D64: core predicates grow to 16 — uses + reports_to promoted from the watchlist

README.md

@@ -154,5 +154,5 @@ and flow downward.
 | [plan/analysis/p3_agent_navigation.md](plan/analysis/p3_agent_navigation.md) | P3 agent navigation — materialized tree vs index-only, the `_index.md` contract, facets/views/fan-out, why directory LLM summaries are rejected (→ e0 §6, F6) |
 | [plan/plans/roadmap.md](plan/plans/roadmap.md) | Build order: phase spine, stack, gate register, WP format (phases 0–8) |
 | [plan/designs/packaging_distribution_design.md](plan/designs/packaging_distribution_design.md) | Delivery artifacts, delivery-only task execution, enforced code architecture (D62) |
-| [decisions.md](decisions.md) | Architecture decision log (D1–D63) with rationale |
+| [decisions.md](decisions.md) | Architecture decision log (D1–D64) with rationale |
 | [questions.md](questions.md) | Open questions to resolve before building |

decisions.md

@@ -424,6 +424,10 @@ supersession (`registries_design.md` §4, extension packs).
 representation of time. A claim's immutable asserted-validity interval (D41) is *claim metadata*, not
 a relation object/predicate or a Date-node, so it is fully compatible — D18 is unchanged.
 
+**Refined by D64.** The seed core is now 8 types + **16** predicates: `uses`
+(Person | Organization → Product) and `reports_to` (Person → Person) promoted from the
+registries §4 watchlist — the first watchlist graduations. Everything else here is unchanged.
+
 ## D19. Coref is satisfied inside the E2 extraction call (no dedicated model)
 
 **Decision.** Coref is the guarantee that no claim leaves E2 with a dangling pronoun — satisfied
@@ -1768,3 +1772,37 @@ three-calls-per-chunk math). E1 spike 8 narrows from "which model" to "which sto
 prefix quality", measured on the golden set. P1 index/parameter choices unblock (dimension now
 bounded). Questions #3 closes; review finding F8 closes. The embedder port gains its two named
 adapters (OpenRouter-hosted; self-hosted weights).
+
+## D64. Core predicates grow to 16 — `uses` and `reports_to` promoted from the watchlist
+
+**Decision.** The D18 seed core gains two predicates, taking the core from 14 to **16**:
+**`uses`** (Person | Organization → Product — adoption/consumption of a product/system/tool;
+change-prone, ordinary supersession; deliberately distinct from `works_on`, which means
+building/active engagement, not using) and **`reports_to`** (Person → Person — the
+organizational reporting line; change-prone). Both move from the predicate watchlist
+(registries §4) into the core table with these tight signatures; their formerly designated
+pack homes (systems; work/HR) no longer apply to them. The watchlist keeps `owns`/
+`acquired_by`, `lives_in`, and the guardrailed `enables`; the D5 `other:` promotion funnel
+remains the default path for everything else — this is an owner promotion, not a change to
+the funnel rule.
+
+**Context.** The watchlist promotes on demonstrated `other:` volume, not intuition. These two
+are promoted ahead of volume because every named deployment (registries §1) needs them
+first-class from day one: "who uses which system/tool" is the backbone of the migration
+deployment's as-is landscape and a bread-and-butter assistant/agency query ("person A uses
+software X"), and `reports_to` is the org-chart backbone of people-centric retrieval. Both
+carry exactly the properties that qualified the original fourteen: tight domain/range over
+core types (the D18 gate bites), natural evidence aggregation (the same usage/reporting fact
+recurs across sources), and clean supersession semantics (tool adoption and reporting lines
+end and change — the bi-temporal model fits). Waiting for the funnel would have meant an
+interim of `other:uses` / `other:reports_to` edges that bypass domain/range validation
+(tier='other' is ungoverned until promotion) for facts already known to be wanted governed.
+
+**Consequences.** Registries §4: the core table has 16 rows (`related_to` stays last as the
+permissive parent); the watchlist shrinks to three entries. p2 §3's seed vocabulary updated;
+extraction prompts pick both up by registry render (D15 — rows, not prompt engineering).
+Core-tier obligations attach: D22 golden-set coverage for both, and the core stability
+commitment (a future split pays the D15-flagged split cost). Signature notes: systems-pack
+subtypes (`System`/`Module ⊂ Product`) inherit into `uses`'s range via D15 inheritance;
+`reports_to` stays strictly person-to-person (a role-based reporting line is modeled through
+the person holding the role).

plan/designs/k_layers_design.md

@@ -699,7 +699,8 @@ between compiled and authored does **not** fall on the as-is/to-be line. The lin
 any other; only content no source yet attests must be authored.
 
