diff --git a/.pql/changelog/ticket_history/2026-07.sql b/.pql/changelog/ticket_history/2026-07.sql
index a2418bf30..c0eded075 100644
--- a/.pql/changelog/ticket_history/2026-07.sql
+++ b/.pql/changelog/ticket_history/2026-07.sql
@@ -1608,3 +1608,15 @@ Scope expansion (Jeroen round-2, 2026-07-21): beyond window-local contrast, eval
 RULING (Jeroen, 2026-07-21, round-3 direction): the smooth-gradient rendering is INTERIM ONLY — he expects not to like it as the destination. The destination is SMALLER DETERMINISTICALLY-CALCULATED PIXELS (sub-district sampling, the sketch above), and the larger vision: SCROLL-ZOOM AS CONTINUOUS INFORMATION LoD — zooming in from the regional level progressively refines the sampled resolution seamlessly (no mode transition), ''zoom in to tile size from this level'' in his words. The design pass must therefore produce: (a) the granularity ladder the zoom traverses (district 2048m -> quarter 512m -> ? ), with derivation modes + per-window costs at each rung; (b) wire implications — finer windows multiply cell counts (quarter = 16x district cells/window); weigh against the windowed-family ceiling and the SS4 budget model (likely per-rung window requests reusing the same district_window carrier at different granularity, or a granularity param — design call); (c) THE FLOOR QUESTION back to Jeroen explicitly: ''tile size'' intersects the D-226(d) ceiling (chunk/tile output never Atlas-mapped; quarter is the current legal floor) — the pass presents the options (stop at quarter vs supersede D-226(d) for the window view) rather than assuming; (d) Araminta: encoding continuity across rungs (one colorizer family, hue stability as resolution refines). The COMPOSITE_SMOOTH linear-filter path ships until this lands, then is expected to be retired.', NULL, '2026-07-21 17:00:31', '2026-07-21 17:00:31.999', '2026-07-21 17:00:31.999', NULL, 'c2418a61461bee5600fd26e7e7081325', 2) ON CONFLICT(hash) DO NOTHING;
 INSERT INTO ticket_history (ticket_record_id, field, old_value, new_value, changed_by, changed_at, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FRA9Z84VKCZ3ECQDEGDW4MSW', 'status', 'in_progress', 'review', NULL, '2026-07-21 17:52:07', '2026-07-21 17:52:07.607', '2026-07-21 17:52:07.607', NULL, '3c63f148664245c3e75cb1db2ee27052', 2) ON CONFLICT(hash) DO NOTHING;
 INSERT INTO ticket_history (ticket_record_id, field, old_value, new_value, changed_by, changed_at, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FRA9Z84VKCZ3ECQDEGDW4MSW', 'status', 'review', 'done', NULL, '2026-07-21 18:02:45', '2026-07-21 18:02:45.248', '2026-07-21 18:02:45.248', NULL, 'c515eb85a2d4ea403834823693a55228', 2) ON CONFLICT(hash) DO NOTHING;
+INSERT INTO ticket_history (ticket_record_id, field, old_value, new_value, changed_by, changed_at, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FR97DPEHANCYX2AC4021BSJC', 'description', 'Follow-on from Jeroen''s first descent (2026-07-21): even with T-1142 fixed, many interior windows are genuinely one morphology zone across 66km, and the composite''s absolute lightness modulation (0.7 + 0.3 x elev_q/100, D-226 T-1124 SS5) spans only ~1-3% lightness over a window whose local elev range is small — the window reads as a dead flat fill rather than terrain. Proposal to evaluate WITH Araminta (SS5 is her encoding): window-local contrast stretch — normalize lightness within the window''s own elev min..max (fixed band, e.g. 0.62..0.92) with a degenerate floor (single-value window stays honestly flat), deterministic per window, mirroring how the aliveness_probe elev panel achieved its 12-level local texture. Alternative/complement: a faint deterministic per-cell variation or contour hinting. Any change to the recorded SS5 encoding gets a one-line amendment note. Assess AFTER T-1142 lands — real clicks will then land on real terrain and the residual flatness can be judged on coastal vs interior windows.
