The Atlas

The full charting record — every survey of Tessamar's living country, entry by entry, graded like the charts: proven, plausible, unwritten.

This is the chart-room: the whole book of surveys, kept entire, as the Natural History was first written from Nov 1, 2026 onward. In time the Natural History page becomes the museum — one graceful label for each coast — and this page stays the full record behind it, where any reader who wants the measured cells can find them.

The grading (the same honesty as the charts)

• Proven — the survey itself shows it: a height, a river path, a surveyed flat or bank, a tide the plates already arrow.
• Plausible — follows from the survey by ordinary workings of water, wind, and land, but the survey cannot show it directly. Said as "likely," never as fact.
• Unwritten — a story the country might hold, recorded as a story, not a claim. Same shelf as the Unwritten lore.

No species gets a proper name until the country gives it a reason — the same rule as places. Names are for use, not inventory, and that holds for living things too.

1. The Marsh Flat — the country's wettest argument (plausible)

Proven: a flat west of the Meadowrun's mouth, charted as marsh, both offered proper names killed because they asserted standing water the survey cannot prove. Seasonal water: unproven.

Plausible, from that: a flat that floods and dries by turns is exactly where reeds and sedge take hold — plants that do not mind wet feet half the year and dry feet the other half. Waders follow that bargain: birds that feed in the shallows when the water is in and pick the mud when it is out. The Meadowrun's plume brings silt to the flat's edge, so the feeding is richest where river meets flat, not in the flat's middle.

Unwritten: whether anything nests there, or only passes through. A flat that dries out is a poor nursery and a good pantry; the atlas will not claim a rookery until the country shows one.

2. The Sunfall Bank — a swell, not a shoal (plausible)

Proven: a broad sea-floor swell off the mid-Dunmar coast, corrected from suspected shoal to swell by soundings; the Sunfall Drift sets south past it, the basin's return road for the Deepgate's flood.

Plausible, from that: a swell lifts the bottom toward the light without breaking the surface, and a current running past a rise stirs the water over it. Stirred water feeds small life, and small life feeds larger: the bank is a likely feeding ground for fish, and where fish gather, the birds and boats that follow fish gather too. It is not a hazard — the survey says swell, not shoal — so it is a working water, not a warning.

Unwritten: what fish, what birds, what boats. The bank earns its use first; its visitors get their names after.

3. The High Table's crown — the empty ground (plausible)

Proven: the crown is the chart's emptiest place on purpose — its ≥0.95 outline dotted, its peak marked, and nothing else drawn inside it, because nothing stands there. Seven river heads rise around its rim and run outward; none rises inside the crown itself. The crown is high ground ringed by water's beginnings.

Plausible, from that: ground that high and that bare is wind-scoured and thin-soiled — the kind of place where only low, tough growth holds on, if any does, and where no tree gets a foothold worth charting. The rivers all leave the crown rather than crossing it, so its middle stays dry even while its rim feeds the whole isle. Birds may cross it, and nothing in the survey says they stop there. The emptiness is not a gap in the natural history; it is the natural history — a high, open place the living country mostly passes over.

Unwritten: whether anything at all roots on the crown, or whether it is bare rock and wind from rim to rim. The atlas will not plant a garden on ground the survey left blank.

4. The Stones' neck — a crossing twice a day (plausible)

Proven: the neck between the two Stones is a saddle at ~0.5207, barely above the sea at 0.52 — charted awash in all but name, a wading crossing, not a walk. The Stones themselves are two small islets, their summits low (0.5502 and 0.5457), with no river, no bay, no landing named on either. The Deepgate's flood runs north past them, fast water at the Gate's northern narrows.

Plausible, from that: ground that is sea and land by turns belongs to neither, and the living country there is the kind that does not mind the bargain — low, salt-tolerant growth clinging to rock that dries between tides, and the small shore-life that feeds in the wet hours and waits out the dry ones. A neck that floods twice a day is a pantry, not a home: waders will work it at low water, picking the exposed ground before the tide takes it back. Nothing tall grows there — the islets are too low, too salt-washed, and too small to hold soil worth the name — and nothing drinks there, since no thread of fresh water runs on either stone.

Unwritten: whether anything nests on the Stones at all, or whether they are only a feeding stop between other shores. The chart says wading, not walking; the natural history says visiting, not living — until the country shows otherwise.

5. The northeast island's three threads — the dome's whole water (plausible)

Proven: the island is one dome — 11,828 land cells, a single flat-topped summit at 0.663, no second summit above 0.60 — and it sheds exactly three threads worth the name. The flow survey is plain about it: median coast accumulation on the island is 16, and these three are the only outlets over 100. The north thread is the island's one real water — summit-born, its head at (472, 1089), 0.652, just below the cap, running almost due north, 36 cells to its outlet, accumulation 606: the dome's main drain, carrying more water than Savrel's unnamed east burn on a shorter thread. Its fall is steady — a drop of 0.133 from head to sea, every step downhill, no flat or pooling step in its whole course, sixteen of its cells at or above 0.60. The east thread is 30 cells, slope-born at 0.596, accumulation 215, with three near-flat steps in its course. The west thread is 31 cells, slope-born at 0.636, accumulation 185, with one near-flat step. All three are drawn in full on Plate VII, and all three are unnamed: the north thread is shorter than Savrel's 59-cell east burn, which already failed the bar, and the other two are shorter still.

Plausible, from that: the three threads do not keep the same hours. A summit-born thread with a steady, unbroken fall is the island's strongest case for water that runs year-round — likely the dome's one reliable drink. The slope-born pair, with their near-flat steps, are likelier fair-weather water: running when the dome is wet, pooling briefly where their courses flatten, and giving out in a dry spell. And whatever life the island holds, if it holds any, will be found along these three lines and nowhere else, because they are the only fresh water the dome makes. Their mouths are the island's three meeting places, where fresh water reaches the salt — the kind of edge where shore-life gathers on every other coast the atlas has walked.

Unwritten: whether the north thread truly runs year-round. The atlas has no rainfall model and no spring the survey can show, so the claim stays plausible and unasserted — and whether anything drinks from any of the three at all. The island has given no sign of a resident, only of water. Three threads, honestly drawn; who they run for is not yet known.

6. The water between the island and the Shiproad Shore — the Gate's own water, not quiet water (plausible)

Proven: the water between the island and the Shiproad Shore is not an unnamed quiet reach at all — it is the Deepgate itself, the strait the sailing model charts as the tide road. Surveyed across the island's latitudes (rows 440–520), the gap runs 136–218 cells of honest open sea — widest, 218 cells, at row 440, narrowing to 136 at row 500 — with a sea floor at 0.42–0.48 in mid-water, no sill and no shoal across it, and pale shallows fringing both coasts (the survey's edge values climb back toward 0.52 at either shore). The tide the plates arrow runs through this water: flood north, hardest at the northern narrows by the Stepping Stones; ebb south, curling into Gathaven farther down the Shiproad Shore. The island's west thread opens straight into this moving water at (476, 1004), so its fresh water meets a tide, not a still pool.

Plausible, from that: a strait this wide does not race the way its northern narrows do — the same tide spread over 136–218 cells of water moves more gently than through the Needle's Eye's 26 — but it is still moving water, twice a day, every day, and moving water is not quiet water. It carries whatever the tide carries past both shores: stirred food for small life, and small life for whatever follows it. The island's west shore, facing the Gate, is therefore a tide-washed edge rather than a still-water one — its meeting places scoured by the ebb and flood, not settled like Gathaven's curl. Nothing here makes the island a landing place — it still has no bay, no shelf, no roadstead — but its water is the charted world's busiest water, and any life on its west side lives on the Gate's schedule, not the basin's.

Unwritten: the current's speed at the island's latitude (the plates arrow direction and relative strength only, and the sailing model leaves speeds blank), and what fish or birds actually work this stretch. The atlas knows the road and its width and its floor; it does not yet know its traffic.

7. The settlement question — Gathaven's promise against the island's silence (plausible)

Proven: the candidates do not stand equal. Gathaven is the one true bay on the Shiproad Shore, and the ebb does its work there in full: the southward water curls into the bay, slows into a back-eddy and settling bowl, and leaves a soft silt bottom with good holding. The price stands in the same fact — the haven silts slowly and must be sounded, not trusted. It also stands on the charted road: the Deepgate passes its mouth, Lissen comes down to the open shore south of the bay, and the Green Terraces lie behind its Strand.

Varra's other sheltered waters each carry a smaller claim. The Pocket, the Dunmar Coast's one bite, catches the southward drift in a gentler back-eddy; sand settles there, giving soft bottom and fair holding for small craft, sheltered from the north and northwest and open to the southwest swell. The Northern Inlet is a cul-de-sac whose west end is a true roadstead, its water pale and its westerlies broken by Ostrel's hill; Ostsen falls from that hill to the inlet. The Cup is the south's one bay, round and south-opening, with ample soundings and three rivers of its own — Gatewater on its west shore, Cupsen at its head, and Colsen on its east shore. Its water turns in a slow shear eddy and flushes as it turns; no settling bowl is claimed, no anchor is inked, and its holding is unproven.

Fresh water and food do not gather in the same places as shelter. Varra's rim-born rivers reach every coast, and the southern fan is the country's richest fall of water. Farther west along the Low Strand, the Marsh Flat spreads for some 950 cells within a hand's breadth of the sea beside the Meadowrun's mouth, its seasonal water unproven. Off the Dunmar Coast, the Sunfall Bank is a broad swell, not a shoal, and Dunsen is the only fresh water between the Pocket and the point. Across the Gate, the unnamed island has three threads and no landing: its coast runs smooth and unbroken, with no bay, inlet, shelf, or roadstead, and its west thread meets moving tide water rather than a still shore. The Stepping Stones have no river, no bay, and no landing, only an awash neck in fast water. The High Table's crown has no water of its own at all; its rivers rise on the rim and run away from it.

Plausible, from that: Gathaven alone joins the whole bargain in one place. It offers shelter that catches, holding that is already judged good, a named burn on the same shore, terrace country behind it, and the Gate's road at its mouth. If any shore has earned a landing, Gathaven has — not because it is perfect, but because none of its essentials is missing. Its costs are maintenance and obedience: sound the silting bottom, work the tides rather than fight them, and remember that the road in runs through pale shallows and, at the north end, the Gate's fastest water.

The Pocket earns the next call, and a smaller one: a small-craft landing. Fair holding and a lee give it rest; Dunsen gives its coast a watering place; the Sunfall Bank offshore is a likely feeding ground where the drift stirs the water over a rise. Together those make a modest, coherent offer. The southwest opening and the small-craft limit keep it below Gathaven — a haven for a call, not the country's port in waiting.

The Northern Inlet earns a roadstead, which is not quite the same thing as a landing. Its sheltered west end and Ostsen's fresh water make it a plausible place to wait and water, with Varra's north coast in reach. But the inlet is a cul-de-sac, and the chart makes no holding claim for its bottom. It offers rest at anchor more clearly than it offers a shore to build a life from.

The Cup is the reverse of the island: land-rich and harbour-poor. Three rivers give it the strongest fresh-water case of any bay in the atlas, and the Marsh Flat's flood-and-dry bargain likely supports reeds, sedge, and feeding waders farther west, richest where the Meadowrun's plume meets the flat. The Green Terraces behind the southern shore are the kind of lowland where cultivation is at least conceivable. Yet a settlement cannot anchor on a maybe. Until the Cup's bottom is sounded and asked to hold, it earns an approach and a question, not a landing.

The island offers water without access. Its north thread is the dome's one strong case for a reliable drink, and its three mouths are the only meeting places of fresh and salt water it has. But a drink does not make a landing when the coast gives no bay, no shelf, and no sheltered shore to receive a boat — and on the west side, even the meeting place is scoured twice a day by the Gate. The Stones offer still less: a low-water pantry on the neck, then the tide takes the ground back, and there is nothing to drink. The crown offers height without water, soil, or foothold. Any use of this country begins at its rims and coasts, not at its empty centre.

So the answer is narrow. Gathaven has earned a landing; the Pocket has earned a small-craft call; the Northern Inlet has earned a roadstead; the Cup has not yet earned a landing; and the unnamed island and the Stepping Stones have earned neither rest nor welcome. No second shore offers the whole bargain.

Unwritten: whether any of Varra's rivers keeps its course through a dry season, and whether Lissen runs when it is most needed — the chart traces water, but it has no rainfall, no season, and no spring to show. The soil's depth and richness, the presence of timber or other building country, the weather that tests every shelter, and the true abundance of fish, birds, and shore-life all remain unmeasured. The Cup's holding is still an unanswered question, and the soundings of the lesser anchorages are thinner than a founding would want. Whether anyone ever comes to any of these shores, from where, or in what numbers, this entry does not decide. It grades only what the country offers: one promise kept in geography, three lesser calls, one rich bay still waiting to be asked, and an island whose water runs in silence behind a coast that does not open.

8. The Green Terraces — the ground every landing points back to (plausible)

Proven: the Green Terraces are Varra's lowland ring, and the chart paints them in its own colours: the lowland green of the heightmap's ramp, heights from 0.55 to 0.64, between the Strand's sand below and the Amber Slopes' dry tawny hills above. On Varra itself — the main landmass, 236,338 cells — the green band holds 85,488 cells, the largest single band on the isle, and it rings the whole of it: a median 31 cells inland from the sea, nine tenths of it within 77 cells, its farthest reach 125 cells deep where the ring runs wide. A Strand's width of sand, a median six cells, lies between the green and the water everywhere, so the terraces are near-shore ground, not shore ground. The ring is gentle: its median gradient is 0.0024 a cell, and more than two thirds of it falls below 0.003 — ground a walker would barely feel. And it is watered, in lines: every river the atlas has traced spends part of its course in the green — the four rim rivers 26 to 35 green cells each, the southern fan's rivers and Savrel's own the same — so the ring is crossed, coast to coast, by fresh water running downhill to the Strand.

Plausible, from that: this is the country's working ground, if it has any. The gazetteer said it from the first walking — the lowland greens, where any farm or town would have to be — and the survey now backs the saying with numbers: the lowest gentle ground on the isle, the broadest band, watered along its river lines, lying behind every landing the settlement question judged. Gathaven's terrace country, the Pocket's, the Cup's, the Marsh Flat's western lowland — all of it is this one ring, seen from different shores. Between the rivers the case thins: the green away from a river line has no water of its own to show, and the west's one burn, Dunsen, is the only fresh water on a long stretch of Sunfall Shore — so the ring's promise is not even, but braided, greenest and wettest where a river crosses it, drier in the stretches between. The east knolls break the ring's clean circle, and Savrel's shoulder rises through it in the southeast, but neither breaks the rule: the terraces are where low, watered, walkable ground gathers, and everything above them climbs toward the crown and everything below them is sand and sea.

Unwritten: the soil itself — its depth, its richness, whether it holds a crop or only a pasture — and the rain that would have to fall on it, and in what season. The atlas can show gentle ground and river lines, but it cannot show a field, and it will not claim one: no tilled ground, no timber stand, no building country is charted on the Terraces yet. What is proven is the stage — broad, low, gentle, river-crossed — and what is plausible is only that, if anyone ever works this country, they work it here first, from Gathaven outward, following the water downhill as the rivers already do.

9. The Amber Slopes — the middle stair, higher but no steeper (plausible)

Proven: the Amber Slopes are the chart's dry tawny band, heights from 0.64 to 0.73, between the Green Terraces below and the Duncrags' brown highlands above — the middle stair of the world stepping down from crown to coast. On Varra itself — the main landmass, 236,338 cells — the amber band holds 53,463 cells, more than a fifth of the isle and its second-largest band, and it sits a clear step inland of the green: a median 61 cells from the sea against the green's 31, nine tenths of it within 98 cells, its farthest reach 138 cells deep where the stair runs wide. It is not one unbroken ring but three gatherings: the great northern and western stair (36,598 cells), Savrel's southeastern shoulder (12,453 cells), and Ostrel's own rise (4,343 cells), with the river valleys cutting between them — the ring breaks where water runs, as a stair should. And here the survey corrects the eye: the amber is higher ground, not steeper ground. Its median gradient is 0.0026 a cell against the green's 0.0024, and nearly two thirds of it falls below 0.003 — ground a walker would barely feel, only higher. The climb from terrace to slope is a long gentle lift, about nine hundredths of the world's height spread over roughly thirty cells of inland ground, not a scarp. It is watered, in lines, like the green below it: every river the atlas has traced spends part of its course in the amber — the four rim rivers 17 to 25 amber cells each, the southern fan's rivers 22 to 35, Savrel's own 42, the longest amber course on the isle, since Savrel's river is born high and runs the whole stair down.

