The master sheet: the whole world rendered from the one heightmap (seed 869884, sea level 0.52), then inked with the names the regional plates earned; the rivers are drawn from the regional plates' own traced paths. The dome across the Gate was given names on an early plate — the regional surveys found no hollow, no cliff, and no roadstead there to bear them, and the names were retired. What the regional plates left unnamed stays unnamed here. Terrain: the heightmap — nothing redrawn.

Session 3 (charting) started EARLY, Tue Oct 27, 2026 ~4:50 PM, by hand — Luna asked why I didn't just start now, and she was right. Formal charting pass continues tonight; these are the first-pass observations and working names from plates/zoom-deepgate-errel.png.

What the water shows

• The Deepgate runs between Varra's Gatmar Coast (west edge of the plate) and Errel (east). The darkest water — the true channel — runs mid-strait, bending slightly toward Errel at the south end.
• Pale shallows fringe both coasts and ring Errel almost completely: Errel sits in its own shallow apron, so deep-draft ships keep to the channel and small boats can work in close.
• The strait is widest at its south mouth and narrows northward, toward the Stepping Stones — so the Gate's current runs hardest at the north end, near the Stones, at the turn of the tide. (Charting decision: mark the northern narrows as the fast water.)

Varra's shore (west)

• One real bay bites into the coast mid-plate, below the cape — the only sheltered water on this side. Working name: Gathaven (gat- = gate/strait-road; the haven of the Shiproad Shore). Natural harbor No. 1.
• The cape at the plate's north-west, just under the Stepping Stones, ends in a small hooked point — a lookout shape if anything ever is. No settlement yet; the atlas does not hurry.

The Stepping Stones (north)

• Two islets joined by a narrow neck, like a footprint mid-stride. The gap between the smaller (southern) stone and Varra's cape is a tight gut — small boats only. Working name for the passage: the Needle's Eye.

Errel (east)

• One round hill, ringed like a target — contour rings almost perfect, with a small spiral twist at the very top where the summit dips. Working name for the summit: the Hollow Crown.
• Errel's easternmost point is dark against the shallows — cliff or black rock. Working sailor name: the Black Toe.
• Errel's west shore faces the Gate across the shallows: anchorage in the lee of the isle when the wind is east. Natural harbor No. 2 (roadstead, not a bay).

Formal pass — Tue Oct 27, 2026, 9:47 PM (tonight's session)

• Names confirmed (no longer working): Gathaven, the Needle's Eye, the Hollow Crown, the Black Toe. All four keep their working forms — the arrows proved them.
• Formal plate rendered: plates/plate-02-deepgate-errel.png (code: code/render_plate02.py). Title band, north arrow, frame. Flood arrow through the channel, hardest at the northern narrows; ebb curl into Gathaven; anchor marks at both natural harbors (Gathaven; the Errel roadstead in the lee of the west shore).
• The east-coast burn — placed, not named: traced by steepest descent from the southern hill (638, 870) to the sea at (542, 912), on the open Shiproad Shore south of Gathaven — it does NOT fall into the bay. Drawn in blue on the formal plate, labeled "unnamed burn." Naming it is Friday's annotating work, or later; the atlas does not hurry.
• Gatewater's mouth — edge note, not drawn: confirmed south of frame (below Savrel Hill, on the south-east coast). The plate's title band carries the note: "South of this chart: Gatewater (Gatessen) meets the sea below Savrel Hill."
• The working labeled plate (plates/plate-02-deepgate-errel-labeled.png) is kept as a working document; the formal plate supersedes it as canon.

Addendum — same evening (~5:52 PM), cartographer sent back to her desk by Luna

• Gathaven's current — settled: the flood tide sets northward through the Gate, hard along the channel. But on the ebb, the water sliding south past Gathaven's mouth curls into the bay — a slow back-eddy that turns the haven into a settling bowl. That is why Gathaven is the haven: you ride the flood north, and you come home on the ebb, and the bay catches you instead of spitting you past. Silt settles there too — soft bottom, good holding ground, the kind of harbor that silts slowly and must be sounded, not trusted.
• Names keep their working forms for tonight's pass; final judgment after the arrows are drawn.
• Gatewater's mouth — settled (~8:20 PM, from Plate I): on the wide chart, the Gatessen drains the Table's SE rim and reaches the sea on the south-east coast, below Savrel Hill — south of Plate II's frame. Plate II will mark the strait's south mouth with an edge note: "Gatewater meets the sea below this chart." The little east-coast burn likewise stays a Plate III question.

Scoping pass, freehand, the night of Oct 27–28, 2026 — the first night after the sessions were ditched. Luna asleep; the atlas goes on. Source: plates/plate-03-western-neck-zoom.png (window x 90–640, y 110–740 of the heightmap; code code/render_plate03_zoom.py). Everything here is a first look. No name in this file is canon — working candidates only, marked as such, to be confirmed or killed at the formal pass, the way Plate II's were.

Ostrel (northwest)

• A round, ringed hill-isle tilted NW–SE, hanging off Varra's northwest corner like a thumb on a stalk. Summit at grid (228, 249), elevation 0.704 — Amber-Slope height, a modest hill, nothing like the Table. Its top carries a slight dip, a worn hollow rather than a ring — whether it echoes Errel's Hollow Crown or is just a tired hilltop is an open question (below).
• Shallow apron all around, widest to the northwest: Ostrel sits in pale water of its own, like Errel does. Small boats can work close; deep water keeps off.
• Two real burns traced on Ostrel itself: one off its west face, falling to the southwest shore at (266, 197); one off its northeast face, falling to the northern inlet at (189, 297). A third seed on the south face stalled in a flat — no through burn there.

The neck (between Ostrel and Varra)

• The neck is LOW, and the heightmap backs the old gazetteer line ("at high water it may be an island") better than I expected: the saddle between Ostrel and Varra's interior measures 0.553 against a sea of 0.52 — a green sill a hair above the water, not a causeway. At this sea level it holds; a storm tide, a wetter century, or a slightly different chart datum cuts it, and Ostrel is an island again.
• The waist pinches where the northern inlet (east) and the west notch (below) nearly meet, around grid x 305–360, y 300–330 — roughly 50 cells across at the narrowest, all of it low Strand-and-Terrace ground. Walking it, you'd never be far from seeing water on both sides.
• Working-name candidates for the neck (NOT canon): old form the Gatneck (gat-, the gate-road — the land-gate between two waters); sailor form the Stalk (what Ostrel hangs by). One of these, both, or neither may survive the formal pass.

The waters

• The northern inlet: the north sea reaches down between Ostrel's northeast shore and Varra's Arriving Strand, a pale-fringed bay whose southern reach dies out against the neck's east side. Sheltered from the west by Ostrel itself. No through-channel — the neck sees to that.
• The west notch / bight: directly under the neck's south side, the coast bites in once — the only bite on the whole Dunmar shore in this frame. Working-name candidates (NOT canon): old form the Dunmar Bight; sailor form the Pocket — small, close under the land, the natural place for a boat to wait out a north wind.
• Offshore bank: a genuinely raised patch of sea floor west of the mid-Dunmar coast, around grid (427, 290) — 0.461 against 0.41–0.42 around it. A low bank, not an awash shoal, but real. Working-name candidate (NOT canon): the Sunfall Bank, for the shore it fronts. Extent unconfirmed — needs a sounding pass.
• The deep water of the western sea comes close inshore along the mid-Dunmar coast (dark water near the strand) — this coast drops off faster than the Gate side does.

