Core/Scripts/DrainageAnalysis.cs is the reference's DrainageAnalysis ported verbatim: the
Barnes priority-flood routing fill (8-connected, seeded from the four borders, index
tiebreak, pit fills ONE ULP above the parent so every filled cell keeps a strictly
descending path to its spill; the terrain heightmap itself is never written - the fill lives
in its own array), terminal-basin qualification (depth >= 2 m AND area >= 10,000 px; the rest
are pits filled through), D8 flow directions on the routing surface - FOR ANALYSIS ONLY, the
reverted-as-carving landmine stated in the file - Kahn accumulation, the memoised destination
walk crediting a terminal basin with TOTAL inflow, sea-reaching outlets ranked by drainage
area with the outlet separation, main stems by max accumulation, mountain exits from the
along-stem grade, lean tributaries, lean endorheic terminals, and the promoted giants
(provisional routes computed as the reference did, not drawn - routing is a later task).
WorldScale-denominated; Dir / Acc / Filled / FullFilled / BasinId / BasinInflow exposed so the
caller can prove the invariants.
"The sea" is the OCEAN body from the region layer: RegionLabeling.OceanMask - the 4-connected
water component touching the border, on the CLASSIFY field (the water-side complement of the
land contract). Enclosed lagoons, lake beds and island-fringe pockets are ordinary terrain to
the router.
DrainageTool (4 task-11 seeds at 8192, the eroded fields bit-identical to the 11 dumps)
renders the log-scaled accumulation map and the promoted-candidates overlay (trunks cyan,
endorheic giants orange, lean terminals red; nothing carved) and writes the accumulation
.f32. Oracle, all passing: render AND classify fields bit-identical before/after the analysis
(zero terrain cells written), no water added, full fill >= original everywhere with a
non-ascending path to the border from every cell, ocean mask all below sea and on the border,
dir/acc/candidates identical across two runs.
Endorheic basins are the expected first-class output: on every seed the top giants out-drain
the top trunks - the biggest drainages pool inland because erosion delivers the upland network
only and cannot cross the flats.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013EY3ZTF6NwzF8ukBHQXSK7
Step 0: tag terrain-shape-v1 on a59e52f (the frag_4-locked, gallery-confirmed state).
Core/Scripts/HydraulicErosion.cs is the reference's pass ported verbatim - the four governors
(250,000 droplets, lifetime 384, carve cap 15 m, deposit cap 6 m) on one net-displacement
ledger read both ways and PROVEN on exit (the CARVE-CAP / DEPOSIT-CAP violations throw), the
sea clamp (spawn rejected below sea, death at sea, both brushes skip below-sea cells, the 0.5 m
margin floor on the erode brush), the cone-weighted normalized brush shared by erode and
deposit, the droplet physics (inertia 0.35, capacity 4, min slope 0.02, erode 0.12, deposit
0.15, evaporation 0.004, gravity 4, brush 2), PCG32 seeded at seed + 9271, the crater
exclusion whole. Engine-free; every metres<->raw conversion through WorldScale (no literal
251 - the same multiply/divide, so bit-identical arithmetic).
Tools/Scripts/ErosionPass.cs is the caller (pass 2b): render field only (copied if aliased to
classify), governors clamped as the reference ConfigManager clamped them, the crater exclusion
passed through INERT (no crater => radius 0 => weight 1 everywhere; the reference's
"core < carve factor" warning dormant), and the FLOOD GUARD - render water pixels counted
before and after, any change throws. TerrainGenConfig gains Erosion (default OFF, the anchor
rule) and the governors/physics/crater fields. Pass2Result.WithHeight hands back the eroded
render field. ShadeRenderer is a pure-hillshade plate (land only) because the palette relief's
0.30 hillshade hides half-metre drainage. ErosionTool carries TerrainShapeV1 (the locked shape's
values, pinned once) and the batch: 4 gallery seeds x off/on at 8192, grayscale + .f32 +
relief + shade per field, a mid-slope 1024-px crop off/on (relief and shade), the stats.
Faithful first, no tune: at 8192 the pass touches ~88 % of land cells at a mean 0.14 m, carve
15.00 / deposit 6.00 m at the caps, 0.5 m mean on the massif with 18 % of its cells moved more
than a metre. It reads as dissected summits with radial gully fans and carve/deposit bands
along the slope breaks, not dendritic networks: droplets die of lifetime (170k of 224k) before
they converge. A lifetime probe at 4096 (scratch) shows 1536 and 4096 identical - every
droplet is dead of evaporation by ~1,300 steps - so lifetime is not the lever; droplet count
is. The mid-slope green->yellow transition is not softened at the faithful tune.