Plausible, from that: this is the country's middle ground in every sense — above the working terraces, below the crags, walkable throughout, but thinner in its promise. Where the green's case rested on breadth and lowness, the amber's rests on height and dry footing: ground above the Strand's sand and the terraces' damp, with a view over both, drained by the same river lines that water the lowland. That is a real offer, and a narrower one. Between the rivers it thins exactly as the green's does, only higher and drier in colour — and the colour is the chart's paint, not a rainfall record, so no dryness is proven, only height. A lookout, a dry camp, a road along the stair — these are the uses the ground honestly suggests, and each is a suggestion, not a claim. What the amber is not is a second terrace: it does not widen the working ground so much as lift it, and anyone who climbed it looking for more farmland would find the same braided bargain as below, with less of the lowland's breadth and none of its nearness to the shore. Savrel's shoulder is the exception that tests the rule — a whole high country of amber in the southeast, cut by Savsen's long course — and even there the ground offers height and a river, not a plain.

Unwritten: the soil of the slopes — whether it is thin over rock or deep enough to hold anything — and the rain that would have to fall on it, and whether anything grows there worth cutting, grazing, or building with. The atlas can show a gentle, river-crossed stair, but it cannot show a wood, a pasture, or a quarry, and it will not claim one: no timber stand, no grazing ground, no building stone is charted on the Slopes yet. What is proven is the stair itself — broad, walkable, higher but no steeper, broken by river valleys into three gatherings — and what is plausible is only that, if anyone ever works this country, the amber is where they go second: up from the terraces for height, dry footing, and a view, following the rivers that already cut the way down.

10. The Duncrags — the high ring, where the climb keeps its grade (plausible)

Proven: the Duncrags are the chart's brown highland ring, heights from 0.73 to 0.95, between the Amber Slopes below and the High Table's crown above — the top stair of the world before the flat heart itself. On Varra itself — the main landmass, 236,338 cells — the Duncrags hold 56,357 cells, edging past the amber to become the isle's second-largest band, and they sit a clear step inland of everything below them: a median 108 cells from the sea against the amber's 61 and the green's 31, nine tenths of them within 145 cells, the farthest reach 176 cells deep where the ring runs wide around the crown. It is very nearly one ring: a great apron of 55,637 cells wrapped around the Table (rows 325–606, cols 403–697), plus Savrel's own cap, 719 cells standing apart in the southeast (rows 631–661, cols 845–883), detached because the col between them falls to about 0.610, below the band — and one lone cell, a single high point the survey notes and does not chase. The paint splits the ring in half, and the survey says so plainly: 28,208 cells of brown, 0.73–0.82, and 28,149 of grey, 0.82–0.95, so "brown highlands" is the chart's name for a band whose upper half the ramp already paints another colour. And here the survey makes its third correction to the eye: the Duncrags are higher ground, barely steeper ground. Their median gradient is 0.0028 a cell against the green's 0.0024 and the amber's 0.0026, and well over half of the band falls below 0.003 — the stair keeps its gentle grade all the way up, from terrace to slope to high ring, and the climb from shore to crown is a long lift, not a wall. It is watered, in lines, like everything below it, and more so: every river the atlas has traced is born at this ring's inner rim or crosses it on the way down — the four rim rivers spend 45 to 62 cells each in the Duncrags, longer than in the green or the amber, because their heads stand here; the southern fan's rivers 55 to 65, save Colsen, which is slope-born at 0.721 and never reaches the band at all; Savsen only 14, because Savrel's cap barely clears 0.73 before its river falls; the east burn 7, a short high thread.

Plausible, from that: this is the country's roof, not its floor. Where the green's case rested on breadth and the amber's on height with dry footing, the Duncrags' case rests on nearness to the water's birth: seven river heads stand at this ring's inner edge, and every river on the isle runs through it before it runs anywhere else. That is a real offer, and the narrowest yet. A walker who climbs here finds the same braided bargain as below, lifted higher still — water along the river lines, and between them a high ring with no water of its own to show, no flat to hold a field, and the crown's emptiness waiting above. The ring's uses, if it has any, are the uses of a watershed: a lookout over the whole stair, a crossing on the way to the rim, a place to stand and see where the water goes. Savrel's cap is the exception that tests the rule again — a small high country of its own, 719 cells with a river born on it, a mountain in miniature rather than a second Table — and even there the ground offers height and a river, not a plain. The one place the ring breaks its own gentle grade is Gatessen's country on the south-southeast rim, where the river zigzags down through the Shattered Stair — but the blocks themselves are lore, not survey: the gazetteer records them as building stones dropped by a fumbling god, and the atlas has not yet decided whether they are real stone below the chart's sight or a story the zigzag invented for itself. Until it decides, the Stair is a tale told about this ring, not a feature of it.

Unwritten: the soil of the high ring — whether it is thin over rock or deep enough to hold anything — and whether anything grows there worth cutting, grazing, or building with, and whether the grey upper half lives any differently from the brown lower half, since the paint splits what the survey has not yet shown to be two countries. The atlas can show a broad, gentle, river-crossed ring at the foot of the crown, but it cannot show a wood, a pasture, or a quarry, and it will not claim one: no timber stand, no grazing ground, no building stone is charted on the Duncrags yet. What is proven is the ring itself — the isle's second band, the rivers' birthplace, the climb that keeps its grade — and what is plausible is only that, if anyone ever works this country, the Duncrags are where they go last: up from the terraces, past the slopes, for height, for the view, and for the heads of the water, following the rivers that already cut the way down.

11. The Strand — the last stair, where every river ends (plausible)

Proven: the Strand is the chart's lowest band, heights from 0.52 to 0.55, between the sea below and the Green Terraces above — the last stair of the world, and the thinnest: only three hundredths of the world's height, yet on Varra itself — the main landmass, 236,338 cells — it holds 30,795 cells, about an eighth of the isle. It rings the whole of it, shore to shore: a median six cells from the sea, the same six the Terraces entry counted from the other side, nine tenths of it within 17 cells, its farthest reach 49 cells deep where the shore runs wide. It is gentle as the green it borders: its median gradient is 0.0024 a cell, and more than two thirds of it falls below 0.003 — ground a walker would barely feel, even at the water's edge. And it is watered at its end, not its middle: every river the atlas has traced crosses the Strand on its way out — the four rim rivers 9 to 15 Strand cells each, the southern fan's rivers 6 to 12, Savrel's east burn 7 — so the band is where fresh water meets salt, at every mouth the survey knows, and nowhere else on the isle does a river end.

Plausible, from that: this is landing ground, not working ground — the threshold of the country, not its floor. Where the green's case rested on breadth, the amber's on height, and the Duncrags' on the rivers' birth, the Strand's case rests on nearness to the water and nothing else: it is the only band every landing must stand on. Gathaven's harbour, the Pocket's anchorage, the Cup's slow shore — all of them are Strand first, terraces second. But its narrowness is its limit, stated in the same numbers: six cells deep at the median is room to beach a boat and step ashore, not room to stay. A camp, a landing, a crossing on the way inland — these are the uses the ground honestly suggests, and each is a suggestion, not a claim. Between the river mouths the case thins to the shore itself: low ground with no water of its own, no breadth behind it, and the sea at its feet. And the band sits only just above the sea — three hundredths of height in all — so whatever tide Tessamar has eats into it first; the Stones' awash neck, barely above the water at 0.5207, is the warning written into the survey itself. The tide's own range is unwritten, but the Strand is where any tide would be felt, before anywhere else on the isle.

Unwritten: the shore itself — whether it is sand, shingle, or mud at the river mouths, what colour it runs, what the tide leaves behind on it — and the water just offshore, which the survey's sea level hides rather than shows. The atlas can show low, gentle, river-ended ground at the water's edge, but it cannot show a beach, and it will not claim one: no sand sample, no tide pool, no shell bank is charted on the Strand yet. What is proven is the ring itself — the isle's last stair, every river's end, the edge between the country and the sea — and what is plausible is only that, if anyone ever lands, they land here first and move inland at once, up through the green, because the Strand holds no room to stay. With it, the stair is walked whole: crown, crags, slopes, terraces, shore.

12. The rivers taken together — one water system, fifteen mouths (plausible)

Proven: the atlas has now traced fifteen distinct threads, 1,207 cells in all, and the inventory that gathers them (rivers-together-survey.md) shows what no single river could: they are cell-disjoint, every one of them. No traced thread shares a cell with another, and no traced thread joins another before the sea — the country has no traced confluence anywhere. Their heads tell two stories and no more: rim-born, on or just inside the High Table's rim — the four rim rivers, 480 cells between them, one per quarter of the rim, plus the Straight and Cupsen of the southern fan — and slope- or summit-born on a single hill — Colsen on the col country's slope, Savsen and the unnamed east burn on Savrel's cap, the island's north thread on its dome. The longest single thread is the Straight, 142 cells, longer than any rim river; the shortest named river is Colsen, 77 cells; the shortest traced thread of all is the island's east thread, 30 cells. And every one of them ends the same way: on the Strand, the last stair, where every river ends (Entry 11). No traced thread ends inland, and none is lost in a mere or a sink. The count is honest about its edges: Ostrel's burns, Lissen the Brightburn, and any lesser thread not yet traced are not in it, and the lengths are cell counts along the traced path, not straight-line distances and not miles — the world has no scale in miles.

Plausible, from that: this is a country that drains outward and never gathers. A rim, a hill, a dome — each sheds its water down its own slope to its own mouth, and no water ever meets another's on the way. So fresh water is everywhere at the rim and nowhere collected: a traveller inland need not seek a great river, because any thread followed downhill leads to the sea, and no thread followed uphill leads to a lake, a gathering, or a second country. The mouths say the same thing from the other side: fifteen small endings, not one great one. Where a world with joining rivers grows a delta and a port at the joining, Tessamar grows fifteen separate landings, each its own small mouth on the Strand — which is why the Strand's case, in Entry 11, rested on many small thresholds and never on one harbour that a river made. And the year's water stays plausible, not proven, for every thread alike: the atlas has no rainfall model, so no thread's year-round flow is claimed, not even the island's north thread, whose steady fall and accumulation the survey already recorded as a likelihood and left there.

Unwritten: the water itself — its volume, its seasons, what it carries down from the crown and the cap to the Strand's mouths — and whether the threads ever ran differently, in a wetter age when the sea stood higher, the same age the neck-flooding lore leaves open. The atlas can show fifteen paths from head to sea, and it can show that they never meet, but it cannot show a river in spate or in drought, and it will not claim one: no flood, no dry bed, no delta is charted on any thread yet. What is proven is the shape of the system — radial, separate, fifteen mouths and no joining — and what is plausible is only that, if anyone ever lives by this water, they live by one thread at a time, at its mouth on the Strand, because the country gives them no river that is more than one thread, and no mouth that serves more than its own valley. With it, the water is walked whole: crown heads, hill heads, dome heads, fifteen separate ways down to the sea.

13. The open sea — the Four-Cornered Sea beyond the Strand (plausible)

Proven: the sea is everything the chart paints below 0.52, on the same 1200-by-900 survey that holds the land — and it is named, as a whole, the Four-Cornered Sea, the water Tessamar itself is named for. It is not empty water around a country: it is the country's other half, and the atlas has charted more of its behaviour than of any single land band. The currents are inked on the plates, not guessed: the Deepgate's flood sets north, fast water past the Stepping Stones and through the Gate's northern narrows; the Sunfall Drift sets south along the Dunmar Coast, the flood's return road, sheds the Pocket's back-eddy, bends east around Duntail, and returns east along the Low Strand; the Cup turns in its slow shear eddy, in by Savrel's side and out by Gatewater's shore, broken by three river plumes; Gathaven holds its tide-curl, the bay that catches a boat coming home. The sea floor is charted in exactly one place and no more: the Sunfall Bank, a broad swell off the mid-Dunmar coast, corrected from suspected shoal to swell by the soundings — a bank, not a hazard, and the only bank the survey claims. The anchorages are charted, and their limits charted with them: Gathaven's natural harbour, the Pocket's anchorage, the Northern Inlet's roadstead — and the Cup, where no anchor is inked, the holding left unproven rather than assumed. The whole-heightmap survey found no hidden land beyond the charted coasts: only the land already drawn, plus two awash cells at 0.5202, rocks that show at low water and vanish at high, described and given no proper name. And the sea's warning is inked at the Stones: the neck's saddle at 0.5207, awash in all but name, a wading crossing, not a walk.

Plausible, from that: the sea is a road, not a wall. The charted currents close into a circuit around Varra: flood north through the Gate, Drift south down the west coast, return east along the south shore, and home — which is why the sailing model's first voyage works at all, flood out and ebb home into Gathaven, the bay catching the boat. A sailor who knows that circuit need not know the open sea to move around the country; the coast itself is the road, and the harbours are its resting places, one to a coast and no more. Beyond the charted coasts the same road runs outward, and here the case thins honestly: the open sea is a way onward, not a place the atlas has been. The two awash rocks are the only thing the survey puts in it, and they are a warning, not a destination — water that hides its floor at high tide and shows it at low. What the circuit suggests, and no more, is that Tessamar's sailors would learn the sea the way they learned the land: one current at a time, one harbour at a time, never assuming a depth or a holding they have not sounded.

Unwritten: everything the sea level hides rather than shows — the depth of the water beyond the one bank, the tide's own range, the weather over it, what lives in it, and what lies beyond the horizon in any direction. The atlas can show a circuit of currents, three anchorages and one bank, and two rocks that come and go with the tide, but it cannot show a sounding, a storm, a fish, or a far shore, and it will not claim one: no depth, no tide table, no sea creature, no outward voyage is charted yet. What is proven is the road itself — flood, Drift, return, eddy, curl — and what is plausible is only that, if anyone ever sails this sea, they sail the circuit first, harbour to harbour, because the circuit is the one part of the open water the survey has actually sounded. With it, the world is walked whole, land and water both: crown, crags, slopes, terraces, shore, rivers, and now the sea that holds them all.

14. The settlement question, in earnest — what a first landing at Gathaven would stand on (plausible)

Proven: Entry 7 already ranked the shores, and the ranking has not moved: Gathaven alone joins the whole bargain in one place. What taking the question in earnest adds is not a new winner but a test — a landing stands on four things, and each one can be asked of the survey in turn. First, ground that the tide does not take back: the Strand at Gathaven, like the Strand everywhere, is the last stair and the thinnest band, a median six cells from the sea, and the Stones' awash neck warns that any tide is felt there first — so a landing cannot stay on the Strand, it must step off it. Second, water to drink: Lissen the Brightburn comes down to the open shore south of the bay, the named burn on Gathaven's own shore, and the Green Terraces lie behind the Strand, braided by river lines. Third, a way inland: the terraces are the country's working ground if it has any, the largest band on the isle, ringing it at a median 31 cells from the sea, with the Amber Slopes above them and the Duncrags above those — a stair a landing can climb without leaving its own coast. Fourth, a way back to the sea: the Deepgate passes the bay's mouth, the flood runs north through the Gate, and the ebb curls home into Gathaven itself, the bay that catches a boat coming home.

