wip(assembler): M3 round 1 -- native FloatingIslands recipe in Core (standalone)

Bounds 517 boxes, 0 missed, negative control caught; capsule route 1044 centers, raised-route control
fails; W1/W8/W16 stable (field bf3ce7da876431e2); 39 ms native vs 64 ms legacy. My review: islands read
as thin flat pads with a radial pinch under them, not rock; next round fixes the look.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MmSoF64BbRFmk9C3p1VsBJ
This commit is contained in:
Fr0zkaandClaude Opus 5.5 committed 2026-10-10 11:36:32 +02:00
1 parent 4661a445be
commit 53a1e69af7
16 files changed
+2971 -1

No files matched your search

@@ -987,3 +987,43 @@ I also regenerated the five MC doorway failure phases with independently sampled
The report has the full metrics, audit logs, skirt captures, source inventory, and failed-run ledger at [WorldAssemblerM2/REPORT.md](../../WorldAssemblerM2/REPORT.md). UBT failed twice while I corrected an undeclared check variable and standalone compatibility indexing, then passed; there was no `0xC0000142` retry, new UECC folder, owner editor/project access, `.uasset` edit, commit, push, or stash. The temporary skirt-export helper was removed; no permanent source file was added.
— Luna
## Luna, M3 round 1 — FloatingIslands standalone Core proof
Hi Claude,
The reusable `FloatingIslandRecipeV1` family now lives in Core, with seeded stable island IDs,
cluster planning, lobe-warped outlines, flat-ish tops, tapered closed undersides, finite influence
boxes, sharp-tag-ready connectors, and a planner-reserved supported route. The frozen owner
FloatingIslands strate was the visual oracle. On its matched legacy probe, native deliberately
uses bounded harmonic contour warp instead of legacy query-space 3D fBM, closes the nominal
underside, guarantees a five-island route backbone, and constrains height changes for capsule
grade. The seed produced six islands and five connectors; consequently this local native slice is
denser and more routeable than the sparse stochastic legacy slice. I am not claiming visual parity.
The final bounds audit passed 517 boxes / 13,932 point samples with zero missed bounds, 108
intersecting boxes correctly left unknown, four void boxes certified as air, and its shrunken-island
negative control detected. The final composed route passed 1,044 capsule centers / 283,968 field
samples, with no rock collisions, unsupported centers, or grade failures. Raising it 8 voxels caused
all 1,044 support checks to fail, as intended.
Plan and field output repeat at W1/W8/W16. The recipe, plan, manifest, and field hashes are
`111825edfebea17f`, `65675554b7aa947e`, `03a4685554edce1a`, and `bf3ce7da876431e2`. Existing
M1 round-1 and SurfaceWorld round-4 manifest hashes reparsed unchanged. On the same 32,768-point
workload, legacy took 63.62 ms and native 39.31 ms (0.62× legacy time); the sign mismatch was
2.512% over the sampled, mostly-air region, with 90 legacy-solid versus 885 native-solid samples.
The CPU-only matched contact sheet includes overview, underside, edge/cliff, tile/LOD reference
planes, and route views; each capture is under 300 KB. The tile/LOD view is a reference-plane
inspection, not a production seam-closure test. No viewer, GPU, owner editor, or owner project build
was used. The Unreal manifest adapter and in-game/cooked route checks remain for later M3 rounds.
The final standalone audit and Core `Target All` build passed. The report and run ledger are at
[`WorldAssemblerM3/Round1/REPORT.md`](../../WorldAssemblerM3/Round1/REPORT.md), with the
[matched contact sheet](../../WorldAssemblerM3/Round1/results/captures/contact_sheet.jpg). The
report lists all compile/harness corrections and five route iterations; no run crashed, no
`0xC0000142` retry occurred, and no build or test approached the 15-minute limit. Four new source
files are listed there. No `Source/` or `Standalone/Source/` file, owner project file, `.uasset`,
commit, push, or stash was touched.
— Luna