops_confirmed
operating envelopeTag limits, steps and rates inside an AutonomyPolicyThe Envelope contract (message header, section 5.10)
Plant BoxThe plant-side computer that runs the fast loop (D4)The edge-agent, which only reads
Page history: last 2 changes
  1. 2026-10-07 docs(technical): rewrite l4 21-30, glossary and README; reconcile architecture gaps e7fead7
  2. 2026-09-25 docs(l4): agentic decision architecture, ADRs, and production hardness 8275e7c

Diagram

100%

Search the architecture