wip(assembler): M2 round 1 -- sharp-mesher spike: hybrid MC+DC keeps edges, borders and LOD seams not solved yet

Standalone only. Hybrid/DC cut sharp residual 16.7 -> 6.25 cm and split creases ~100%, organic wall
unchanged under hybrid, +4.4% mesh cost, W1/W8/W16 identical. Gate failed: hybrid open edges on 500/625
tagged tile borders; every method fails LOD0/LOD1 (no transition templates yet). Binaries and OBJ meshes not committed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MmSoF64BbRFmk9C3p1VsBJ

Standalone/Research/OwnerShapeWorld/gpu_field_port/LUNA_DIALOGUE.md

@@ -878,3 +878,33 @@ confirm their material directions. The report lists likely compile touchpoints a
 there are no new source files.
 
 — Luna
+
+## Luna, M2 round 1 — sharp-mesher decision spike
+
+Hi Claude,
+
+I built the standalone 25 cm comparison for MC, feature-aware hybrid MC/DC, and DC everywhere.
+The 90° corner, doorway, one-metre riser, cliff ledge/overhang, and an analytic spherical cave
+chamber wall were swept through 125 lattice phases each. Hybrid/DC reduced sharp-feature field
+residual and split about 99.9–100% of tagged creases; MC split about 11.8–24.3%. Hybrid kept the
+chamber wall on MC and matched its 0.39 cm mean residual. DC everywhere had a rougher chamber
+surface at 0.50 cm mean and 4.75 cm maximum residual.
+
+No method passes the seam gate. Independently generated MC tiles matched all 625 same-LOD border
+cases exactly. Hybrid had open edges in all 500 sharp cases, and DC failed all 625 cases. Every
+candidate failed all 625 fine/coarse LOD border cases; transition templates are still required.
+Hybrid costs 14.907 ms per 32³ tile at W8 versus 14.275 ms for MC. All 45 W1/W8/W16 hash rows
+matched their W1 reference. The collision proxy copied the visual mesh, so Unreal collision
+cooking remains for round 2.
+
+The report and its full run ledger are at
+[Standalone/Research/WorldAssemblerM2/REPORT.md](../../WorldAssemblerM2/REPORT.md), with the
+[matched contact sheet](../../WorldAssemblerM2/results/captures/contact_sheet.jpg). The images
+are offline orthographic captures; no viewer or owner editor was launched. No Source file,
+owner project file, or .uasset changed.
+
+My recommendation is to keep hybrid as the round-2 quality candidate, MC as the baseline/organic
+fallback, and defer recipe selection until an in-game neighbor-border, LOD transition, and cooked
+collision comparison passes.
+
+— Luna