Oracle, all passing on 4 seeds: erosion OFF bit-identical to terrain-shape-v1 (the task-10
gallery dumps, 67 M cells each); classify bit-identical off vs on and == the pass-1 field;
region labeling + island tag identical; flood guard (27.5 M water pixels unchanged, every
seed); caps proven; tag/coastline; classify == raw; centre is land; eroded field
bit-identical across two runs.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013EY3ZTF6NwzF8ukBHQXSK7
FragGalleryTool pins every frag_4 value explicitly (amp 0.5, freq 12, window 0.66 +- 0.18,
zero-mean, stretch 2 with band 0.70/0.05 and the sinker stretched, speck revert 2.5e-7, offshore
and shelf off, labeling on) and proves it is the 09 setting: at 4096 the frozen field is
bit-identical to the 09 frag_4 dump on both anchors (a9, 16.8 M cells each), the frag-off field
to the 08 stretch_3 dump (a8), the interior locked against it (r), ids deterministic (o), the
curve untouched (a1). No generation code changed.
Eight seeds at 8192 - the anchors 1063685222 and 999999937 plus 20260822, 31415926, 27182818,
16180339, 14142135, 17320508 (constants chosen before any render) - each with grayscale, .f32,
relief and the labeled-regions overlay, and the hemisphere-split count/size table with a stated
numeric read rule (too solid < 12 big islands across both hemispheres; shredded if big < 25 % of
all; else good spread). Seven read good spread; 14142135 trips the ratio rule (17 big of 69) and
is the most solid mainland of the eight (99.3 % of land). Graduation held.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013EY3ZTF6NwzF8ukBHQXSK7
A dedicated term, not the edge noise scaled: pass 1's edge noise is positive-only, squircle-
modulated, 2.5x the base frequency - the coastline's jitter; scaling it would roughen the whole
rim and bias the coast inward. CoastalFragment adds, to the PRE-power falloff, amp * window(f) *
noise(x + off, y + off) with its own deterministic field (seed offset 9109, coordinate offset
0.37 map widths - D-059), zero-mean, at 12 periods per map width (the lobe/neck scale, chosen by
probe against 20 and 32: 20 climbs into smalls, 32 is fuzz), weighted by a smooth window on the
falloff value itself, centred 0.66 (the coast's falloff for a median base noise, from the 08
diagnosis), half-width 0.18, EXACTLY ZERO beyond - so the interior, the massif and the deep sea
are bit-identical by construction. Thin necks of barely-land flip first; nothing is detected,
nothing is stamped. FragmentBitesOnly is exposed (probed: it erodes the coast inward - mainland
-3/-8/-14 % at 0.06/0.15/0.30 - rather than detaching pieces; off by default).
Batch: BatchRoot(9, "coastal_fragment") - 4 amplitudes (0.06, 0.15, 0.30, 0.50) x task 08's two
seeds at 4096, stretch fixed at 2 (08's stretch_3 rung, whose dump is the bit-identical baseline,
a8), sinker untouched, speck revert at < 4 cells, offshore/shelf off; lean render (regions overlay
+ relief + .f32), the hemisphere-split count/size table with largest-three and histograms.
Oracle, all passing: a1 (the curve untouched), a8 x2 (frag OFF at stretch 2 == the 08 field,
16.8 M cells each), interior locked (r: every cell clear of the window bit-identical, 8/8),
high-ground report (s, informational), centre is land, tag/coastline, classify == raw, ids and
heights deterministic.
The read: both hemispheres fragment; the north more readily (N 10 -> 28, 13 -> 40 islands across
the ladder; S 4 -> 31, 14 -> 28) and the south's big stretch pieces are the zero-mean noise's
other edge: on 999999937 the 83k-cell southern fragment is BRIDGED back onto the mainland at amp
0.50 (mainland +183k). The working range is amp 0.15-0.30; 0.50 is the bookend. Graduation held.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013EY3ZTF6NwzF8ukBHQXSK7
Diagnostic first (SouthernStretchTool, ISLA_DIAG_ONLY; scratch/southern_diagnosis.md). Along
the central south profile the mask blend climbs 0.018 per 0.01 N, edge noise adds 0.03-0.06,
and the southern sinker is zero until 0.75 N then adds 0.024 per 0.01 N before the 2.5 power;
the coast sits at 0.76-0.82 N, ~0.06 N north of where the blend alone would end. The stretch
compresses the southward distance the mask sees inside a fixed feathered latitude band,
y' = yB + (y - yB)/(1 + s*ramp), texture (base noise, edge noise, latitude) untouched, so the
extended mass keeps the elevation of the rows it came from. With the sinker reading the REAL y
it cancels the stretch outright: median coast saturates at 0.83-0.84 from s = 1 to 16 and the
south stays 1-10 tiny nubs on both seeds. The lever is therefore two things - extend reach AND
hold the sinker back - and this pass couples them 1:1 (StretchSinker = true; the real-y mode is
kept as the config's other setting). Ladder chosen from the sweep, non-linear: 0.5, 1, 2, 3, 5
(peninsula / onset / few big pieces / big pieces consolidating / over-stretched sheet); 8 and 16
are indistinguishable from 5.
Topography: cells above the band take the untouched code path (fy == y), so the classify field
north of the band is bit-identical by construction. SouthernStretch.cs holds the band constants
(start 0.70 N, feather 0.05 N - fixed for the batch) and the ramp. TerrainGenConfig: SouthStretch
(the swept axis, 0 = off), SouthBandStartFrac, SouthBandFeatherFrac, StretchSinker.
