--- title: "Implementation plan: answers (World 14)" date: 2026-07-04 status: executed tags: [world, walker-narrative, grief-arc, bargaining, depression, destiny-entity, keyword-cost-modifiers, concede, plan] modules: [core-effects, data-worlds] related: - .lore/work/specs/answers.md - .lore/work/design/answers-card-design.md - .lore/work/plans/questions.md - .lore/work/specs/the-beginning.md - .lore/reference/worlds/authoring/theme-authoring.md - .lore/reference/worlds/catalog/endworlds-trilogy-concept.md --- # Implementation plan: answers (World 14) Source spec: `.lore/work/specs/answers.md` (REQ-W14-1..29). Concrete card templates, costs, and deck counts are authored in `.lore/work/design/answers-card-design.md` (produced during this plan-prep pass, mirroring `questions-card-design.md`'s role for World 13) — copy its JSON directly rather than re-deriving numbers during implementation. **Ground truth confirmed before writing this plan (all verified directly against code, not assumed from spec prose):** - World 13 (`questions`) is fully implemented and registered (`src/data/worlds/questions/`, `src/data/worlds/registry.ts`). `answers` is a clean slate — no `src/data/worlds/answers/` directory, no `Bargaining`/`Depression` anywhere in `src/`. - `ClearCostPerOtherKeyword` exists as a full `PersistentModifier` kind with working switch-arms in `src/core/engine/effectiveCards.ts` (both `activeKeywordCostModifiers` and `extraWorldCardCost`) and a render arm in `src/game/view/CardView.ts:539`, but **zero** `KEYWORD_COST_MODIFIERS` entries use it today — Bargaining will be its first real activation, exactly as the spec claims. - `RemoveKeyword` only ever strips `card.appliedKeywords`, never authored/static `keywords` — confirmed in `src/core/effects/appliedKeywords.ts`'s own code comment. Decay (`tickAppliedKeywordsAtTurnStart`) is likewise `appliedKeywords`-only. - `.lore/reference/worlds/catalog/endworlds-trilogy-concept.md` is **already current** (rewritten during World 13's plan execution, Step 21) — REQ-W14-26's "shared blocker" is already resolved; no action needed here beyond a small addendum (Slice 5) for the Destiny-reuse decision below. - World 15 (`the-beginning`, spec-only, not yet planned) hard-depends on this world's Act III palette (its own REQ-W15-23: World 15 Act II reprises World 14's Act III palette) and confirms signature verb `concede` doesn't collide with World 15's `unburden`. - **Update (2026-07-04, post plan-prep): the user committed `5b2e0e5` ("Updated destiny.") directly against `master`, which changes two of this plan's premises before implementation started.** First, `WORLD_THREAT_BY_WORLD_ID`'s missing `"questions"` entry — the pre-existing bug this plan originally set out to fix for both worlds — is now already fixed upstream (`questions: "Destiny"` is live in `src/core/effects/gainCard.ts`). Slice 1 Step 3 now only needs to add `"answers"`. Second, and much more significant: the same commit rewrote the shared `Destiny` card template itself. See the fully-rewritten deviation note directly below — every cost-formula and behavior claim elsewhere in this plan that references the old `Denial:2` / `ExileTopWorldCards` version of `Destiny` is stale and has been corrected in place. **Deviation from the ratified spec, confirmed with the user during plan-prep — read `answers-card-design.md`'s "Decision superseding REQ-W14-12/13" section in full before touching Slice 2's Act 3 work. Then read this paragraph, not that section's numbers, for what `Destiny` actually does today.