Plausible, from that: a first landing at Gathaven would stand, in order, on the sounding, the step, the burn, and the road. Sound the silting bottom first, because the same settling bowl that gives good holding fills slowly and must not be trusted unsounded. Step off the Strand onto the terraces behind it, because the shore itself is too narrow at the median to hold anything that means to stay — the landing's ground is the green, not the sand, and its door is the Strand, not its floor. Drink from Lissen and from the terrace threads, and remember that no thread's year-round flow is proven, so the first season's true work is learning which water keeps running when it is most needed. And work the road rather than fighting it: in through the Gate's pale shallows, out on the flood, home on the ebb, the bay catching the boat — the sailing model's first voyage, run backwards, is the landing's supply line. The other shores keep their lesser calls and no more: the Pocket a small-craft call, the Northern Inlet a roadstead, the Cup still an approach until its bottom is asked. Nothing in this entry moves a boat or builds a house; it only says what the ground, the water, and the road would have to give before either could be attempted in earnest.

Unwritten: everything a founding would actually need and the survey cannot give — the soil's depth and richness under the terraces behind the bay, timber or other building country within reach, the weather that tests the shelter, the rainfall that decides whether Lissen runs in a dry season, and the numbers and the whence of anyone who might ever come. The bay's own shape at closer scale is unmeasured too: no surveyed point fixes its mouth, no sounding fixes its inner bottom beyond the silt and the holding already judged. Whether anyone ever comes to Gathaven, from where, or in what numbers, this entry does not decide. It grades only what a landing would stand on: a sounded bottom, a step off the Strand, a named burn on the shore, terrace ground behind it, and a road that runs home on the ebb.

15. The Cup's holding, asked — what the bottom gives, and what it withholds

Proven: The Cup's question was always its bottom, and the survey can now ask it in full — shape, if not substance. The bay opens due south and has no sill: the soundings deepen steadily from its head to the open sea, 0.496–0.510 at row 650, through a body of 0.43–0.50, to 0.406–0.416 at its mouth at row 690, and 0.313–0.326 in open water at row 730. A fresh survey of the bay's box (rows 640–700, cols 600–720, on code/heightmap.npy, no terrain touched) finds the same slope, not a shelf: sea cells there run a median 0.458, deepening southward row by row, at a mean gradient of 0.0027 — gentle, unbroken, and without a flat step, a pocket, or a bowl anywhere in it. The water over that slope is the slow shear eddy the plates already inked: in by Savrel's side, west across the head, out by Gatewater's shore, broken by three river plumes — Gatewater down the west shore, Cupsen across the head, Colsen down the east shore — so the bay turns and flushes in the same motion. No anchor is inked in the Cup, and no sounding in the survey changes that: depth is ample throughout, and holding is still unproven.

Plausible, from that: depth was never the Cup's problem, and asking the bottom properly only sharpens what is. A harbour's bottom must do two things at once: give water enough to float in, and give ground an anchor can bite and stay in. The Cup gives the first everywhere and, on the evidence of its shape, promises the second nowhere in particular. Gathaven earns its anchor from a settling bowl, where the ebb's curl lets silt fall and a soft bottom form; the Pocket earns its small-craft call from a back-eddy that does the same, gentler. The Cup's eddy does the opposite work: it turns and flushes, and the three rivers plume out through the turn, so whatever the water carries is carried onward rather than laid down. A smooth slope under moving water is ground that sheds, not ground that holds. What a sounding there would have to find, before the Cup earns a landing, is therefore not more depth but a break in the slope — a flat step, a soft pocket, a shelved reach where an anchor bites and the eddy's turn does not drag it free — and it would have to find it twice over: holding at the turn, and holding under the southern weather the atlas has never seen. Until such a place is found by a sounding, not by a survey, the Cup keeps exactly the call Entry 7 gave it: an approach, a bay to enter and leave by, not a harbour to trust a boat to.

Unwritten: everything a sounding, and only a sounding, could give — the substance of the bottom itself, sand or silt or mud or rock, in any reach of the bay; whether a flat pocket hides below the chart's resolution on that smooth slope; whether the slow turn ever runs backward under southern weather, as Entry 5 left open; the tide's range over a bottom this open; and what the three river plumes lay down, if anything, where they meet the eddy. The heightmap gives elevation, never substance, and this entry does not pretend otherwise: the Cup's holding was asked, and the honest answer to the asking is a narrower question, not a harbour. No boat is moved, no anchor inked, no landing earned — only the terms on which one could ever be, stated plainly, so that a future sounding knows what it is looking for.

16. The Marsh Flat, asked in earnest — wet ground that drains, not a mere

Proven: The Marsh Flat's question was always whether water ever truly stands on it, and the survey can now ask that in full — shape, if not substance. The flat west of the Meadowrun's mouth is the flattest low ground on the south coast, and a fresh survey of its core block (rows 650–685, cols 390–450, on code/heightmap.npy, no terrain touched) confirms the Plate IV figures rather than revising them: 1,925 land cells in the block, 972 of them within 0.04 of the sea (0.52–0.56), a land median of 0.5597, and a slope so gentle it reads as a tilt rather than a fall — row medians of land dropping steadily from 0.5957 at row 650 to 0.5256 at row 680, with no land at all left by row 685, where the shore takes over. The point that started the whole story, (672, 420), stands at 0.549, exactly as Plate IV found it, and its southward profile falls cell by cell — 0.549, 0.547, 0.543, on down to 0.520 at row 683 and open water beyond — reaching the sea in about a dozen cells, as Plate IV said. The decisive test is the one Plate IV could only imply: every one of the 625 lowest cells in the core (0.52–0.545), followed by steepest descent, drains to the sea. None is trapped. There is no bowl here, no hollow, no lip holding water back — and that agrees with the full-island survey, which found no enclosed below-sea basin anywhere on Varra. The Meadowrun's mouth, eight cells east of the core's edge at (673, 458), stands at 0.522, a broad flat opening that spreads into the sea rather than cutting to it, and the flat itself sits west of the river's last reach, not touching its channel. The dark mark on Plate III was shallow sea plus this flat, and that correction stands, now with the drainage to back it.

Plausible, from that: the Marsh Flat is wet ground, not a mere waiting for its name back. A mere needs containment — a basin, a lip, a hollow that holds — and this ground has none; it is a slope, gentle to the point of looking flat, that runs everywhere downhill to the shore. Rain falling on it would sheet south slowly, the row medians falling only about 0.07 over thirty rows, and the lowest fringe would stay damp longest, but damp is not standing, and slow is not held: the water has somewhere to go, and the survey watched every low cell go there. What wetness the flat honestly has comes from its lowness and its neighbours, not from any trap of its own — a hand's breadth above the sea at its lowest, beside a river mouth that spreads rather than cuts, open to whatever the southern weather pushes up the Strand. A storm surge or a high tide could wash the lowest fringe, and a wet season could keep the whole flat sodden; neither makes it a mere, and the atlas will not call it one on the strength of either. The two names Plate IV killed stay killed, for a sharper reason now than the one that killed them: the Low Mere and Sylvemere did not merely assert standing water the survey could not prove — they asserted a shape the survey can now disprove, a hollow where there is only a slope. The flat keeps its descriptive name, the Marsh Flat, and its chart caveat, seasonal water unproven — and the caveat now means something narrower: unproven, and with no basin to prove it in.

Unwritten: everything that would have to come from outside the heightmap, because a third elevation pass cannot settle this question and this entry does not pretend otherwise. The substance under the flat — clay that seals or sand that drains — is unmeasured, and it matters more than another decimal of height: sealed ground could puddle where open ground only sheets. The rainfall that would test it, the water table beneath it, the tide's range over a fringe this low, and whether the Meadowrun in flood ever spills westward out of its last reach onto the flat, are all unknown. What would prove standing water, then, is not a survey but a sighting: water seen standing on the flat after rain, or through a season, on ground the survey says should have drained — observed, dated, and held against this drainage. Until such a sighting, the Marsh Flat is what the numbers say it is: the country's wettest argument, settled as far as a heightmap can settle it, and honestly open beyond that. No mere is restored, no name is minted, and the flat is left exactly where the plates left it — stippled, low, and draining south.

17. The broken shore east of the Meadowrun's mouth, asked in earnest — a fringe, a curve, and no cut

Proven: The broken shore's question was always which of three things it is — distributary channels of the Meadowrun, tidal cuts through the Strand, or simply drowned low ground — and a fresh survey of the shore east of the mouth (rows 600–740, cols 450–700, on code/heightmap.npy, code/survey_broken_shore.py, no terrain touched) answers all three, and the answer is mostly none of them. First, the river: Sylvassen is traced as a single channel, 132 cells from its head to the one mouth at (673, 458), standing at 0.5219, and no second mouth appears anywhere in the window — the traced path in code/river-paths-plate04.json forks nowhere, and the ground east of the mouth rises gently inland cell by cell rather than dipping into a parallel channel, so there is no distributary to name. Second, the cuts: the shoreline, taken column by column as the southernmost land cell, runs 673 at the mouth, bulges gently south to 679 at cols 486–489, then recedes in a long steady sweep to 640 at cols 635–636, and advances again to 654–655 by cols 685–700, on the Cup's western side — and no column in the whole window sits more than four cells north of its neighbours two columns either way, which is the test a tidal cut would have to pass, and none passes it. The only breaks in the line are at cell scale: on the row-675 transect the land runs 463–526 with sea notches no wider than four cells, at cols 505–506, 508, and 512–515 — a ragged waterline, not an inlet. Third, the ground itself: in rows 640–720, cols 459–700, 2,870 of the 5,101 land cells sit within 0.04 of the sea (0.52–0.56), and the waterline cells stand at 0.520–0.524 throughout — a hand's breadth above the water, everywhere, for two hundred and forty columns. What Plate IV's scoping notes called frayed resolves, at this scale, into an apron, not a breakage: of the 9,661 sea cells in the band, 1,549 sit within 0.02 below sea level (0.50–0.52), and at every sampled column four to nine of the first fifteen sea cells south of the shore are that shallow — low land on one side of the line and shallow water on the other, at nearly the same tone, so the edge looks torn where it is only thin.

Plausible, from that: the broken shore is not broken at all, and keeping the phrase is honest only as a description of how the edge looks, never as a claim about how it got that way. A walker east from the Meadowrun's mouth walks the Strand the whole way, low and unbroken, with no channel to cross and no cut to step around; the river's plume, which Entry 16 found spreading rather than cutting at its own mouth, lays no fan here either, because there is no second mouth for a fan to build from. The long recession to row 640 is the coast itself bending — the Low Strand curving in toward the Cup's western approach, not a drowned line, since no former line survives anywhere in the survey to be drowned: the shore is one line, gently curved, with a shallow apron offshore, and the fray is the apron showing through. Whatever tide Tessamar has will be felt here first, before anywhere else on the south coast but the Marsh Flat, because the ground is the same hand's breadth above the sea — the fringe cells at 0.520–0.524 go under at the slightest rise and show again at the fall, twice a day if the Stones' neck is any guide to the rhythm, and the apron offshore goes awash with them. That is the shore's whole drama, and it needs no cut to make it: a threshold that breathes, rather than a coast that breaks.

Unwritten: everything a survey of heights cannot give about a shore this low — whether the fringe is sand, shingle, or mud underfoot, and whether the Meadowrun's plume lays silt along it or the Low Strand return sweeps it clean; the tide's range over ground only 0.004 above the water at the line, which decides whether the fringe is a twice-daily flat or a permanently awash apron; and what the shore does beyond this window, where the Cup's western side runs on toward the bay's mouth. What would settle the fringe's character is not another elevation pass but a sighting and a sample: water seen over the fringe at high tide and off it at low, and a handful of the shore itself, held against the survey's line. Until then the broken shore keeps its working description and earns no proper name — a low, smooth, breathing edge, drawn as the plates drew it, and no more broken than the water is deep.

18. Gatewater's mouth from the water, and the Shattered Stair as a sailor would meet it — no bar, and a stair of turns, not falls

Proven: A sailor coming up on Gatewater's mouth from the south meets open water that only deepens, and a shore that bends rather than breaks. The survey (rows 600–740, cols 590–700, and the traced path in code/river-paths-plate04.json, code/survey_gatewater_mouth.py, no terrain touched) finds no bar to cross: down col 597 the sea falls steadily from 0.5154 two cells south of the mouth to 0.3392 at row 730, the medians 0.4412 for rows 660–700 and 0.3656 for rows 700–730, with no sill, no step, and no shallow shelf anywhere on the approach — the same smooth slope Entry 15 found under the Cup, running on past the mouth into open water. The mouth itself stands at the datum exactly, 0.5203 at (653, 597), on the Cup's western shore and not at the bay's head: east of it the shoreline, taken as the southernmost land cell, recedes steadily from 653 to 641 at cols 632–637, the Cup's headward curve, and then advances again to 655 at col 687 — so a boat feeling its way east along the Strand finds the mouth by the coast's own bend, not by any fan or delta, since Gatewater arrives the way Plate IV's notes said it does, walking: only nine of its 128 traced cells stand below 0.55, and the last reach climbs cell by cell, 0.5203, 0.5251, 0.5279, 0.5302, without a plunge anywhere in it. And the Shattered Stair, met as a passage rather than read from above, keeps its count exactly: 37 changes of direction in 128 cells, confirmed turn by turn on the traced path, densest where a climber would first meet it — 16 turns in the first 42 cells above the mouth — with a knot of near-consecutive turns at cells 70–78 and a steadier rhythm above. What the passage does not contain is any fall: the whole climb rises 0.4464 from mouth to head at a mean of 0.0035 a cell, and no single cell anywhere on the path rises more than about 0.0015. There is no step in the Stair for a boat to shoot or a walker to climb over; the turns are the whole of it.

Plausible, from that: the Stair is a stair in the only sense a sailor's legs would understand — not steps cut in rock, but a slope that refuses to run straight. A boat worked up Gatewater from its mouth would spend its effort on steering, not on lifting: every few cells the channel shoulders the boat around another bend, sixteen times before it has gone forty cells inland, and the water, which from above reads as a zigzag pattern, from the water reads as work — a passage that makes you earn the height by distance rather than by height itself, since the climb is spread so evenly that no one bend is steeper than its neighbours. That evenness is what keeps the legend honest in a new way: Plate IV said the Stair from above is a pattern in the water, not a scar on the hill, and from the water it is a pattern in the going, not a fall in the river — the tumbled Table-rock blocks, if they exist at all, lie below the survey's resolution and below the passage's own evidence, and the name survives not because the river breaks but because it will not run true. And the mouth rewards the same plain reading: deepening water all the way in, no bar, no fan, a shore that curves into a bay on the boat's left hand — a river that can be entered from the sea without ceremony, and then, a few cells in, starts turning, and keeps turning, all the way to the High Table's rim.

Unwritten: everything the heights cannot give about a mouth and a stair met from the water — what Gatewater's fresh water does where it meets the Cup's turning water, whether it lays a plume the Low Strand return then shears away or simply thins into the bay; the soundings of the last reach itself, which the chart knows only as a line of heights, not as depth under a keel; whether the turns are forced by rock the survey cannot see or cut by water the survey cannot watch; and the tide's range at a mouth that stands at the datum, which decides whether the mouth cell itself is wadeable ground or open water twice a day. What would settle the Stair's character is not another elevation pass but a walk up the channel with eyes open: rock underfoot at the bends, or none. Until then the Stair keeps its name as a description of the going — thirty-seven turns, no fall — and the mouth keeps its plain welcome: deep water in, and a river that arrives walking.

