SeymourBETA

Update 2.64

6/10/2026

v2.64.0BETA

Seymour

Aeon growth/customization in the Spira Data Atlas (read-only, `proved-candidate`, Aeon-only, compiled)

Plugs sum_grow.bin (FFX's Aeon customization table) into the provider as a read-only detail kind: new SpiraDataAtlasDetailKind.AeonGrowthRecipe + value-guarded accessor SpiraDataAtlasCatalog.TryGetAeonGrowthRecipe(entryIndex, currentResultRaw, currentItemRaw). Aeon-ONLY: sum_grow.bin is 636 bytes (0x14 header + 77 × 8B entries, Target always 0x007F=Aeon, jppc==inpc byte-identical) → 67 ability recipes (Result cat 3 → Commands) + 10 stat recipes (Result cat 0 → StatOptions, Secondary=1). No PC rows and no Sphere-Grid rows (PcAeonFineStat is a separate corpus; grid topology lives in sphere.bin/panel.bin). Non-work/ source: the 77 rows come from a byte-grounded compiled table (FfxLib/SpiraDataAtlas/SumGrowAeonRecipeData.cs, generated by build_sum_grow_aeon_fine_index.ps1 -EmitCSharp from sum_grow.bin), so they exist in ANY build without work/ (unlike Mix/Sphere/Shop). Names/labels are resolved at runtime by the editor's own AeonCustomizationEntry+dictionaries (single source of truth, RT0-proved by --customization-rt0), never re-invented. Stat-recipe cost = partial (the editor renders Primary× item; in-game cost is 1 item/application and Primary is the increment) — surfaced in the evidence column (;cost-semantics-partial), not just prose. Gate --spiraatlas-rt0 21219 → 21296 details (+77 Aeon; all unique) + 13 new asserts (count 77/67/10; Aeon-only; no PC; no Sphere; ability lookup 66→Ultima/Supreme Gem; stat lookup 67→HP+100/Power Sphere/cost-partial; value-guard 0x9999 hides; out-of-range index miss; domain search; BIBLE domain aeon-growth). Read-only: no writer/UI/runtime/SIN; does not touch RIKKU/WAKKA/JECHT/BAHAMUT/AURORA, ANIMA's Common badge engine, or WriteAeon/save paths. Unblocks YUNA backlog Q1 (was blocked). Verified by a 4-lens adversarial workflow in the scout. Build 0 errors / 361 warnings; --customization-rt0 + --spiraatlas-rt0 + --aicmdmeta-rt0 PASS. Docs: docs/ai/ATLAS_BRASKA_SUM_GROW_FINE_SCOUT_RESULT_2026-06-10.md + docs/ai/HANDOFF_BRASKA_SUM_GROW_FINE_INDEXER_NEXT_2026-06-10.md

v2.63.0

v2.64.1BETAPATCH

Seymour

Atlas evidence badge in the Mix Table Editor (read-only, value-guarded, hide-on-edit)

Read-only UI plugging the SpiraDataAtlasCatalog.TryGetMixCombination accessor (shipped in v2.63.0) into the MixTableEditor: the "Combination Editor" card (detail panel) now shows the selected mix combination's <common:AtlasEvidenceBadgeStrip> — one strip per combination (not per row; avoids 6,328 lookups and clutter), an exact mirror of Treasure/Shop. MixTableEditor_DataModel gained [ObservableProperty] AtlasEvidenceInfo? SelectedMixEvidence + RefreshSelectedMixEvidence() (TryGetMixCombination(origin.Index, result.PartnerIndex, result.RawResult) → AtlasEvidenceInfo.ForDetail), called in OnSelectedResultChanged (covers combination AND origin switch via cascade) and in ResultChanged when RawResult changes (covers editing the outcome via ResultRef→SyncBack→NotifyComputedChanged and "Set Empty"). Hide-on-edit: re-pointing the result to another item, "Set Empty" (raw 0), an empty/upper-triangle mirror cell, or a pair outside the corpus → accessor false/null → the strip self-hides (IsVisible=false); (origin,partner) order irrelevant (max/min canonicalization in the accessor). REUSE of the v2.63.0 capability + ANIMA's Common infra (consumed only) → (precedent: Shop badge UIs v2.61.2/v2.61.5 were PATCH). No new writer: --mixtable-rt0 stays byte-identical (25108/25108). Does not touch ANIMA's Common engine, the provider, Treasure/Shop (RIKKU), or BRASKA/WAKKA/AURORA/BAHAMUT/SIN. Reviewed by a 5-lens adversarial workflow (hide-on-edit, writer/read-only, ANIMA-boundary, XAML-binding, honesty) + synthesis → ship, 0 defects. Build 0 errors / 361 warnings; --spiraatlas-rt0 (Mix 6328) + --mixtable-rt0 (byte-identical) + --aicmdmeta-rt0 PASS. Doc: docs/ai/ATLAS_JECHT_MIX_BADGE_UI_RESULT_2026-06-10.md