** Summary: rather than authoring a second, differently-keyed `Destiny`-flavored card template (blocked by global template-id uniqueness — `"Destiny"` is already claimed by World 13's card), this world **reuses World 13's existing `Destiny` template unmodified** in its own Act 3 deck composition. The template-id stays `"Destiny"`, but as of commit `5b2e0e5` ("Updated destiny.", 2026-07-04) its authored content is no longer what REQ-W14-12/13 or the design doc's superseding-decision section describe. Ground truth, read directly from `src/data/allCards.json`: | Field | Old (what the design doc/spec assume) | Current (as of `5b2e0e5`) | |---|---|---| | `name` | `"Destiny"` | `"The Destiny"` (display only — templateId key is still `"Destiny"`, so nothing else in this plan that references it by id needs to change) | | `cost` | `6` | `15` | | `keywords` | `["Denial:2"]` | `[]` — **no authored keyword at all** | | `onCleared` | `ExileTopWorldCards{amount:3}` | `SurviveWorld` | | `onEndOfTurn` | `None` | `Sequence[Damage{amount:2}, ExileTopWorldCards{amount:1}]` | Consequences that ripple through the rest of this plan: - **`onCleared: SurviveWorld` is the same effect `Door` uses** (`src/core/effects/worldCards.ts`'s `SurviveWorldHandler` sets `state.status = "won"` unconditionally, ending the entire run in victory). Clearing `Destiny` is no longer a hazard-clear that exiles future world cards — it is a second, more expensive win condition. This is a materially different mechanical and narrative role than either the spec's original intent ("quietly picked it up and kept walking") or this plan's original superseding-decision framing ("cut off future moments"). Neither reading applies cleanly anymore; this needs fresh narrative framing, not a shrug that the old ambiguity still covers it. Flagged again, more urgently, in "Open items carried forward" below. - **`onEndOfTurn` is now an active, escalating punishment** (2 damage + exile 1 top world card, every turn `Destiny` sits undealt), not inert. This matters directly for Step 6's `A Deal Too Easy` beat (see that step's corrected description) — an early, uncleared copy seeded into Act 1 is no longer a passive Chekhov's-gun, it's actively costing the player resources every turn until they discard or clear it. - **`Destiny` carries no keyword at all**, so there is no `ClearCostPerSelfKeyword` contribution from it whatsoever (not `Denial`, not `Depression`) — that entire piece of the original cost-formula discussion no longer applies. The only remaining tax vector is `Bargaining`'s `ClearCostPerOtherKeyword`, which (confirmed directly in `src/core/engine/effectiveCards.ts`) applies to *any* priced card based on `Bargaining` value summed across *other* cards in hand, regardless of whether the priced card itself carries any keyword. So `Destiny`'s corrected effective-cost formula for validation is simply: **`15 + costPer(Bargaining) × sum(Bargaining values on other cards in hand)`** — see Step 22.5's corrected assertion. - `WORLD_THREAT_BY_WORLD_ID["answers"]` still needs to be added (Step 3) and will still resolve to the same `"Destiny"` template, still letting an Act 1 card seed an early, uncleared copy of it (still realizes REQ-W14-5's "glimpsed in Act I, carried into Act III" beat structurally) — but see Step 6 for why the stakes of that beat changed.
Legend: steps in the same slice can mostly proceed in file order; a step marked (depends on N) must wait for step N's edit to land first. Slices execute in order 1→5 — Slice 2's card authoring uses the Bargaining/Depression keyword strings and the WORLD_THREAT_BY_WORLD_ID fix Slice 1 adds, so Slice 1 must fully land before Slice 2 begins.
## Slice 1 — Core-engine keyword slice (REQ-W14-7..11, REQ-W14-28, REQ-W14-29.3/4)
Step 1. Add "Bargaining" and "Depression" to the KeywordName union in src/core/model/types.ts (currently 10 members ending "Anger").
Step 2 (depends on 1). In src/core/model/keywords.ts:
Step 3. The "questions" entry in src/core/effects/gainCard.ts's WORLD_THREAT_BY_WORLD_ID map was already added by the user directly (commit 5b2e0e5), ahead of this plan's implementation. Only add the new entry: 
"answers": "Destiny",
Both worlds' Destiny card is the same template (see the superseding-decision note above), so both entries point at the identical string. No separate PR callout needed for the "questions" half — that fix already shipped outside this plan.