Oracle, all passing: a1/a3 stretch OFF bit-identical; a3b stretch 5 north of the band
bit-identical to the task-03 dump (2.93 M cells), a4b the same at 8192 against the
terrain-curve-v1 gallery dump (46.97 M cells north of the band, 20.1 M differ below - the
relaxation); per field north bit-locked classify and render vs stretch 0 (p, p2), northern
island set invariant (q), centre is land, tag/coastline, classify == raw; ids and heights
deterministic. Batch: BatchRoot(8, "southern_stretch_explore") - 5 levels x 2 seeds
(1063685222, 999999937 - the two 07 seeds with the most mainland south of the band), lean render
(regions overlay + relief + .f32), the hemisphere-split count/size table. Region labeling on,
offshore / shelf / speck revert off: the pure fragmentation signal.
The read at 4096: s 0.5 peninsula (coast +0.04 N); s 1 onset (999999937 6 -> 12 southern islands,
11 substantial); s 2 few big pieces (35k / 86k-cell fragments); s 3-5 a solid sheet to the trench
whose southern edge is the trench wall - a ruler-straight coast - and whose island count falls as
pieces re-merge. Gravel never dominates. Graduation held, per the task.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013EY3ZTF6NwzF8ukBHQXSK7
Core/Scripts/RegionLabeling.cs is the shared-infra contract, built to the letter: it runs on
the CLASSIFY (raw) field; land is 8-connected, the deliberate complement of water's 4 (a
diagonal isthmus joins; the water either side stays separate); a component is a maximal
8-connected set of land cells; the MAINLAND is the component containing the map centre —
not merely the largest, which a later fragmentation step could flip — with a flagged
fallback to the largest if the centre were ever water (asserted, never needed: oracle m);
every other component is an island; per component id / sizeCells / centroid / hemisphere
(by centroid, one label per island) / isMainland. Ids come from a fixed scan order and are
proven stable across two generations (oracle o, 16.8M cells). It knows nothing about
offshore or stamped. Engine-free, in Core as C++-candidate math; the hemisphere convention
moved there with it, OffshoreAnalysis aliases it.
Tools/Scripts/RegionPass.cs is pass 1c: label, revert, relabel, tag. The island tag
(renamed IsIsland; IslandHemisphere from the component's centroid; Pass1Result.Regions
carries the whole table) is now a CONSEQUENCE of labeling — every non-mainland component.
That is the fix for the chat2/06 overlay, which tagged only what the offshore pass raised:
1063685222 has 11 natural islands including a 94,511-cell detached mass, 20260821 has 19,
all grey in 06's tags.png and all coloured now. The offshore pass itself is untouched; its
internal Tag stays for its own guards and is no longer exported.
The speck revert (TerrainGenConfig.SpeckRevert / MinLandComponentFrac) lowers every
non-mainland component below the threshold to the mean of its ring of adjacent sea cells,
held strictly below sea. Origin-blind: a natural nub goes the same way as an offshore dot
(6 natural components / 273 cells on the bare 1063685222 field at threshold_mid — reported
as a3r, informational). Lower-only and component-only are asserted cell by cell in the
pass and re-proven on the finished fields by oracle n (mainland bit-identical filter OFF
vs ON; every changed cell in a sub-threshold island, lowered below sea); the mainland is
never a candidate and its size is asserted unchanged across the revert. A reverted
offshore island leaves its submerged skirt as a shoal — not this component, by the rule.
Classify/render consistency is by construction (pass 1, curve identity at sea) and
asserted by oracle k. Deliberately OFF in the bare TerrainGenConfig for the reason the
shelf and islets are: the raw field has natural specks, so default-ON would move the
calibration pool and every regression dump; the batch turns it on.
Thresholds swept on 8 seeds at 4096 (1e-5 / 3e-5 / 1e-4 of the map = 168 / 503 / 1,678
cells): low removes 0–6 nubs per seed, mid (the config default, equal to the offshore
guard) 2–11, high 26–36 — most of the offshore islands, the "fewer, bigger" bookend. The
count/size table carries natural / pre / post counts per hemisphere, min/median/mean/max
and a log-spaced size histogram — the instrument for the southern-stretch step.
Oracle, all passing: a1, a3, a4 (8192, 67M cells) with labeling ON + revert OFF; a6 NEW —
labeling ON + revert OFF on the 06 preset bit-identical to the 06 batch's render field
(labeling is pure analysis); j0; m, n, o, i, j, k, l, b per field. Batch:
BatchRoot(7, "region_labeling") — exactly 4 plates (three thresholds on 1063685222,
threshold_mid on 20260821, the table's most-natural-islands seed), each with grayscale /
.f32 / relief / the labeled-regions overlay / the tag overlay, plus count_size_table.md/.csv.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013EY3ZTF6NwzF8ukBHQXSK7
The seeded floor from chat2/05 — FloorNorth/FloorSouth, the PCG32 placement, StampWeight,
StampRadiusFrac/StampCoreFrac/StampEdgeJitter, the centre separation and land gap, the
floor self-check and oracle (h), the centre rings on the tag overlay — is reverted out
whole, as a forward commit. It looked stamped. It is one checkout away at 3b96e06. What
remains is ONE island mechanism: the organic noise-field layer, with chat2/05 stage 2's
shape untouched (freq 16, crest 24 m pre-curve, core 0.25, sharpness 2.5, corners on,
moat + falloff test + outer bound unmoved), now OffshoreSettings.Organic(). The count per
hemisphere is a statistical outcome of the tuning, read off a table, never guaranteed.