19. Gathaven's settling bowl from the water — the landing itself, asked

Proven: Gathaven is a wide, shallow bite, and the survey can now say how wide and how shallow. A fresh survey of the bay's box (rows 386–427, cols 799–818, on code/heightmap.npy, code/survey_gathaven_bowl.py, no terrain touched) finds the coast receding from 818 at row 380 to 800 at row 410 and advancing again to 817 by row 430 — a mouth some forty rows across between a north headland near row 384 and a south headland near row 428, and a head no deeper than col 799, reached only around rows 410–418, where even col 802 is sea for just nine rows. The bowl's floor is a saucer, not a pocket: the sea cells in the box run a median 0.5139, deepest at the mouth itself, 0.4938 at (410, 818), and shoaling steadily toward the head, where the floor sits within three thousandths of the datum — 0.5174–0.5181 at col 802, rows 410–418. The mouth's own cross-section, down col 815, is a clean U: 0.5188 at the north headland, deepening to 0.5013 at row 410, rising again to 0.5184 at the south headland. There is no sill anywhere on the approach: east of the mouth the water only deepens, 0.4957 at col 820, 0.4878 at col 830, about 0.469 by col 850, the Gate's channel floor taking the bay's water without a step. And there is no flat step, no shelved reach, no hidden pocket anywhere inside the bowl — the floor falls and rises smoothly, mouth to head, headland to headland, the way a saucer does.

Plausible, from that: the settling bowl earns its name and pays for it in the same figure. The ebb curl the plates inked — southward water sliding past the mouth and turning inward — is exactly the motion that lays silt down, and a saucer is exactly the shape that keeps it: soft bottom, good holding, the harbour that catches. But the same survey that gives Gathaven its anchor takes away its depth. A boat does not anchor in the bowl's middle, and it certainly does not anchor at the head, where the floor stands within a whisper of the datum and any tide worth the name would lay it bare — the head is wading water, not anchorage, and the landing's ground is not the water's edge but the mouthward half of the saucer, where the floor drops toward the channel and a keel still floats. That is also why Entry 14's warning sharpens rather than softens on a closer look: a bowl this shallow fills. The silt that holds the anchor is the silt that shoals the anchorage, so the sounding is not a courtesy paid once but a duty repeated — sounded before trusting, and sounded again, because the bottom a boat trusted last season is not quite the bottom this season brings. The wide mouth cuts both ways too: forty rows of open water is an easy entrance and a poor wall, shelter from the headlands' sides and open east to the Gate itself, so a boat rides out weather in Gathaven the way the bay itself does — caught, held, and still facing the road it came in by. None of this unmakes the call Entry 7 gave the bay. It only gives the landing its terms: anchor mouthward, sound before trusting, sound again, and step off the Strand — the door is the shore, the floor is the green behind it, and the water keeps only what it can hold.

Unwritten: everything a sounding, and only a sounding, could give — the substance of the floor itself, silt or mud or sand, in any reach of the saucer; whether the silt the ebb curl is supposed to lay down is actually there, or whether the bowl's smooth floor is swept clean by the same water that fills it; the tide's range at a head that stands within three thousandths of the datum, which decides whether the head bares twice a day or only in weather; what the wide mouth does under southern weather the atlas has never seen; and whether the anchorage mouthward of the saucer's middle holds under a flood running north past the bay's mouth. The heightmap gives elevation, never substance, and this entry does not pretend otherwise: Gathaven's bowl was asked from the water, and the honest answer is a harbour with terms — a saucer that catches and fills, an anchor that holds where the water is deep enough to float in, and a head that belongs to waders, not to boats. No boat is moved, no anchor inked beyond the one the plates already gave the bay — only the terms on which a landing could ever stand there, stated plainly, so that a future sounding knows what it is looking for.

20. The Strand at Gathaven — the landing's door, and a ramp, not a threshold

Proven: Entry 19 ended at the door, and the door, asked properly, turns out to be no door at all in the sense of a threshold — it is a slope, and a gentle one. A fresh survey of the shore at Gathaven's head and headlands (on code/heightmap.npy, code/survey_gathaven_strand.py, no terrain touched) finds no step anywhere the water meets the land. The fine profile down row 410 climbs cell by cell without a break: 0.5099 at col 810, 0.5181 at col 802, 0.5204 at col 800 — the datum crossed without a jump — 0.5326 at col 790, 0.5475 at col 780, 0.5704 at col 760, and on inland, 0.6015 at col 740, 0.6631 at col 710, 0.6967 at col 700. The mean rise from the coast twenty cells inland is 0.0014 a cell at row 410, 0.0008 at row 405, 0.0018 at row 420 — the same quiet grade as the bowl's own floor, running on through the datum as if the datum were not there. The headland shores keep the same grain: the north shore at row 390 rises from 0.5181 at col 810 to 0.5519 at col 780, the south shore at row 427 from 0.5192 at col 810 to 0.5644 at col 780. No flat shelf, no bank, no cliff, no beach step appears anywhere sampled — the shore is one ramp, bowl to green, and the survey cannot find the line where the landing ends and the country begins except the datum itself. What the ramp does give, at Gathaven, is breadth the Strand does not give everywhere: Entry 11 counted the Strand band (0.52–0.55) a median six cells deep around the isle, but at Gathaven's head the 0.55 line stands twenty-one cells inland at row 410 (col 778), twenty-five at row 400 (col 774) — and it narrows southward, to ten cells at row 425 (col 789), as the south headland steepens toward the bay. Behind the low ground the green runs long: 0.60 is reached fifty-nine cells inland at row 410 (col 741), 0.65 eighty-five cells inland (col 715), so the terraces behind the door are some sixty cells of rising ground before the amber begins.

Plausible, from that: a landing at Gathaven does not step ashore so much as walk up out of the water. The bowl's head is wading water, Entry 19 said, and the shore answers it in kind: wading deepens into walking without a moment when a foot leaves the water for dry ground, because there is no dry ground at the line, only ground a hand's breadth above the sea, rising so gently that a boat beached at the head sits on a slope, not on a shelf. That breadth is the door's whole offer, and its whole limit in one figure: twenty-odd cells of low ground is room to beach, to unload, to stand a boat's crew ashore together — room the median Strand, at six cells, does not give — and it is still not room to stay. The green begins where the low ground ends, twenty cells in, and the landing's ground, as Entry 14 already judged, is that green, not this sand: the Strand is the door precisely because it leads somewhere, sixty cells of terrace rising behind it before the country climbs in earnest. The narrowing to the south is a small practical reading, not a new rule: the door's easy side is the head's middle and north, where the low ground runs widest, and the south headland's steeper shore is the side a beaching boat would rather not choose. And the ramp cuts the other way too, as every low shore in this atlas does: ground this gentle, this close to the datum, is the first ground any tide takes back, so the door that lets a boat walk ashore is also the door the sea walks in by, and a camp left on the Strand would wake to water before any camp left on the green.

Unwritten: everything a heightmap cannot give about a shore it can only slope — whether the ramp underfoot is sand, shingle, or mud, and whether it changes from the head to the headlands; the tide's range over ground that stands within a few thousandths of the datum at the line, which decides how much of the twenty-odd cells of low ground bares and floods twice a day; whether the silt the bowl is supposed to lay down reaches this shore too, or whether the ramp is swept clean by the same ebb curl that fills the bowl; and what Lissen the Brightburn does where it meets this shore south of the bay, the named burn on the landing's own coast, which no entry has yet followed to its mouth. The heightmap gives elevation, never substance, and this entry does not pretend otherwise: Gathaven's Strand was asked as a door, and the honest answer is a ramp with breadth — a slope to walk up, a green to step onto, and a sea that takes the lowest of it back, twice a day or in weather, until a sighting says otherwise. No boat is beached by this entry, no camp is pitched on the Strand — only the terms on which a landing could ever walk ashore there, stated plainly, so that a future sounding knows what it is looking for.

21. Lissen the Brightburn where it meets the shore — the landing's drink, a journey south, not a step

Proven: Entry 20 ended by asking what Lissen does where it meets this shore, and the first thing the survey corrects is the question's own wording: Lissen does not meet Gathaven's shore at all. It meets the open Shiproad Shore south of the bay. A fresh survey of the burn as Plate II charted it (on code/heightmap.npy, code/survey_lissen.py, no terrain touched) re-traces it by steepest descent from (638, 870) at 0.7391 — the traced head stands on Savrel's cap edge, twelve rows north of Savrel's summit itself, 0.7414 at (650, 871), so there is almost no higher ground left above it — north-east to its mouth, last land (543, 911) at 0.5202, outlet sea (542, 912) at 0.519. Ninety-eight cells, a fall of 0.2201 at a mean 0.00227 a cell, the steepest single step 0.0054: a steady river, no fall anywhere in it, nine near-flat steps under 0.001 and none of them a pooling step, since every step still falls. Band by band it runs 7 cells of Duncrags, 33 of Amber, 41 of Green, 16 of Strand — the longest stretch in the green, like the other rivers, but this is no Table-rim river: it is a cap river, born on Savrel and running off its north-east flank, while Savrel's other summit-born thread, the unnamed east burn, starts a few cells away at (640, 874) and runs south-east to the south coast instead. The two threads never touch — zero cells in common — a cap shedding two ways from almost one point. And the accumulation is what makes Lissen matter out of proportion to its ninety-eight cells: 2,437 at the last land cell, 2,439 at the outlet, against 569 for the next outlet anywhere on the coast between rows 430 and 570. The catchment gathers fast on the amber-and-green flank — 146 already twenty cells down the thread, 1,139 by the sixtieth, 2,123 by the eightieth — so the thread a chart draws thin is, by the water it gathers, the coast's main freshwater outlet south of the bay. At the mouth itself there is no bay, no inlet, no pocket, nothing to hold a boat: the coast there only advances eastward as it runs south, 875 at row 520, 890 at row 530, 903 at row 540, 928 at row 550, 954 at row 570, and the water off the mouth deepens steadily east, 0.5185 at col 920, 0.5005 at col 932, 0.4845 at col 952, 0.4486 at col 1002 — no bar, no sill, no step. The mouth sits 115 rows south of Gathaven's south headland near row 428, and six rows south of the unnamed island's last land at row 537: just beyond the island's shadow, where the Gate's south mouth opens and the water is open sea, not the narrow channel.

Plausible, from that: Entry 14 said a landing stands on four things, and counted a drink among them — Lissen on the same shore — and this entry separates what that sentence joined. The drink is on the same coast, not at the same place. A landing at Gathaven does not step from its boats to fresh water; it rows or walks some 115 rows of open shore south to get it, along a Strand that is open Gate water all the way, with no second bay to duck into on the journey. That distance is the entry's practical finding, and it cuts both ways. Against the landing, it means harbour and drink are not one site but two, and a crew watering at Lissen's mouth is watering in the open, at a mouth with no anchorage of its own — fill and go, not fill and stay, with the Gate's south-mouth water at its back. For the landing, it means the water, when reached, is the real thing: a summit-born thread with the coast's largest gathered catchment, a steady fall with no pooling reach to go stale in, and a mouth that runs clear into deepening water instead of spreading over a bar. Year-round flow stays plausible, not proven, for Lissen as for every thread Entry 12 counted — the atlas still has no rainfall model and no spring to show — but if any burn on this coast runs when the others thin, the survey's money is on this one, because a catchment of 2,437 does not gather by accident. The name, meanwhile, the survey cannot cash at all: “Brightburn” is the sailors' name, kept as a name, and a heightmap gives elevation, never brightness, clarity, or colour — whether the water runs bright, and why, is not a thing this survey can see, and this entry does not pretend it can.

Unwritten: everything a visit, and only a visit, could give — the substance of the bed where the burn meets the sea, sand or shingle or mud, and whether the mouth's clear run stays clear under a flood running north past it; the tide's range over a last land cell at 0.5202, within two thousandths of the datum, which decides whether the mouth bares and floods twice a day or only in weather; whether Lissen runs in a dry season, and whether its plume shows in the Gate's south-mouth water at all; what, if anything, drinks from it besides a landing's crew, along ninety-eight cells of green and amber flank; and whether the mouth is fordable at low water, a crossing a foot party could use on the journey south. The heightmap gives elevation, never substance, and this entry does not pretend otherwise: Lissen was asked where it meets the shore, and the honest answer is a correction and a bargain together — not Gathaven's burn but the coast's, 115 rows south of the bay, open water at its mouth, the largest drink on the Shiproad Shore, and a drink a landing must travel for. No boat is watered by this entry, no cask filled — only the terms on which a landing could ever drink there, stated plainly, so that a future sounding knows what it is looking for.

22. The open shore — the 115 rows between harbour and drink

Proven: Entry 21 ended with a journey, not a place: 115 rows of open shore between Gathaven's south headland and Lissen's mouth, which a landing must travel for its drink. A fresh survey of that strip (on code/heightmap.npy, code/survey_open_shore.py, no terrain touched) walks it row by row, from the headland near row 428 to the mouth at row 543, and finds it open in the fullest sense the chart allows. Varra's own shoreline runs col 814 at row 428, 816 through rows 433–448, 821 at row 453, 830 at 458, 847 at 468, 853 at 478, 857 at 483, 859 at 488, holding 859–861 through rows 488–508, 864 at 513, 874 at 518, 883 at 528, 897 at 533, 901 at 538, and 918 at row 543, Lissen's mouth — a net eastward advance of 104 cells over the 115 rows, coast leaning steadily seaward as it runs south. The largest westward row-to-row step anywhere in the strip is a single cell; no row recedes ten cells west of both its neighbours; no run of sea reaches west of the shoreline at all. There is no bay, no inlet, no pocket, no second Gathaven anywhere on the journey — nothing to duck into, nothing to beach in, nothing that holds a boat. Two cells at the datum's edge (col 890 at row 530, 0.5201; col 917 at row 542, 0.5203) sit awash amid the sea's fringe, but they are fringe, not islets: no land stands east of the shoreline anywhere in the strip except the island across the channel, below. The water off the shore deepens steadily east at every sampling — at row 470, 0.5129 five cells out, 0.4955 at fifteen, 0.468 at thirty, 0.4488 at sixty; at row 510, 0.511, 0.4881, 0.4591, 0.4562 at the same distances — no bar, no sill, no step offshore either. The Strand underfoot, walked west from the line, runs 26 cells wide at row 430, 19 at 440, 16 at 450, 26 at 460, 19 at 470, 15 at 480, 11 at 490, 8 at 500, 12 at 510 — then widens as the journey nears its end, 41 cells at row 520, 61 at 530, 80 at 540, the low ground fanning out toward Lissen's mouth. And for the middle of the journey the walker is not alone: across the water, the unnamed island's west shore stands 141 cells off at row 484, 138 at 490, 136 at row 500, 138 at 509 — a channel floor of 0.44–0.46 mid-water at rows 490 and 500, a clean U with no sill — and south of row 509 the island's shore recedes eastward, 1002 at row 510, 1013 at 515, 1024 at 520, 1045 at 525, 1068 at 535, the channel widening as the island tapers toward its southern tail, until at Lissen's mouth the water is open sea, as Entry 21 said.

Plausible, from that: the journey is what Entry 21's bargain costs in ground. Harbour and drink are two sites, and this entry measures the road between them: a shore that never shelters, never narrows to a crossing, never offers a second harbour, but never bars the way either. A crew walking it walks the Strand — low ground, mostly a dozen to two dozen cells wide in the north and middle, widening to eighty as Lissen's low country opens — with the green immediately behind it wherever the Strand ends, so the walk is never far from the terraces Entry 8 counted the country's working ground, and never on them either: the shore road is the shore's own, not the country's. A crew rowing it rows open Gate water all the way, with no bay to run for if weather comes up from the south mouth, and with the island across the channel for company through the middle miles — near enough to see, 136 cells at the narrowest, too far and too water-cut to be a refuge, a neighbour shore, not a shelter. The widening Strand toward the mouth reads as the journey easing at its end: the last rows give room to spread, to set casks down, to work without standing in the water's edge — room Entry 20's door gave at Gathaven's head, here given at the drink instead. But the widening cuts the same way Entry 20's ramp did: ground this low, this wide, this close to the datum is the first ground any tide takes back, and a party that lingers at the mouth lingers on ground the sea reclaims twice a day or in weather. And the shore's very openness is the journey's discipline: there is nowhere on it to leave a boat unattended, nowhere to cache casks for a return, no halfway house between harbour and drink — the 115 rows must be travelled whole, there and back, or the drink is not had.

