passages: dense centripetal spline + open-air half-pipe attempt (L111, failed audits)

Route resampled to ~1 m (310 -> 2,790 segments); open air writes a 0.75 m floor slab plus 1.0 m
side walls, no roof. The 3D spline overshoots between planned route points, raising floor grades
above 15 deg; a grade-limited variant left 2 unsupported samples on passage 0-1, and the shell/
ripple audits still model the old chord route. Stopped after two designs. Not merged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016dhxM43dnWvUu2B5Ph1iG1

Standalone/Research/WorldAssemblerM3/Linux111/REPORT.md

@@ -0,0 +1,42 @@
+# Linux111 report
+
+## Design
+
+Version 218 resamples passage controls with a centripetal Catmull–Rom curve at about 1 m spacing (310 control spans became 2,790 prepared segments). Open-air shells use a 0.75 m floor slab and side walls up to 1.0 m above the floor; there is no roof. Passage interior remains higher priority than shell. The second design constrained the floor route to a 14.5° grade ceiling.
+
+The per-tile candidate lists were not measured. The 310→2,790 figures are prepared route segment counts, not tile candidate list sizes.
+
+## Speed
+
+L109 medians below are the candidate column from the saved five-pair timing matrix. L111 timing was not run after the audits failed.
+
+| Stratum | L109 W8 near-ready (s) | L109 W12 near-ready (s) | L111 W8/W12 |
+|---|---:|---:|---|
+| SurfaceWorld | 0.737 | 0.761 | Not run |
+| FloatingIslands | 0.911 | 0.863 | Not run |
+| TunnelNetwork | 0.956 | 0.895 | Not run |
+| VerticalShafts | 0.789 | 0.738 | Not run |
+
+## Audits
+
+- Clang standalone build completed. The second design removed all grade failures. Passage 0–1 still had 2 unsupported walk samples; passages 1–2 and 2–3 passed. The combined resolved-passage audit failed.
+- Mouth-profile audits passed 3/3. Naturalness and ripple audits passed 2/3.
+- The half-pipe shell audit passed 2/3; passage 0–1 reported 11,544 floor and 6,945 wall violations. A targeted trace showed at least one reported floor sample was 0.101 m inside the passage after relief, where interior correctly overrides the shell. The audit's smooth-route overlap model still disagreed at that point, so the remaining shell counts are not conclusive.
+- Fast tier, Clang/GCC W1/W8/W12 hash matrix, readiness timing, and Unreal Linux build were not run. Canonical v218 hashes were not generated.
+
+## Captures
+
+No Linux111 capture images were generated. Mouths, 25/50/75% stations, open-air sections, and programmatically verified 4 m outside/interior close-ups remain outstanding.
+
+The candidate audit found these open-air sections:
+
+- passage 0–1: 6.408–16.782 m; 19.500–24.439 m; 1,044.046–1,498.708 m
+- passage 1–2: 81.654–85.646 m; 104.092–104.829 m
+- passage 2–3: 7.701–26.008 m; 372.697–377.363 m
+
+## Failed attempts
+
+1. Dense spline plus half-pipe: walk grades exceeded 15° and passage 0–1 failed shell, naturalness, and ripple checks.
+2. Dense spline plus a grade-limited floor: grade failures cleared, but passage 0–1 retained 2 unsupported walk samples and its shell/naturalness/ripple audits still failed. Work stopped after this second design as requested.
+
+Logs: `Saved/Linux/l111/build-clang.log` and `Saved/Linux/l111/audit-owner.log`.