Step 4 (depends on 2, 3). No change needed in src/core/engine/effectiveCards.ts — extraWorldCardCost/effectiveWorldCardCost already iterate KEYWORD_COST_MODIFIERS generically, including the ClearCostPerOtherKeyword arm. Confirm this by reading the function before writing tests, don't assume from the spec text alone.
Step 5 (depends on 2, 4). New unit tests in src/core/tests/effectiveCards.test.ts, in a new describe("effectiveWorldCardCost — Bargaining/Depression (World 14 Slice 1)", ...) block mirroring the existing Denial/Anger block:
## Slice 2 — World content authoring (REQ-W14-4..6, 12..20) All templates, exact JSON, and deck counts are in `.lore/work/design/answers-card-design.md` — copy directly. That document also resolves the authored-vs-applied split (REQ-W14-8a: every Bargaining/Depression-bearing card except `Destiny` and its self-transform escalation partner `It Calcified` gains its keyword via `ApplyKeyword{target:"self"}` on `onDraw`, staying strippable) and the Destiny-reuse deviation (see this plan's header note).
Step 6 — Act 1 (depends on Slice 1; REQ-W14-4, 15, 16). Author Act 1 card templates into src/data/allCards.json's cardTemplates map: The Ledger Never Closes (signature Bargaining threat), A Broker Who Owes Nothing (Obstructed tool-fetch, grants Call In The Favor), What Would You Give Up (the required Modal genuine branch: destroy a hand card or take 2 damage), The Archive Has a Price (Obstructed hazard, onCleared: AddPlayerCardToTop queuing the lore card A Page From the Ledger), A Deal Too Easy (onCleared: AddThreatToWorldDeck, now resolving to Destiny per Slice 1's fix — this seeds an early copy at the front of Act 1's own draw pile, which is the intended mechanism for REQ-W14-5's "glimpsed in Act I, carried into Act III" beat. **Corrected per the Destiny redesign in commit `5b2e0e5`:** this early copy is no longer inert. Every turn it sits undealt in Act 1 it fires `onEndOfTurn: Sequence[Damage 2, ExileTopWorldCards 1]`, so leaving it unresolved now has a real, escalating cost rather than being a free Chekhov's-gun. `Destiny` is still `discardable: true`, so the player can shed it without clearing it — likely the only realistic Act 1 outcome, since clearing it costs 15 and now triggers `SurviveWorld` (instant full-run win) rather than the old `ExileTopWorldCards{amount:3}`. An Act-1 player clearing a cost-15 card is extremely unlikely given typical early resources, but this is a placeholder-numbers plan and hasn't been sim-checked; flag to the user whether an early, low-probability instant-win path through Act 1 is acceptable or needs an explicit guard. Whether the beat reads as "carried onward" rather than "cleared and forgotten" is still unconfirmed until Step 22.10's manual playtest.), and A Reading of the Ledger (OfferBoon reflavor per REQ-W14-15, drawing from the new pool-ledger-reading). Exact JSON in the design doc's Act 1 section.
Step 7 — Act 2 (REQ-W14-17). Author Act 2 card templates: A Fracture Opens → Another Fracture (Self-Transform Pattern — AddWorldCardToDeck{bTop:true} then DestroySelf, explicitly not ReturnWorldCards) and The Point of No Return (ExileTopWorldCards{amount:2}, foreshadowing Act 3's larger commitment). Neither card carries Bargaining/Depression — Act 2 is deliberately keyword-free per the spec, carrying the beat through accumulating hazard visibility, not tax. Exact JSON in the design doc's Act 2 section.
Step 8 — Act 3 (depends on Slice 1; REQ-W14-6, 12, 13, 18). Author Act 3 card templates: The Weight Doesn't Lift (signature Depression threat — self-tax only, no spreading escalation, per REQ-W14-6's "plainest, least showy shape" framing), It Won't Go Away → It Calcified (Depression's Self-Transform pair, mirroring She Says She's Fine → He Isn't Coming Back exactly — the escalated card authors Depression:2 statically, is discardable:false, deliberately uncounterable), A Reason to Keep Moving (Obstructed tool-fetch, grants Let It Sit), and Just Keep Walking (grants the Brace-as-endurance reward Keep Walking). Do not author a new Destiny template here — it already exists from World 13; reference it by name only in deck composition (Step 9). The fixed { "templateId": "The Walker", "count": 1 } closer is likewise referenced only in deck composition, never redefined (REQ-W14-2/3).