Unwritten: everything a heightmap cannot give about a road it can only slope — whether the Strand underfoot is sand a foot can walk, shingle that turns an ankle, or mud that holds a boot, and whether it changes from the headland's narrow low ground to the mouth's wide fan; the tide's range over a shoreline that stands within a few thousandths of the datum at the line, which decides how much of the Strand bares and floods, and whether the journey is walkable at low water or only ever a row; what weather from the Gate's south mouth does to a crew on open water with no bay to run for; whether anything crosses this shore besides a landing's crew — anything that drinks at Lissen's mouth from the landward side, or anything that walks the Strand for its own reasons; and whether the island across the channel, near enough to see for the middle of the walk, is ever visited from this shore, or only ever looked at. The heightmap gives elevation, never substance, and this entry does not pretend otherwise: the open shore was asked as a road, and the honest answer is a road with no shelter and no bar — 115 rows of low ground between a harbour and a drink, an island for company through the middle, and room at the end to set a cask down. No crew walks it in this entry, no cask is filled — only the terms on which a landing could ever make the journey, stated plainly, so that a future sounding knows what it is looking for.

23. The Gate's south mouth — where the strait dissolves, not where it shuts (plausible)

Proven: Entry 22 left the island tapering toward its southern tail and promised the mouth itself: the open water Lissen's mouth faces, where the Deepgate stops being a strait. A fresh survey (on code/heightmap.npy, code/survey_gate_south_mouth.py, no terrain touched) reads every land run in cols 750–1200, row by row — a correction worth recording, because a first pass that searched only a cols-900 window mistook Varra's own eastward bulge south of row 684 for island land, and this entry uses the runs, not the window, so the island is never confused with the shore it faces. On that honest reading the channel does not pinch shut at its south end; it widens as it dies. Its floor holds 0.44–0.46 mid-water through rows 484–509 (narrowest 135 cells at row 504, 136 at row 499), then the island's west shore recedes eastward — 1000 at row 509, 1010 at 514, 1021 at 519, 1043 at 524, 1056 at 529, 1066 at 534 — and the channel widens with it, 138 cells at row 509, 143 at 514, 163 at 524, 169 at 529, 168 at 534, its floor rising only slightly as it widens, 0.4517 at row 509 to 0.4675 at row 529. The island itself narrows the same way: 130 cells wide at row 509 (1000–1129), 109 at 517, 78 at 521, 59 at 525, 44 at 529, 29 at 533 — and at row 537 a single cell, col 1077, at 0.5203, three thousandths above the sea, the tail's last toe touching water. At row 538 there is one land run only, Varra's shore at col 901: the island is gone, the channel is open sea, and it is gone six rows north of Lissen's mouth at row 543. No sill marks the passing, no bar, no shallow ridge across the water the strait becomes: the transect east of Varra's shore at row 538 runs 0.5089 ten cells out, 0.4922 at thirty, 0.4731 at sixty, 0.4611 at a hundred — open water deepening steadily, as it does at Lissen's mouth itself (0.5077, 0.485, 0.4657, 0.4534 at the same distances). South of the mouth the deepening steepens and never checks: at 150 cells east of the shore, 0.484 at row 543, 0.4266 at 553, 0.3439 at 563, 0.2784 at 573, 0.231 at 593, 0.2009 at 603, 0.1677 at 613 — the sea floor falls away eastward and southward without a step back up, and Varra's own shore, freed of the channel, advances east to col 973 at rows 633–643 before falling back to 932 by row 693, a broad southward bulge with no bay cut in it anywhere in the survey.

Plausible, from that: a south mouth that is not a mouth at all, in the sense a sailor means the word. The Gate's north end has a needle to thread — the Needle's Eye, small boats only, as Plate II inked it — but its south end has no gate, no narrows, no bar to cross: a strait that widens, shallows by a fraction, and then simply runs out of island, dissolving into open sea six rows before the river a landing drinks from. That changes what the south mouth means for a landing. Weather does not arrive through a gap here; it arrives over open water, from the south and east alike, with nothing to funnel it and nothing to break it — the same openness Entry 22 found along the shore road, extended seaward. A boat coming up from the south does not enter the Gate so much as find itself already in it, the island's tail slipping past to starboard and Varra's shore opening northward, and a boat leaving southward does not cross a threshold either: the channel floor under it stays 0.44–0.47 to the last island row, then the water deepens under it without ceremony, and the Deepgate is behind it before any lookout could call the passing. The last island cell matters in the same honest way the two awash rocks of the Unwritten mattered: at 0.5203 it is land by three thousandths, awash in all but name at any honest tide, a toe, not a terminus — the island ends, as it lived on its own plate, unnamed and unusable, giving no reason for a name even at the place where it ends. And Lissen's mouth, six rows south of that ending, faces water that is already and entirely open sea: the drink a landing walks 115 rows for is drawn at a shore the Gate no longer bounds, which is why Entry 21's journey ended at open water, not at a gate's door.

Unwritten: everything a heightmap cannot give about a mouth that is a dissolving, not a door — the tide's run through water that widens instead of narrowing, whether the flood still sets north here with anything like the Gate's northern speed or spreads and slackens as the channel dies; what the last island cell looks like at low water, whether it ever shows as a rock a boat must mind or stays a wash no chart need mark; what weather from the open south does to a crew working Lissen's mouth with no bay behind them and no island beside them anymore; and whether any sailor, finding no gate at the Gate's south end, ever bothered to call it a mouth at all, or simply called the whole southern water the sea, and the strait something that only exists while the island lasts beside it. The heightmap gives elevation, never current, and this entry does not pretend otherwise: the south mouth was asked as a place, and the honest answer is that it is not one — it is the row where an island ends, a channel widens past use, and open sea takes over without a mark. No gate stands here, no bar is crossed — only the terms on which a boat passes from strait to sea, stated plainly, so that a future sounding knows there is no threshold to look for.

24. The shore south of Lissen's mouth — a bulge, a fall-back, and a turn (plausible)

Proven: Entry 23 left Varra's shore advancing east south of Lissen's mouth and promised the road itself: the southward shore from the drink, surveyed on code/heightmap.npy with code/survey_shore_south_lissen.py, no terrain touched. The survey reads every land run in cols 750–1200, row by row, so the island — gone since row 538 — can never be confused with the shore again. From Lissen's mouth at row 543, shore col 918, the coast advances steadily east: 930 at row 553, 948 at 563, 956 at 573, 960 at 583, 967 at 593, 969 at 603 — then holds its ground through a shallow dip, 966 at 613, 968 at 623 — and reaches its farthest east at row 640, col 974, fifty-six cells east of the mouth's shore. From there it falls back, steadily and without a bay cut anywhere in the run: 966 at 653, 957 at 663, 948 at 673, 943 at 683, 932 at 693, 915 at 703, 893 at 713 — and by row 723 the eastern run has shrunk to cols 782–827 alone, and by row 733 there is no land left in this eastern window at all: the coast has turned west onto the south shore the southern plates charted. The southmost land confirms the turn from the other side: row 728 at col 800, row 724 at 825, row 716 at 850, row 710 at 900 — the shore's southern edge, met and turned. No bay, no inlet, no pocket interrupts the bulge or the fall-back; the largest row-to-row steps are the turn itself, not a cut. The water off the shore deepens steadily east at every sampling, with no bar and no sill: at the bulge's peak, row 640, 0.4807 ten cells out, 0.3846 at thirty, 0.2910 at sixty, and the same clean deepening at rows 600, 650, 680, and 700. The Strand underfoot is narrow through almost the whole road — 8 to 14 cells through the bulge, against 37 at row 550 in Lissen's mouth fan — widening only to 26 at row 710, where the low ground gathers before the turn. One meeting is worth recording exactly: at row 693 the shore stands at col 932, the very outlet Savrel's unnamed east burn was traced to (693, 932), so the burn meets the sea on the falling shore, in the open, with no bay or pocket made for it — and it stays unnamed, as Plate VI judged it.

Plausible, from that: the road south from the drink is Entry 22's road continued, and lonelier. North of Lissen's mouth a walker had the island across the channel for company through the middle miles; south of it there is no companion shore at all — open sea to the east from the first row, and the bulge only pushes the walker further out into it, fifty-six cells east of the mouth, with nothing to break weather from the south or east and nowhere to duck if it comes. The bulge is a shoulder, not a shelter: it advances and falls back without ever cutting a bay, so a boat has no more harbour at row 640 than at row 543, and a crew walking it walks the same narrow Strand, eight to fourteen cells wide, with the green immediately behind it and the water immediately beside it — ground the tide takes back first, as Entry 20 warned at Gathaven's door. The one easing on the road is Savrel's unnamed east burn at row 693, a second water a hundred and fifty rows south of Lissen's mouth: a crew that has walked this far need not walk back thirsty, but the burn earns no harbour with its water — it arrives at open shore, its flow unproven beyond its traced thread, and the road past it keeps falling back toward the turn. At the turn itself the road changes its kind: the east shore is done, the south shore begins, and the southward road from the drink meets the country's southern country, where Savrel's own river reaches the sea on its own shore. A landing that walks south from its drink does not find a second Gathaven waiting; it finds the same open terms, carried further, until the coast itself turns its back on the Gate's water.

Unwritten: everything a heightmap cannot give about a road it can only slope — whether the narrow Strand through the bulge is sand, shingle, or mud, and whether it changes where the mouth's fan gives way to the bulge's narrow line; the tide's range over a shoreline that stands within a few thousandths of the datum at the line, which decides how much of those eight-to-fourteen cells bare and flood, and whether the southward road is walkable at low water or only ever a row; what weather from the open south and east does to a crew on a shore with no bay for a hundred and eighty rows, and to a boat left at Lissen's mouth while its crew walks; whether Savrel's unnamed east burn runs when a crew would need it, or only in the fair weather that also lets a crew walk; and what the south shore, once the coast has turned, offers a road that has travelled this far — whether it opens, or only continues the same openness westward. The heightmap gives elevation, never substance, and this entry does not pretend otherwise: the shore south of the drink was asked as a road, and the honest answer is a bulge with no bay, a fall-back with no shelter, and a turn — a hundred and eighty rows of low ground between a drink and a south coast, a second water near the turn, and open sea all the way. No crew walks it in this entry, no cask is filled at the burn — only the terms on which a landing could ever walk south, stated plainly, so that a future sounding knows there is no harbour to look for.

25. The south coast westward — Savsen's mouth, and a shore that keeps no harbour (plausible)

Proven: Entry 24 left the coast turning west at the south and promised the road itself: the south coast from the turn westward, surveyed on code/heightmap.npy with code/survey_south_coast_west.py, no terrain touched. The turn's toe is the survey's southmost ground: row 728 at col 800, with 724 at col 825 and 716 at col 850. Savsen's mouth sits just east of that toe, at (716, 857), exactly where Plate IV traced it: Savsen runs 82 cells from its summit-cap head at (643, 870), 0.740, straight down Savrel's south face to the sea at 0.5221. West of the toe the coast steps back north without ever cutting a bay — 715 at col 775, 693 at col 750, 675 at col 725, 654 at col 700, easing to 643–652 through cols 625–650, then gently south again to 676 at col 475 — and the line continues west of the Meadowrun's mouth (673, 458) at 677–683 through col 400: one coast, unbroken, for 550 columns. The steepest step is 16 cells between cols 755 and 750, a shoulder receding, not a cut. The water off the shore deepens steadily south at every sampling, no bar and no sill, and the Strand underfoot is narrow all along it, five to nine cells.

Plausible, from that: the south coast answers the settlement question the same way the east shore did, and more plainly. Savsen gives the south a second named drink, but a drink is not a harbour, and this coast keeps no harbour anywhere in the survey. A boat could not wait off Savsen's mouth: deep enough water, clean of bars, but open southern water with weather unbroken by any headland, and only five to nine cells of low ground behind it. A crew walking west walks the same terms Entry 22 measured, with one difference: here there is no companion shore at all, only the Low Strand return working this water from the west and open sea to the south. Savsen's mouth is a fill-and-go water, not a fill-and-stay one — and here there is not even a Gate's channel to go home by, only the long south coast back eastward and the long east shore back north to Gathaven.

Unwritten: everything a heightmap cannot give about a shore it can only slope — whether the narrow Strand is sand, shingle, or mud; the tide's range over a shoreline within a few thousandths of the datum; whether summit-born Savsen runs year-round; what southern weather does to a boat left off an open south mouth; and what the coast does west of the Meadowrun's mouth, where the south shore must turn north toward the Sunfall Shore.

26. The south coast west of the Meadowrun's mouth — Duntail, and the turn north (plausible)

Proven: Entry 25 left the south coast running west of the Meadowrun's mouth and promised the turn: surveyed on code/heightmap.npy with code/survey_south_coast_turn_north.py, no terrain touched. West of the river (673, 458) the southmost land runs 674–683 through col 350, unbroken, then leans south into a low lobe — 692 at col 340, 703 at col 320, 693 at col 300 — and west of col 295 there is no land in the south window at all. That lobe is Plate III's Duntail, tip (703, 314) at 0.520, a spit at sea height with a shallow tail trailing southwest, charted awash — keep off. North of it the Sunfall Shore runs straight: westmost 300–307 at rows 685–700, leaning east to 355 at row 500, stepping west to 338 at row 490 and holding 304–313 down to row 310, where the reading stops rather than mistake Ostrel for Varra's coast. No bay cuts either line. Offshore water deepens steadily on both lines, shallower off the lobe itself (0.5074 ten rows out at col 300) than off the plain shore (0.4853 at col 400). The low apron widens southward: 45 cells eastward at row 650, 13 at row 600, six at rows 550 and 500.

Plausible, from that: the south road ends in a warning, not a harbour. A crew walking west from the Meadowrun walks Entry 25's terms for a hundred and ten columns more, then meets a spit that barely counts as land — and a current that takes over where the road stops: the Sunfall Drift rounds Duntail and bends east as the Low Strand return, so a boat turning north at the point meets the Drift head-on, with no bay to wait it out in.

Unwritten: whether Duntail's tip bares as dry sand or only as a wash; the tide's range over ground at the datum itself; what the Drift does at the point; whether the westward shoulder near rows 500–490 is a burn's mouth or only a shoulder; and what the Sunfall Shore offers north of the turn.

27. The Sunfall Shore northward from Duntail — the Drift's home water, the Bank, and the Pocket (plausible)

Proven: Entry 26 left the Sunfall Shore at its turn and promised the road itself: the west coast northward from Duntail, surveyed on code/heightmap.npy with code/survey_sunfall_shore_north.py, no terrain touched. The shore is Varra's own for a long run, and it runs in three characters. From the point northward it first leans east: westmost land 307 at row 700, 300 at 690, 303 at 680, 308 at 670, 319 at 660, 320 at 650, 331 at 640, 337 at 630, 340 at 620, 342 at 610, 339 at 600, 340 at 590, 342 at 580, 343 at 570, 349 at 560, 348 at 550, 349 at 540, 350 at 530, 349 at 520, 351 at 510, 355 at 500 — a shore stepping gently seaward as it runs north, with no step in it larger than eleven cells. Then the one real shoulder on the whole reach: 352 at row 495 to 338 at row 490, a fourteen-cell westward step in five rows, the shoulder Entry 26 left open — and this survey traces no burn's mouth at it, so the shoulder stays a shoulder, not a watering place, until a tracing says otherwise. North of the step the shore leans west again: 330 at 480, 325 at 470, 323 at 460, 320 at 450 and 440, 317 at 430, 314 at 420, 312 at 410, 310 at 400, 313 at 390, 310 at 380, 307 at 370, 304 at 360, 309 at 350, 313 at 340, 330 and 320, 311 at 310 — a long, nearly straight coast, its largest steps three cells, closing toward the neck. At row 305 the reading changes hands: westmost land 285, then 253 at row 300, 249 at 295, 242 at 290 — and the land runs show why, a single run from col 253 to 599 at row 300, Ostrel's land merged with Varra's by the Gatneck, so the westmost line north of about row 310 is Ostrel's shore, not the Sunfall Shore's. The survey stops there rather than mistake one land for the other, and the Pocket — the bight Plate III inks under the neck's south side — is the price of that honesty: a row-wise westmost reading cannot draw its bite, because Ostrel's merged land stands west of it. Its anchorage claim therefore rests, for this entry, on Plate III's own soundings, not on this pass: pale water at 0.494–0.495, a slow back-eddy shed by the Drift, soft bottom, fair holding for small craft, sheltered from the north and northwest, open to the southwest swell.