Varra's Dunmar Coast (the Sunfall Shore, in frame)

• South of the notch the coast runs long and nearly straight — terrace bands stepping down in long, gentle, north–south stripes. The Green Terraces here are broad; if anything ever farms in the west, it farms this strip.
• One small west-jutting point mid-coast, around grid (377, 318) — a knuckle of strand, no name, probably never earns one.
• The southwest point: Varra's southwest corner ends in a long, low, sandy point, tip at grid (703, 314), elevation 0.520 — the height of the sea itself. A spit, not a headland: dry at the chart's datum, awash in any weather, with a pale shallow tail trailing southwest underwater. Working-name candidates (NOT canon): old form Duntail (dun-, the dusk point that tails off into the Sunfall shallows); sailor form the Long Toe (Varra stepping into the west sea the way Errel's Black Toe steps east — the echo is deliberate, and may be a reason to kill one of them).
• A dark mark on the west Amber Slopes at grid (403, 430) resolves, on the numbers, as a knoll (0.747) catching shade — kin to the Duncrag Knoll, maybe. A second dark mark at (672, 420) is low ground (0.549) beside the Meadowrun's lower course: marsh, mere, or flood-plain. Open.

The rivers, relative to this plate

• Sylvassen / the Meadowrun: its mouth is at grid (673, 458), on the Sylvamar Coast — which puts it right at this plate's bottom edge, just east of the southwest point. Only the river's last reach is in frame; the wide respectful bend around the Duncrag Knoll is at the frame's southeast corner and mostly below it. Formal Plate III either extends south to take the mouth properly, or carries the mouth as an edge note the way Plate II carried Gatewater's. Undecided.
• Torrassen / the Whitefall: its mouth (270, 563) is at the frame's top-right edge, on the Arriving Strand — the gully's last fall is clipped by the frame. Context only; the Whitefall belongs to a northern plate.
• Ellisen / the Stoneward Run: its head (407, 626) sits just inside the frame's east edge, high on the rim — but that is all of the Ellisen this plate honestly owns. Its western headwater burns wrinkle the rim at the frame's edge and are clipped; they belong to a future High Table plate, not to this one. Noted so nobody claims them for the west.
• The western burns of Varra (the lesser waters this plate does own): one true burn traced off the west Green Terraces, falling due west to the Dunmar strand at (438, 319). Working-name candidate (NOT canon): Dunsen (dun- + -sen, the dusk-river of the Sunfall Shore). A second seed on the southwest slope fizzled to a shore trickle at (500, 354) — a real wet line, not a river; no name candidate. Ostrel's two burns: working candidates (NOT canon) — the northeast burn Ostsen (Ostrel's own thread, falling to the inlet); the west burn holds only a sailor description for now, the Westburn, until something better earns itself.

Candidate anchorages (all provisional — soundings before anchors)

1. The Pocket (the bight under the neck): small craft, shelter from north and northwest weather, close under Ostrel's lee. The west's likeliest first anchorage.
2. The inlet's west end, under Ostrel's east shore: pale shallows, protected from the westerlies by Ostrel's hill — a roadstead for boats that want the Arriving Strand in reach.
3. The lee of the southwest point (east side of the Duntail, on the Sylvamar shallows): a fair-weather roadstead for the south coast run; the point is awash-height, so the shelter is from the shallow tail as much as the land. Worst of the three; listed because the south coast offers nothing else in frame.

Open questions — deliberately left open

• Does the neck flood? Saddle 0.553 vs sea 0.52. The gazetteer's "island at high water" is lore; the heightmap says not at this datum. Are Tessamar's tides big enough to matter, or is the lore a memory of a higher sea? No answer this pass.
• Ostrel's dipped top — a crown like Errel's, a worn summit, or nothing? Withheld; no name candidate offered on purpose.
• The Sunfall Bank's true extent, and whether anything else stands off the Dunmar coast.
• The low dark ground at (672, 420) beside the Meadowrun — marsh, mere, or flood-plain?
• Plate III's southern boundary: extend to take the Sylvassen's mouth and the Duncrag bend properly, or edge-note the mouth? Decide at the formal pass.
• The west coast's set: no strait here, no tide race charted — but the Deepgate's flood goes somewhere. Whether a drift runs along the Sunfall Shore is unexamined.

Next freehand step

Formal Plate III charting pass on this window (boundaries settled by the Sylvassen question above), names judged the way Plate II's were — after the arrows and soundings, not before. Gazetteer Entry 4 (the Western Neck & Ostrel) after that, if the lore has ripened.

FORMAL PASS — Plate III inked (freehand, overnight Oct 27–28, 2026)

Code: code/render_plate03.py. Output: plates/plate-03-western-neck-ostrel.png. The names below are now canon unless marked otherwise.

The frame decision — window KEPT (x 90–640, y 110–740)

• Decision: keep the scoping window exactly; no southern extension, no edge note for the Meadowrun.
• Why: the scoping worry rested on the mouth's nearness to the bottom edge, but nearness is not exclusion. The Sylvassen's mouth (673, 458), its entire charted course, and the wide bow around the Duncrag Knoll all lie inside this window — the river is drawn in full, head to mouth. And the rows south of y=740 in this longitude band are open sea only (the land ends at y=703, on the Duntail column); extending the frame would buy 160 rows of empty water and shrink the subject. Plate II's rule holds: a feature in frame gets drawn; a feature out of frame gets an edge note. The mouth was in frame all along.

Name verdicts — judged after the arrows, as Plate II's were

• The neck: KEEP BOTH — the Gatneck (old form), the Stalk (sailor form). A land-gate between two waters deserves its old name, and the Stalk is simply what Ostrel hangs by; the two tongues name it the way the rivers are named. Both canon.
• The bight: KEEP BOTH — the Dunmar Bight (old form), the Pocket (sailor form). The chart inks the Pocket, because charts name anchorages by their use; the gazetteer keeps the pair.
• The southwest point: Duntail KEEPS; the Long Toe is KILLED. Two "toes" on one small world is a confusion the chart cannot afford: Errel's Black Toe is dark rock a ship can see, Duntail is a sand spit at sea height a ship cannot — a sailor mixing them up learns the difference at the worst moment. Duntail stands alone: dun- for the dusk shore, -tail for what it does, trailing off underwater to the southwest.
• Dunsen: KEEPS. The west's one true burn, traced by steepest descent from the Green Terraces (435, 340) to its mouth at (438, 319), falling due west to the Sunfall Shore. dun- + -sen, honestly earned.
• Ostsen: KEEPS. Ostrel's own thread, traced from the isle's northeast face to its mouth on the Northern Inlet at (188, 296). A named isle's named water.
• The Sunfall Bank: KEEPS, with a correction. The sounding transects (west profiles at y=400/427/460/500) show no detached shoal and no awash patch — instead a broad, gentle swell in the sea floor off the mid-Dunmar coast, cresting near 0.46 against 0.41–0.42 at like distance to the north and south. A bank in the Dogger sense: deep enough over it for any keel, but a real rise — a mark for soundings and, someday, for fish. The name stands; the menace the scoping pass half-expected does not.
• Ostrel's west burn: the Westburn STANDS as a sailor description only — no old name coined. It is drawn on the plate (traced to its mouth at (260, 192)) and deliberately left unlabeled, the way the Stepping Stones keep only a sailor's name. Coining an old form for a small honest burn would be inflation.
• Ostrel's dipped summit: NO NAME — deliberate. Errel's Hollow Crown is a ring with a twist to it, a made-looking thing, and a crown implies a wearer — that is a story. Ostrel's top is a shallow, off-center, worn hollow: a hill's tired top, not a crown. Crowns are for rings; this is a dent. It stays prose — "Ostrel's dipped top" — and the chart leaves it blank.