v2.64.0

v2.64.2BETA

Seymour

Blitzball save writer LAB + prize offsets SAVE-PROVED against a real save (no tournament)

The Blitzball prize offsets went from RE to save-proved: I located the real saves (Documents\SQUARE ENIX\...\FINAL FANTASY X\ffx_000..006, 7 slots of 0x6900 = 0x40 header + 0x68C0 SaveData) and ffx_000 decoded correctly (league = Megalixir/Mega-Potion/Speed Sphere, tournament = Elixir/Super Goalie/Potion, etc.) — the save is plaintext, no encryption/compression (gil/story/battle_count sane), --base 0x40 confirmed. The 0x40 header has a "cxs" signature + play-time/location; no field matches a SaveData checksum (likely no blocking checksum, but one in-game load is still needed). New LAB writer Tools/BlitzballSaveWrite.cs (--blitz-save-write <in> <field> <value> <out>): writes one prize-index u16 to a NEW file (NEVER overwrites the source; refuses in-place overwrite), with before/after resolved via the Atlas catalog. Validated on a COPY of ffx_006: tournament0 25 (Elixir) -> 50 (Saturn Crest), diff = exactly 2 bytes, the rest byte-identical. BlitzballSaveRead.Resolve became internal ResolvePrize (reused by the writer). save (ffx_NNN) ≠ game file: editing the save changes THIS playthrough; changing the pool (takara.bin 220..320) is Treasure/RIKKU. No Atlas UI/Shop/Treasure/Mix/Common/Aurora/Bahamut/SIN changes. Build 0 errors / 361 warnings. Doc: docs/ai/BLITZBALL_SAVE_RUNTIME_WRITER_SCOUT_2026-06-09.md

v2.64.1

v2.64.3BETA

Seymour

Aeon Growth evidence badge in the Customization Editor (read-only, value-guarded, hide-on-edit)

Plugs <common:AtlasEvidenceBadgeStrip> into the "Aeon Grow / Teach" tab of CustomizationEditor ("Recipe Editor" card), consuming the already-built accessor SpiraDataAtlasCatalog.TryGetAeonGrowthRecipe(entryIndex, currentResultRaw, currentItemRaw) (v2.64.0). New AtlasEvidenceInfo? SelectedAeonEvidence in CustomizationEditor_DataModel, recomputed on selection change and when the selected row raises RawResult (AeonRecipeChanged) → hide-on-edit: editing the taught ability/stat OR the cost item (raw mismatch) or an out-of-range index makes the accessor return null and the strip hides (no stale badge), an exact mirror of Mix (v2.64.1). The 10 stat recipes show a partial badge (cost-semantics: the editor renders Primary× item, but in-game cost is 1 item/application). No writer or cost-display fix (out of scope). Reuses ANIMA's Common engine (consumed only). PC stays blocked (no PC rows) and Sphere metadata-only (no join) — nothing to show. No provider/gate change (--spiraatlas-rt0 intact, 21296 details / 77 Aeon). Build 0 errors / 361 warnings; --customization-rt0 + --spiraatlas-rt0 + --aicmdmeta-rt0 PASS. Does not touch JECHT/Mix, TIDUS/PC, AURON/Sphere, RIKKU/WAKKA/VALEFOR/AURORA/SIN. Doc: docs/ai/ATLAS_BRASKA_AEON_GROWTH_BADGE_UI_RESULT_2026-06-10.md

v2.64.2

Back to updates

SeymourBETA

Update 2.64

6/10/2026

v2.64.0BETA

Seymour

Aeon growth/customization in the Spira Data Atlas (read-only, `proved-candidate`, Aeon-only, compiled)