+
+Scope expansion (Jeroen round-2, 2026-07-21): beyond window-local contrast, evaluate SUB-DISTRICT SAMPLING to de-block the composite at the data level — the D-227 invention fields (invent_primitives: warped envelope, scatter, character) are continuous at metre resolution and quarter granularity (512m, 4x4 per district) sits WITHIN the D-226(d) Atlas ceiling (the cap rules out chunk/tile mapping, not quarters). But classification (derive_district) is per-district today — finer sampling means a new derivation mode (e.g. sample the continuous primitives at quarter spacing and reuse the district''s classification for hue while modulating lightness/moisture per quarter, or classify per quarter at 16x the derive cost — n=32 window = 16384 quarter cells x ~7us = ~115ms background). Needs Dudley/Tyre on data feasibility + cost, Araminta on encoding. INTERIM already shipping in T-1145: linear-filtered texture rendering (presentation smoothing, no data change) — judge the residual need after both T-1145 and real-aim clicking land.
+
+RULING (Jeroen, 2026-07-21, round-3 direction): the smooth-gradient rendering is INTERIM ONLY — he expects not to like it as the destination. The destination is SMALLER DETERMINISTICALLY-CALCULATED PIXELS (sub-district sampling, the sketch above), and the larger vision: SCROLL-ZOOM AS CONTINUOUS INFORMATION LoD — zooming in from the regional level progressively refines the sampled resolution seamlessly (no mode transition), ''zoom in to tile size from this level'' in his words. The design pass must therefore produce: (a) the granularity ladder the zoom traverses (district 2048m -> quarter 512m -> ? ), with derivation modes + per-window costs at each rung; (b) wire implications — finer windows multiply cell counts (quarter = 16x district cells/window); weigh against the windowed-family ceiling and the SS4 budget model (likely per-rung window requests reusing the same district_window carrier at different granularity, or a granularity param — design call); (c) THE FLOOR QUESTION back to Jeroen explicitly: ''tile size'' intersects the D-226(d) ceiling (chunk/tile output never Atlas-mapped; quarter is the current legal floor) — the pass presents the options (stop at quarter vs supersede D-226(d) for the window view) rather than assuming; (d) Araminta: encoding continuity across rungs (one colorizer family, hue stability as resolution refines). The COMPOSITE_SMOOTH linear-filter path ships until this lands, then is expected to be retired.', 'Follow-on from Jeroen''s first descent (2026-07-21): even with T-1142 fixed, many interior windows are genuinely one morphology zone across 66km, and the composite''s absolute lightness modulation (0.7 + 0.3 x elev_q/100, D-226 T-1124 SS5) spans only ~1-3% lightness over a window whose local elev range is small — the window reads as a dead flat fill rather than terrain. Proposal to evaluate WITH Araminta (SS5 is her encoding): window-local contrast stretch — normalize lightness within the window''s own elev min..max (fixed band, e.g. 0.62..0.92) with a degenerate floor (single-value window stays honestly flat), deterministic per window, mirroring how the aliveness_probe elev panel achieved its 12-level local texture. Alternative/complement: a faint deterministic per-cell variation or contour hinting. Any change to the recorded SS5 encoding gets a one-line amendment note. Assess AFTER T-1142 lands — real clicks will then land on real terrain and the residual flatness can be judged on coastal vs interior windows.
+
+Scope expansion (Jeroen round-2, 2026-07-21): beyond window-local contrast, evaluate SUB-DISTRICT SAMPLING to de-block the composite at the data level — the D-227 invention fields (invent_primitives: warped envelope, scatter, character) are continuous at metre resolution and quarter granularity (512m, 4x4 per district) sits WITHIN the D-226(d) Atlas ceiling (the cap rules out chunk/tile mapping, not quarters). But classification (derive_district) is per-district today — finer sampling means a new derivation mode (e.g. sample the continuous primitives at quarter spacing and reuse the district''s classification for hue while modulating lightness/moisture per quarter, or classify per quarter at 16x the derive cost — n=32 window = 16384 quarter cells x ~7us = ~115ms background). Needs Dudley/Tyre on data feasibility + cost, Araminta on encoding. INTERIM already shipping in T-1145: linear-filtered texture rendering (presentation smoothing, no data change) — judge the residual need after both T-1145 and real-aim clicking land.