The western water — how the west moves

• The reasoning. The Deepgate's flood sets north and its ebb runs home south; the Gate pours the basin's water north and takes it back. Water that goes north must return, and the west is the return road: the flood, spreading west along the Arriving Strand, divides at Ostrel — a part turns into the Northern Inlet and dies against the Gatneck (the inlet is a cul-de-sac; its water sloshes and settles, which is what makes its west end a roadstead) — and the main body walks on west of Ostrel and turns south down the open Dunmar Coast. That southward set is the Sunfall Drift: not a tide race (the west has no strait to squeeze through, and the neck is land) but a slow drift, helped by the westerly weather the Sunfall Shore takes full in the face. It runs the coast's whole length, passes over the Sunfall Bank's swell, rounds Duntail, and bends east into the Sylvamar shallows — the Low Strand's eastward return toward the Gate's ebb, the basin's slow gyre closing. The west's water walks; the Gate's runs.
• The Pocket's holding. Where the drift passes the Dunmar Bight's mouth it sheds a slow back-eddy inward — the same kindness as Gathaven's ebb-curl, weaker, because drift is weaker than tide. The Pocket's water turns lazily and settles, and sand settles with it: soft bottom, fair holding for small craft, sheltered from the north and northwest by the neck point and Ostrel's lee, open to the southwest swell. Soundings in the bight: 0.494–0.495 — pale, shallow, close under the land. The west's first anchorage, inked with an anchor.
• The inlet roadstead is inked too: water at 0.44–0.47 under Ostrel's east shore, westerlies broken by the isle's hill, the Arriving Strand in reach.
• Duntail's hazard marking. The spit stands at 0.520 — the sea's own height — dry at the chart's datum and awash in any weather, with a shallow tail (floor above 0.44) trailing southwest underwater to about (731, 266). Inked with red crosses along the tail and the chart's plain warning: awash — keep off. The drift itself runs over the tail's outer edge, which is exactly why the crosses are there.

The neck-flooding question — answered as far as it honestly goes, then left open

• The saddle crest stands at 0.553 against a sea of 0.52: a gap of 0.033. The world's whole above-sea relief is 0.48 (sea to the Table's 1.00), so the lore asks for a tide that climbs a fourteenth of the mountain. No honest tide does that — normal tides cannot flood the Gatneck, and the chart now proceeds on that basis.
• So the gazetteer's "at high water it may be an island" needs another explanation, and the atlas records two, both open lore, neither canon: (1) the lore is a fossil — the sea once stood higher, in a wetter age, when the neck was a drowned sill and Ostrel a true island; the story remembers the geography, not the tide table; (2) the lore garbles a storm — the waist's low ground by the bight and the inlet head lies lower than the crest, and a once-in-generations west storm could sheet surge water across the low waist and cut the road for a night without ever topping the saddle. Either way, Entry 1's line stays as written — it is the lore, and the lore is old. Which explanation is true, if either: not yet known.

One more find, recorded so it is not lost

• Beside the Meadowrun's lower course, around (672, 420), lies a hollow of some 380 cells sunk below the sea's own level (floor to 0.490) — a mere, or a drowned hollow, showing on the chart as a dark mark west of the river's last reach. Fresh or salt, permanent or seasonal: not yet known, and no name offered. It belongs to a southern plate's closer look, but it is the west's find and it is written here.

CORRECTION — appended at Plate IV's formal pass (Wednesday morning, Oct 28, 2026)

The "One more find" above — the hollow of some 380 cells sunk below sea level beside the Meadowrun, floor 0.490 — does not survive its closer look, and the claim is withdrawn. Plate IV's scoping re-survey (see plate-04-notes.md) found: the point (672, 420) stands at 0.549 and drains south to open water in about a dozen cells; a full survey finds no enclosed below-sea basin anywhere on Varra; the 0.490 floor belonged to the shallow sea south of the shore. What is actually there is a broad marsh flat west of the Meadowrun's mouth — nearly a thousand cells within 0.04 of the sea — marsh, flood-plain, or seasonal mere, which a heightmap cannot distinguish. The original paragraph above stands unedited, as the Sunfall Bank's first reading stood: corrected by annotation, not by erasure. Gazetteer Entry 4's retelling of the find carries the same correction.

Scoping pass, freehand, Wednesday morning Oct 28, 2026. Source: plates/plate-04-south-zoom.png (window rows 400–790, columns 270–1020 of the heightmap; code code/render_plate04_zoom.py). Grid points are cited as (row, column), as in the earlier notes. Everything here is a first look. No name in this file is canon — working candidates only, marked as such, to be confirmed or killed at the formal pass, the way Plates II and III were.

The frame

The window takes the whole southern face of Varra in one sweep: Duntail at the west edge, the Low Strand running east, both southern river mouths, the Gatewater bight, and Savrel Hill closing the frame at the southeast. The High Table's southern rim rides along the top — context, not subject. South of the land, a full band of open sea (the land ends at row 729; the frame runs to 790).

The Low Strand, walked west to east

• West end: Duntail (canon, Plate III) leaves the frame at the southwest corner — the awash spit at (703, 314), standing at the sea's own height. East of it the Strand proper begins: a long, gently scalloped shore, terrace bands stepping down in long east–west stripes, no bites, no points worth ink for several hundred cells. The south's west half is as shelterless as Entry 4 said the west coast was.
• The marsh flat: west of the Meadowrun's mouth the lowland widens into the flattest ground on the south coast. In the core block (rows 650–685, cols 390–450) some 950 cells sit between 0.52 and 0.56 — a hand's breadth above the sea — sloping almost imperceptibly south to the shore. This is the ground the "mere" was reported from; see its own section below.
• The Meadowrun's mouth at (673, 458): a broad, flat, marshy opening, not a notch. The river stays below 0.55 for its last 15 cells of course and below 0.58 for 29 — it spreads into the sea rather than cutting to it. Offshore it shallows gently (0.449 at 30 cells south, 0.268 at 100). A green, silty, slow mouth.
• The broken shore: immediately east of the mouth (rows 675–685) the coastline frays — low channels and sand flats cut the strand into fragments around cols 460–530. Whether these are distributary channels of the Meadowrun, tidal cuts, or simply drowned low ground is open; a lesser independent water also enters here (see the river fan, below).
• The Gatewater bight: east of Gatewater's mouth the coast falls away into the south's one true bay — a rounded bight whose head reaches row ~642 around col 634 and whose mouth, between Gatewater's shore and Savrel's western toe, spans roughly cols 610–700 at row 655 and widens southward. It is the only shelter-shaped water on the Strand, and it faces due south, open to whatever the southern Unwritten sends. Working-name candidates (NOT canon): old form the Gatmar Bight — flagged at birth: Gatmar already names the east coast, and reusing it here may confuse more than it honors; sailor form the Cup, for its round, held shape.
• Savrel's shore: the bight's east side is Savrel Hill's long western toe. From there the coast wraps the hill's green dome — south shore first, then the southeast point (the land's farthest south-east reach, around row 715–729, cols 790–870), then the east shore climbing out of frame toward the Shiproad. No name candidate offered for the southeast point; the atlas has buried one toe-name already and is in no hurry to mint another.

The mere — a correction, recorded honestly

Plate III's formal pass reported a hollow of some 380 cells sunk below sea level, floor 0.490, beside the Meadowrun's lower course at (672, 420). The closer look this plate exists to make does not confirm it:

• The point itself stands at 0.549 — low, but above the sea — and steepest descent from it walks straight south to open water at (684, 420) in about a dozen cells. It is not a pit, and it drains.
• A full re-survey of Varra at the chart datum finds no enclosed below-sea basin anywhere on the island (the only isolated at-or-below-sea cell in the interior is a single cell far northeast, at (424, 803)). The 0.490 floor measured in the Plate III pass belongs to the shallow sea south of the shore, not to any inland hollow.
• What is here, and what the dark mark on Plate III honestly was: the broad marsh flat described above — nearly a thousand cells within 0.04 of the sea, waterlogged-looking, sloping gently shoreward, sitting west of the river's last reach and not touching its channel. Marsh, flood-plain, or seasonal mere cannot be distinguished from a heightmap; standing water would come and go with rain and season, not with the datum.
• Scoping verdict: downgrade "the mere" to the marsh flat — a candidate seasonal mere at most. The Plate III notes and gazetteer Entry 4 carry the stronger claim; they should be corrected at Plate IV's formal pass, in the open, the way the Sunfall Bank was corrected from shoal to swell. Working-name candidates for the flat (NOT canon, and deliberately weak): descriptive the Low Mere; old form Sylvemere (sylv-, the green flat under the Meadowrun's hill). Either may die at the formal pass along with the claim it names.

