WORLD OF GONZ WIKIOpen game ↗

Engineering, operations & design history

Native lake study: actual context findings

Full repository source: docs/art-direction/display-cache-history-20261104/lake-windpatch-native-20261104/context-review.md

Technical documentation includes historical decisions, proposals and verification reports. For playing instructions and practical rules, open Gameplay.

Native lake study: actual context findings

The selected native wind-patch artwork remains a study. Its asymmetric arrangement preserves fine painted marks, but direct comparison in the actual client shows similar bright diagonal bands in Region on desktop and phone. The pattern gate fails; no original is replaced and no source is admitted. The current study remains useful input for a source-specific sampler investigation.

The two retained desktop context/lake-before-after-1440.png and phone context/lake-before-after-390.png boards contain original and study Region/Hex crops at native screenshot pixel size, without resizing. All twelve frames compare three exact native sources at unchanged canonical lake cell 262807 and its fields shoreline, using normal client controls and explicit private response substitutions. The study preserves the observed Region contact contour without a new visible gap. Hex crops show the water interior and cannot qualify Hex shore contact. Palette fit, recurrence, other biomes and shores, relevant state, lifecycle and performance remain open.

Five complete GPU output comparisons preserve 12,793,743 pixels relative to the existing native algorithm, including nonsquare source aspect and fractional extent; cancelled staging publishes nothing. This is technical parity, not proof of visual quality. All 9,210 catalogued native files (14,656,936,570 bytes) match recorded hashes in a full streaming check. Originals, selected finals and both Citadel hamlets are retained.

The first two rejected generation intermediates were removed after selected-output and original preservation checks. The exact retirement receipt records 3,729,573 reclaimed bytes. Only two bounded final context boards are retained; raw pixels and complete map screenshots were compared in memory. No production save, geography, source permission or public renderer changed.

See the source generation and final prompt generation.json, initial native study, source-bound visual findings context-visual-review.json, actual context results context/results.json, GPU parity pixel-validation.json, complete native preservation catalog-preservation.json and intermediate retirement intermediate-retirement.json.