Do not redefine `The Walker`, `Summon Door`, `Door`, or `Destiny` anywhere in this world's data — all four already exist from starter source or World 13; duplicate template ids throw in `buildWorld`.
Step 9 (depends on 6, 7, 8). Reward player cards (REQ-W14-14): Call In The Favor (strips Bargaining), Let It Sit (strips Depression), A Page From the Ledger (minor lore-delivery card, queued not granted), Keep Walking (Brace endurance reframe), Let It Go (DiscardThenDraw cycling reward), and the two pool-ledger-reading boons Ask For More Time / Take What's Owed. Also add the new pool entry to src/data/boonPools.json: "pool-ledger-reading": ["Ask For More Time", "Take What's Owed"]. Exact JSON in the design doc's Reward Player Cards and New Boon Pool sections.
Step 10 (depends on 6, 7, 8, 9). Deck composition (REQ-W14-20/21): Act 1 — 8 cards (heaviest, two Modal copies plus the boon-choice card); Act 2 — 6 cards (moderate); Act 3 — 9 cards including the reused Destiny and the fixed { "templateId": "The Walker", "count": 1 } closer. Exact per-template counts in the design doc's Deck Composition section.
## Slice 3 — World bundle wiring (REQ-W14-1, 22..24) No `src/data/worlds/answers/` directory exists yet — from-scratch bundle, structurally mirroring the shipped `src/data/worlds/questions/` bundle (`cards.json`, `index.ts`, `meta.ts`, `theme.ts`).
Step 11 (depends on 10). Create src/data/worlds/answers/cards.json (deck composition only, referencing Slice 2 template ids — copy directly from the design doc).
Step 12. Create src/data/worlds/answers/theme.ts exporting a VisualTheme (mirrors QUESTIONS_THEME's shape): worldId: "answers", intrusionHue trending cool/muted, desaturating further toward Act 3's grey-blue (REQ-W14-22/23 — the opposite direction from World 13's warming-toward-hot arc, and the specific palette World 15's own Act II is specced to reprise). backdrop keys "answers-bg"/"answers-overlay".
Step 13. Create src/data/worlds/answers/meta.ts exporting WorldDisplayData/WorldHelpData (mirrors QUESTIONS_DISPLAY/QUESTIONS_HELP). Help text must not spell out the grief-stage metaphor (REQ-W14-24) — describe mechanics (Bargaining taxing every other card, Depression taxing itself, Destiny's cost as a running total) without naming "bargaining"/"depression" as emotional stages. Also note plainly, mirroring how questions's own help/flavor disambiguates the Destiny/DestinyScene naming pun, that this world's Destiny card is the same card carried over from questions, not a new one — this is now load-bearing documentation given the superseding decision, not optional flavor.
Step 14 (depends on 11, 12, 13). Create src/data/worlds/answers/index.ts assembling the WorldDataBundle (mirrors QUESTIONS_BUNDLE), then register it in src/data/worlds/registry.ts's worldDataRegistry array (14th entry).
Step 15. Wire src/game/worlds/assetBindings.ts: no reality/overlay/cardfront/inset art exists yet for this world. **Correction from an earlier draft of this plan, caught during fresh-eyes review:** only bird-building and highway-volcano actually shipped without card inset art — verified directly against assetBindings.ts and src/data/worlds/questions/theme.ts, questions itself has a full worldCardfrontKey: "questions-cardfront" plus a named inset binding for essentially every card, including the shared Destiny template (questions-inset-destiny). Shipping answers without art would make it the **first end-world in this trilogy** to ship art-less, not a continuation of existing practice for this specific comparison point — flag this distinction explicitly to the user rather than citing questions as settled precedent. If art generation isn't feasible before this world ships, that may still be the right call, but it should be made on accurate information.