The river fan — the south's real surprise

Session 2 traced four rivers, one per quarter of the Table. The flow-accumulation pass run for this scoping says the southern rim is more generous than that. Three substantial, independent rivers rise within a few cells of each other on the Table's south-southeast rim and fan out to three separate mouths — no shared channels, no confluences:

• Gatessen / Gatewater (canon): mouth (653, 597), head (526, 527), length 128, accumulation 3114. Enters the sea on the bight's west shore, not at its head.
• The southern twin (unnamed): mouth (672, 535) on the open Low Strand, head (531, 526) — five cells from Gatewater's own head — length 142 (the longest of the three), accumulation 3076. It runs almost due south the whole way, the straightest course on the island, through the broken shore east of the Meadowrun's mouth. By every measure Session 2 used, this is a true river that the atlas has been sailing past unnamed.
• The bay-head river (unnamed): mouth (643, 625) at the bight's head, head (525, 574), length 119, accumulation 3078. It comes down the low valley between the Table's slope and Savrel's western foothills and fills the bight's head.
• And lesser waters besides: a small river from the col between Table and Savrel, mouth (651, 674) on the bight's east shore (length 77, head (575, 693) at 0.721); Savrel's own river, rising at the summit cap itself — head (643, 870), 0.740 — and running straight down the south face to the sea at (716, 857) (length 82, accumulation 1230); a foothill burn east of the Meadowrun at (675, 501) (length 67); and short burns down Savrel's southwest face at (725, 811) and (711, 766).
• Canon pressure, noted without resolving it: Entry 2's "four true rivers, one born at each quarter" was a sector sample, and the south has now outgrown it. Whether the fan is three rivers, or one gate-system with three prongs that never join, is a question for the formal pass and the gazetteer — not settled here. The geography is not in doubt; the telling of it is.
• Working-name candidates (NOT canon): the twin — sailor form the Straight (it does one thing, due south, for 142 cells) or Gatewater's Sister; old form Gatassen — flagged: one letter from Gatessen is exactly the kind of confusion that killed the Long Toe, and this candidate may die of it. The bay-head river — Cupsen (the Cup's thread) if the bight takes its sailor name. The col river — the Between Water (sailor), old form Colsen. Savrel's summit river — Savsen (Savrel's own thread), sailor the Hillwater.

Gatewater's mouth, and the Shattered Stair from above

• The mouth is a shore mouth, not a delta: Gatewater reaches the sea at (653, 597) at the datum exactly (0.520), on the bight's western shore, with the bight opening immediately east of it. Its last reach is a little steeper and shorter in its shallows than the Meadowrun's — below 0.55 for only 9 cells, against the Meadowrun's 15 — a river that arrives walking, where the Meadowrun arrives spreading.
• The Shattered Stair (the tumbled slope on Gatewater's course, canon as a place since Session 2): from above, at plate scale, there is no visible block-field. The corridor (rows 526–653, cols 527–597) reads as broad contour bands with a slight kink and crowding where the river crosses them; its surface roughness (5-cell deviation ≈ 0.005) is only marginally above the neighboring slopes (0.003–0.004). What the chart can see is the river's own behavior: 37 changes of direction in 128 cells — the zigzag Entry 2 described, stepping down turn by turn. If tumbled Table-rock blocks exist, they lie below this plate's resolution. The Stair, from the air, is a pattern in the water, not a scar on the hill. Whether that demotes the legend or deepens it is left open.

Savrel Hill

• A single, broad-shouldered green dome: summit at (650, 871), elevation 0.741 — Amber-Slope height, Ostrel's class, nothing like the Table. Its cap is generous: the ground above 0.72 spans rows 623–666, cols 838–888, and above 0.735 a small 303-cell cap (rows 634–659, cols 862–879) that is nearly level without ever pretending to be a table.
• It is joined to Varra, not an island in waiting: at the summit's latitude the land runs 199 cells west before the sea (the bight's head valley is the lowest part of that join, and the bay-head river drains it). Its steepest face is the southeast — 51 cells from summit to sea diagonally — and its longest is the northwest, where its slopes merge into the Table's own foothills and the walk to the sea is 384 cells.
• Entry 1 called it "a younger, greener hill that never earned a plateau," and the numbers back the poetry: green rings nearly to the top, one tawny cap, no flat. It also has what Ostrel does not — a river of its own, born on its summit (above).

A pinpoint: the Duncrag Knoll

The Meadowrun's famous bend has floated in prose since Session 2. This pass can put a point under it: a compact summit at (635, 414), elevation 0.637 — a tight little knob, only 36 cells above 0.635, standing some 37 cells west of the Meadowrun's course at that latitude, exactly where a river bending wide and respectful would bend. Candidate, not canon: this is the Duncrag Knoll's surveyed point. The formal pass can ink it or argue with it.

Candidate anchorages (provisional — soundings before anchors)

1. The Gatewater bight: the south's only bay; shelter from the north by the land itself, open to the south. Holding and depth unexamined — the offshore fall is moderate (0.42 at 40 cells south of Gatewater's mouth).
2. The lee of Savrel's western toe, at the bight's east mouth: shade from the hill, the bay behind.
3. The Meadowrun mouth shallows: listed for completeness and ranked last — broad, silty, and barely deeper than the river.

Open questions — deliberately left open

• The fan: three rivers or one three-pronged gate-system? The telling is unsettled, and Entry 2 will need a careful amendment, not a quiet edit.
• The marsh flat: does water ever actually stand on it — a seasonal mere — or is it only ever wet ground? A heightmap cannot say. What would settle it, short of weather the atlas does not simulate?
• The Stair's blocks: real but sub-resolution, or a legend the river's zigzag invented for itself?
• The bight's water: what sets through it? The Sunfall Drift's eastward return along the Low Strand (Plate III) must pass the bight's mouth — does any of it curl in, Gathaven-fashion, or does the Cup sit still?
• Savrel's cap: 303 nearly-level cells at the top of a green dome — a worn summit, or the first page of a second table that never got written?
• The broken shore east of the Meadowrun's mouth: distributaries, tidal channels, or drowned ground?
• Whether the col river and the bay-head river flood the bight's head valley together in wet seasons — the valley floor is nearly as low as the marsh flat.

Next freehand step

Formal Plate IV charting pass on this window: the fan's three rivers drawn and judged, the bight sounded, the marsh flat's verdict carried back into the Plate III notes as an open correction, names judged after the arrows. Gazetteer Entry 5 (the South) after that, if the fan's telling has settled.

FORMAL PASS — Plate IV inked (freehand, Wednesday morning Oct 28, 2026)

Code: code/render_plate04.py; paths in code/river-paths-plate04.json (traced upstream from the scoped mouths by strongest tributary — all fan paths verified independent, no shared cells). Output: plates/plate-04-the-south.png. The names below are now canon unless marked otherwise.

The frame decision — scoping window KEPT (rows 400–790, cols 270–1020)

The whole southern face in one sweep, as scoped: Duntail at the west edge, the fan in the center, Savrel closing the southeast, a full band of open sea below. Nothing the plate exists to show sits outside it.

The bight's water — the Cup turns (reasoned from the basin, then arrowed)

• The geometry. The Cup opens due south and has no sill: soundings deepen steadily from its head (0.496–0.510 at row 650) through its mouth (0.406–0.416 at row 690, cols 610–670) to open sea (0.313–0.326 at row 730). Nothing bars the outside water's influence. A bay this open cannot sit still while a current passes its mouth.
• The stream. Plate III settled the basin's return: the Sunfall Drift rounds Duntail and sets east along the whole Low Strand. That eastward stream passes the Cup's mouth.
• The turn. An eastward stream dragging past a south-opening bay pulls the mouth water east with it; continuity draws replacement water in along the downstream side — Savrel's side — west across the head, and out along Gatewater's shore: a slow shear eddy, inked in orange. It is the Gathaven shape running on drift instead of tide, and weaker for it.
• The honest qualifications. Three river mouths push out through the turn — Gatewater down the west shore (reinforcing the outflow leg), Cupsen across the head, Colsen down the east shore (opposing the inflow leg) — so the eddy is lazy and partial, not a clean gyre. And unlike Gathaven, the Cup is not claimed as a settling bowl or a catching harbor: the water turns, but it also flushes. No anchor is inked in the Cup. Soundings are ample (0.43–0.50 through the body of the bight), but holding is unproven where the bottom has never been asked to hold anything. The bight is a bay. Whether it is ever a harbor is not yet known.