+
+RULING (Jeroen, 2026-07-21, round-3 direction): the smooth-gradient rendering is INTERIM ONLY — he expects not to like it as the destination. The destination is SMALLER DETERMINISTICALLY-CALCULATED PIXELS (sub-district sampling, the sketch above), and the larger vision: SCROLL-ZOOM AS CONTINUOUS INFORMATION LoD — zooming in from the regional level progressively refines the sampled resolution seamlessly (no mode transition), ''zoom in to tile size from this level'' in his words. The design pass must therefore produce: (a) the granularity ladder the zoom traverses (district 2048m -> quarter 512m -> ? ), with derivation modes + per-window costs at each rung; (b) wire implications — finer windows multiply cell counts (quarter = 16x district cells/window); weigh against the windowed-family ceiling and the SS4 budget model (likely per-rung window requests reusing the same district_window carrier at different granularity, or a granularity param — design call); (c) THE FLOOR QUESTION back to Jeroen explicitly: ''tile size'' intersects the D-226(d) ceiling (chunk/tile output never Atlas-mapped; quarter is the current legal floor) — the pass presents the options (stop at quarter vs supersede D-226(d) for the window view) rather than assuming; (d) Araminta: encoding continuity across rungs (one colorizer family, hue stability as resolution refines). The COMPOSITE_SMOOTH linear-filter path ships until this lands, then is expected to be retired.
+
+D-166 amendment 2026-07-21 (Jeroen, from the Groombridge/Lendel zoom captures): the seamless zoom ladder this ticket designs is now a HARD Phase-4 deliverable condition — T-750 does not close until the ladder stands (planetary map -> tile scale, every level deterministically derived at native granularity, never magnified interpolation; BHAG itself unchanged). Evidence baked into the amendment: at Lendel scale one heightmap px spans ~19 districts, so the current 16-district window is sub-pixel vs its source — beyond fit-zoom the viewer can only magnify the same composite, and the real waterline sat ~36 districts from the pixel-edge estimate (all window-scale-and-below information is D-227 invented detail and must be derived, not smoothed). This raises the design pass''s stakes: the granularity-rung ladder, wire budget, and the D-226(d) floor question are now phase-gating.', NULL, '2026-07-21 19:01:45', '2026-07-21 19:01:45.387', '2026-07-21 19:01:45.387', NULL, '0b7b95e7418d0b279a5e4e115085a453', 2) ON CONFLICT(hash) DO NOTHING;
+INSERT INTO ticket_history (ticket_record_id, field, old_value, new_value, changed_by, changed_at, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FR97DPEHANCYX2AC4021BSJC', 'priority', 'medium', 'high', NULL, '2026-07-21 19:01:54', '2026-07-21 19:01:54.239', '2026-07-21 19:01:54.239', NULL, '7d0d4a6b25d6430273ce7df42230dc3a', 2) ON CONFLICT(hash) DO NOTHING;
diff --git a/.pql/changelog/tickets/2026-07.sql b/.pql/changelog/tickets/2026-07.sql
index c3c6aa06e..1647bb0a6 100644
--- a/.pql/changelog/tickets/2026-07.sql
+++ b/.pql/changelog/tickets/2026-07.sql
@@ -2371,3 +2371,17 @@ RULING (Jeroen, 2026-07-21, round-3 direction): the smooth-gradient rendering is
 INSERT INTO tickets (record_id, type, parent_record_id, title, description, status, priority, assigned_to, team, decision_ref, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FRBK72KQBDCBPEV4CF9M0PA0', 'task', '06FBPPMZNNEV052DBYYY3A897C', 'Hunt leaked scene-tree pollers contaminating gdUnit full-suite runs', 'Evidence from the T-1146 fix round (branch implant-input-gate, commit 4af3dd260): at least one node leaked into the persistent gdUnit scene tree by an earlier suite (a) consumes SimBridge.poll_snapshot() every frame — starving any later live test that polls directly (poll_snapshot is consume-and-clear), and (b) while SimBridge sits in ambient test_mode=true + CONNECTED, applies TestHarness mock snapshots into GameState.current_tick every frame for the REST OF THE RUN (observed current_tick ~15-18 before the input-gate suites even connected; a freshly spawned server then panicked with ''InputQueue: tick ordering violated, last=16 new=3''). The T-1146 live suite now defends itself (signal capture + SimBridge neutralization in before_test, see test_input_gate_live.gd comments), but the leak itself remains: it burns CPU across ~3.4k tests and will bite the next live-pipeline suite that doesn''t know to defend. Hunt: instrument the gdUnit root''s children between suites (a one-off print of get_children() after each suite, or a before_test tripwire