Current Checkpoint

Authority

Product and player contract: docs/01_GAME_DESIGN.md
Gameplay, balance, failure and Save authority: docs/02_GAMEPLAY_AND_BALANCE.md
Page plan and complete state map: docs/03_PAGES_AND_UX.md
Visual, animation and audio authority: docs/04_ART_ANIMATION_AUDIO.md
Complete content authority: docs/05_CONTENT_CATALOG.md
Program Goal, active repair contract, implementation plan and ledger: docs/06_DEVELOPMENT_AND_REVIEW.md

Current State

Stage: DEV-3_GUILD_FIRST_MARKET_IN_PROGRESS
Active implementation step: DEV-3A_CONTRACT_FOUNDATION
Planning baseline: d59555c5a3cf95fc44de79ba06c02e650eb3df3b
Documentation consolidation baseline: 53824d2394e7facdc8dca19aa20a690a5c960aca
Current review set: exactly six validated Markdown files under docs/.
Rejected pixel Gallery item: 5vl2EqsYWkey
Latest functional evidence: combined RG-03/RG-04 midpoint systems, Save-v11, five-profile simulation, runtime regression and Web transport passed.
Human visual evidence: FAIL; the Owner rejected the pixel-art presentation as a contradiction of the accepted Japanese 2D hand-drawn direction.
Superseded DEV-1 functional source: 43e5bacbf88953ae99676400532189f02143be20.
DEV-1 functional evidence: source-bound entry suite and complete preserved FG-4 midpoint regression both PASS. On 2026-08-17 the Owner rejected its generic translucent entry chrome as substantially different from the accepted Japanese 2D fantasy UI reference.
DEV-1F local evidence: six focused DEV-1 headless checks, real parsed touch and 20 normal-render development frames pass after the Title gained enamel, ivory, gold-trimmed, heraldic and icon-forward runtime chrome. This is not a renewed candidate or human visual pass.
DEV-1G local evidence: the same six checks, parsed touch and final 20-frame development matrix pass after Boot, first launch, Settings and Credits adopt the same chrome. The first compressed setup-grid attempt was rejected and repaired before this pass.
Revised DEV-1 candidate source: d3b93268776b87b7bfb407c79165011facc438f8.
Revised source-bound DEV-1 suite and complete preserved FG-4 regression PASS with strict diagnostics 0. Source-bound entry frames reach agent_visual_reviewed; only the Owner can grant human visual acceptance.
Latest Owner product decision, 2026-08-17: New Game's choose-two Recruits form the persistent initial Expedition Company; Contract runs are bounded expeditions inside continuous incremental growth, do not discard the roster, and zero-HP Recruits return for the next stage instead of dying permanently.
DEV-2A authority correction is complete: all six current documents place that rule above frozen FG-4 lifecycle evidence. The exact next-stage HP amount remains a later balance decision; DEV-2 guarantees return and persistence but does not invent a full-heal value.
DEV-2B implementation is locally complete: one saved seed creates exactly four complete deterministic Recruit offers in fixed positions; zero-to-two selected IDs survive restart; duplicate, unknown, third-selection, corrupt replica and interrupted-candidate boundaries are explicit; no Company or Contract state exists before commit.
DEV-2C implementation is locally complete: Title reaches the real four-offer page; Inspect/Compare preserve selection; exact 0/2..2/2 gating, visible third-selection rejection, Back/restart restoration and parsed touch pass. Sixteen normal-render development frames cover required sizes, languages and 100/130% text; this is not source-bound candidate or human acceptance.
DEV-2D implementation is locally complete: Review names the selected two and unchosen two; Revise preserves the draft; Create Company atomically writes exactly the selected persistent Recruits with a deterministic receipt, then reaches the bounded Guild handoff. Duplicate commit, candidate recovery, lingering-draft cleanup, restart/Continue, Guild Settings return and visible disabled-action reasons pass. Twenty-four normal-render development frames cover all six DEV-2 page states at the required layout matrix; this remains dirty-tree guidance, not source-bound candidate or human acceptance.
DEV-2D implementation source: 71f49a4926499ec8aa4911f9d162df709ecf132f; pushed, clean and equal to upstream before DEV-2E activation.
DEV-2 candidate source: d1990538d5d70059f1203bf4817d1b0e9bf49a74.