Add a music entry only if a track exists; otherwise flag as a follow-up rather than fabricating a placeholder.
Step 16 (depends on 14, 15). Confirm src/core/tests/worldRegistry.test.ts's existing conformance suite (extended during World 13's plan to check duplicate template ids and all-five-hooks) picks up answers automatically since it iterates the registry generically — no new assertions needed there, just confirm it passes. Add a new src/core/tests/answers.test.ts, mirroring newDerelict.test.ts's shape: registry/build sanity (selectTheme("answers") resolves, buildWorld assembles, no duplicate ids), a hook/keyword validation loop over all of this world's WORLD_CARDS against the current VALID_KEYWORDS set, and isolated effect tests for the Self-Transform pairs and the reused Destiny card's cost formula (see Slice 5's validation gate for the exact expected value). This satisfies REQ-W14-27.
## Slice 4 — Grief-support interstitial (trilogy-level: no new work)
Step 17 — verification only. This scaffolding is already fully built and already includes answers: confirmed directly in src/game/scenes/griefSupportGate.ts, whose exported GRIEF_SUPPORT_WORLD_IDS constant is ["questions", "answers", "the-beginning"], and in WorldSelectScene.ts's constructor, which already threads a GriefSupportStore. No code changes needed for REQ-W14-27a. Add this to the manual validation pass only: on a fresh profile, select answers first (before ever touching questions) and confirm the interstitial shows; dismiss it; confirm selecting questions afterward skips straight to Table (proving the flag is genuinely trilogy-wide, not per-world, in both directions).
## Slice 5 — Unlock gating, documentation, and final validation (REQ-W14-25, 26, 29)
Step 18. Add a worldUnlock Destiny Blessing entry to src/data/unlocks/catalog.json, mirroring the existing world-questions entry exactly in shape: 
{
  "id": "world-answers",
  "name": "Answers",
  "description": "Opens Answers in World Select.",
  "cost": 25,
  "destinyWeight": 0,
  "effect": { "type": "worldUnlock", "worldId": "answers" }
}
Cost 25 is the next escalation step above questions's 20 (existing costs step 5→10→15→20 in three-world tiers) — a placeholder for balance tuning, not a resolved final value, same caveat as every other authored number in this plan. Without this entry, isWorldUnlocked would default answers to unlocked from game start, which is very likely wrong for a narrative-closer world reached only after progressing through the trilogy's earlier entries.
Step 19 (depends on 18). Extend src/game/tests/unlockAssetBindings.test.ts's or the relevant unlock-catalog test's coverage to confirm the new world-answers entry resolves and gates answers correctly (mirroring however world-questions is currently covered, if at all — if no existing test asserts world-gating behavior per catalog entry, flag this as a pre-existing test gap rather than silently skipping coverage for the new entry).
Step 20. .lore/reference/worlds/authoring/theme-authoring.md: add the signature-verb table row answers | concede | Bargaining/Depression keywords tax cost from two different directions — spending elsewhere (Bargaining) and settling into one thing (Depression) — every act gives something up (REQ-W14-6), immediately following the questions row. Document Bargaining/Depression as C2 prose (extending the existing Denial/Anger paragraph), not a new Signature Effects table row (REQ-W14-25) — call out Bargaining explicitly as ClearCostPerOtherKeyword's first real user. Also add a short note alongside the existing "naming collision, intentional (questions)" callout recording that answers reuses the same Destiny template rather than authoring its own — this is now the second place (with Step 13's help text) that needs to carry the superseding-decision note forward so it doesn't silently drift out of documentation.
Step 21. No update needed to .lore/reference/worlds/catalog/endworlds-trilogy-concept.md's core structure — confirmed already current (rewritten during World 13's execution). Add only a small addendum noting the Destiny-reuse decision (one shared entity across all three end-worlds, not three reskins), so this reference doc doesn't drift out of sync with what Slice 2/3 actually ship.