asserting no unexpected siblings) to identify which suite leaks the polling node — candidates are suites instantiating main.tscn or harness-driven scenes (test_checklist, test_client_p2/p3, test_free_camera, test_audio_sprint13, test_p0_regressions, test_hub_teleport, test_camera_anchor). Fix at the source (auto_free/queue_free discipline or explicit teardown), then consider removing the now-redundant defensive layers or keeping them as belt-and-braces (document either way).', 'backlog', 'medium', NULL, 'client', NULL, '2026-07-21 17:50:24.925', '2026-07-21 17:50:24.925', NULL, 'f071dd295ce09bdd1070dcd6d8c724e9', 2) ON CONFLICT(record_id) DO UPDATE SET type=excluded.type, parent_record_id=excluded.parent_record_id, title=excluded.title, description=excluded.description, status=excluded.status, priority=excluded.priority, assigned_to=excluded.assigned_to, team=excluded.team, decision_ref=excluded.decision_ref, updated_at=excluded.updated_at, deleted_at=excluded.deleted_at, hash=excluded.hash, canonical_version=excluded.canonical_version WHERE excluded.updated_at > tickets.updated_at OR (excluded.updated_at = tickets.updated_at AND excluded.hash > tickets.hash);
 INSERT INTO tickets (record_id, type, parent_record_id, title, description, status, priority, assigned_to, team, decision_ref, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FRA9Z84VKCZ3ECQDEGDW4MSW', 'bug', '06FBPPMZNNEV052DBYYY3A897C', 'Gameplay movement input leaks to the server while an implant screen is open', 'Found by Stig during T-1145 (2026-07-21), pre-existing: InputMapper._process() polls the D-054 move_north/west/south/east actions unconditionally and the queue flush path (input_mapper.gd -> main.gd flush_queue -> sim_bridge.send_input) never checks HudGroups.gameplay_occluded / is_app_active — so pressing WASD while ANY implant screen is open (Atlas, browser, wiki...) queues real movement commands to the server and walks the character blind behind the fullscreen overlay. Affects every implant screen equally; the companion is immune only because readers cannot send inputs (T-1130 matrix). Fix shape: gate the InputMapper poll (or the queue flush) on not HudGroups.gameplay_occluded — decide whether stance/other gameplay actions need the same gate while there (they do — audit the full polled action set). Regression test: occluded -> WASD poll queues nothing; unoccluded -> queues normally. Note: T-1145''s regional-window pan deliberately reads raw physical keycodes, architecturally independent of this fix.', 'review', 'high', NULL, 'client', NULL, '2026-07-21 14:50:12.902', '2026-07-21 17:52:07.607', NULL, '7ad09fbe0ce12206745a1d08785822e8', 2) ON CONFLICT(record_id) DO UPDATE SET type=excluded.type, parent_record_id=excluded.parent_record_id, title=excluded.title, description=excluded.description, status=excluded.status, priority=excluded.priority, assigned_to=excluded.assigned_to, team=excluded.team, decision_ref=excluded.decision_ref, updated_at=excluded.updated_at, deleted_at=excluded.deleted_at, hash=excluded.hash, canonical_version=excluded.canonical_version WHERE excluded.updated_at > tickets.updated_at OR (excluded.updated_at = tickets.updated_at AND excluded.hash > tickets.hash);
 INSERT INTO tickets (record_id, type, parent_record_id, title, description, status, priority, assigned_to, team, decision_ref, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FRA9Z84VKCZ3ECQDEGDW4MSW', 'bug', '06FBPPMZNNEV052DBYYY3A897C', 'Gameplay movement input leaks to the server while an implant screen is open', 'Found by Stig during T-1145 (2026-07-21), pre-existing: InputMapper._process() polls the D-054 move_north/west/south/east actions unconditionally and the queue flush path (input_mapper.gd -> main.gd flush_queue -> sim_bridge.send_input) never checks HudGroups.gameplay_occluded / is_app_active — so pressing WASD while ANY implant screen is open (Atlas, browser, wiki...) queues real movement commands to the server and walks the character blind behind the fullscreen overlay. Affects every implant screen equally; the companion is immune only because readers cannot send inputs (T-1130 matrix). Fix shape: gate the InputMapper poll (or the queue flush) on not HudGroups.gameplay_occluded — decide whether stance/other gameplay actions need the same gate while there (they do — audit the full polled action set). Regression test: occluded -> WASD poll queues nothing; unoccluded -> queues normally. Note: T-1145''s regional-window pan deliberately reads raw physical keycodes, architecturally independent of this