The south was measured before anything moved (OffshoreDiagnosis, over task 01's pool at
2048): the premise that the south generates fewer islands is not what the field says.
The NORTH has 1.75x less island-eligible ocean (56 % of its ocean is zone vs 69 %; 233
zone rows per column vs 407) and under half the island candidates (1,183 over-threshold
peaks in zone vs 2,422). Per cell the south is richer too (484 vs 414 candidates per
Mcell). The gate that loses candidates is the moat in the north and the falloff test in
the south — the bulging south coast pushes the 0.72 contour out — but neither binds
enough to suppress it. So the fix is density, and the binding constraint for "a couple
north, consistently" is the north.
The knobs: Density (the main one, the calibration quantile) and SouthWeight (south density
= Density x weight — applied to the quantile, so a south island looks exactly like a north
one; there are just more). Three levels batched on 12 seeds at 4096, SouthWeight 1.25:
0.016 N 2/4.4/10 S 8/13.8/21 (every seed clears N >= 2, S >= 3, but N's minimum IS 2),
0.022 N 3/6.3/12 S 15/19.9/29 (the preset — the first level whose north minimum sits
comfortably above the target), 0.030 N 3/7.8/15 S 19/25.2/33 (the too-many bookend). The
south weight is the developer's stated preference, not a fix; the SouthWeight 1.0 tables
(S 6/10.9/16 at 0.022) are in the batch's scratch/ for the comparison.
The guards, so more density does not buy slop, each a revert by component membership:
specks (< 3e-5 of the map's area, ~500 cells at 4096 — 70-100 per field at these
densities, the noise caps the reference also surfaced), clusters (the smaller of two
islands whose shores are within 0.8 % of the map width — 0-7 per field), blobs (> 6e-4 —
never bit). And the reference's submerged humps, now attributed by HUMP rather than by
bounding box: the first 4096 plate probe showed hollow ghost rings beside islands — a
guard-dropped island's cap went back to seabed but its rim survived inside its
neighbour's bbox keep-region. A hump (one connected raised region) now stays only if it
holds a kept island, and a dropped island's own cap goes even inside a kept hump.
Oracle, all passing: a1 (Phase-1 dump), a3 (task-03 dump), a4 NEW — offshore OFF at 8192
bit-identical to the terrain-curve-v1 tag's own 04 gallery dump over 67,108,864 cells —
j0 shelf inert on land, and per field moat / mainland unmoved / tag-coastline / HMaxSeed
(unchanged, 1.289169) / classify on all 36 table fields and 4 plates. Batch:
BatchRoot(6, "offshore_organic_tune") — exactly 4 plates (three densities on 1063685222,
density_mid on 424242, the table's sparsest south) + count_table.md/.csv + diagnosis.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013EY3ZTF6NwzF8ukBHQXSK7
Stage 1 fills the IslandFalloff.cs stub from the reference, verbatim: the coast shelf
(below-sea only, depth-preserving, held strictly below sea by MathF.BitDecrement — the
clamp that makes "cannot move the waterline" exact rather than statistical) and the islet
layer (OffshoreBlob, CalibrateThreshold against the field's ACTUAL distribution, the
OffshoreZoneWeight moat + pre-Trench-falloff test). Both run as pass 1b — OffshorePass, a
second sweep over the finished pass-1 arrays with the same per-pixel arithmetic in the
same order — and HMaxSeed is retaken AFTER them, as the reference did. That closes
chat2/00 Drift §2. The value did not move on any of 15 runs; the order is now right by
construction and oracle (l) prints it every time.
Stage 2 is the reshape, OffshoreSettings.Hybrid(): a seeded floor of ≥2 N / ≥4 S islands
placed by a PCG32 off the world seed — min-separated, clear of ALL existing land by a gap
so each stamp is its own connected component by construction, fully inside the zone so
the moat and falloff protections gate the floor exactly as they gate the organic layer —
plus the noise layer on top, smaller (freq 16), lower (24 m pre-curve, which the
preserved toe squashes to ~5 m and keeps there across curve tweaks), flatter (core 0.25),
crisper (edge sharpness 2.5, stamp rim jittered so it is rigid without being a compass
disc), south-weighted (0.007 N / 0.012 S, blended across the midline), corners allowed
(trench mask 0.90→0.97). Every lifted cell is tagged with its hemisphere — NORTH is rows
[0, N/2), y runs south, read off the spine's southern fade and the southern sinker, not
invented — and carried through Pass1Result → Pass2Result for a consumer that does not
exist yet.
