# C03_r4 vs INT R02 room fit (measured on the frozen assembly)

Model: `C03_r4.blend`, SHA-256 `f24ebb01a4f9fabf21518c4dde5b2de61d97da360eb0f676903b7c008b610dc4` (104 objects, 146,170 triangles). Rooms: INT R02 `INTERIOR_LAYOUT.json`, SHA-256 `aa14f38b150b966670397a3a12cd9bf054dafb387645a34149ebb5ed2a2942ff` (81 rooms; L00-L03, L25, L26). Proposed perimeter allowance 1.2 m (a proposal, not an approved wall assembly). No INT R02 room was changed. Labels: measured / proposed / interpreted.

Method: slice the enclosure at each room's floor level and clear-height top, worst signed margin from the room rectangle to the outline (shapely), as in the C02 re-fit. Produced by a subagent (scripts in `scripts/`); the L25/L26 count was checked independently by the lead with a simpler corner-in-superellipse test (no bulge, no allowance): 20 of 32 rooms outside the glass, matching.

## Results (measured)

| Group | Rooms | FITS (>= 1.2 m) | INSIDE_BELOW_ALLOWANCE | OUTSIDE |
|---|---|---|---|---|
| L25 + L26 (tower skin) | 32 | 10 | 2 (L25-RM081/082, +0.19 m) | **20** |
| L00-L03 (podium wall lines, interpreted) | 49 | 38 | 1 (L00-RM005, +1.15 m) | **10** (8 outside wall and deck, 2 wall only) |

Glass footprint (measured on the mesh): 51.0 x 39.6 m at z = 126 against the proposed 60 x 40 m block; 50.2 x 39.0 m at z = 130.5 against 60 x 48 m.

Worst shortfalls: L26-RM001/002/003/006 -14.35 m (corners about +-30, +-24); L25-RM002/003 -13.97 m; L26-RM008 -9.19 m. Podium: L00-RM003 and L03-RM001 -22.2 m; L00/L01/L02-RM001/002 -19.4 m. Full per-room list: `fit_C03_r4.csv` / `.json`.

## Podium findings

- The C03 front wall sits about 27 m further back than C02's, because the entrance lobby back wall is at y = -32.6 to -33.25 m and the front tiers are recessed so the basins show. INT R02 front rooms (arrival atrium and galleries) lie in front of that wall. Interpreted: the front bay is bounded by the lobby back wall rear face; closing the gap at the wall chord or at the lobby glazing gives the same statuses.
- The 5.4 m L02 clear height (room top z = 17.4 m) runs about 0.9 m into the tier-1 slab, whose underside in C03 is at z = 16.5 m (1.5 m lip). This affects all 11 L02 rooms. Cause: C03 slab thickness (builder choice), not INT R02; C02 used 0.6 m (underside 17.4 m).

## C02 regression (method check)

The same script reproduces the C02 result: 75 FITS / 5 INSIDE_BELOW_ALLOWANCE / 1 OUTSIDE_WALL_AND_DECK, all statuses matching both published C02 files, margins within 0.0004 m (`regression_C02.csv`).

## Options to clear the L25/L26 conflict (proposed, not decisions)

Each was rebuilt and measured by the reviewer (`sweep_options.json`, `variant_results.json`):
- The glass plan is a superellipse with exponent 2.5, so even an untapered 68 x 54 m body leaves the L26 block corner 5.44 m outside.
- Squarer plan (N >= 7.8, set in `tower.py`) with the glass held at 68 x 54 m: +1.30 m margin measured.
- N = 10 with top 67.8 x 51.7 m: +1.23 m.
- Knobs only (`CL_TOP_W`/`CL_TOP_D`/`CL_FLARE_K`, N = 2.5): the glass must be about 82.8 x 65.8 m wide, which is far wider than the R06 68 x 54 m base.
- Keeping the top at 42 x 32 m with N = 10 and K = 0.08 leaves 4 rooms about 4 mm under the allowance.
All of these change the silhouette; their silhouette effect was not measured. Taper, top size, plan shape and room-versus-shell choices are Legend's decisions.

## Limits

- The working-tree `podium.py` and `ribs.py` hash differently from `FREEZE_C03_r4.json` (rounds after the freeze); results use the frozen blend. Rib intrusions in rebuilt variants are indicative only.
- Not tested: silhouette impact of the options, engine collision, door alignment, intermediate floors L04-L24 (no INT R02 layout).
- INT R02 is unchanged; nothing here approves a shell, scale, placement or engine behaviour.