Step 22 — Final validation gate (depends on all prior steps). Per REQ-W14-29 (with the Destiny formula corrected per the superseding-decision note):
  1. bun run test — all existing tests plus new Slice 1/3 tests pass.
  2. bun run lint && bun run typecheck && bun run build succeed.
  3. Confirm via the Slice 1 unit test that one other card carrying Bargaining:3 produces the same tax as three other cards each carrying Bargaining:1 (both sum to 3) — the count-vs-value trap.
  4. Confirm via the Slice 1 unit test that a card carrying Bargaining itself receives zero tax contribution from its own value, while every other card's cost rises (REQ-W14-10).
  5. Verify Destiny's effective cost in the answers world, with a separate Bargaining:3-carrying card elsewhere in hand and no other modifiers active, equals 15 + 3×costPer(Bargaining) — not a formula involving Denial or Depression. As of commit 5b2e0e5, Destiny carries keywords: [] (no authored keyword at all), so it contributes zero to ClearCostPerSelfKeyword and the only active modifier is Bargaining's ClearCostPerOtherKeyword, which taxes it same as any other card in hand. This supersedes both the spec's original Depression-based formula and this plan's earlier Denial-based correction (see the rewritten superseding-decision note; this is the second, current correction of REQ-W14-29.5).
  6. Confirm no engine change touched The Walker/Door/SurviveWorld wiring (REQ-W14-2/3) — byte-identical before and after.
  7. Verify a signature-threat card's applied Bargaining/Depression (via ApplyKeyword{target:"self"}) decays to zero after enough turns with no reapplication, and that the matching RemoveKeyword reward strips it before then.
  8. Verify a RemoveKeyword reward targeting Destiny is a no-op. As of commit 5b2e0e5, Destiny authors no keyword at all (keywords: []), so this is now trivially true rather than a demonstration of authored/applied asymmetry against a specific keyword (the old Denial:2 case this check originally targeted no longer exists) — keep the assertion for regression safety, but don't read a passing result as proving the asymmetry point it originally did.
  9. Manually verify the grief-support interstitial per Slice 4's Step 17 check (fresh profile, select answers first, confirm trilogy-wide flag behavior in both directions).
  10. Manually play through answers once via bun run dev to confirm the Act 1 A Deal Too Easy → early-Destiny beat is visible and legible, and that the world isn't unlocked from a fresh profile without the Step 18 Blessing purchased.
## Open items carried forward (flag to the user, not silently resolved) - Final `worldId`/title (`answers`) and the signature verb `concede` are still spec-level placeholders per the spec's own Open Questions — this plan proceeds with them as working names, not confirmed final. - Whether `Bargaining`'s "the carrier pays nothing, everyone else pays" asymmetry reads clearly without an explicit UI callout is unresolved (spec's own open item) — flag during Slice 5's manual playtest. - `cardfront`/inset art for this world does not exist yet (Step 15) — flagged as a gap, not silently deferred, same as `questions` shipped. - The unlock cost of `25` (Step 18) is an unvalidated placeholder pending a `bun run sim` balance pass, same caveat the design doc raises for every other authored number. - The Destiny-reuse decision (see header note) is the single biggest deviation from the ratified spec in this entire plan, and commit `5b2e0e5` made it bigger still. `Destiny` no longer reads as either "cut off future moments" (this plan's original superseding-decision framing) or the spec's intended "quietly picked it up and kept walking" — its `onCleared` is now `SurviveWorld`, the same instant-win effect `Door` uses, with a `15`-cost gate and an active per-turn punishment (2 damage + exile 1 world card) while it sits undealt. Reusing this template in `answers` means an Act 3 (and, via the Step 6 seed, theoretically an Act 1) copy of `Destiny` is now a second way to end the entire run in victory, not a hazard-clear. Flag this to the user explicitly when the plan is reviewed — narrative framing for what it means to "clear Destiny" in `answers` needs to be decided fresh, not inherited from either the original spec or this plan's now-outdated superseding note. It should also be revisited before World 15's own Destiny handling is planned, since the same template-id collision (and now this same instant-win behavior) will recur there.