# Tutorial verification — 18 October 2026

## Checked

- Re-summarized all 49 existing file reports: zero extraction errors, 41,878,112,224 bytes, 19,320,233 records, 648,058 eligible footprints; 31 feet and 18 millimeter files. Reports cover the supplied archive, not the whole city.
- Ran the portable extractor's known-answer helper assertions inside Rhino, then read the 22,186,708-byte SI parks model independently. Reproduced 6,486 object records, 213 eligible footprints and millimeter units. SHA-256 `80c7403600dd52d5b4888e298d33f6509f0f6c735596268f0d86ae49a61229cb` matched the original report; size and modification time were unchanged during reading.
- The generator's numeric self-check covers deterministic seed reset, changed seed, n=4, default n=7, maximum n=20 and footprint scale, paired area/aspect, slider responses and invalid-input rejection.
- Ran RhinoCommon geometry generation at default, n=4, n=20/scale=1.3/spacing=24/height_scale=3, and seed43. All resulting Breps were valid closed solids.
- Saved a separate static 3DM, then reopened it: meters, 99 objects, all valid closed solids. Runtime: Rhino 8.35.26251.13001.
- The city image is a mathematical diagram generated from the same `layout()` records, not a Rhino screenshot or a claim of photorealism.

## Limits

The full 49-file extraction is the existing research run, not a second traversal during tutorial packaging. The portable wrapper was repeat-checked on one small source file; larger-file memory behavior was not benchmarked again. Geometry was tested through RhinoCommon; the reader's end-to-end manual Grasshopper component creation and UI-specific Python editor settings still need user confirmation. No `.gh` file is included; the editable component is created by following the instructions. No active Rhino document or existing Grasshopper canvas was changed. No Boolean union, printability, structural or planning-compliance validation is claimed.

The Markdown research snapshot retains references to large original evidence files that are not bundled. Regenerate those with the extractor; the article's direct download links point to included files. Source-derived cohort records are deliberately filtered and do not support district-wide statistical inference.