Plugs sum_grow.bin (FFX's Aeon customization table) into the provider as a read-only detail kind: new SpiraDataAtlasDetailKind.AeonGrowthRecipe + value-guarded accessor SpiraDataAtlasCatalog.TryGetAeonGrowthRecipe(entryIndex, currentResultRaw, currentItemRaw). Aeon-ONLY: sum_grow.bin is 636 bytes (0x14 header + 77 × 8B entries, Target always 0x007F=Aeon, jppc==inpc byte-identical) → 67 ability recipes (Result cat 3 → Commands) + 10 stat recipes (Result cat 0 → StatOptions, Secondary=1). No PC rows and no Sphere-Grid rows (PcAeonFineStat is a separate corpus; grid topology lives in sphere.bin/panel.bin). Non-work/ source: the 77 rows come from a byte-grounded compiled table (FfxLib/SpiraDataAtlas/SumGrowAeonRecipeData.cs, generated by build_sum_grow_aeon_fine_index.ps1 -EmitCSharp from sum_grow.bin), so they exist in ANY build without work/ (unlike Mix/Sphere/Shop). Names/labels are resolved at runtime by the editor's own AeonCustomizationEntry+dictionaries (single source of truth, RT0-proved by --customization-rt0), never re-invented. Stat-recipe cost = partial (the editor renders Primary× item; in-game cost is 1 item/application and Primary is the increment) — surfaced in the evidence column (;cost-semantics-partial), not just prose. Gate --spiraatlas-rt0 21219 → 21296 details (+77 Aeon; all unique) + 13 new asserts (count 77/67/10; Aeon-only; no PC; no Sphere; ability lookup 66→Ultima/Supreme Gem; stat lookup 67→HP+100/Power Sphere/cost-partial; value-guard 0x9999 hides; out-of-range index miss; domain search; BIBLE domain aeon-growth). Read-only: no writer/UI/runtime/SIN; does not touch RIKKU/WAKKA/JECHT/BAHAMUT/AURORA, ANIMA's Common badge engine, or WriteAeon/save paths. Unblocks YUNA backlog Q1 (was blocked). Verified by a 4-lens adversarial workflow in the scout. Build 0 errors / 361 warnings; --customization-rt0 + --spiraatlas-rt0 + --aicmdmeta-rt0 PASS. Docs: docs/ai/ATLAS_BRASKA_SUM_GROW_FINE_SCOUT_RESULT_2026-06-10.md + docs/ai/HANDOFF_BRASKA_SUM_GROW_FINE_INDEXER_NEXT_2026-06-10.md

v2.63.0

v2.64.1BETAPATCH

Seymour

Atlas evidence badge in the Mix Table Editor (read-only, value-guarded, hide-on-edit)

Read-only UI plugging the SpiraDataAtlasCatalog.TryGetMixCombination accessor (shipped in v2.63.0) into the MixTableEditor: the "Combination Editor" card (detail panel) now shows the selected mix combination's <common:AtlasEvidenceBadgeStrip> — one strip per combination (not per row; avoids 6,328 lookups and clutter), an exact mirror of Treasure/Shop. MixTableEditor_DataModel gained [ObservableProperty] AtlasEvidenceInfo? SelectedMixEvidence + RefreshSelectedMixEvidence() (TryGetMixCombination(origin.Index, result.PartnerIndex, result.RawResult) → AtlasEvidenceInfo.ForDetail), called in OnSelectedResultChanged (covers combination AND origin switch via cascade) and in ResultChanged when RawResult changes (covers editing the outcome via ResultRef→SyncBack→NotifyComputedChanged and "Set Empty"). Hide-on-edit: re-pointing the result to another item, "Set Empty" (raw 0), an empty/upper-triangle mirror cell, or a pair outside the corpus → accessor false/null → the strip self-hides (IsVisible=false); (origin,partner) order irrelevant (max/min canonicalization in the accessor). REUSE of the v2.63.0 capability + ANIMA's Common infra (consumed only) → (precedent: Shop badge UIs v2.61.2/v2.61.5 were PATCH). No new writer: --mixtable-rt0 stays byte-identical (25108/25108). Does not touch ANIMA's Common engine, the provider, Treasure/Shop (RIKKU), or BRASKA/WAKKA/AURORA/BAHAMUT/SIN. Reviewed by a 5-lens adversarial workflow (hide-on-edit, writer/read-only, ANIMA-boundary, XAML-binding, honesty) + synthesis → ship, 0 defects. Build 0 errors / 361 warnings; --spiraatlas-rt0 (Mix 6328) + --mixtable-rt0 (byte-identical) + --aicmdmeta-rt0 PASS. Doc: docs/ai/ATLAS_JECHT_MIX_BADGE_UI_RESULT_2026-06-10.md

