Separate wetland ground painting study

One current diagnostic study is retained beside the unchanged approved original WOG-A-003578. The new painting uses the preserved wetland for identity and approved meadow WOG-A-003579 for detail/style reference; approval does not transfer. The built-in image_gen tool generated one native 1536 by 1024 RGB image. Its complete native image and original were directly inspected. It reduces long root loops but introduces more prominent grass/leaf/stem marks and a warmer/darker appearance.

Twelve isolated component cases cover both sources, two scales and three current/private samplers. CPU/GPU differences are at most three channel levels, with exact alpha and stationary/pan-return pixels. The complete 1024 by 1920 comparison board was directly viewed.

Twenty normal local-client frames compare the approved control with one explicitly private wetland-source substitution at a naturally occurring empty wetland site, span2. The public sampler and on-disk manifests remain exact. All four Region returns and Hex reloads have exact geometry and canvas pixels. Both cross-source Hex foreground comparisons are exact; Region foreground was not separately isolated. All eight complete saved desktop/phone Region-close and Hex scenes were directly viewed. The study retains visible directional grouping and recognizable clumps, with dry warmth, physical plant/grain size and weak wetland/reed contact still unresolved. It is not an accepted improvement or admitted replacement.

The sole existing disposable QA world was reused without world commands, terrain edits, copied saves or fabricated entities. Browser contexts and server closed normally; no raw readback files are retained. Exact tool-output/workspace copies and one previously retained control witness now share native bytes through hardlinks, preserving all paths and reclaiming 11,548,115 duplicate physical bytes.

The study and findings are registered as WOG-A-009265 in the town/wetland batch and await its focused publication. The earlier coastal/town release excludes this later investigation. Historical pre-registration proof receipts remain unchanged; current registration and working state supersede their catalog-registration flags. Complete nonrepetition, physical grain/detail, wetness/contact, cost/pressure/lifecycle and public admission remain open. Keep the original, this single current diagnostic study, exact hashes, compact findings and bounded decisive witnesses.

Evidence: generation provenance generation.json, final prompt final-prompt.txt, visual findings visual-review.json, validation validation-summary.json, normal-world frames world-study/world-results.json, terminal resources world-study/owned-resources-terminal.json, storage sharing storage-sharing.json.