fix.', 'done', 'high', NULL, 'client', NULL, '2026-07-21 14:50:12.902', '2026-07-21 18:02:45.248', NULL, '4d07095ff6a24a0243c850c607d97929', 2) ON CONFLICT(record_id) DO UPDATE SET type=excluded.type, parent_record_id=excluded.parent_record_id, title=excluded.title, description=excluded.description, status=excluded.status, priority=excluded.priority, assigned_to=excluded.assigned_to, team=excluded.team, decision_ref=excluded.decision_ref, updated_at=excluded.updated_at, deleted_at=excluded.deleted_at, hash=excluded.hash, canonical_version=excluded.canonical_version WHERE excluded.updated_at > tickets.updated_at OR (excluded.updated_at = tickets.updated_at AND excluded.hash > tickets.hash);
+INSERT INTO tickets (record_id, type, parent_record_id, title, description, status, priority, assigned_to, team, decision_ref, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FR97DPEHANCYX2AC4021BSJC', 'task', '06FB0TNSRZXCHGS16BFHSSGSV4', 'Regional window legibility on quiet interiors — window-local elevation contrast', 'Follow-on from Jeroen''s first descent (2026-07-21): even with T-1142 fixed, many interior windows are genuinely one morphology zone across 66km, and the composite''s absolute lightness modulation (0.7 + 0.3 x elev_q/100, D-226 T-1124 SS5) spans only ~1-3% lightness over a window whose local elev range is small — the window reads as a dead flat fill rather than terrain. Proposal to evaluate WITH Araminta (SS5 is her encoding): window-local contrast stretch — normalize lightness within the window''s own elev min..max (fixed band, e.g. 0.62..0.92) with a degenerate floor (single-value window stays honestly flat), deterministic per window, mirroring how the aliveness_probe elev panel achieved its 12-level local texture. Alternative/complement: a faint deterministic per-cell variation or contour hinting. Any change to the recorded SS5 encoding gets a one-line amendment note. Assess AFTER T-1142 lands — real clicks will then land on real terrain and the residual flatness can be judged on coastal vs interior windows.
+
+Scope expansion (Jeroen round-2, 2026-07-21): beyond window-local contrast, evaluate SUB-DISTRICT SAMPLING to de-block the composite at the data level — the D-227 invention fields (invent_primitives: warped envelope, scatter, character) are continuous at metre resolution and quarter granularity (512m, 4x4 per district) sits WITHIN the D-226(d) Atlas ceiling (the cap rules out chunk/tile mapping, not quarters). But classification (derive_district) is per-district today — finer sampling means a new derivation mode (e.g. sample the continuous primitives at quarter spacing and reuse the district''s classification for hue while modulating lightness/moisture per quarter, or classify per quarter at 16x the derive cost — n=32 window = 16384 quarter cells x ~7us = ~115ms background). Needs Dudley/Tyre on data feasibility + cost, Araminta on encoding. INTERIM already shipping in T-1145: linear-filtered texture rendering (presentation smoothing, no data change) — judge the residual need after both T-1145 and real-aim clicking land.
+
+RULING (Jeroen, 2026-07-21, round-3 direction): the smooth-gradient rendering is INTERIM ONLY — he expects not to like it as the destination. The destination is SMALLER DETERMINISTICALLY-CALCULATED PIXELS (sub-district sampling, the sketch above), and the larger vision: SCROLL-ZOOM AS CONTINUOUS INFORMATION LoD — zooming in from the regional level progressively refines the sampled resolution seamlessly (no mode transition), ''zoom in to tile size from this level'' in his words. The design pass must therefore produce: (a) the granularity ladder the zoom traverses (district 2048m -> quarter 512m -> ? ), with derivation modes + per-window costs at each rung; (b) wire implications — finer windows multiply cell counts (quarter = 16x district cells/window); weigh against the windowed-family ceiling and the SS4 budget model (likely per-rung window requests reusing the same district_window carrier at different granularity, or a granularity param — design call); (c) THE FLOOR QUESTION back to Jeroen explicitly: ''tile size'' intersects the D-226(d) ceiling (chunk/tile output never Atlas-mapped; quarter is the current legal floor) — the pass presents the options (stop at quarter vs supersede D-226(d) for the window view) rather than assuming; (d) Araminta: encoding continuity across rungs (one colorizer family, hue stability as resolution refines). The COMPOSITE_SMOOTH linear-filter path ships until this lands, then is expected to be retired.