Two things the probes taught, both now in the code: the first organic densities produced
183 blobs of which 174 were noise debris (a six-config sweep fixed that), and the
reference's lerp-to-crest leaves shallow humps across the seabed wherever a blob fails to
surface (a guard reverts them outside any surviving island's skirt; specks by component
membership, bumps by location). The faithful control keeps both, because it is the
reference — its islets measure min 1 cell, median 28.
Oracle, all hard checks passing: floor met on every hybrid seed (N 2–4, S 9–16); zero
land bridges; every offshore-OFF land cell bit-identical with offshore ON; tag ↔ coastline
consistent in both fields; curve-off still bit-identical to Phase 1's dump and
continuous_restored to task 03's. Both gates default OFF, deliberately: flipping them is
the act that retires the Phase-1 regression dumps, and that belongs in a task that
re-baselines the oracles.
The faithful control on seed 8675309 produced one north island. That is the gap the
floor exists to close.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DCWNaDZPfTiAy3meGNGgqt
Renders the committed curve (tag terrain-curve-v1 = 639dc5f) across 8 draws to answer
two questions the anchor seed alone could not: does the curve produce good islands
generally, and is the mid-slope abruptness a curve trait or a lucky/unlucky draw.
No curve code is touched — `git diff terrain-curve-v1` over ClimbCalibration,
ContinuousCurve, HeightCurve, CurveAnchors, CurveKnots and Shaping is empty.
SeedGalleryTool pins mountainLift/peakSharpness/lowlandCeiling as CONSTANTS with no
env override, unlike every other tool in this phase. A gallery must not be knob-tunable:
a stray ISLA_* left over from a probe would render eight plates of a curve nobody chose
and they would look exactly like the real thing. It re-measures the calibration on task
01's pool at the iteration size, so the plates are the gated curve rather than a
look-alike.
The eight seeds were fixed before the first render and never screened or replaced. Seven
are fresh; none is in the calibration pool, so none shaped the curve it is testing.
Two things worth recording about the metric:
- The curve's normalized mid-slope CANNOT discriminate between seeds. It is dv/du on a
fixed control polygon and each seed's denormalization is an affine rescale of both
axes, so it is the same number on every draw by construction. The first cut of the
tool measured exactly that and dutifully printed an identical 1.86 eight times.
- What ships instead is the spatial height gradient through the 100-220 m band, in
metres per pixel — the thing the eye is actually reacting to, and genuinely
seed-dependent through both spikeMax and the terrain's own gradient.
The character note per seed is an eye call, so the tool refuses to invent it: it reads
scratch/character_notes.tsv and prints "(pending)" when absent. ISLA_INDEX_ONLY=1
rebuilds the contact sheet from scratch/metrics.tsv without re-rendering a pixel.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DCWNaDZPfTiAy3meGNGgqt
chat2/02 dissolved the terraces and lost two thirds of the mountain with them, then
concluded the loss was structural and needed a Phase-1 noise change. That conclusion
was wrong, and this commit is the refutation.
A monotone curve is a free reparametrization: it may be gentle in one place and steep
in another, and can lift bottom-heavy input into a substantial massif without ever
going flat. The area above a height is set by where the percentile->height mapping
crosses it, and that mapping is entirely ours to choose. The 02 sweep that "proved"
the loss structural varied climbFeather, which shapes the JOIN, not the mass
distribution — the wrong knob, and too strong a conclusion drawn from it.
Core/ClimbCalibration — the climb's control points are now MEASURED off the staircase
instead of invented from shape knobs. For p in {10,30,50,70,85,95} of above-ceiling
land, take that percentile's raw height and its staircase output height; PCHIP through
the pairs. That reproduces the staircase's elevation envelope, so the mountain mass
returns, while MinNormalizedSecant floors every grade so the flat bench and plateau
interiors become slope. The floor bites on exactly one segment — the plateau — which
is precisely where the staircase was flat.
ContinuousCurve.BuildCalibrated joins it to the same pinned lowland handover, the same
C1 join and the same per-seed spikeMax. The 02 analytic path survives unchanged as the
"before" contrast, and deliberately keeps its strictly-increasing-secant rule: a
calibrated curve is WAVY by design, so convexity is the wrong invariant for it and the
secant floor is the right one.
peakSharpness replaces summitDrama and fixes its bad trade. Drama steepened the peak by
pulling the summit ONSET down, dragging the whole massif with it (p99 199 -> 121 m).
Sharpness reshapes only above the last measured percentile, leaving that height fixed,
so peak and massif are independent: raising it leaves p90, >100 m and >220 m untouched
and only moves land within the summit.
Measured, both seeds, 2048:
variant >100 m >220 m p90
staircase (target) 16.05% 4.53% 127.4 m
continuous_02default 4.80% 0.63% 58.4 m
continuous_restored 14.25% 3.45% 123.2 m
continuous_bigger 19.70% 5.80% 163.5 m
Oracle all hard checks pass, including (a2) staircase still bit-identical to task 01's
dump and (d) lowlands bit-identical across every calibrated variant. New (g) reports
land above 100/220 m per variant and is deliberately NOT gated — it is a taste target
the developer tunes, and gating it would make mountainLift unusable. What it must never
do is stay silent, which is how 02 lost the mountain unnoticed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DCWNaDZPfTiAy3meGNGgqt
rev 3 of the curve redesign. The developer's verdict on the 01 baseline was that the
LOWLANDS ARE GOOD; the fault is the terracing above them. So this adds a second curve
mode that preserves the low plain bit-for-bit and replaces everything above the flood
line with one smooth monotone climb.
