02 — AX-TOPIC Count Reliability

Registration

axis_registry AX-TOPIC: status CANDIDATE (owner GOV-COUNCIL, pending FAC-08 root ratification, Dieu 32/39). node_source=taxonomy (facet FAC-08), substrate_resolver=fn_topic_node_substrate. NOT official.

Counts

• axis_assignment AX-TOPIC: 25 rows.
• UI nodes (v_rp_universal_node_ui_contract axis=AX-TOPIC): 7 (TOPIC-CAND:*).
• v_axis_topic_pivots: 14 · v_axis_topic_decision_queue: 7 · v_axis_topic_automation_candidates: 49 · v_axis_topic_governance_gap: 9.

Reconciliation — PASS (exact)

The 7 topic node counts (assignment_backed): architecture 5 · knowledge_graph 10 · governance 3 · dot_trigger 3 · workflow 2 · cut_pipeline 1 · render_pipeline 1 = 25 == axis_assignment AX-TOPIC rows. The 7 nodes partition the 25 assignments exactly.

Honest labeling

All 7 nodes: governance_status = UNGOVERNED_CANDIDATE, warning_flags ["CANDIDATE_NODE","ORPHAN_TOPIC_NODE"]. count_status assignment_backed. substrate_available true. → every topic node is correctly badged candidate + orphan (not yet anchored to a ratified taxonomy root).

Reliability assessment

• Count chain is internally consistent and exact.
• Source = axis_assignment (assignment_backed). Because the axis is CANDIDATE and all nodes ORPHAN, the 25 assignments are candidate-quality, not authoritative — but this is labeled, not hidden.
• There is a 49-candidate automation funnel (v_axis_topic_automation_candidates) vs only 25 adjudicated assignments → topic coverage is partial (expected for a pilot). The 49 are AI-proposed (Dieu 39), not counted as official.
• topic_pivots(14) ≠ nodes(7) ≠ assignments(25) ≠ automation_candidates(49): these are different denominators (pivot defs vs distinct topics vs links vs proposals) — not a contradiction; the UI must label which denominator each surface uses.

Classification

• 25=25 reconciliation → OK.
• All candidate/orphan → CANDIDATE_EXPECTED (owner-blocked, not bug).
• 49 vs 25 funnel → CANDIDATE_EXPECTED (partial coverage).

Score: 80/100

Exact reconciliation, fully honest labeling. Docked because the underlying topic taxonomy is unratified (candidate-quality source) and coverage is partial — correct for a pilot, but not yet trustworthy as official topic data (0 official, correctly).