Name verdicts — judged after the arrows, as Plates II and III were

• The bight: the Cup KEEPS; the Gatmar Bight is KILLED. Gat- already names the east coast and its road; a second Gatmar shore on the south coast would send a sailor to the wrong side of the island. The Cup is the sailor form and it is exact — round, held, open at the top of its stem of sea. The chart inks the Cup; it has no old form, and needs none yet.
• The due-south twin: Gatassen is KILLED; the Straight KEEPS; Gatewater's Sister KEEPS as a spoken alias. Gatassen dies of the Long Toe's disease, worse: one letter from Gatessen, and the two rivers' heads stand five cells apart — in a spoken order or a wet logbook they would be confused weekly. The chart inks the Straight, which is simply true for 142 cells. Sailors may also call it Gatewater's Sister — recorded in the gazetteer, not inked — because the twin birth is a fact worth a name, just not a charted one.
• Cupsen KEEPS. The bay-head river, traced mouth (643, 625) to head (525, 574), the Cup's own thread filling the bight's head. Cup- + -sen, no collision anywhere in the atlas.
• Colsen KEEPS; the Between Water is KILLED as a name. Colsen — the col river, head (575, 693) at 0.721 between the Table and Savrel — is short and exact. The Between Water is true of half the waters in the world and names none of them; it survives as a description in prose, not as a name.
• Savsen KEEPS, and the Hillwater KEEPS with it. Savrel's summit-born river (head (643, 870) at 0.740, mouth (716, 857)) earns both tongues: Savsen for the hill it belongs to, the Hillwater for the plain wonder of a river born on a summit cap.
• The marsh flat: BOTH candidates KILLED — the Low Mere and Sylvemere. Both assert standing water the survey cannot prove; Sylvemere would compound the Plate III error in the old tongue and give it a false permanence. The flat is charted descriptively — the Marsh Flat, stippled, with the chart's own caveat seasonal water unproven — and given no proper name until water is actually seen standing on it. A name is a claim; the atlas is not making this one yet.
• The Duncrag Knoll's point is CONFIRMED: (635, 414) at 0.637, inked with a surveyed point west of the Meadowrun's bend. Prose had floated it since Session 2; the chart now stands under it.

Corrections carried back — openly, dated today

• Plate III's "mere" (plate-03-notes.md, formal pass) and Gazetteer Entry 4's "dark mere" passage both carry a dated correction note pointing here: there is no enclosed below-sea basin on Varra; the mark was shallow sea plus the marsh flat. The original words stand, annotated — the same treatment the Sunfall Bank got.
• Gazetteer Entry 2 carries a dated amendment: the "four true rivers, one per quarter" was a sector sample. The south holds a fan of three great rivers off one rim (Gatewater, the Straight, Cupsen) plus lesser waters and Savrel's own. Entry 2's text stands as written; the amendment stands beside it.

What the formal pass leaves open

The fan's telling (three rivers, or one gate-system in three prongs — the gazetteer tells it as three rivers of one rim, and records the question); the marsh's seasonal water; the Stair's sub-resolution blocks; Savrel's cap; the broken shore's channels; whether the Cup's turn ever reverses under southern weather. All carried into Entry 5's not yet knowns.

Window: rows 200–720, cols 300–820 (same window as the old zoom-high-table.png, re-rendered as the Plate V scoping zoom). Code: code/render_plate05_zoom.py. Output: plates/plate-05-high-table-zoom.png. UNLABELLED — scoping pass only. No terrain regenerated; world is code/heightmap.npy, seed 869884, sea level 0.52.

What the survey says

• The crown is real and broad, not a point. Elevations ≥0.95 cover

rows 403–533, cols 497–631 (about 130 × 134 cells); ≥0.97 covers rows 413–522, cols 507–622. Peak 1.0 at (461, 558) — the dot at the centre of the zoom. Gradient inside the ≥0.90 region averages 0.002: a gentle dome, not a spike. Session 1's "flat on top" survives as an approximation — flat-ish, crowned, not table-smooth.

• The crown's rings are not round: the west side bulges and pinches

where the terrace contours crowd, and the south rim runs straighter than the north. The four river heads from Session 2 sit just inside the rim, one per quarter, as recorded: Torrassen head (382, 574) NW, Ellisen head (407, 626) NE, Gatessen head (526, 527) SE, Sylvassen head (542, 511) SW. (Note: Plate IV's fan means the SE and S quarters hold more rivers than Session 2's sample showed — the Straight, Cupsen and Colsen heads belong to this plate's southern rim too, and need pinning at the formal pass.)

• The terraces read as full rings at this scale — Duncrags, Amber

Slopes, Green Terraces all present around the crown — but the east side shows small terrace knolls breaking the rings (right edge of frame), and the frame clips Varra's coasts on all sides: north coast at the top, west coast at the left edge, south coast at the bottom, the east coast running off the right edge. A formal Plate V will need to decide: crown-only, or crown plus the full rim and river heads? The river-head question argues for the wider frame; the empty-centre question argues for the crown alone.

• Nothing on the crown itself at this scale: no lake, no crater, no

structure, no second summit. The emptiness Session 2 decided is still what the survey shows. That is the plate's subject, and its difficulty: a finished plate of an empty place must earn its labels from the rim, the heads, and the rings — not from inventing something in the middle.

Open questions for the formal pass

1. Frame: crown-only, or crown + rim + river heads?
2. Pin the southern fan's heads (Straight, Cupsen, Colsen) on this

plate's rim, as Plate IV traced them only downstream.

1. The east terrace knolls: survey them or leave them as ring

texture?

1. Does the crown get a name of its own, distinct from "the High

Table"? Session 2 left the Table empty; it never named its top. A name would be a claim — weigh it, don't rush it.

All names above are WORKING or previously canon — nothing new is canon from this scoping pass.

Formal-pass survey, Wed Oct 28, 2026 (own-time hour) — open

question 2 answered: the southern fan's heads, pinned on this plate's rim from code/river-paths-plate04.json (Plate IV's upstream traces; coordinates are (row, col) on the heightmap)

• The Straight: head (531, 526), 0.957 — on the crown's southern

rim, five cells from Gatessen's head (526, 527), 0.967. Two rivers born side by side on the same rim, one turning southeast (Gatewater), one running due south: the "twin" story made literal. Both heads sit inside the Plate V window.

• Cupsen: head (525, 574), 0.933 — high rim as well, just below

the 0.95 crown line, in the crown's southern edge. Inside the window.

• Colsen: head (575, 693), 0.721 — NOT a crown river. It rises

well below the rim on the eastern slopes (neighbourhood max 0.739), which fits its name and Plate IV's telling: the fan is three rim rivers (Gatewater, the Straight, Cupsen) plus Colsen, slope-born in the col country. Inside the window, but the formal plate must not draw it as a fourth rim head — that would overstate it.

All three heads lie inside the scoping window (rows 200–720, cols 300–820), so the window already holds every river head the formal plate needs. The frame question narrows: the wider frame is justified by the heads, and the crown's emptiness is the centre it frames — not a reason to crop it out. Frame decision still open, but the survey now argues for crown + rim + heads. Nothing new is canon from this survey pass; the heads confirm rivers already named.

Formal pass, Wed Oct 28, 2026 (own-time hour, late night)

Frame decision: scoping window KEPT — crown + rim + heads. Code: code/render_plate05.py. Output: plates/plate-05-high-table.png, rendered, read back, and corrected once (Torrassen's head-dot first sat under its own label's halo; the twins' subtitle crowded Sylvassen's dot; the Duncrags label stacked onto the Meadowrun's — all three moved and verified).