Source-bound DEV-2 candidate suite and complete preserved FG-4 regression PASS with strict diagnostics 0. The candidate covers the complete parsed Boot/setup/Title/choose-two/Review/Revise/restart/Create/Guild path plus 24 normal-render Initial Company frames. Actual-size source-bound frames reach agent_visual_reviewed; only the Owner can grant human visual/comprehension acceptance.
Owner visual review, 2026-08-17: FAIL_CHARACTER_ABSENCE. The candidate is functionally correct but still reads as repeated cards and text. Reopened DEV-2 must make generated Recruits the dominant visual choice through real person-bearing controls, larger Inspect/Compare/Review portraits and the persistent pair standing in Guild; decorative text panels cannot substitute.
DEV-2F implementation is locally complete: all 12 base Professions now have one-to-one non-pixel Recruit stickers; four person-bearing candidate controls, portrait-led Inspect/Compare/Review, the persistent Guild pair and illustrated Contract Board/Market props render over a visible shared environment rather than opaque repeated cards. The eight missing stickers used exactly two 2x2 gpt-image-2 calls costing $0.028; four accepted sources were reused.
Controlled worktree evidence passes all six applicable DEV-1 and all six DEV-2 behavior/parsed-touch checks with strict diagnostics 0 at /private/tmp/aetherbound-guild-test-evidence/dev2-character-worktree-suite/20260817T130943Z-23198. Twenty-four normal-render frames pass the required 760x360, 844x390 and 1280x720, zh_CN/en and 100/130% matrix at /private/tmp/aetherbound-guild-test-evidence/dev2-character-captures/20260817T130637Z-20886. The six representative runtime frames were sent to the Owner; human visual acceptance remains open.
Source-bound DEV-2F implementation commit 9b8a533d86e041c4543e94a10ce42fd7438e9077 passes the complete DEV-2 candidate suite at /private/tmp/aetherbound-guild-test-evidence/dev2-character-candidate/20260817T132045Z-34538: six DEV-1 checks, six DEV-2 checks, parsed input and 24 normal-render frames, with strict diagnostics 0. This does not override the open human gate.
Owner follow-up, 2026-08-17: FAIL_NOT_YET_ALIGNED. The first sticker pass improves subject presence but the four tall portrait-like actors and vertical choice slots still do not read as the accepted compact Japanese 2D unit/HUD system. A worktree graybox now removes the four full-height plates and places all candidates in one visible painted scene with HUD strips and selection rings; 24-frame evidence passes at /private/tmp/aetherbound-guild-test-evidence/dev2-open-stage-captures/20260817T135149Z-71812. Actor proportion/style alignment remains open. The Owner then explicitly required one consistent actor style and questioned the current size; DEV-2F is authorized to replace all 12 mixed sources with one unified compact unit sheet and reduce choice-page display scale while preserving larger Inspect.
The first unified 4x3 sheet was internally consistent but still failed the accepted ordinary-unit boundary with 6.5-7-head anatomy and key-art detail. The final authorized repair call produced all 12 together at compact 4.5-5-head tactical proportions with one face, line, cel-shadow, material and light contract. All 12 now use one-to-one runtime textures; choice actors are reduced, Guild actors are further reduced, and Inspect/Compare stay larger. Focused behavior PASS: /private/tmp/aetherbound-guild-test-evidence/dev2-unified-actor-choice/20260817T144520Z-42716. Twenty-four normal-render frames PASS: /private/tmp/aetherbound-guild-test-evidence/dev2-unified-actor-captures/20260817T144542Z-42958. Agent review finds complete silhouettes, coherent style and usable scale at 760x360, 844x390 and 1280x720; Owner acceptance remains open.
The first source-bound suite from 5700a84 passed all 12 DEV-1/DEV-2 behavior checks, then its single normal-render process accumulated resources after 18 frames and timed out before the 1280x720 fixture. Runner cleanup completed at /private/tmp/aetherbound-guild-test-evidence/dev2-unified-actor-candidate/20260817T145351Z-58238; this is a harness timeout, not a product PASS or visual failure. The isolated 1280x720 fixture passes six states in 3.7 seconds at /private/tmp/aetherbound-guild-test-evidence/dev2-split-capture-fixture-3/20260817T151341Z-77767. Candidate capture now owns one normal-render process per fixture while retaining the same 24-frame matrix and strict diagnostics.