 - The systems extension pack (registries §4) makes the landscape *entities*:
-  `System`/`Module ⊂ Product`, `BusinessProcess ⊂ Concept`, predicates `uses`, `depends_on`,
+  `System`/`Module ⊂ Product`, `BusinessProcess ⊂ Concept`, predicates `uses` (core since D64,
+  inherited by the pack's subtypes), `depends_on`,
   `part_of`. Workshop notes, emails, and specs from different people all resolve onto the same
   module entities — that is plane E doing its job.
 - **`as-is/` is compiled.** The ordering-subsystem page holds an `entity_subtree` rule on the

plan/designs/overall_design.md

@@ -2,7 +2,7 @@
 
 The architecture that satisfies `plan/requirements/requirements_v3.md`. This document is the
 map; per-layer designs (this directory) are the territory. Decision rationale lives in
-`decisions.md` (root, cited as D1–D63); supporting research in `plan/analysis/`.
+`decisions.md` (root, cited as D1–D64); supporting research in `plan/analysis/`.
 
 ## 1. System overview: three planes (D14)
 

plan/designs/p2_graph_design.md

@@ -136,9 +136,10 @@ both entity-keyed blocking and graph queries. So:
   hatch.
 - A periodic job reviews frequent `other:` values and promotes them (or maps them to existing
   predicates). The ontology evolves by governance, not by accretion.
-- Seed vocabulary = the **14 D18 core predicates** (`works_for, member_of, affiliated_with, founded,
-  located_in, part_of, authored, created, about, knows_about, knows, participated_in, works_on,
-  related_to`); the authoritative list + domain/range signatures live in `registries_design.md` §4.
+- Seed vocabulary = the **16 core predicates** (D18, extended by D64: `works_for, member_of,
+  affiliated_with, founded, located_in, part_of, authored, created, about, knows_about, knows,
+  participated_in, works_on, uses, reports_to, related_to`); the authoritative list +
+  domain/range signatures live in `registries_design.md` §4.
   Extend per K2 domain via packs. (The graph stores `predicate` as a property — D18 domain/range is
   enforced upstream by the E3 normalizer, not in the graph.)
 