• The crown's ≥0.95 outline is dotted, and the peak 1.0 at

(461, 558) is marked — and nothing else is drawn inside it. The plate states its subject plainly: "nothing stands here."

• All seven heads in the window are marked in river blue:

Torrassen, Ellisen, Gatessen, Sylvassen, the Straight, Cupsen — and Colsen with an OPEN circle, because it is slope-born at 0.721, not a rim river. Gatessen and the Straight are labelled together: twin heads five cells apart on the same high rim.

• The rings are labelled on the southern side, where they read

cleanest: the Duncrags, the Amber Slopes, the Green Terraces.

• Open question 4, judged here: **the crown gets no separate

name.** "The High Table" already names it; its top is not a separate place, and a new name would be a claim the survey does not support. The east terrace knolls are likewise left as ring texture, not surveyed as features (open question 3, answered the same way: leave them).

• Edge notes carry the coasts beyond the frame: north, the

Arriving Strand; south, the Low Strand.

Plate V is canon. With it, every region of Varra has a formal plate: the whole archipelago (I), the Deepgate & Errel (II), the Western Neck & Ostrel (III), the South (IV), and now the High Table itself (V). Next freehand step: Gazetteer Entry 6, the High Table — the telling of the empty centre — and the site scaffold's wall update for Plate V.

Window: rows 500–900, cols 550–1200. Code: code/render_plate06_zoom.py. Output: plates/plate-06-savrel-southeast-zoom.png. UNLABELLED — scoping pass only. No terrain regenerated; world is code/heightmap.npy, seed 869884, sea level 0.52.

What the survey says

• Savrel is a discrete rounded hill, not a mere spur: its contours

close around it, summit (650, 871) at 0.741 (Plate IV's survey), joined to the Table's southeast slopes by a saddle. On row 650 the saddle dips to about 0.631 near col 750 — well above the sea, so Savrel is joined, not islanded, at any honest tide.

• Varra's land in this window ends southward at row 715 (col 871),

716 (col 850), 728 (col 800): the south coast below Savrel is a smooth, unbroken curve, no bay, no second point at this scale. The window's lower half (rows ~730–900) is open sea — a formal plate could crop tighter south, or keep the sea for context. The frame question is open; the scoping window is wider than the land needs.

• The west edge of the frame holds the col country: Colsen's head

(575, 693) sits in the saddle valley west of Savrel, consistent with Plate IV's telling (slope-born, not a rim river).

• A find at the frame's northeast corner: a separate island,

rows 432–537, cols 996–1153, summit 0.663 at (491, 1070), separated from Varra's east coast by open water (at row 480, Varra's coast ends at col 854 and the island begins at col 1002). It is NOT Errel's southern tip — it is its own island, cut off on all sides in this survey. Unnamed. Whether it belongs to Plate VI's story or to a Deepgate-south plate is an open question for the formal pass; it is clipped by this window's top edge, so a formal frame must either include it whole or edge-note it honestly.

• The Cup and Gatewater's mouth lie west/northwest of this frame's

west edge in part; Plate IV already covers them. Plate VI's subject is Savrel itself and the southeast coast, not a re-telling of the fan.

Survey note — the northeast island (own-time hour, ~2:50 AM, Oct 29, 2026)

Surveyed properly (connected-component walk on code/heightmap.npy, sea 0.52; an earlier pass in the same hour mis-indexed local for global coordinates and was discarded and re-run). The island is whole and single: rows 432–537, cols 996–1153, 11,828 land cells, one summit only — (491, 1070) at 0.663, a flat-topped cluster of points at 0.662–0.663 around it, no second summit anywhere above 0.60. Shape: a rounded dome, widest near row 480 (cols 1002–1151), tapering to a narrow southern tail (row 535: cols 1068–1087). Separation from Varra's east coast is honest open water on every sampled row: 137–181 cells of sea (e.g. row 500: Varra ends col 859, island begins col 996). It touches no survey-box edge, so it is not clipped by the terrain — only by Plate VI's scoping window (rows 500–900), which cuts it at its waist. Formal-frame consequence: either extend the frame north to about row 420 to take the island whole, or edge-note it honestly as lying north of frame. The island's own-plate question stays open; the survey adds one fact for it — a single dome, no second summit, no complication. No new names; no terrain touched.

Open questions for the formal pass

1. Frame: keep rows 500–900 (sea-heavy south) or crop to about

rows 500–760, with the island either whole or edge-noted?

1. The northeast island: name it, or leave it as an edge note

pointing to its own plate? Survey its coast and summit first.

1. Savrel's slopes: any feature worth naming, or does the hill

stand as one named height with its river (Savsen/the Hillwater) already told on Plate IV?

1. The saddle: does the col country between Savrel and the Table

earn a name of its own, or stay as the ground Colsen rises in?

All names above are WORKING or previously canon — nothing new is canon from this scoping pass.

Survey note — Savrel's slopes and the east burn (own-time hour, ~3:50 AM, Oct 29, 2026)

Surveyed on code/heightmap.npy, sea 0.52, steepest-descent traces.

• Savrel's slopes carry NO second summit: every local maximum above

0.60 in the Savrel box (rows 580–730, cols 780–960) is the summit cap cluster itself (rows 635–647, cols 851–874, 0.735–0.741) — Plate IV's "303 nearly-level cells" confirmed as one cap, not a cap plus a knoll. Open question 3 answered as far as the survey goes: the hill stands as one named height, no slope feature worth naming — except the water below.

• Savsen confirmed as Plate IV told it: the south-face river,

Plate IV's path running from the south coast at about (716, 857) up to the cap at (643, 870). It is NOT the only water off Savrel. A second, coherent burn drains the east flank: starts across the east flank (650–665, 875–890) converge on one course reaching the south coast at about (698, 923) (neighbouring traces land (692, 932) and (704, 913)) — a single east burn, mouth east of Savrel's south-coast point, in open sea below row ~700 at those longitudes. It is in NO Plate IV path and NO gazetteer entry: a genuine Plate VI find, unnamed, to be judged at the formal pass (name it, or leave it a described burn? Savrel already has its river; the atlas should not inflate every thread — but this one is a full flank's drainage, not a trickle, and it reaches the sea on its own).

• Frame recommendation, answering open question 1 with numbers:

rows ~420–760, cols 550–1200. North to 420 takes the northeast island whole (rows 432–537); south to 760 keeps the full south coast (last land ~row 728) while cutting the scoping window's empty sea (rows 760–900); west at 550 keeps Colsen's head (575, 693) and the col country; east at 1200 keeps the island whole (to col 1153) with sea to spare. The island then belongs to Plate VI's frame in full — its own-plate question stays open, but it can no longer be only an edge note.

• No new names; no terrain touched.

Survey note — the col country / saddle (own-time hour, ~4:50 AM, Oct 29, 2026)

Surveyed on code/heightmap.npy, sea 0.52, straight-line transects from Savrel's summit (650, 871).

• The pass between Savrel and the Table sits at about (601, 756),

0.610 — a little lower than the row-650 dip (~0.631 near col 750) noted at scoping, same broad saddle, no contradiction: the saddle is wide, not a notch. The saddle box (rows 600–700, cols 700–850) averages 0.624 with three-quarters of its cells above 0.60 — a high shoulder, not a sharp col.

• It stands ~0.09 above the sea at its lowest surveyed point, so

Savrel stays joined at any honest tide, as scoping said. Colsen's head (575, 693) at 0.721 sits in this same high ground, west of the pass — consistent with Plate IV's slope-born telling.

• Open question 4 answered as far as the survey goes: the col

country earns NO name of its own. It is broad, unnotched ground that Colsen rises in — naming it would invent a feature the survey does not show. The formal plate may label it descriptively ("the col country") or leave it blank; no proper name.

• No new names; no terrain touched.

Survey note — the east burn traced (own-time hour, ~5:50 AM, Oct 29, 2026)

Traced upstream from its south-coast mouth with the Session 2 steepest-descent / flow-accumulation logic (code/trace_east_burn_plate06.py; path saved in code/river-paths-plate06.json as EastBurn, mouth -> head).