Owner follow-up rejects the current DEV-2 button and outer-frame treatment: repeated double-gold cut-corner boxes, generic diamond medallions and one continuous enclosing border still read as utility chrome rather than an authored Japanese fantasy command layer. DEV-2F remains active for a narrow procedural-chrome repair; all button behavior and DEV-3 scope stay frozen.
Owner requires every review correction to remain durable and prevent repeat failures. The active plan now has a DEV-2F recurrence checklist covering character priority, unified actor style/scale, open framing, differentiated semantic commands, focused visual iteration and explicit human-gate status.
Owner also rejects repeated opaque long-running checks during visual review. DEV-2F now limits each visual iteration to the focused current-page behavior check and six representative frames. Any check expected to exceed 60 seconds must be named with purpose and duration before launch; full DEV-2/FG-4 regression waits for visual-direction acceptance.
DEV-2F command-chrome candidate v2 is implemented in real Godot UI: Initial Company pages now use open corner ornaments plus header/footer rails; parchment return tabs, compact A/B markers, plus/check selection commands, compare arrows and confirmation checks replace generic diamond medallions. Guild footer and disabled destinations now use the same lighter hierarchy. Focused behavior passes in 1.6 seconds at /private/tmp/aetherbound-guild-test-evidence/dev2-command-chrome-choice/20260817T153507Z-918; six 844x390 zh_CN 130% frames pass in 2.4 seconds at /private/tmp/aetherbound-guild-test-evidence/dev2-command-chrome-review-v2/20260817T153537Z-1291. Human visual acceptance remains open. Owner follow-up rejects v2's content spacing and frame finish: content and ornament share one rectangle, so titles, state copy, actors and footer controls sit too close to the edge; sparse corner strokes read as unfinished rather than intentionally open. DEV-2F remains active for a separate outer-frame/content-inset repair.
DEV-2F frame/spacing candidate v3 now renders frame and content in separate safe rectangles. One complete cut-corner hairline, four constructed corner fittings and aligned header/footer rails sit outside a stable inner inset for headings, actors, copy and commands. Focused behavior passes in 1.5 seconds at /private/tmp/aetherbound-guild-test-evidence/dev2-frame-spacing-choice/20260817T154715Z-19139; six 844x390 zh_CN 130% frames pass in 2.4 seconds at /private/tmp/aetherbound-guild-test-evidence/dev2-frame-spacing-review/20260817T154743Z-20697. Human visual acceptance remains open.
Owner accepted DEV-2F frame/spacing candidate v3 for continuation on 2026-08-17. DEV-2 source 4352e253a83ee9f3ce3bca6eac99c5e70691f989 passes the source-bound candidate suite at /private/tmp/aetherbound-guild-test-evidence/dev2-frame-spacing-candidate/20260817T155546Z-29597 and the preserved FG-4 regression at /private/tmp/aetherbound-guild-test-evidence/dev2-final-fg4-regression/20260817T155717Z-30828, both with strict diagnostics 0 and complete runner cleanup. DEV-2 is complete.
DEV-3 is activated under one durable five-substep plan. DEV-3A owns only the authoritative Company-store migration and deterministic first Contract/ Market persistence: v1 Company compatibility, one active Contract, 52 Coin, Line capacity four, exact 3/4/3 offers and retry-safe receipts. Guild UI, Market review and hiring remain later DEV-3 substeps.
Broader human comprehension/fun, device, package and release gates remain OPEN.

Preserve

Verified DEV-1 runtime, behavior and source-bound evidence.
Existing FG-4 runtime and Save-v11 evidence as historical implementation until their newly superseded casualty/Contract-close semantics reach their named later conversion step; do not delete or rewrite them inside DEV-2.
Touch, localization, accessibility and audio-event routing.
Rejected pixel assets and H5 only as Git/evidence history.

Frozen

Runtime changes outside active DEV-3A.
Pixel asset expansion.
FG-5 and later Regions.
New image generation during planning review.
Native packaging, devices, signing, stores and release.
Multi-Agent development.

Next Action

Implement and narrowly verify DEV-3A: migrate the verified Company store into one authoritative Contract/Market state without changing its two Recruit records, create exactly one retry-safe first Contract with 52 Coin, Line capacity four and a stable 3/4/3 board, then push the focused boundary before activating Guild UI work.

14 KiB Raw Blame History