+
+D-166 amendment 2026-07-21 (Jeroen, from the Groombridge/Lendel zoom captures): the seamless zoom ladder this ticket designs is now a HARD Phase-4 deliverable condition — T-750 does not close until the ladder stands (planetary map -> tile scale, every level deterministically derived at native granularity, never magnified interpolation; BHAG itself unchanged). Evidence baked into the amendment: at Lendel scale one heightmap px spans ~19 districts, so the current 16-district window is sub-pixel vs its source — beyond fit-zoom the viewer can only magnify the same composite, and the real waterline sat ~36 districts from the pixel-edge estimate (all window-scale-and-below information is D-227 invented detail and must be derived, not smoothed). This raises the design pass''s stakes: the granularity-rung ladder, wire budget, and the D-226(d) floor question are now phase-gating.', 'backlog', 'medium', NULL, 'client', NULL, '2026-07-21 12:19:16.212', '2026-07-21 19:01:45.386', NULL, '25333e2ddba335263d66e44c12e48c09', 2) ON CONFLICT(record_id) DO UPDATE SET type=excluded.type, parent_record_id=excluded.parent_record_id, title=excluded.title, description=excluded.description, status=excluded.status, priority=excluded.priority, assigned_to=excluded.assigned_to, team=excluded.team, decision_ref=excluded.decision_ref, updated_at=excluded.updated_at, deleted_at=excluded.deleted_at, hash=excluded.hash, canonical_version=excluded.canonical_version WHERE excluded.updated_at > tickets.updated_at OR (excluded.updated_at = tickets.updated_at AND excluded.hash > tickets.hash);
+INSERT INTO tickets (record_id, type, parent_record_id, title, description, status, priority, assigned_to, team, decision_ref, created_at, updated_at, deleted_at, hash, canonical_version) VALUES ('06FR97DPEHANCYX2AC4021BSJC', 'task', '06FB0TNSRZXCHGS16BFHSSGSV4', 'Regional window legibility on quiet interiors — window-local elevation contrast', 'Follow-on from Jeroen''s first descent (2026-07-21): even with T-1142 fixed, many interior windows are genuinely one morphology zone across 66km, and the composite''s absolute lightness modulation (0.7 + 0.3 x elev_q/100, D-226 T-1124 SS5) spans only ~1-3% lightness over a window whose local elev range is small — the window reads as a dead flat fill rather than terrain. Proposal to evaluate WITH Araminta (SS5 is her encoding): window-local contrast stretch — normalize lightness within the window''s own elev min..max (fixed band, e.g. 0.62..0.92) with a degenerate floor (single-value window stays honestly flat), deterministic per window, mirroring how the aliveness_probe elev panel achieved its 12-level local texture. Alternative/complement: a faint deterministic per-cell variation or contour hinting. Any change to the recorded SS5 encoding gets a one-line amendment note. Assess AFTER T-1142 lands — real clicks will then land on real terrain and the residual flatness can be judged on coastal vs interior windows.
+
+Scope expansion (Jeroen round-2, 2026-07-21): beyond window-local contrast, evaluate SUB-DISTRICT SAMPLING to de-block the composite at the data level — the D-227 invention fields (invent_primitives: warped envelope, scatter, character) are continuous at metre resolution and quarter granularity (512m, 4x4 per district) sits WITHIN the D-226(d) Atlas ceiling (the cap rules out chunk/tile mapping, not quarters). But classification (derive_district) is per-district today — finer sampling means a new derivation mode (e.g. sample the continuous primitives at quarter spacing and reuse the district''s classification for hue while modulating lightness/moisture per quarter, or classify per quarter at 16x the derive cost — n=32 window = 16384 quarter cells x ~7us = ~115ms background). Needs Dudley/Tyre on data feasibility + cost, Araminta on encoding. INTERIM already shipping in T-1145: linear-filtered texture rendering (presentation smoothing, no data change) — judge the residual need after both T-1145 and real-aim clicking land.