The west's one true burn stands on the line where Plate III put it: Dunsen's mouth at (438, 319), 0.5197, a thousandth below the sea, with westmost land at row 438 at col 320 — the mouth on the shoreline itself, the only fresh water the survey names on this whole reach, its head inland at (435, 340) on Plate III's tracing. Off the mid-coast the Bank keeps its corrected shape: in the box rows 380–480, cols 240–320, sea cells run a median 0.4209, and the crest near (427, 290) stands at 0.4614, about twenty-six cells west of the coast at row 427 (westmost 316) — a rise of about four hundredths above the surrounding floor, a swell in the Dogger sense, not a shoal and nowhere near awash. The water off the shore deepens steadily westward everywhere it was sampled, with no bar and no sill: at row 650, 0.504 ten cells out, 0.4578 at thirty, 0.3787 at sixty; at 600, 0.4908, 0.4177, 0.3458; at 550, 0.4797, 0.4071, 0.323; at 450, 0.4872, 0.4325, 0.3479; at 350, paler near the neck, 0.5085, 0.4863, 0.3918. The Strand underfoot tells the reach's shape in width: 45 cells eastward at row 650, the wide low apron inland of Duntail's lobe; 13 at 600; only six at rows 550 and 500; 10 at 450; nine at 400; 16 at 350, widening toward the neck's low waist. No bay, no inlet, no pocket cuts the Varra line anywhere between Duntail and the neck — the Pocket is the one bite the chart claims, and it sits exactly where this survey's line changes hands.

Plausible, from that: this is the Drift's home water, and the entry's plainest fact is how long a boat must live in it. The Sunfall Drift is the basin's return road — the Gate's flood, spread west along the Arriving Strand, turning south down this open coast, over the Bank's swell, round Duntail, and east as the Low Strand return — and on this reach it has no strait to squeeze through, so it walks rather than runs: a slow southward set against a northbound course for some four hundred rows, with Dunsen the only drink on the way and the Pocket the only anchorage at the end of it. A boat beating north from Duntail meets the Drift head-on the whole way, on a shore that offers no bay to wait it out in until the Pocket, and a boat running south rides the same water homeward, past the Bank's stirred water — Entry 2's likely feeding ground, still only likely — and round the point into the return. The shore's narrowness sharpens the bargain: six to sixteen cells of Strand through the middle miles is room to beach and step ashore, not room to stay, and the one wide apron belongs to Duntail, the point itself a warning. So the Sunfall Shore earns, as a road, what the south coast earned as a road: passage, not residence — a coast to work along, with one small-craft call at its north end and one drink in its middle, and the current itself as the country it must be reckoned with.

Unwritten: everything a heightmap cannot give about a drift it can only slope — the Drift's actual speed on this reach, whether it runs faster over the Bank's swell or only stirs there, and what it does at Duntail's tail, where Entry 26 left the meeting open; the Pocket's exact bite, its mouth's width and its inner bottom beyond Plate III's pale soundings, which this survey could not separate from Ostrel's merged land and therefore did not re-measure; whether the shoulder at rows 495–490 hides a valley's foot or a shore trickle no tracing has yet followed, the burn question Entry 26 left open and this entry leaves open still; what the Bank's stirred water actually feeds, beyond Entry 2's likelihood; and what Ostrel's own west shore offers north of the junction, where this survey stopped rather than claim another land's coast for Varra's. The heightmap gives elevation, never current, and this entry does not pretend otherwise: the Sunfall Shore was asked as the Drift's home water, and the honest answer is a long, straight, shelterless coast, one swell, one drink, one anchorage at the far end that this pass could not re-draw — a road four hundred rows long, walked against the water going north and with it going south. No boat sails it in this entry, no anchor is newly inked — only the terms on which a boat could ever work it, stated plainly, so that a future sounding knows what it is looking for.

28. The Pocket, asked in earnest — a lee, not a bay (plausible)

Proven: Entry 27 left the Pocket unmeasured, because a row-wise westmost reading cannot draw a bight Ostrel's merged land stands west of — and promised it asked in earnest: surveyed on code/heightmap.npy with code/survey_pocket.py, no terrain touched, following the coastline row by row through the neck-to-Varra step instead of reading one line past it. The step is the Pocket. From Ostrel's south flank the coast stands at col 253 at rows 300–301 and 254 at row 302, then steps east in four rows: 273 at row 303, a nineteen-cell step; 276 at row 304; a frayed fringe at row 305, single cells at 285, 287, and 289, each barely land at 0.5201–0.5207; and 310 at row 306, a twenty-five-cell step. Fifty-seven cells east in all, from the neck's tip to Varra's own shore. South of the step the inner shore runs straight: westmost 310–313 for rows 306–340, only three cells of movement in thirty-five rows. South of that it fades, not closes: 311 at row 341, 310 at row 345, 309 at row 350, leaning west into the Sunfall Shore proper — no southern headland, no closing point, no second bite. So the question Entry 27 left open answers itself, and the answer is a correction: the Pocket has no mouth's width to measure, because it is not a bay cut into the land. Its opening is its whole west side, some thirty-five rows of straight inner shore, sheltered from the north only by the neck's fifty-seven-cell westward protrusion standing above it.

The water inside that lee keeps Plate III's pale shape, now re-measured rather than borrowed. In the box rows 305–345, cols 250–312, sea cells run a median 0.4933 — Plate III's 0.494–0.495, within the sounding's own grain: 0.4927 at (315, 290), 0.4924 at (320, 290), 0.4987 at (315, 300) and (320, 300). Off the inner shore the water is a shallow shelf before it is anything else: at row 320, shore col 313, 0.5115 five cells out, 0.5001 at ten, 0.4936 at twenty, 0.4837 at forty, 0.458 at sixty; at row 330 the same slow fall, 0.5103, 0.5012, 0.4992, 0.4673, 0.4468. The first water below 0.48 at row 320 lies fifty-one cells west of the shore, below 0.46 sixty cells west. Row by row the bight's own water stays pale: medians 0.504 for rows 305–315, 0.4891 for 315–325, 0.4768 for 325–335, 0.4809 for 335–346, deepest within the bight 0.4328 in the 325–335 band. The box's deepest cell, 0.4046 at (345, 250), sits south of the bight's fade, in open sea, and is not counted as the Pocket's. Ashore the landing rises without a step: at row 320, 0.5305 five cells inland, 0.5382 at ten, 0.5531 at twenty, 0.5926 at forty — a ramp out of the water, Gathaven's door in miniature, not a threshold crossed. And the step itself carries small print: the fragments at rows 301–305, maxima 0.5201–0.5331, are awash ground in all but name — a close-in approach under the neck threads dots that are land only at the chart's datum.

Plausible, from that: the Pocket keeps its anchor, narrowed to what the survey actually supports. It is a small-craft call and nothing larger: pale water, a shelf some twenty to forty cells wide before the floor falls away, a straight inner shore to lie off, and shelter that is real but one-sided — the neck and Ostrel break the north and northwest weather, and the southwest swell walks straight in, as the gazetteer said. A boat anchors out on the shelf, not close under the shore, where five cells out is still wading water at 0.5115; a boat beaches on the ramp, but the ramp climbs to 0.55 within twenty cells, so beaching is a step ashore, not a stay. What the heightmap cannot do is prove the rest of Plate III's claim, and this entry does not pretend otherwise: the slow back-eddy shed by the Drift, and the soft bottom the settling sand implies, are a mechanism and a hope, not a sounding — elevation gives depth and shelter, never current and never sand. The Pocket's other plain term is thirst: no drink belongs to it. The nearest named water is Dunsen, its mouth at (438, 319), some ninety-seven rows south of the bight's fade — a boat that waits out weather in the Pocket waits on carried water, or walks south for it. So the Pocket earns, as an anchorage, what the Sunfall Shore earned as a road: a call, not a home — the one place on a shelterless coast where a small boat may lie under a lee, in pale water, with nowhere to fill and nothing to stay for.

Unwritten: whether the Drift's back-eddy truly turns inside the lee, and how hard, if it turns at all; what the shelf's bottom is — sand, as the settling story wants, or rock the heightmap cannot feel; whether anything has ever anchored here, and whether the awash dots at the step dry at the lowest honest tide or stay wet in every season; whether the neck's south flank itself offers a landing above the frayed step; and what Ostrel's own west shore offers north of the junction, where this survey stopped rather than claim another land's coast for Varra's. The Pocket was asked in earnest, and the honest answer is smaller than its name: not a bay with a mouth to measure, but a corner where the coast steps east and the north wind stops — a lee, kept, and kept small.

29. Ostrel's west shore — a smooth coast, and no second Pocket (plausible)

Proven: Entry 28 stopped at the neck-to-Varra step rather than claim another land's coast for Varra's, and promised this coast in earnest: surveyed on code/heightmap.npy with code/survey_ostrel_west_shore.py, no terrain touched, following the westmost coast row by row from the step northward to Ostrel's tip. The coast is smooth in a way the Sunfall Shore, for all its straightness, never quite is. Its tip is a rounded point, not a spike: first land a single cell at (163, 251), widening to 237 at row 165, 221 at row 170, with northmost land at rows 164 for cols 250–260. From there it bulges west to its shoulder, 174 at row 230, then leans east without a break: 185 at row 250, 203 at row 270, 232 at row 285, 253 at row 300, where the step to the Pocket begins. North of that step the largest single-row step in the whole coast is five cells, at row 166, the tip's own flare; everywhere else no step exceeds three cells. No bite, no cove, no second headland, no closing point: the west shore is one long curve, out to a shoulder and back in to the neck. The junction itself is dated, not guessed: at row 280 the land still runs in two pieces, (220, 447) and (469, 597), with the Northern Inlet's water between them; at row 281 the gap narrows to cols 452–461; at row 282 the runs join, (224, 599), and the inlet is shut. South of row 282 the coast this entry walks is the Gatneck's west flank, not Ostrel proper — low ground leaning east to the step — and north of it, Ostrel's own west face.

The water off that face is a fringe, not a shelf, and the distinction matters. In the twenty cells of sea nearest the shore, medians run 0.4901 for rows 165–200, 0.4724 for rows 200–235, 0.4838 for rows 235–270, and 0.4925 for rows 270–301 — pale water, close in, all along the coast, palest at the north and at the neck flank. But it does not last: at row 220, shore col 177, the water falls 0.5000 five cells out, 0.4730 at ten, 0.4186 at twenty, 0.3534 at forty, 0.294 at sixty, 0.241 at a hundred; at row 260 the same shape, 0.5038, 0.4842, 0.4607, 0.401, 0.3316, 0.2679. The first water below 0.48 lies nine cells west at row 220, twelve at row 260, fourteen at row 180; below 0.46, thirteen, twenty-one, and nineteen. A boat off Ostrel's west shore is in pale water for a dozen cells and in open water a dozen more — the Pocket's twenty-to-forty-cell shelf, this is not. Ashore the ramp climbs without a threshold, as at the Pocket, but steeper on Ostrel's face and gentler on the neck: at row 220, 0.5469 five cells inland, 0.5647 at ten, 0.5991 at twenty, 0.6534 at forty, 0.6846 at sixty; at row 280, on the flank, 0.5331, 0.5469, 0.5602, 0.5761, 0.5994. Ostrel's summit, re-found in passing, stands at (228, 249), 0.7036, and its sampled cap — rows 222–234, cols 242–258 — runs 0.686 to 0.7036, a flat top at this grain; the worn dip the gazetteer records does not resolve as a hollow here, and this entry does not place it. And the Westburn, Ostrel's west water, leaves no mark on the coast at all: through rows 250–275 the westmost coast moves one cell a row, shore elevations 0.520–0.524, ten cells inland 0.55–0.56 throughout — no notch, no mouth-bay, no widening where Plate III puts its mouth at the shore line. A burn you have to know, not one you can see from the water.

Plausible, from that: Ostrel's west shore is a coast to pass, not a coast to stop on — the mirror image of the Pocket, and the reason the Pocket matters. The Drift walks this whole length southward, but with no bight to shed an eddy into, there is no trap for it and no settling water: a boat southbound rides past, a boat northbound works against the set in open water, and neither finds a lee, because the coast never turns its back to the weather. The west wind, which the Pocket escapes under the neck, meets this shore full in the face for a hundred and thirty-five rows. What the shore does offer is close water and a quick step ashore: a boat may lie in the pale fringe a few cells out, beach on the ramp, and walk up — and the Westburn is drink, if you know where its mouth hides, which the coast itself will not tell you. That is a watering stop, not a landing: no shelter to wait weather in, no shelf to anchor out on, no bay to name a settlement after. Ostrel keeps its one anchorage, and it is not on this side. The hill behind the shore explains the shape — a single dome, 0.7036 at its top, falling straight to the sea on the west, its water running off in two small threads and nothing larger — and the coast keeps the dome's honesty: nothing invented, nothing hidden, nothing to stop for except the water itself.

Unwritten: whether the Westburn runs year-round, or dries to the trickle its small catchment suggests; what the pale fringe's bottom is — sand, as a landing story would want, or rock the heightmap cannot feel; whether the Drift's southward set is felt close in under the shore, or only in the deeper water a dozen cells out; whether the rounded tip itself offers a lookout or a landing, which elevation alone cannot say is a use; and what Ostrel's east shore and the Northern Inlet's west end offer in the same measured detail, to close the isle's circle. Ostrel's west shore was asked in earnest, and the honest answer is a clean one: a smooth coast, a fringe of pale water, a hidden drink, and no second Pocket — the isle's open side, kept open.

30. Ostrel's east shore and the Northern Inlet — a lee, and a roadstead kept (plausible)

Proven: Entry 29 closed the isle's open side and promised this one in earnest: surveyed on code/heightmap.npy with code/survey_ostrel_east_shore.py, no terrain touched, following the eastmost coast row by row from the tip southward to the neck, and the inlet's water with it. The east shore is as smooth as the west, and smoother in its middle reach: from the tip's flare — a single cell at row 165, eastmost 237, widening 38 cells to 275 by row 170, the tip's own flare, the largest single-row step of the whole coast at 31 cells in row 166 — it leans east without a break: 286 at row 180, 296 at 190, 303 at 200, 305 at 210, 318 at 230, 332 at 240, 344 at 250, 349 at 260. Through the middle reach no five-row step exceeds six cells, and no bite, cove, or second headland interrupts it. At row 264 the coast steps: 29 cells in one row, 355 to 384, and the steps keep coming — 387 at 265, 425 at 270, 438 at 275, 447 at 280. That is not Ostrel's dome anymore. Inland the ramp tells the change: rows 190 to 250 climb the dome's flank, 0.533 to 0.54 five cells in, 0.547 to 0.56 at ten, 0.58 at twenty, 0.60 to 0.66 at forty; at row 270 the ground goes flat, 0.5212, 0.5236, 0.5312, 0.534, 0.5372 — low ground barely above the sea. The step at row 264 is the coast leaving the dome for the neck's east shoulder, and the shoulder is the inlet's southern shore: low, flat, and running east until the inlet shuts. The shutting itself is dated, not guessed: at row 280 the land still runs in two pieces, (220, 447) and (469, 597), with the inlet's water between them; at row 281 the gap narrows to cols 452–461; at row 282 the runs join, (224, 638), and the inlet is shut. South of row 282 the box cols 350–480 holds no sea at all — the cul-de-sac is closed, as the gazetteer said, and no water passes through.