Core/ContinuousCurve — piecewise, and the pieces have different loyalties:
- at/below sea: identity, as ever.
- above sea to K2: DELEGATES to HeightCurve's own toe+red branches. Not "equivalent" —
the same code path, so the same floats. Oracle (d) holds it to that.
- above the ceiling: a Fritsch-Carlson (PCHIP) monotone spline to the 420 m cap,
C1-joined to the red band's exit slope. Monotone by construction for any ordered
control points, which retires the 24-corner sweep; a 10k strict-increase sample runs
per seed anyway, because "cannot fail" is worth a millisecond.
- Build() REFUSES rather than degrades: a ceiling near the old bench, a drama that folds
the summit under its own onset, control-point secants that are not strictly increasing
(the no-magnet rule, enforced rather than hoped for).
Only BENCH_*/PLATEAU_* are dropped. SEA/ORANGE_CEIL/RED_CEIL survive because they are
the storm-ladder FLOOD TIERS and they live inside the preserved lowland; PEAK_CAP and
the per-seed spikeMax normalization survive as the summit.
Shelf detail is forced off in continuous mode: the flat benches it de-slabbed no longer
exist, and painting noise on the climb now would pre-judge what erosion should carve.
Oracle, all hard checks passing:
- (a1) curve off is bit-identical to Phase 1's dump.
- (a2) staircase mode is bit-identical to TASK 01's dump — the control is provably the
control, not a re-derivation. (CurveBaselineTool is pinned to Staircase so the config
default moving to Continuous cannot drift it.)
- (d) lowlands bit-identical to the staircase over 3.6M cells, every continuous variant,
both seeds. The lifted_WRONG bookend fails it on 1.6M cells, as intended.
- (f) sea identity per CELL, not per count, including 67M cells at 8192.
The finding, measured and recorded in the batch scratch: the massif SHRANK. Land above
100 m goes 14.9% -> 4.8%, above 220 m 4.5% -> 0.6%. A feather sweep to the practical
floor recovers ~1.3 points, so this is structural, not a tuning miss: the staircase's
highland area was an artifact of the bench and plateau acting as magnets, and a curve
with no magnets preserves the raw distribution's bottom-heavy shape. "No terraces" and
"the same land up high" are not both available from curve work alone.
Exploration batch, not convergence. A tuning pass follows once a direction is picked.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DCWNaDZPfTiAy3meGNGgqt
Ports the reference's v5 height curve and shelf-detail passes onto Phase 1's shape
and re-calibrates them against this repo's actual pass-1 distribution. This is the
BASELINE the reshape gets judged against, not the reshape.
Core (engine-free, D-060):
- WorldScale — THE vertical yardstick. One metres/raw number (251), replacing the
prototype's three duplicate M_PER_UNIT constants and ~20 bare literals. The
chunk-height coupling it had there is recorded as a DEFERRED vault decision, not
inherited. RawFromMetres divides, matching the reference bit-for-bit.
- HeightCurve — the 7 bands, the frozen corner-fix blends, the per-seed spike
normalization, the 24-corner monotonicity sweep that throws and refuses.
Identity at and below sea, which everything downstream rests on.
- CurveKnots / CurveAnchors — input knots (measured percentiles) and output anchors
(storm ladder) split apart and both made parameters, so the anchors are A/B-able
without editing source. The reference's shipped knots are kept beside the measured
ones as the fidelity yardstick.
- TerrainDetailPass — micro-relief skin plus the shelf-edge KNOT warp (which slides
K3/K4/K5, not height — that is what keeps monotonicity structural). The crater
exclusion is ported and inert until the carve lands.
Tools:
- Shaping — pass 2a, producing the two height fields. classify is bit-for-bit the
raw pass-1 field; render is curved and detailed. Aliased when the curve is off,
as the reference did. Pass1Result is left immutable so the oracle can compare.
- LandHistogram — the calibration engine AND the diagnostic. The reference shipped
six knot literals and threw the measuring instrument away; this rebuilds it.
- ShapingOracle + CurveBaselineTool — four automatic checks before anything is
looked at, and the batch that runs them.
Measured, not assumed:
- Knots re-measured over a 6-seed / 12.8M-sample pool. They differ from the
reference's by at most 5.6 m of world height, against a 44.7 m per-seed spread —
the pass-1 port is faithful.
- Oracle all pass, including pass 1 bit-identical to Phase 1's own .f32 dump.
- Band shares land on 60/13/10/5/8/3/1 to 0.00 pp.
- Knots hold across map size: the 8K delta (5.8 m) sits inside seed noise.