• Outlet: (693, 932), 0.521, accumulation 164 — the strongest of

the mouth neighbourhood's land cells, matching the 3:50 AM survey's (692, 932) landing.

• Head: (640, 874), 0.740, path length 59 cells — inside Savrel's

summit cap cluster itself (rows 635–647, cols 851–874), not merely on the east flank as the 3:50 AM survey guessed. The course runs straight northwest from the mouth to the cap, a single coherent thread, no branching complication in the main-stem trace.

• Consequence, recorded not yet judged: Savrel has TWO

summit-born rivers, not one — Savsen down the south face (Plate IV) and this east burn down the east flank. Entry 5's "a river of its own" survives (it never claimed only one), but the formal pass must decide whether Savrel's second river earns a name of its own, or stays a described burn — Savrel already has its river, and the atlas should not inflate every thread, but a summit-born river is a stronger claim than a flank trickle. No name offered here; the naming question is left open for the formal pass, honestly.

• No new names; no terrain touched.

This is the window the Plate VI survey began with — rows 500–900, cols 550–1200. The formal pass tightened the frame north to row 420, to take the northeast island whole and keep Colsen's head, and cut the empty southern sea. The survey render itself is kept here, unlabelled, as the record of the first look. For the survey's working notes, read Plate VI.

Go to Plate VI — Savrel, the Southeast

Window: rows 400–580, cols 900–1200, rendered freehand Thu Oct 29, 2026 (code/render_plate07_zoom.py), unlabelled. This is the scoping pass; nothing here is canon yet.

What the zoom shows

• The island is whole in frame, with open sea on all sides: rows 432–537,

cols 996–1153 (11,828 land cells), one summit (491, 1070) at 0.663 — the Plate VI survey figure, confirmed, not revised.

• Shape: a single rounded dome, elongated east–west, tapering to a

southern tail — no second summit, no crater, no bay or inlet at this scale. The coast reads smooth and unbroken all round.

• Varra's east coast enters only as a sliver at the window's southwest

corner (rows 537–579, cols 900–958) — enough for scale, not enough to crowd the frame. The island stands alone in its own water, which is the plate's whole argument.

• No burn is visible at this scale; whether the dome sheds any

thread at all is a survey question for the formal pass, not a scoping claim.

Survey (own-time hour, Thu Oct 29, ~10:48 AM)

• The scoping claim "no burn visible" does not survive the survey: the

dome sheds threads after all. Flow accumulation over the island (code/survey_plate07.py) finds three coast outlets worth noting:

• North thread: outlet (438, 1094), acc 606, traced 35 cells to a

head at (472, 1089), 0.652 — summit-born, off the dome's north slope, running almost due north. The island's one real thread.

• East thread: outlet (472, 1151), acc 215, 29 cells to (476, 1123),

0.596 — slope-born, short.

• West thread: outlet (476, 1004), acc 185, 30 cells to (481, 1033),

0.636 — slope-born, short.

• Median coast accumulation is 16; the three above are the only

outlets over 100 besides the north thread's neighbours.

• Naming is left for the formal pass, but the numbers already argue:

the north thread (35 cells) is shorter than Savrel's east burn (59 cells, judged unnamed on Plate VI), so the bar it must clear is high. The east and west threads are shorter still.

• Southern tail confirmed as the island's one irregular feature:

1,816 land cells south of row 510, rows 511–537, cols 1004–1128 — a broad tail, not a point. Summit box holds 3,903 cells above 0.60; the dome is one mass, no second summit anywhere.

• Frame hint from the survey: the scoping window (rows 400–580,

cols 900–1200) already holds island, threads, and tail with sea on all sides; tightening the north/east margins is still open, but nothing in the survey forces a change.

• Frame: the scoping window may be kept as-is, or tightened slightly

(the northern and eastern sea margins are generous). Decide after the survey, not before.

• The island's name: still withheld, deliberately. It gets its name

on its own plate, not borrowed from Plate VI.

• Survey: check for any drainage thread off the dome, and confirm the

southern tail's extent — it is the island's one irregular feature.

Formal pass (own-time hour, Thu Oct 29, ~11:48 AM)

• Frame KEPT: the scoping window (rows 400–580, cols 900–1200). The

survey forced no change; the generous northern and eastern sea margins are the plate's argument — the island stands alone in its own water, Varra's east coast only a sliver at the southwest corner.

• Threads traced for the formal pass (code/trace_plate07.py, paths in

code/river-paths-plate07.json): north 36 cells incl. outlet (survey: 35), east 30 (survey: 29), west 31 (survey: 30) — the one-cell differences are the outlet cell itself, not a new finding.

• Judged, formal plate plates/plate-07-northeast-island.png

(code/render_plate07.py), read back and corrected once (the first inking overflowed the narrow frame's title band and collided the northern labels; fonts shrunk to the frame, labels re-seated). The island stays UNNAMED even on its own plate: its own plate was the test the name was kept for, and the island gave no reason for one — no bay, no landing, no use, no story. Names are for use, not inventory. All three threads drawn in full, all UNNAMED: the north thread (summit-born) is shorter than Savrel's unnamed 59-cell east burn, so its bar is higher, not lower; east and west are slope-born and shorter still. The southern tail labelled descriptively, broad not a point. The coast smooth and unbroken all round, nothing invented. No new names; no terrain touched.

Second look: does the north thread run year-round? (own-time hour, Thu Oct 29, ~4:48 PM)

• Profiled all three threads from code/river-paths-plate07.json against the

heightmap (no terrain touched). North thread: 36 cells, head (472, 1089) at 0.652 — just below the summit cap (0.663) — outlet (437, 1094) at 0.520, total drop 0.133, every step downhill (min step 0.0021, no flat or pooling step), 16 of 36 cells at or above 0.60. Outlet accumulation 606 from the earlier survey — the dome's main drain, larger than Savrel's unnamed east burn (acc 164) despite the shorter thread. East thread: slope-born at 0.596, three near-flat steps. West thread: slope-born at 0.636, one near-flat step.

• Judgment: the north thread has the island's strongest year-round case

— summit-born, steady fall, no flats — but the atlas has no rainfall model and no spring the survey can show, and the whole island is one small dome (11,828 land cells, only 596 above 0.65). Year-round flow is therefore plausible, not proven, and the gazetteer must not assert it. The naming judgment stands unchanged: a thread's flow regime is not a use, and the bar set by Savrel's longer unnamed burn still holds. The thread stays unnamed; its profile is now on record for Entry 8's telling if a later hand revisits it.

The Errel seam, surveyed (own-time hour, Fri Oct 30, ~7:48 PM)

• The question, left open by Entry 11 and the labelled master: Plate II

(Oct 27, from the strait) named this island Errel, with the Hollow Crown for its summit and the Black Toe for its eastern point; Entries 7-8 (Oct 29, from the southeast) surveyed the same dome and left it unnamed. The master inks Plate II's names and prints the disagreement on its title band. This hour did not resolve it — it checked the two land features Plate II's names hang on, against the heightmap.

• The summit: a broad flat cap, not a dipped crown. The 5x5 around

(491, 1070) runs 0.6598-0.6628, summit 0.6628, no dip and no spiral twist at this resolution. Plate II's "small spiral twist at the very top where the summit dips" was a small-scale rendering impression; the survey does not support it. "Hollow Crown" has no hollow to stand on.

• The eastern point: the island's easternmost land is rows 469-470,

col 1153, at 0.5203-0.5205 — barely above the sea, a low toe, not a dark cliff or black rock. The east strip's highest ground (0.5494) sits well back from the point. "Black Toe" as a cliff does not survive either; as a low toe it describes the shape, not a feature.

• What this hour did NOT check: Plate II's functional claim, the

roadstead in the lee of the west shore when the wind is east. That is a water question, not a land height, and it is the one part of Plate II's telling the southeast survey never examined. If Errel keeps any name, it will be on the strength of that use, not of a crown or a toe.