The inlet's water deepens northward and shallows to its head, which is the shape a roadstead wants. In the box rows 240–290, cols 350–480, the sea medians run 0.4385 for rows 240–255, 0.4912 for 255–270, and 0.5069 for 270–282, deepest 0.3843. Off the east face the floor falls fast in the north: at row 190, shore col 296, the water reads 0.5098 five cells out, 0.4955 at ten, 0.461 at twenty, 0.4015 at forty, 0.3566 at sixty, 0.2799 at a hundred; at row 230 the same shape, gentler — 0.5148, 0.5076, 0.4855, 0.4333, 0.3848, 0.3427. But in the inlet's middle reach the fall holds: at row 250, shore col 344, 0.505 at five, 0.4902 at ten, 0.4711 at twenty, 0.4709 at forty, 0.4443 at sixty, 0.4255 at a hundred — Plate III's roadstead water, 0.44–0.47, re-found forty to sixty cells off the shore, where a boat can lie. South of that the head shallows away: along cols 380 to 440 the water reads 0.4704, 0.448, 0.4432, 0.4239 at row 250, and by row 270 the head's water is a narrowing gutter against Varra's side, 0.5055 at col 440 with land to its west — pale, shallow, and closing. At the mouth there is no sill to cross: north of the tip the water deepens eastward without a bar, 0.4249 at col 300 to 0.2314 at col 450 on row 165, and the inlet opens cleanly to the north sea. And Ostsen, Ostrel's named thread, leaves no mark on the coast at all: through rows 184–199 the eastmost coast steps one cell a row past its traced mouth at (189, 296), shore elevations 0.520–0.523, ten cells inland 0.544–0.551 throughout — no notch, no mouth-bay, no widening. Like the Westburn, a burn you have to know, not one you can see from the water — but this one earned its name on Plate III, and keeps it.

Plausible, from that: Ostrel's east shore is the mirror image of its west: a coast to stop on, not to pass. The hill that leaves the west face open to the weather breaks the westerlies here, and the inlet does the rest — a cul-de-sac takes what enters it and settles it, no through-current to drag an anchor or hurry a boat onward. The roadstead Plate III inked is real, not just inked: pale water 0.44–0.47 in the middle reach, forty to sixty cells off a smooth shore, the Arriving Strand in reach across the gutter, the neck at your back. A boat that wants Varra's north coast without rounding Ostrel's tip into open sea can lie here and wait weather out — the one anchorage the west shore could not offer, and the reason the gazetteer's line holds: what enters the inlet settles. The Drift's dying arm belongs here too: the flood's return, dividing at Ostrel, sends a part into the inlet to die against the neck, and that dying is the roadstead's calm — not a current to ride, but the absence of one. The head is the limit of the promise: south of row 265 the water shallows to a gutter and then to land, so the roadstead is the middle reach, not the whole bay — a call, not a harbour town, and the isle keeps it that way.

Unwritten: whether the roadstead's bottom is sand, as an anchorage story would want, or rock the heightmap cannot feel; whether Ostsen runs year-round, or dries like its west-face twin may; whether the Drift's dying arm is felt as a set inside the inlet at all, or only as the stilling the cul-de-sac implies; what Varra's Arriving Strand spur offers as the inlet's east wall, in the same measured detail; and whether the low neck shoulder floods at high water, or stands dry in every season. Ostrel's circle is closed now — west shore passed, east shore stopped on, the inlet's water measured — and the honest answer is a paired one: the isle's open side and its sheltered side, one dome, two coasts, and the roadstead between the hill and the Strand.

31. Varra's Arriving Strand facing the inlet — the east wall, a Strand ramp, not a headland (plausible)

Proven: Entry 30 closed Ostrel's circle and left one wall of the inlet unmeasured: this one. Surveyed on code/heightmap.npy with code/survey_arriving_strand_inlet.py, no terrain touched, following Varra's west edge row by row from its first cell to the inlet's closure. The wall is short. Varra first stands against the inlet as a single cell at (257, 535), 0.5202 — the tip of a low northward nub — and its west edge leans west in small steps: 533 at row 258, 530 at 260, 528 at 262 and 263, 525 at 264, 524 at 265, no step larger than three cells. Then the edge appears to jump — 18 cells at row 266, 20 at row 267, the two largest steps of the whole face — and the jump is a merger, not a bay: at row 266 a low fragment stands separate in the gap, cols 488–498, every cell between 0.5200 and 0.5215, barely above the sea; by row 267 it has joined Varra's run, whose west edge is 486. South of the merger the face is smooth in the same grain as Ostrel's two shores: 485 at row 268, 483 at 269, 482 at 270, 480 at 272, 477 at 275 and 276, 473 at 278, 471 at 279, 469 at 280, steps of one or two cells the whole way, no bite, cove, or headland anywhere on it. Shore elevations along the entire wall stay between 0.5201 and 0.5226 — this face never rises out of the Strand at its edge. The closure is the one Entry 30 dated from the other side, confirmed: at row 281 the gap is ten sea cells, cols 452–461, and at row 282 the runs join and the inlet is shut.

The inlet narrows against this wall like a funnel closing. Between Ostrel's east edge and Varra's west edge the sea gap runs 185 cells at row 258, 178 at 261, 140 at 264, then 78 at 267 — the merger rows — 56 at 270, 43 at 273, 36 at 276, 25 at 279, and the final ten at 281. The water against the wall shallows with the narrowing: in the east half of the inlet, cols 440–480, the sea medians run 0.4276 for rows 240–255, 0.4751 for 255–270, and 0.5053 for 270–282, the head's band, whose deepest cell is only 0.4827. Off the tip the shallows are gentler and the opening clean: west of the row 260 edge the water reads 0.5162 five cells out, 0.5116 at ten, 0.5062 at twenty, 0.4979 at forty; and north of the tip, down col 535, it deepens without a sill — 0.5014 at row 250, 0.4737 at 240, 0.4152 at 220, 0.3617 at 200. Inland, the wall's face is a ramp: at row 275 the ground reaches 0.55 twenty-eight cells east of the edge, at row 280 in seventeen cells; at row 270 it stays low for a long way, 0.521 at the edge and only 0.5452 sixty cells in — the Torrassen's flat valley ground behind the shore. The tip itself climbs southward, not at its face: down col 535 the land reads 0.5202 at row 257, 0.5272 at 260, 0.5368 at 265, 0.5459 at 270, 0.5615 at 280, 0.6123 at 300, 0.6654 at 320. And the Torrassen's traced mouth, (270, 563) at 0.5217 with open water a row north of it, sits on the north shore eighty-one cells east of this wall's row 270 edge: the river belongs to the Arriving Strand's north face, round the tip — no traced thread crosses the inlet wall at all.

Plausible, from that: the east wall is a shore, not an anchorage — the far side of Entry 30's promise, now measured. The roadstead lies across the water in the inlet's middle reach; what this wall offers a boat is a landing, not lying-room. Its face is the Strand's own grain, a ramp a landing could walk up the way Gathaven's door is walked up, but the water against its lower half is the closing gutter, median 0.5053, too thin to lie in — a boat that crossed to this shore would be beaching, not anchoring. Entry 30 said the Arriving Strand was "in reach across the gutter," and the reach is literal and shrinking: fifty-six sea cells at row 270, twenty-five at 279, ten at 281, none at 282, where the two shores become one piece of land and the inlet simply ends. Nor does the wall offer a drink: no thread crosses it, and the nearest named water is the Torrassen's mouth, round the tip on the open north shore. The wall's whole character is plainness — twenty-five rows of low, smooth Strand between the open sea and the closed head, the quiet side of the inlet's mouth and the plain end of its funnel.

Unwritten: whether the wall's bottom is sand, as a beaching story would want, or rock the heightmap cannot feel; whether the low fragment at the head, cols 488–498 at row 266, stands dry at high water or goes awash like the Stones' neck, its 0.5200–0.5215 being barely a claim on dryness; whether any landing ever had reason to choose this wall over the roadstead across the water; and what the Arriving Strand does east of the Torrassen's mouth, where the north shore falls back into a shallow scoop the first-land rows already show — cols 575–600 not standing until rows 273–281, against the tip's 257. The inlet is measured on both sides now: Ostrel's dome and lee to the west, Varra's low ramp to the east, and between them a roadstead that narrows to a gutter and shuts.

32. The Arriving Strand eastward from the Torrassen's mouth — the shallow scoop, a scoop in plan only (plausible)

Proven: Entry 31 ended at the Torrassen's mouth and named what lies east of it: a shallow scoop in the north shore. Surveyed on code/heightmap.npy with code/survey_arriving_strand_scoop.py, no terrain touched, following the first land row column by column from the inlet tip (257 at col 535) to beyond the Ellisen's mouth. Between the two mouths — the Torrassen's at (270, 563) and the Ellisen's at (302, 703), one hundred and forty columns apart — the shore never once steps more than a single cell from one column to the next: of the 140 column-to-column moves, 92 change the first land row not at all and the other 48 change it by exactly one. No bite, cove, notch, or headland interrupts the whole segment; even the two river mouths are invisible in the shore line, each a plain crossing — the Torrassen's traced mouth cell at 0.5217 with open water a row north, the Ellisen's at 0.5221 likewise, the coast stepping by ones straight through both. The fall-back itself is gentle and soon flattens: 270 at col 563, 271 at 575, 277 at 585, 281 at 600 — and there the retreat nearly stops. Across the sixty columns from 595 to 655 the first land row stays between 278 and 288, median 281: a plateau of shore, only eleven rows south of the Torrassen's mouth line and twenty-four south of the tip. East of the plateau the fall resumes at the same single-cell pace — 287 at col 650, 293 at 665, 295 at 680 and again at 690 — reaching 302 at the Ellisen's mouth, where it levels a second time: cols 703 to 733 all stand at rows 302–303 before the coast falls away again, 312 at col 743 and 335 at 813.

The water in front of the scoop is open-sea water, deepening without hesitation. In the twenty cells north of the shore, sea medians run 0.4781 for cols 563–582, 0.4822 for 590–609, 0.4812 for 620–639, 0.4935 for 650–669 — the palest band, off the plateau's east end — and 0.4892 for 680–699. The profiles fall steadily at every sampled column: at col 600, 0.4978 five cells north, 0.4828 at ten, 0.452 at twenty, 0.3912 at forty, 0.2914 at a hundred; at col 640, 0.5043, 0.4874, 0.4509, 0.3851, 0.3289, 0.2754. No bar, no sill, no shelf anywhere along the segment. Ashore, the Strand is narrow — five cells to 0.55 at col 600, thirteen at 625, fifteen at 650 and 700 — and the ground behind it climbs fast and keeps climbing: at col 600, 0.5552 five cells inland, 0.5837 at ten, 0.6268 at twenty, 0.7709 at sixty; down col 630 the land reads 0.5493 ten cells south of the shore, 0.6547 at forty, and 0.8166 by row 360. The scoop's back is not a low bay-head plain; it is the Table's own northern slope, rising straight off the shore. No traced thread crosses the segment between its two mouths.

Plausible, from that: the scoop is a scoop in plan only, and the plan flatters it. Eleven rows of retreat across a hundred and forty columns, with no western rim but the Torrassen's mouth shore and no eastern rim at all — the coast does not come back north to close it, it simply levels at the Ellisen and falls away again later — is not a bay in any sense a boat can use. It offers no lee: the north wind and the northern sea meet it as fully as they meet the tip, and the Drift's westward spread along this coast finds no eddy to shed here, no settling water, no roadstead. What the shore does offer is the Arriving Strand's ordinary gift, held evenly for a hundred and forty columns: deep water close in, a narrow Strand, and a ramp a landing can walk up — with a named drink at each end, the Torrassen at the west and the Ellisen at the east, and none in between. If the inlet's east wall was a landing shore, this is a landing coast: longer, smoother, and no more a harbour at its middle than at its ends. The scoop's real character is its smoothness — a shore so even that two rivers cross it without leaving a mark on its line.

Unwritten: whether the scoop's bottom is sand, as a beaching story would want, or rock the heightmap cannot feel; whether the palest nearshore band, off cols 650–669, is a bottom feature or only the profile's accident; whether any lesser burn crosses the plateau untraced, as Ostsen and the Westburn hid on Ostrel's shores; and what the Arriving Strand does east of the Ellisen's mouth, where the flat run at rows 302–303 gives out after col 733 and the coast falls south again toward the Stoneward country. The scoop was asked in earnest, and the honest answer is a shallow one: a flattening in a falling shore, measured, and no harbour found in it.

33. The Arriving Strand eastward from the Ellisen's mouth — the flat run, and the turn at the cape (plausible)

Proven: Entry 32 ended at the Ellisen's mouth and named what lies east of it: a flat run, and then the coast's turn toward the Stoneward country. Surveyed on code/heightmap.npy with code/survey_arriving_strand_east.py, no terrain touched, following the first land row column by column from the Ellisen's mouth to the cape, and Varra's eastmost land row by row around the turn. The flat run is real, and nearly as flat as its first reading: across cols 703 to 733 the first land row stays between 301 and 304 — seventeen columns at 302, seven at 303, the rest at 301 or 304 — and the Ellisen's mouth is as invisible in the line here as it was from the west, its traced mouth cell at (302, 703) reading 0.5221, the shore stepping straight through it. East of the run the fall resumes gently: 305 at col 735, then the segment's only real steps, two cells at col 737 and three at 738, 311 at 740, 312 at 745 and 750, 313 at 755 and 760, 314 from 765 to 780, 313 at 785, 315 at 790, 318 at 795, 320 at 800. Of the ninety-seven column-to-column moves from 703 to 800, sixty-six change the first land row not at all, twenty-nine change it by one, and only those two at 737–738 change it by more. No bite, cove, notch, or headland interrupts the whole run from the Ellisen to the cape's shoulder, and no traced thread crosses it east of the Ellisen's mouth.

The water in front of the run is the scoop's water continued — open sea, deepening steadily, and shallowing as the cape nears. In the twenty cells north of the shore, sea medians run 0.4931 for cols 703–722, 0.4986 for 730–749, 0.4931 for 760–779, and 0.5068 for 785–804, the palest band of the whole north coast, off the cape's shoulder. The profiles fall without a bar, sill, or shelf: at col 703, 0.5093 five cells north, 0.4725 at twenty, 0.4159 at forty, 0.2586 at a hundred; at col 800 the water stays thinner close in — 0.5121 at five, 0.509 at ten, 0.496 at twenty — before falling to 0.3309 at a hundred. Ashore, the character changes under the even shore: at col 703 the ground still climbs like the Table's slope, 0.5692 twenty cells inland and 0.6601 at sixty; at col 770 it stalls, 0.5624 at twenty and 0.5454 at forty; at col 800 it barely climbs at all, 0.5341 at twenty and 0.5301 at sixty. The Strand itself stays narrow to 0.55 — nine to twenty-four cells at the sampled columns — but the distance to 0.60 blows out eastward: thirty-six cells at col 710, 59 at 730, 108 at 750, 126 at 770, and 291 at 790. The cape's shoulder tops out at 0.5624, at (336, 785): low ground, barely out of the Strand's own band.

East of col 800 the fall steepens and the coast turns. The first land row runs 321 at cols 801–802, 323 at 803–804 — the hooked point's cell, (323, 804), reads 0.5205, as Plate VIII inked it — 326 at 805, 329 at 807, 332 at 810, 335 at 813, 342 at 815, and 347 at 817, in steps of up to five cells, against the ones and zeroes of the flat run. Then the first-land reading gives out: cols 818–820 carry only stray fringe cells — rows 370–380 at col 818, a single cell at (376, 819), a single cell at (373, 820) reading 0.5202 — and cols 821–826 carry no land at all until row 453. That gap is the Gate's northern channel: the box at cols 818–826, rows 300–450, has a sea median of 0.5094 and only thirteen land cells in it, every one of them the cape's eastern fringe. Read by rows instead, the cape's east face bulges and falls back: Varra's eastmost land stands at col 800 at row 320, 807 at 330, 814 at 340, 817 at 350, 812 at 360, 818 through rows 370–380, then recedes — 807 at 390, 806 at 400, 800 at 410. The water off that face is shallow close in and deepens steadily, with no sill: at row 370, 0.513 five cells east, 0.5079 at ten, 0.4933 at twenty, 0.4737 at forty. And in the cape's northeast shoulder the survey found a small sea slot: at cols 811–817 a gap of sea separates a northern fragment of the shoulder from the main run below it — one row wide at cols 811–812, five rows at 813, eight at 815, and fifteen rows at 817, rows 354–368, where it opens into the channel. The slot's cells read 0.5151–0.5200, barely sea; west of col 811 the fragment joins the main land again, in the unbroken run at col 810, rows 332–387. Across the channel stands the southwest Stone: the Needle's Eye's closest approach, some twenty-six cells between (310, 827) and the hook at (323, 804), stands as Plate VIII measured it.

