settled-reach

Files

T

jpmschweitzerandClaude Fable 5 cf6ef2bc31 feat(simulation): district-window zoom-ladder renderer — aliveness_probe --render/--window/--anchor (T-1123)

Renders true-2km district maps (derive_district per cell, NOT the coarse eager grid) as PNGs: morphology (17-zone palette transliterated from atlas_marker_overlay.gd, sync-commented), elev/temp/moisture/vegetation panels, anchor crosshair, km-honest filenames + manifest. believability.rs gains cascade_snapshot_for_body (cascade_for_body now a thin wrapper); probe re-derives the D-203-freed TerrainAnalysis via run_layer1 only under --render. Determinism proven: 5/5 panels byte-identical across passes AND cross-process (md5s reproduced in a fresh run). ~8us/district debug. First-ever look at the district layer produced the T-1125/T-1126/T-1127 findings (D-227 seam ends at ~500km; ocean-blind vegetation; unrendered glaciation).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-07-16 22:26:57 +02:00

content

refactor(simulation): strip archetype trace + HeritageRoot per cascade (#877 , #878 )

2026-04-22 08:55:48 +02:00

data

fix(db): M1 — re-tag 5 steel-complex/shipyard corps to Standalone-shaped values; regen (PR #178 review)

2026-07-16 17:03:54 +02:00

examples

feat(simulation): cache TerrainAnalysis + derive basin_direction from the D8 thalweg (T-1044, T-1047)

2026-06-17 09:46:42 +02:00

sr-voice

fix(engine): offload Gemma to GPU by setting n_gpu_layers=999 (#833 )