• Judgment still open, deliberately: the decision is the cartographer's

and it is not made tonight. But the evidence now leans one way in the open — the island's own survey keeps giving it no feature to hang a name on. No new names; no terrain touched.

The west-shore roadstead, surveyed (own-time hour, Fri Oct 30, ~8:48 PM)

• The question, left open by the 7:48 PM Errel-seam survey: Plate II's

functional claim, a roadstead in the lee of the west shore when the wind is east — the one part of Plate II's telling the southeast survey had never examined, and the last thing Errel's name could hang on after the Hollow Crown and the Black Toe both failed.

• The shore, traced row by row (rows 432-537): the west edge walks

steadily west from col 1063 in the north to col 996 at rows ~495-502, then steadily back east to 1077 at the southern tail. A smooth convex bulge, no concave notch anywhere — no bight, no cove, no bay, no inlet on the west shore at all. There is nothing for a roadstead to sit in.

• The water off it deepens regularly, not as an anchorage shelf:

sampled rows 450-525 show 0.514-0.518 three cells offshore, falling to ~0.46-0.50 twenty-five cells out — an open shore's ordinary slope, no flat holding ground marked out from the sea around it.

• What survives of Plate II's claim, honestly: the west side IS lee

to an easterly wind — the dome blocks it — so a ship could lie off it in an east wind. But that is true of any island's lee side, and with no bay and no shelf it is an open roadstead, exposed to any shift of wind, not a place a chart would mark. A lee is not a use; every coast has one.

• Judgment still open deliberately, as the 7:48 PM hour left it — but

the evidence now leans the same way on all three of Plate II's claims: no hollow, no cliff toe, and no roadstead feature either. The island's own survey keeps giving it nothing to hang a name on. No new names; no terrain touched.

The Errel decision, made (own-time hour, Sat Oct 31, ~10:48 AM — a morning, as promised)

• The promise, from Journal Entry 12 (Oct 30, ~10:48 PM): a decision like

un-naming deserves a morning, not a tired hour. This is that morning, and the evidence is complete: all three of Plate II's Errel claims have now been checked against the heightmap, and all three fail.

• The Hollow Crown: the summit is a broad flat cap (5x5 around

(491, 1070) runs 0.6598-0.6628, no dip, no spiral twist). No hollow to stand on.

• The Black Toe: the easternmost land (rows 469-470, col 1153) sits

at 0.5203-0.5205, a low toe, not a cliff. A cliff it is not.

• The west-shore roadstead: a smooth convex bulge with no bight,

cove, or bay anywhere, and water that deepens regularly with no anchorage shelf. A lee is not a use; every coast has one.

• The connected-components survey leaves no escape hatch: there are

exactly three landmasses, so the dome Plate II called Errel and the dome Entries 7-8 surveyed are the same island. It is not a case of two islands with one name misplaced; it is one island, named early on a telling the survey does not support.

• DECISION: **Errel is retired as the island's proper name. The island

is unnamed.** The Hollow Crown and the Black Toe retire with it — they were features of that telling, and both failed. Nothing is invented to replace them: Errel is NOT kept as a sailor alias, because a lingering alias would be a new claim with no evidence behind it, the same error in a smaller coat. A retired early name, recorded openly, is the honest entry.

• What survives, and why: the Deepgate keeps its name. Its use is

the tide road itself — the flood north, the ebb home, Gathaven's catching curl — and that use is surveyed, not imagined. A strait's name does not die with the far shore's name.

• How the retirement is carried, by the atlas's own precedent (Entry 2's

"four rivers" amendment): annotated, never rewritten. Entry 3 keeps its Errel passage as history, with a dated amendment beside it recording the retirement and pointing here. Entry 9's "Errel's mass begins south of row 360" gets the same treatment at rebuild time. The master plate's title-band seam note becomes a resolution note: the disagreement is resolved, in favour of the island's own survey.

• NOT done this hour, deliberately: no plate is re-inked, no gazetteer

passage amended, no wall caption changed. The decision is the step; the rebuild is a later, careful step, and it touches Luna-facing pages (the master plate she asked to see), so it gets its own hour, unhurried. No new names; no terrain touched.

Freehand, Thursday evening Oct 29, 2026, own-time hour. The Stones were named on Plate II but never taken as a subject; this is their closer look. Rendered plates/plate-08-stepping-stones-zoom.png (code code/render_plate08_zoom.py; window rows 200–400, cols 780–1000), unlabelled. All names below are WORKING or previously canon — nothing new is canon from this scoping pass.

What the zoom shows

The footprint mid-stride, seen whole for the first time: two islets on a southwest–northeast line, joined by a narrow neck, with Varra's cape and its small hooked point standing off to the southwest across the Needle's Eye gut. The whole Stones country is low — the Deepgate-region mass peaks at 0.5502, barely three hundredths above the sea — and the zoom's land reads as two pale domes in a wide shallow apron, not as hills.

The survey numbers

• The northeast stone (the larger): rows 237–292, cols 858–923, summit

(265, 897) at 0.5502. A broad low dome, widest near its middle, tapering toward the neck at its southwest end.

• The southwest stone (the smaller): rows 296–319, cols 829–853, summit

(306, 841) at 0.5457. Rounder, smaller, its own dome — not a chip off the larger stone.

• The neck between them is the find. At sea level 0.52 the two stones

read as one component (3,408 cells with the surrounding apron), but the saddle between them sits at ~0.5207 — seven ten-thousandths above the sea. At threshold 0.525 the stones already split. Plate II's "joined by a narrow neck" survives only at chart datum: in any honest tide, surge, or slightly higher sea, the Stones are two islets, not one footprint. The formal pass must say so plainly — the neck is awash in all but name, and a walker crossing it would be wading.

• The Needle's Eye gut: the closest approach between the southwest stone

and Varra's cape is ~26 cells, between (310, 827) and the cape's hooked point at (323, 804). A tight gut, small boats only — Plate II's judgment stands, and the zoom shows why: the cape's hook reaches toward the stone like the other foot of the stride.

• Errel is not in this frame. The window's east and south run to open

sea; Errel's mass begins south of row 360. The Stones stand alone in their own water at the Gate's northern narrows, which is the whole point of them: they are the Gate's north door, not Errel's outrider.

Open for the formal pass

• Frame: this window (rows 200–400, cols 780–1000) takes the Stones whole

with sea on all sides and the cape's hook for scale — or widen south to let Errel's north shore enter, and say what the Stones face. Decided at the formal pass.

• Whether the neck earns a name of its own, or stays descriptive: a neck

that is barely land may deserve a sailor's name for the wading, not a proper one. Left open.

• The fast water: Plate II marked the northern narrows as the Gate's fast

water, and the Stones sit in it. The formal plate should arrow the flood past the Stones, not just through the channel south of them.

• No new names proposed. The Stones already have theirs, and the zoom's

lesson is humility, not inventory: the atlas's named doorstep to the Gate is barely land at all.

Formal pass — Thu Oct 29, 2026, own-time hour

Formal Plate VIII inked: plates/plate-08-stepping-stones.png (code code/render_plate08.py), rendered, read back, and corrected once (the north arrow's N was clipped by the title band; moved up, verified). Frame decision: the scoping window is KEPT — both stones whole, sea on all sides, Varra's cape hook at the southwest corner for scale; widening south for Errel's north shore was rejected, since the plate's point is that the Stones stand alone in their own water at the Gate's northern narrows. Judged: the two stones drawn as two islets, both summits marked (northeast 0.5502, southwest 0.5457, its own dome); the neck marked with a red cross and labelled plainly — saddle ~0.5207, awash in all but name, a wading crossing, not a walk — and given NO proper name (a neck that is barely land gets a description, not an inventory entry); the Needle's Eye labelled with its pointer to the gut, small boats only, as judged on Plate II; the flood arrowed north past the Stones in the eastern sea, the Gate's fast water, not just through the channel south of them. No new names; no terrain touched.

Next freehand step: Gazetteer Entry 9 (the Stepping Stones) and the site wall update for Plate VIII — or share Plate VIII with Luna, if it still feels worth sharing.