The finding the histograms deliver: 83% of land ends below 100 m and 96% below
220 m, with the median column at 13 m. That is the share targets doing exactly what
they say, not a bug — and it is the developer's call, which is why nothing here
reshapes it and the palette was deliberately left mis-fitted rather than recalibrated
to disguise it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DCWNaDZPfTiAy3meGNGgqt
chat1/05 renamed 01_pass1_port to 02_pass1_port, bringing it into line with the
authoring-task naming convention. ReviewBatchTool and ReliefRenderTool both
defaulted ISLA_SOURCE to the old name, so running either without an explicit
override found no .f32 dumps.
That failed quietly, which is the reason to fix it rather than document it:
HeightField.Load returns null on a missing file, so ReliefRenderTool silently
regenerated instead of loading, and ReviewBatchTool skipped 1_noise_grayscale
entirely with only a printed warning. Nothing corrupted; it just cost generation
time and dropped a view, both easy to miss.
Both defaults now name 02_pass1_port — where TerrainGenTool writes by default,
BatchRoot(task 2, "pass1_port"). Generator output and renderer source agree again.
ISLA_SOURCE stays overridable and unchanged in shape. Note in passing that it is a
FULL folder name, prefix included, not a bare descriptor: it names an existing
folder to read rather than composing a new one, so it does not go through BatchRoot.
Commented at both sites so the asymmetry with ISLA_BATCH reads as deliberate.
Verified end to end, not just by build: with no ISLA_SOURCE set, ReliefRenderTool
reports "loaded in 173 ms" against 02_pass1_port, and ReviewBatchTool produces
1_noise_grayscale again.
Presentation only. No generation file is touched — Topography, TerrainNoise,
IslandFalloff, Pass1Result, TerrainGenConfig and GenerationScale are all unchanged.
Batch naming, fixed in code. The prefix is the AUTHORING TASK number, not a running
counter: 04_review means "the batch task 04 authored", not "the fifth batch". It had
already drifted — tasks 02 and 03 produced 00_smoke through 05_trophy_10240 across
two tasks, so no folder name said which task made what. Now the task number is an
explicit argument to ToolingPaths.BatchRoot(taskNumber, descriptor), which composes
the prefix itself and REFUSES a descriptor that carries its own. All three tools
take it as ISLA_TASK. Verified: ISLA_BATCH=05_foo is refused with a stated reason
and exit 2, and creates no folder.
Also fixed: an exception out of _Ready does not stop Godot — it logs and the process
sits there with no main loop, so a misconfigured run HUNG rather than failing. A
hang looks like slow work, which is worse than a crash. The batch tools now catch,
print what was refused, and exit non-zero.
Grayscale mode: normalize a field to its own [min,max]. This is the one place
per-image normalization is correct — everywhere else anchors are fixed so images
compare, but here the point is to see one field at full contrast. The range is
printed and indexed so a shade reads back to a height. You cannot judge noise
through a palette: a ramp bends the distribution, a hillshade adds shape the data
does not have.
The wide Costa-Rica palette, and the reframing the developer asked for: the pretty
map is the WIDE GRADIENT RENDERED FLAT, and relief is no longer the hero. On raw
pass-1 noise a hillshade has nothing coherent to shade, so it renders fine fractal
bumpiness as fuzz that actively hides the elevation the colour is showing. Strong
hillshade is retained as diagnostic_relief and labelled a dev view — its bumpiness
is exaggerated slope, not extra terrain. Subtle relief is kept for comparison, with
the honest note that it still fuzzes until Phase 2 carves coherent landforms.
Legend: a colour bar with ticks drawn from the palette's own ramp, so it cannot
drift from the map beside it. Ticks are RELATIVE height and the image says NOT
METRES — the conversion is Phase 2's, and labelling it "m" would invent a fact in
the artefact a reader most trusts. Text comes from a 5x7 bitmap font written for
this, specifically so a legend does not drag in the SubViewport capture path.
Presentation only. Nothing here touches Topography, the noise, or any generation
constant; the renderer is handed a height field LOADED from a .f32 dump and has no
way to produce one, so a look change provably cannot move the terrain.
Hillshade: Horn 3x3, cell size 1, edges clamped (never wrapped — the Trench border
is a wall, not a seam). ZExaggeration is required rather than optional: measured
median land slope is 0.22 degrees unexaggerated, so the island shades as a flat
plane. Chosen from a measured slope table — zex 75 gives 16/28/37 deg at land
median/p90/p99, which reads as natural relief. It is a look dial; raw units are not
metres and no code may read it as if they were.
Palette: stops placed on the MEASURED height distribution, not spread linearly.
Land is bottom-heavy (median 0.456, p99 1.113, max 1.415 over 8.6M land columns
across four seeds), so a linear ramp would spend 99% of its range on 1% of the land
and the map would read as green with a few white dots. Bathymetry is compressed by
d/(d+0.35) so the shallow shelf gets the range and the featureless abyss flattens.
Anchors are absolute and fixed across the batch — a map coloured on its own min/max
cannot be compared with its neighbour, and comparison is the point of a batch.