plan/designs/postgres_schema_design.md

@@ -494,8 +494,8 @@ COMMENT ON TABLE scope_interests IS
   'Per-scope interest list (D16): the predicate/type footprint that defines the scope''s PROJECT_GRAPH_CYPHER view and what its K2 compilation selects. A query/compile-time selection over fully-extracted facts — never a promotion trigger (D28 withdrawn).';
 ```
 
-The **seed core** (D18) — 8 entity types and 14 predicates with signatures — is data inserted by a
-migration, not schema. Its authoritative list is in `registries_design.md` §4; the seeding
+The **seed core** (D18, extended by D64) — 8 entity types and 16 predicates with signatures — is
+data inserted by a migration, not schema. Its authoritative list is in `registries_design.md` §4; the seeding
 migration cites that section rather than duplicating it here (avoids drift).
 
 ---

plan/designs/registries_design.md

@@ -130,10 +130,10 @@ One canonical cascade. Stop at the first confident match. **Registry-self-contai
 
 ## 4. Ontology — universal core + anchored extensions (D15, D18)
 
-- **Seed core (D18):** 8 entity types — `Person, Organization, Place, Document⊂CreativeWork,
-  Event, Concept, Project, Product` — and the 14 core predicates below. `related_to` is the
-  predicate-side core parent (the extend-never-fork anchor + permissive escape). Time is
-  bi-temporal edge metadata, never a predicate/Date-node.
+- **Seed core (D18, extended by D64):** 8 entity types — `Person, Organization, Place,
+  Document⊂CreativeWork, Event, Concept, Project, Product` — and the 16 core predicates below.
+  `related_to` is the predicate-side core parent (the extend-never-fork anchor + permissive
+  escape). Time is bi-temporal edge metadata, never a predicate/Date-node.
 
 **Core predicates (the authoritative starting set — domain/range is the enforced signature):**
 
@@ -152,7 +152,9 @@ One canonical cascade. Stop at the first confident match. **Registry-self-contai
 | 11 | `knows` | Person → Person | social graph |
 | 12 | `participated_in` | Person \| Organization → Event \| Project | involvement |
 | 13 | `works_on` | Person \| Organization → Project \| Product | active engagement — change-prone |
-| 14 | `related_to` | any → any | permissive core parent (escape + extend-never-fork anchor) |
+| 14 | `uses` | Person \| Organization → Product | adoption/consumption of a product/system/tool — change-prone; distinct from `works_on` (building ≠ using) — promoted from the watchlist, D64 |
+| 15 | `reports_to` | Person → Person | organizational reporting line — change-prone (supersession) — promoted from the watchlist, D64 |
+| 16 | `related_to` | any → any | permissive core parent (escape + extend-never-fork anchor) |
 
 Multi-signature predicates list each allowed `(subject_type, object_type)` pair (Graphiti
 `edge_type_map` shape); subtypes inherit a parent's signatures (D15). Schema.org property
@@ -246,10 +248,13 @@ into the appropriate pack with a tight signature. Adding one is a registry row;
 premature core predicate (prompt space, golden-set coverage, a split if it's wrong) is the
 reason to wait.
 
+**Two graduations (D64):** `uses` and `reports_to` were promoted into the core (rows 14–15
+above) by owner decision — the named deployments made the demand case without waiting for
+`other:` volume (an as-is system landscape runs on `uses`; people-centric retrieval runs on
+`reports_to`). The rows below remain held back.
+
 | Candidate | Tight signature | Future home |
 |---|---|---|
-| `uses` | Organization \| Person → Product | systems pack |
-| `reports_to` | Person → Person | work/HR pack |
 | `owns` / `acquired_by` | Organization → Organization \| Product | business pack |
 | `lives_in` | Person → Place | personal pack |
 | `enables` | Concept → Concept (tight only) | research scope — guardrailed (see below) |

questions.md

@@ -4,7 +4,7 @@ The living register of **what is not settled yet** — open decisions, unwritten
 designs, known risks, and concrete inconsistencies to fix. It is the one place to look for "what's
 still open"; it cross-links the two specialized trackers:
 
-- **`decisions.md`** — what *is* decided (D1–D63).
+- **`decisions.md`** — what *is* decided (D1–D64).
 - **`plan/analysis/objections.md`** — the step-back critique (O1–O6) with accept/reject status.
 - The **design-doc index** in `plan/designs/overall_design.md` — which design docs are written
   (`current`) vs. `planned`.
@@ -161,7 +161,7 @@ Keep this current: when something here is decided, move it to a decision and pru
     still says** "row counts are contingent on the value gate — size against *gated* volume." Update
     D23.
 27. ~~**P2 graph design ontology is stale vs D18.**~~ **RESOLVED (D44).** `p2_graph_design.md` §2/§3 now
-    use the D18 seed core (8 types + 14 predicates), `DOC_CROSSREF(kind)` (generalizing the old `CITES`),
+    use the D18/D64 seed core (8 types + 16 predicates), `DOC_CROSSREF(kind)` (generalizing the old `CITES`),
     and the `IS_DOCUMENT` bridge — see the Postgres→LadybugDB translation analysis
     (`plan/analysis/ladybug_translation_research/SYNTHESIS.md`) and the `v_graph_*` projection views
     (`postgres_schema_design.md` §10.A).
@@ -191,7 +191,7 @@ Keep this current: when something here is decided, move it to a decision and pru
   belief tier. Open remainders stay tracked above (#5 whose-beliefs, #6 cadence, #24 hard-delete
   residuals) and in the design's §11 spikes.
 
-- **Ontology seed** (was "what seeds the ontology") → **D18** (8 core types + 14 predicates with
+- **Ontology seed** (was "what seeds the ontology") → **D18** (8 core types + 14 predicates, since grown to 16 by D64, with
   domain/range; extension packs; `other:` promotion).
 - **Multi-tenant / ID scoping** (was "single user or multi-tenant") → **D16** + the deployment model
   (`registries_design.md` §1): separate deployments = separate Postgres instances + entity spaces.