+
+RULING (Jeroen, 2026-07-21, round-3 direction): the smooth-gradient rendering is INTERIM ONLY — he expects not to like it as the destination. The destination is SMALLER DETERMINISTICALLY-CALCULATED PIXELS (sub-district sampling, the sketch above), and the larger vision: SCROLL-ZOOM AS CONTINUOUS INFORMATION LoD — zooming in from the regional level progressively refines the sampled resolution seamlessly (no mode transition), ''zoom in to tile size from this level'' in his words. The design pass must therefore produce: (a) the granularity ladder the zoom traverses (district 2048m -> quarter 512m -> ? ), with derivation modes + per-window costs at each rung; (b) wire implications — finer windows multiply cell counts (quarter = 16x district cells/window); weigh against the windowed-family ceiling and the SS4 budget model (likely per-rung window requests reusing the same district_window carrier at different granularity, or a granularity param — design call); (c) THE FLOOR QUESTION back to Jeroen explicitly: ''tile size'' intersects the D-226(d) ceiling (chunk/tile output never Atlas-mapped; quarter is the current legal floor) — the pass presents the options (stop at quarter vs supersede D-226(d) for the window view) rather than assuming; (d) Araminta: encoding continuity across rungs (one colorizer family, hue stability as resolution refines). The COMPOSITE_SMOOTH linear-filter path ships until this lands, then is expected to be retired.
+
+D-166 amendment 2026-07-21 (Jeroen, from the Groombridge/Lendel zoom captures): the seamless zoom ladder this ticket designs is now a HARD Phase-4 deliverable condition — T-750 does not close until the ladder stands (planetary map -> tile scale, every level deterministically derived at native granularity, never magnified interpolation; BHAG itself unchanged). Evidence baked into the amendment: at Lendel scale one heightmap px spans ~19 districts, so the current 16-district window is sub-pixel vs its source — beyond fit-zoom the viewer can only magnify the same composite, and the real waterline sat ~36 districts from the pixel-edge estimate (all window-scale-and-below information is D-227 invented detail and must be derived, not smoothed). This raises the design pass''s stakes: the granularity-rung ladder, wire budget, and the D-226(d) floor question are now phase-gating.', 'backlog', 'high', NULL, 'client', NULL, '2026-07-21 12:19:16.212', '2026-07-21 19:01:54.239', NULL, '6b60196e54ea21056fef844d4655176b', 2) ON CONFLICT(record_id) DO UPDATE SET type=excluded.type, parent_record_id=excluded.parent_record_id, title=excluded.title, description=excluded.description, status=excluded.status, priority=excluded.priority, assigned_to=excluded.assigned_to, team=excluded.team, decision_ref=excluded.decision_ref, updated_at=excluded.updated_at, deleted_at=excluded.deleted_at, hash=excluded.hash, canonical_version=excluded.canonical_version WHERE excluded.updated_at > tickets.updated_at OR (excluded.updated_at = tickets.updated_at AND excluded.hash > tickets.hash);
diff --git a/CLAUDE.md b/CLAUDE.md
index 0d429c802..99390c903 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -41,7 +41,7 @@ Development follows a strict cascade. Each phase has a concrete deliverable. **D
 | 1 | Wiki content complete — all planets, moons, stations, heightmaps, artwork | Implant-ready Godot map of the Reach with click-throughs + wiki/GTTR popups |
 | 2 | Economics layer — supply/demand, transport, political/social pressure, corporations, supply chains | Economics spreadsheets/graphs with runtime-tweakable simulation |
 | 3 | Planetary/moon maps & station layouts — cities, rivers, mountains, roads, biomes, rail | Atlas of the Reach (implant app) |
-| 4 | World generation (tile/chunk/block) — deterministic multilayer seed→tile cascade; each layer above the tile fill viewed as a **map in the implant Atlas** (no in-world rendering; chunk/tile/voxel output is harness-verified, never Atlas-mapped — D-226 amendment 2026-07-13) | Deterministic walkable-world data + per-layer Atlas maps + asset catalog |
+| 4 | World generation (tile/chunk/block) — deterministic multilayer seed→tile cascade; each layer above the tile fill viewed as a **map in the implant Atlas** (no in-world rendering; chunk/tile/voxel output is harness-verified, never Atlas-mapped — D-226 amendment 2026-07-13) | Deterministic walkable-world data + per-layer Atlas maps + asset catalog + **seamless Atlas zoom ladder** (planetary map → tile scale, each level deterministically derived at native granularity, never magnified interpolation — D-166 amendment 2026-07-21) |