v2.64.0

v2.64.2BETA

Seymour

Blitzball save writer LAB + prize offsets SAVE-PROVED against a real save (no tournament)

The Blitzball prize offsets went from RE to save-proved: I located the real saves (Documents\SQUARE ENIX\...\FINAL FANTASY X\ffx_000..006, 7 slots of 0x6900 = 0x40 header + 0x68C0 SaveData) and ffx_000 decoded correctly (league = Megalixir/Mega-Potion/Speed Sphere, tournament = Elixir/Super Goalie/Potion, etc.) — the save is plaintext, no encryption/compression (gil/story/battle_count sane), --base 0x40 confirmed. The 0x40 header has a "cxs" signature + play-time/location; no field matches a SaveData checksum (likely no blocking checksum, but one in-game load is still needed). New LAB writer Tools/BlitzballSaveWrite.cs (--blitz-save-write <in> <field> <value> <out>): writes one prize-index u16 to a NEW file (NEVER overwrites the source; refuses in-place overwrite), with before/after resolved via the Atlas catalog. Validated on a COPY of ffx_006: tournament0 25 (Elixir) -> 50 (Saturn Crest), diff = exactly 2 bytes, the rest byte-identical. BlitzballSaveRead.Resolve became internal ResolvePrize (reused by the writer). save (ffx_NNN) ≠ game file: editing the save changes THIS playthrough; changing the pool (takara.bin 220..320) is Treasure/RIKKU. No Atlas UI/Shop/Treasure/Mix/Common/Aurora/Bahamut/SIN changes. Build 0 errors / 361 warnings. Doc: docs/ai/BLITZBALL_SAVE_RUNTIME_WRITER_SCOUT_2026-06-09.md

v2.64.1

v2.64.3BETA

Seymour

Aeon Growth evidence badge in the Customization Editor (read-only, value-guarded, hide-on-edit)

Plugs <common:AtlasEvidenceBadgeStrip> into the "Aeon Grow / Teach" tab of CustomizationEditor ("Recipe Editor" card), consuming the already-built accessor SpiraDataAtlasCatalog.TryGetAeonGrowthRecipe(entryIndex, currentResultRaw, currentItemRaw) (v2.64.0). New AtlasEvidenceInfo? SelectedAeonEvidence in CustomizationEditor_DataModel, recomputed on selection change and when the selected row raises RawResult (AeonRecipeChanged) → hide-on-edit: editing the taught ability/stat OR the cost item (raw mismatch) or an out-of-range index makes the accessor return null and the strip hides (no stale badge), an exact mirror of Mix (v2.64.1). The 10 stat recipes show a partial badge (cost-semantics: the editor renders Primary× item, but in-game cost is 1 item/application). No writer or cost-display fix (out of scope). Reuses ANIMA's Common engine (consumed only). PC stays blocked (no PC rows) and Sphere metadata-only (no join) — nothing to show. No provider/gate change (--spiraatlas-rt0 intact, 21296 details / 77 Aeon). Build 0 errors / 361 warnings; --customization-rt0 + --spiraatlas-rt0 + --aicmdmeta-rt0 PASS. Does not touch JECHT/Mix, TIDUS/PC, AURON/Sphere, RIKKU/WAKKA/VALEFOR/AURORA/SIN. Doc: docs/ai/ATLAS_BRASKA_AEON_GROWTH_BADGE_UI_RESULT_2026-06-10.md

v2.64.2