Blend: the naive tint x hillshade darkens everything by 29% before any slope is
involved, because flat ground shades to sin(altitude). So shade is normalized by
its flat-ground value first — flat terrain keeps its true tint and only SLOPE moves
the colour — then shadows multiply while highlights screen toward white. That
asymmetry is the difference between a colour ramp and a map you would frame.
Relief fades to zero with depth below sea. Not only taste: the deep floor is the
Trench, a synthetic wall whose gradient is ~1.5x the p99 land gradient, so shading
it faithfully draws a bright rim around the map and lights the abyss with noise
mottle. The fade puts relief on the near-shore shelf where the bathymetry is real.
Three named looks (atlas / relief / dusk), gated like any other change and kept
deliberately few — iteration fatigue on a subjective gate is a real failure mode.
Each pair isolates one question.
HeightMapRenderer.cs is retired: its raw I/O moved to HeightField.cs and its ramp
to ReliefPalette.cs, so there is one palette everywhere and the generator's own
quick-look is coloured identically to a beauty render.
No biome words, no classification, no water. The 0.15 line is a colour boundary.
Ports MapGenerator.GenerateTopography's pass-1 loop (reference ~:553-619 at tag
pre-rewrite-reference / ab78883) into Tools/. Working code and its tuned constants
carried over verbatim, read from source rather than re-derived from a design
summary (D-050). Six elements, in the reference's execution order — the order is
load-bearing: the sinker lands before the Pow (superlinear), the Trench after it
(a raw additive wall), preTrenchFalloff captured between them.
The fractal config is now PINNED. The reference set only NoiseType/Seed/Frequency,
so the island's entire fractal character was Godot 4.7.1 constructor defaults that
nothing recorded. Measured on 4.7.2 (NoiseDefaultsProbe): Fbm / 5 / 0.5 / 2.0 / 0.0
— identical to the 4.7.1 values, so pinning reproduces the look with no delta. The
terrain no longer depends on an engine default, and the probe makes a future drift
visible instead of silent.
Two deliberate departures from the reference's literals, both flagged in-code:
- the latitude noise offset is expressed in map widths, not the reference's raw
+1000 px, so a resize no longer samples a different slice of the noise field
(chat1/00 §4.2). Pinned to 1000/8192 so the 8K realization is unchanged.
- the `if (temperature < 0.65f)` spine gate is dropped as the provable no-op it
is — southernFade already reaches exactly 0 at 0.65. Bit-identical.
The reference's "temperature" is ported as a latitude field and deliberately NOT
named temperature: it is stage-1-local input to terrain geometry, and climate is
stage 3 (D-049 §2, D-056). The ±0.1 wobble is kept — it is what stops the spine's
southern terminus being a ruler-straight line.
Deferred to Phase 2 with the seam open: coast shelf and offshore islets (below-sea,
judged once water renders). Pass1Result exposes PreTrenchFalloff at their exact
consumption point, plus HMaxSeed for the seed-dependent curve.
Render is a direct Godot.Image, not the SubViewport capture path — that machinery
exists to composite vector overlays and is why --headless hangs; Phase 1 has none.
Hypsometric above 0.15, bathymetric below, fixed ramp anchors so seeds and ablation
rungs are comparable. No biome words anywhere.
Verified: land fraction 51.5% identical at 1024, 2048 and 10240 — the scaling
discipline holds across a 10x resize. Full-size 10240 pass: 36.3 s.
No curve, erosion, rivers, water, crater, biomes, roads, or mesher.
Stands up the CODE_REPO the rewrite is written into (D-049, D-058). Godot 4.7.2 /
.NET 8 / Godot.NET.Sdk 4.7.2. Nothing is generated, meshed, or ported — this is the
shell and the two data contracts.
Four layers, each with its boundary stated in a directory README:
Core/ math + data only, and engine-free — depends on nothing above it
Server/ authoritative logic (empty this phase)
Client/ rendering (empty this phase)
Tools/ the offline generator — may NOT reference Client (Phase 1 fills it)
Contracts (compiling stubs, no algorithms):
- Column model (D-053): a 2D grid of columns, each a stack of (material, thickness)
runs, any material at any depth. Air is the TOP RUN — there is no air-vs-solid
height branch, and no surface-height accessor exists to reintroduce one. Water is
not a band; it stays an overlay.
- Material schemas (D-054): terrain and building as two append-only registries,
bridged by a recipe seam that is deliberately EMPTY. Identity is a registry key,
never a serialized ordinal. Mesh style is per-ORIGIN and is not a material field.
Rails, so Phase 1 inherits them rather than rediscovering them:
- GenerationScale: MapSize is a parameter, everything derives from it, nothing uses
a raw pixel number. Tightening over the reference — decorrelation offsets are
declared in map widths, not pixels (chat1/00 §4.2).
- ToolingPaths: config/blueprint/output paths all env-overridable, resolved in one
place, so a batch cannot touch the developer's live files.
- FileSafety: the permanent no-deletion rules as throws rather than sentences.
user:// isolation: project name and use_custom_user_dir both pin the runtime dir to
~/.local/share/islaApocalypse-v2/, away from the old prototype's preserved seeds and
batches. Tools/Scenes/UserDirProbe.tscn confirms it rather than assuming it.