Plausible, from that: this is where the Arriving Strand ends — not with a headland but with a lowering. The flat run is the Ellisen's parting gift, a landing coast in the scoop's own grain, with a named drink at its west end and none east of it; but the backland tells the truer story: the Table's northern slope, which rose straight off the shore behind the scoop, gives out column by column until the shore is backed by a low apron that reaches 0.60 only after hundreds of cells, if at all. A landing here walks onto low, open ground, not up a ramp to the Table. The cape itself is no strongpoint — 0.5624 at its shoulder is a swell, not a hill — and its importance is purely positional: it is the corner of the world's great isle, the north face behind it, the east face ahead, the Stones across the Needle's Eye to the northeast, and the Gate's channel opening south of its fringe. The slot in its shoulder is a gutter, not a harbour: at most fifteen rows wide and seven columns long, shallow, and shut at its west end — a crack a small boat could be caught in, not a call it could choose. Neither face offers a lee: the north face east of the Ellisen is the scoop's verdict repeated, and the east face's thin fringe is the Gate's own water at its narrows, which Plate II already judged fast. The turn is the feature: measured, it is a bulge to col 818 and a fall back to 800 in ninety rows — a corner rounded, not a point speared.

Unwritten: whether the run's bottom is sand, as a beaching story would want, or rock the heightmap cannot feel; whether the slot drains or stands at low water, its cells lying between 0.5151 and 0.5200, barely below the sea's own line; whether the cape's eastern fringe cells, at 0.5202–0.5233, stand dry at high water or go awash like the Stones' neck; whether any lesser burn crosses the flat run untraced, as Ostsen and the Westburn hid on Ostrel's shores; and what the Shiproad Shore does south of the cape, where the eastmost line already bulges again — 817 at row 430, 816 at 440–450 — below the bay Plate II called Gathaven. The north coast is measured whole now, from the Northern Inlet's wall to the cape's fringe: one long landing shore, two named drinks upon it, no harbour anywhere on it, and a corner at its end.

34. The Shiproad Shore southward from the cape — Gathaven's shore and its two headlands, re-measured (plausible)

Proven: Entry 33 ended at the cape's fringe and named what lies south of it: the Shiproad Shore, and the one bay on it. Surveyed on code/heightmap.npy with code/survey_shiproad_shore_south.py, no terrain touched, reading Varra's eastmost land row by row from the cape's east face to below Gathaven's south headland. The cape's face holds the line the turn promised: eastmost land stands at col 800 at row 320, eases out to 817 by row 350, and reaches the segment's northern extreme, 819 at row 376, without a single step larger than two cells. Then the shore falls back by ones — 818 at row 380, 813 at 385, 807 at 390 — into the bay. The recess is as Entry 19 measured it from the water, and the line confirms the bowl: eastmost land bottoms at col 799, at rows 412, 414, 416 and 417, with the whole head between 799 and 807 from row 395 to 425 — a bite nineteen cells deep against the headlands' 817–819, and some fifty rows wide between the falls. The return south is the sharpest move on the whole segment: from 808 at row 427 the line steps six cells to 814 at 428 — the largest single step in a hundred and sixty rows — and stands at 816–817 by rows 430–436, holding 815–817 through row 450. Of the hundred and sixty row-to-row moves from row 320 to 480, sixty-eight change the eastmost column not at all, sixty-four change it by one, and only twenty-eight by more, the rest of the larger ones all south of the bay, where the shore begins its run to Lissen.

The water in front of the shore is the Gate's water, thin and steady. In the twenty cells east of the shore, sea medians run 0.5064 for rows 330–349, 0.5067 for 380–399, 0.5076 for 430–449, and 0.5083 for 460–479 — the palest band the farthest south, as the coast leans toward the channel's narrowing. The profiles fall without a bar, sill, or shelf: at row 350, 0.5113 five cells east, 0.5029 at twenty, 0.4192 at a hundred; at the bay's head, row 410, 0.517 at five, 0.4914 at twenty, and about 0.47 at a hundred, over the bowl Entry 19 sounded. Ashore, neither headland is a height: within twenty-five cells of the shore the north headland tops at 0.5537 (row 387) and the south at 0.5729 (row 445) — swells in plan, not rises. The Strand runs nineteen to twenty-three cells to 0.55 at the bay and the south headland, but broadens behind the cape's face to fifty-two cells at row 350, where 0.60 is ninety-three cells inland: the cape's back is the broadest low apron on this coast. South of the south headland the shore advances without interruption — 825 at row 455, 841 at 465, 848 at 470, 854 at 480, 859 through rows 490–500, 875 at 520, and 903 at 540 — the long southeast run toward Lissen's mouth at (543, 911), which Entry 22 walked from the other end. No traced thread crosses the segment between the cape and that run's end.

Plausible, from that: Gathaven's two headlands, re-measured in earnest, are headlands the way the scoop was a scoop: true in plan, modest in fact. Nineteen cells of recess against a shore that otherwise moves by zeroes and ones is a real bite — the only one on this coast — but the arms that hold it rise barely out of the Strand's own band, and the south arm returns to the line in a single six-cell step, a corner rather than a climb. That is the bay's whole bargain restated from the landward side: the shelter is in the water's motion, the ebb curl Entry 19 sounded, not in any wall of ground. What the shore itself gives a landing is the same gift the north coast gave, widened at one place: behind the cape's face a low apron fifty-two cells deep to 0.55, room the bay's own Strand, at twenty-two cells, does not match — the cape's back, not the haven's head, is this coast's broadest standing ground, and it stands at the corner where the Gate's fast water begins, not in the haven's lee. South of the bay the coast commits to its southeast run early and evenly, with no second bite, notch, or landing-place offered between the south headland and the open shore Entry 22 measured: Gathaven is not the best harbour on the Shiproad Shore by a shade, it is the only break in a hundred and sixty rows of smooth, advancing coast.

Unwritten: whether the six-cell step at row 428 is a rock shoulder or only the profile's accident, the heightmap giving shape and never substance; whether the palest nearshore band south of the bay, rows 460–479 at 0.5083, shoals further toward Lissen or is the channel's own thinning; whether any lesser burn crosses the southeast run untraced, as Ostsen and the Westburn hid on Ostrel's shores; and what the Gate's east side does opposite this shore, where the unnamed dome's west coast — surveyed once for a roadstead it did not have, in the Errel retirement — has never been walked row by row in its own right. The Shiproad Shore is measured whole now, from the cape's fringe to Lissen's mouth: one bay, two low arms, a broad apron at the corner, and a long smooth run south.

35. The unnamed dome's west shore — the Gate's east side, walked in its own right (plausible)

Proven: the dome's west shore has been surveyed once before, for a roadstead it did not have, in the Errel retirement; this entry walks it row by row in its own right. Surveyed on code/heightmap.npy with code/survey_dome_west_shore.py, reading only the island's own land component (peak cell (491, 1070)) — a plain column window catches Varra's shore south of the island's last row and was corrected for in the making, no terrain touched. The island stands on rows 432–537, and its westmost land makes one single arc: 1063 at the northern tip (row 432, the island eleven cells wide there), an eighteen-cell flare to 1045 at row 433, then a smooth westward lean — 1039 at row 437, 1012 at 460, 1007 at 470, 998 at 489 — to the shore's westernmost stand at col 996, held without a single step for nine rows, 495–504, and then the retreat east: 1006 at 512, 1024 at 520, 1050 at 527, 1062 at 532, ending in a single land cell at (537, 1077). Along the main face, rows 460–520, the line never once turns back on itself: of sixty moves, twenty-nine change the westmost column not at all, twenty-two by one, and only nine by more, none by more than three. There is no recess anywhere in the arc — no row where the shore falls back and returns the way Gathaven's head falls nineteen cells behind its arms. The west thread's outlet, which Entry 6 placed at (476, 1004), sits exactly on this line: westmost 1004 at row 476, between 1005 above and 1003 below, the thread opening through the shore without a notch, a delta, or a step of any kind.

The water in front of it is the Gate's, and it falls away honestly. Twenty-cell nearshore medians west of the shore run 0.4977 for rows 440–459, 0.4877 for 465–484, 0.5009 for 490–509, and 0.5084 for 515–534 — the palest band over the southern tail, where the shore itself is retreating. The profiles fall without a bar, sill, or shelf to the channel's floor: at row 476, 0.5011 five cells west, 0.4843 at ten, 0.4678 at twenty, and about 0.45 at sixty to a hundred; at the westernmost stand, row 495, 0.4717 at twenty and 0.4362 at a hundred. The gap this shore faces is the one Entry 6 measured: 218 cells of open sea at row 440, narrowing to 136 at row 500, widening again to 148 at 520. Ashore, the dome rises straight off its own shore, with none of the cape's broad apron: the Strand is six cells to 0.55 at rows 470 and 476, eight at 495, eleven at 450, and 0.60 stands only seventeen to eighteen cells inland on the face — at row 476 the ground is already 0.6106 twenty cells east of the shore and 0.6383 at forty; within twenty-five cells of the shore the ground tops at 0.6275 at (475, 1030). Only the southern tail relaxes: at row 520 the Strand broadens to twenty-six cells to 0.55, and 0.60 is a hundred and seventy-five cells inland, the tail's low flat running out to the single last cell at row 537.

Plausible, from that: the retirement's verdict stands, now measured rather than inherited: this shore keeps no roadstead because it keeps nothing at all — a convex arc that bulges toward the Gate instead of cupping away from it, a face the tide washes along rather than a bowl it settles in. What the west side offers is exposure in its purest surveyed form: the charted world's busiest water in front, a six-cell Strand, and the dome's own slope climbing straight behind, with no flat worth standing on between the water and the rise except the southern tail's low run, which faces the Gate's south mouth and its dissolving current, not any lee. The west thread makes the same point in fresh water: the island's second drink reaches the sea as a plain crossing of the shore line, gathering no pool, cutting no gully the heightmap can show, so that even the dome's own water leaves it without making a place. The one softening is at the arc's waist: the nine held rows at col 996 are the shore's straightest reach, the closest this coast comes to a wall, and a wall — even a low, smooth one — is where the Gate's wash would strike most evenly and where anything clinging to the shore would hold the firmest footing the west side has.

Unwritten: whether the eighteen-cell flare at row 433 is the northern tip's true shoulder or a fringe the survey rounds into the island, the heightmap giving shape and never substance; whether the southern tail's broad low flat ever holds standing water behind its Strand, as the Marsh Flat question was asked and answered on Varra — here unasked, and no mere claimed; whether the west thread runs year-round where it crosses the shore, its inland course already judged plausible, not proven, in the Plate VII notes; and what the dome's north shore does between this arc's tip and the north thread's mouth, the one side of the island no entry has yet walked in its own right. The Gate's east side is measured now: one arc, no bite, a narrow Strand, and the dome rising straight behind it.

36. The unnamed dome's north shore — the crown's flat edge, and the thread's one-cell nick (plausible)

Proven: Entry 35 left this side unwritten — what the dome's north shore does between the west arc's tip and the north thread's mouth. This entry walks exactly that segment in its own right. Surveyed on code/heightmap.npy with code/survey_dome_north_shore.py, reading only the island's own land component (peak cell (491, 1070)), no terrain touched. The island's northernmost land is not a point but a held line: row 432 stands across eleven columns, 1063–1073, before the edge begins its drift south — 433 at col 1074, 435 at 1079, 436 at 1083, 437 at 1088 — reaching row 438 at col 1094, where the north thread opens. Across the whole segment, thirty-one column-to-column moves: twenty-five change the northmost row not at all, six by one, and none by more. There is no recess anywhere in it. The north thread's mouth sits almost on that line, but not quite: its traced path ends at (437, 1094), 0.5198, a breath under the sea, and the first land cell, (438, 1094) at 0.5239, stands one cell behind neighbours that hold row 437 on both sides — a one-cell nick in the edge, where the west thread crossed its shore exactly on the line. No notch, no gully, no delta the heightmap can show. Read whole, the north edge behind the segment runs cols 996–1153 in the same temper: of a hundred and fifty-seven moves, eighty-one step 0, fifty-three step 1, twenty-three larger, and none larger than four, those at the far diagonal ends where the edge turns down the west and east sides.

The water in front of this shore is the Unwritten's own. No other land stands north of the island in the sampled columns — at cols 1070, 1094, and 1120 the grid holds nothing above the sea all the way off the chart. Twenty-cell nearshore medians north of the shore run 0.4805 for cols 1045–1064, 0.4745 for 1065–1084 — the deepest band, off the flat edge itself — and 0.4828 for 1085–1104, deeper water than the west shore faces anywhere. The profiles fall without a bar, sill, or shelf: at col 1075, 0.4975 five cells north, 0.4749 at ten, 0.4335 at twenty, 0.3485 at a hundred; at the thread's own column, 1094, 0.4586 at twenty and 0.3283 at a hundred. Ashore, the dome climbs off this shore as straight as off its west: the Strand is six cells to 0.55 at cols 1063 and 1075, seven at 1085, eight at the thread's column, and 0.60 stands sixteen cells inland at col 1063 and nineteen at cols 1075, 1085, and 1094 — at twenty cells south the ground is already 0.6156 at col 1063 and 0.6056 at the thread's column. Within twenty-five cells of the shore the ground tops at 0.6309 at (458, 1062). And the climb does not level: sixty cells south of col 1063 the ground is 0.6578, within a breath of the summit itself, 0.663 at (491, 1070) — which stands only fifty-nine rows behind this edge. The north shore backs directly onto the cap's own slope.

Plausible, from that: this is the crown's edge, not a coast the dome happens to have. The summit cap the Errel seam found — a broad flat top, no dip — reaches the sea on this side as a straight rim: eleven held columns is the cap arriving at its own edge, and the six-row drift to the thread's mouth is the gentlest fall any shore in this atlas has yet shown, smoother even than the west arc's main face. What it faces is the emptiest water in the charted world: no land, no bank, no awash rock to the chart's edge, the Long Outward's own water. The west shore at least fronts a road — the Gate's traffic passes it by. The north shore fronts nothing passing at all. And its one break is the smallest a shore can have: the dome's main drain, its one real thread, leaves through a nick one cell deep, gathering no pool and cutting no gully, so that even here the island's water makes no place. A landing on this side would step from the deepest nearshore water yet measured on the dome straight onto the climb to its top — a shore that is all threshold, the cap's doorstep with no room in front of the door.

Unwritten: whether the eleven held columns at row 432 are the cap's true rim or a fringe the survey rounds into one line, the same question the west arc's eighteen-cell flare left open at row 433, the heightmap giving shape and never substance; whether the north thread runs year-round where it nicks the shore, its inland course already judged plausible, not proven, in the Plate VII notes; what the water north of the island does — no current there has any survey behind it, only the Unwritten's lore, and no set is claimed; and what the dome's east shore does from this edge's east end, where the northmost line turns down through cols 1140–1153, past the east thread's mouth to where the southern tail begins — the one side of the island no entry has yet walked in its own right. The dome's north is measured now: a flat crown edge, a one-cell nick, deep open water, and the cap climbing straight behind it.

Next, whenever it pulls: the unnamed dome's east shore — from the north edge's east end down past the east thread's mouth to the southern tail, walked in its own right.