 | 5 | Player control & in-world rendering — character, walls/stairs/doors, lighting, drawn on **generated** tiles (no test map) | Player viewport with final-version assets on the generated world |
 | 6 | Detail coloring — room-level NPC population, cultural room grammar | Only when the world is walkable |
 
diff --git a/governance/decisions/architecture.md b/governance/decisions/architecture.md
index def314f4c..994bfccd1 100644
--- a/governance/decisions/architecture.md
+++ b/governance/decisions/architecture.md
@@ -601,6 +601,7 @@ Technical foundation decisions that constrain implementation: engine, client-ser
 
 - **Amendment (2026-05-22):** Phases 4 and 5 swapped — **world generation now precedes player control**. The rule: no player-control or in-world rendering work begins until the generator can deterministically seed-generate every tile of every world via the full multilayer cascade. The original Phase 4 "2-floor test map" is **dropped** — test layers are produced by the generator itself once layer-drawing begins; we start drawing the world only when generation knows what to draw. Generation progress is viewed as **per-layer maps in the implant Atlas** (the Phase 3 deliverable, already built), not via an in-world renderer. The existing in-world rendering code is **left as-is until Phase 5** — neither built upon nor removed before then. Rationale: building player systems against a throwaway test substrate means rebuilding them against real generated tiles later; gating player work on deterministic generation avoids that waste. Epics T-749 (now Phase 5) and T-750 (now Phase 4) and the CLAUDE.md cascade table are updated to match.
 - **Amendment (2026-06-12):** Build order **within and after** the cascade clarified (Jeroen). The world is built **outside-in**: (1) geo layer; (2) the economic layer drawn onto the world — cities, buildings, streets, blocks, and the inter-settlement road/rail network; (3) building templates; (4) **door boundary contracts** defining the seam between exteriors and building interiors. Only then (5) the background NPC layer and building interiors (Phase 6), and only after that does gameplay/scenario design resume ("making it a game again" — tracked in the post-cascade gameplay parking epic, outside any phase). **Player control and visual rendering (Phase 5) ramp in parallel once outside generation produces walkable exteriors** — the Phase-5 trigger is "outsides generate deterministically and are walkable", not "all Phase-4 polish complete"; T-962 holds the gate and is updated to this trigger. Guiding statement: *"Before this is a game I want it to be a Reach a character can travel through."* NPCs, scenarios, and gameplay systems are conceived, designed, and built only after the traversable world exists. This amendment absorbs the v0.2 scope cluster — D-114, D-115, D-117, D-118, D-120 are marked superseded by this record, with each one's surviving design substance noted on the record itself.
+- **Amendment (2026-07-21):** **Seamless zoom ladder is a hard Phase-4 deliverable condition** (Jeroen, reviewing the Groombridge/Lendel Atlas zoom-ladder captures). The guiding statement (BHAG) is deliberately **unchanged** — *"Before this is a game I want it to be a Reach a character can travel through."* What changes is the Phase 4 exit bar: the Atlas must provide a **continuous zoom ladder from the planetary map down to tile scale**, where every zoom level shows **deterministically calculated information at that level's native granularity** ("LoD on the information in view" — the T-1143 ruling), never magnified interpolation of a coarser composite. Evidence that forced the condition: at Lendel's scale one heightmap pixel spans ~19 districts, so the 16-district window is *sub-pixel* relative to its source — beyond fit-zoom the current viewer can only linearly magnify the same 16×16 composite (captures 2026-07-21, `.cache/screenshots/groombridge/`), and the real waterline sat ~36 districts from the pixel-edge estimate, i.e. **all** information at window scale and below is invented detail (D-227) that must be *derived*, not smoothed. T-1143 is the design pass for the ladder (granularity rungs, wire budget, the D-226(d) floor question); the Phase-4 epic (T-750) does not close until the ladder stands. CLAUDE.md's cascade table is updated to carry the condition.
 - **Rationale:** The pattern of negotiating pragmatic v0.2 cuts while discussing room-level detail repeatedly produced superseded decisions, confused agents, and distracted from building the actual product. The cascade enforces a first-things-first discipline: each layer of the game is grounded in the layer below it before detail is added.
 - **Raised by:** Jeroen, established 2026-03-24 during world generation workshop.
 - **Dissent:** None.