The Long Game: why Zingers ends after 40 minutes, and how it stops

Status: shipped Stages 0–6 (July 2026). Design companion to the bible. Player-facing Flight fiction stays in bible/10-ascent.md; naming in vocabulary.md. Companions: design-vision.md, essence.md, flight-first-plan.md.

This doc answers one question: how long can a player play before they run out of reasons, and what do we change? Stages below are live unless marked deferred.

1. The honest clock

Cold-player estimate for the build as it stands:

That reads like a thin game. It is not a thin game. See §2. It is a legible- surface problem: the content exists and the player never learns it exists.

2. What is actually built (the surprise)

An audit of the live code, not the docs, found:

• Flight: 100 sectors across 10 Reaches, hazards, camps, first-light chests, Crown caches, scout, ghosts, share links, a 2500-Crown Hundred purse. (components/grounds/climb/*, lib/climb-campaign.ts)
• Champions: First Minds + baked dex (Stage 6 wave 1), type pentagon, 5 tiers, career-derived bodies + phenotype, Saga ledger, Imprints. (lib/engine/roster.ts, lib/minds/baked.ts, lib/imprints.ts, lib/evolve/*)
• Return hooks: Daily fight, daily caches, seasonal goals, 28-day seasons, Clan war, prediction streaks, a global rating board that self-plays via cron, Director guiding, unlock track, wing traits, Conditions, expeditions. That is on the order of 15 hours of content being served in the first 30 minutes, mostly invisibly. The problem was never volume.

3. Diagnosis. Four structural faults

1. Nothing points at the next thing. No function anywhere answers what should I do now? Every hook is discoverable only by already knowing about it.
2. Everything is open at minute one. All regions, all scenarios, all 100 sectors, all 8 recruits. A 15-hour reveal spent instantly. Games that feel long mostly just withhold better.
3. Raising has no visible effect on the verb you perform. Training costs 60 Crowns and moves three invisible dials. The player never sees a run or a fight go differently because of what they taught.
4. The repeated verb has no variance. Every run is the same run; every fight is the same fight with different prose.

4. The yardstick (stop measuring against Pokémon)

Pokémon and Zelda are content-volume games. Hand-placed worlds built by hundreds of people. We lose that race permanently and the comparison only produces despair.

Our actual peers are Balatro (one card game, ~150 modifiers, hundreds of hours), Vampire Survivors (one mechanic plus an unlock tree), and Football Manager (no world at all. Pure career friction). All three extract enormous playtime from a fraction of the assets Zingers already has, using three tools: run variance, drip-fed unlocks, and a career state that carries consequences.

Target: 10–15 hours across the first month, with a real 8–12 minute daily reason to open. Not Zelda. Five to ten times where we are, reachable without new art.

5. The direction decision (founder call, July 2026)

Flight is the game. The champion is the save file. Battles are punctuation.

Reading LLM debate transcripts is a novelty that dies on repetition. Depth belongs where the player's hands are, which is Flight. This is not a retreat from the champion. It is an inversion of the causality:

Once raising a champion is a build system for the verb the player actually performs . handling, scout range, a second life, gold pull, hazard reads. "raise a mind" stops being a side tab and becomes the reason to keep raising. That is the Pokémon feeling (team building, meaningful choices, legible consequence) relocated to the right verb.

Battles stay: "you fly, it fights" is the differentiator, and the async league is content that runs while players sleep. But they get compressed into resolved, glanceable, stake-carrying moments. Never a wall of text to read.

6. The plan

Ordered by return on effort. Each stage is shippable alone.

Stage 0. The Director (1–2 days) ✅ shipped

A pure function over existing save state that answers "what now?" in one line and one button, rendered on every surface a player lands on. Creates no content; it makes built content perceivable. lib/director.ts + components/director/*.

Stage 1. The unlock track (3–5 days) ✅ shipped

One Trainer track that every activity feeds, where each rank opens something by name: Reaches in blocks, then tribunal, gauntlet, broker, the second region, recruit slots, scout. Not new content. Rationed content. Retro-fits the existing build into a 15-hour reveal and gives the Director a spine to point at.

Live module: lib/unlock-ladder.ts. Enforced on Flight ceilings (mobile + desktop), Hub gates, Broker, Scout, recruit slots; Director primary/also can point at the next named door. Reach II prove gate unchanged. Depth already flown is grandfathered so existing saves aren't soft-locked.

Stage 2. Champion → Flight causality (1–2 weeks) ✅ shipped (v1)

The core of §5. Each champion grants wing traits that measurably change a run; imprints and milestones teach new ones; the player picks a loadout before launch. Turns 8 minds into 8 × loadout-space and makes every raising action pay out in the verb that matters.

Live module: lib/wing-traits.ts. One innate trait per Force type + up to one earned trait (axes / camps / temperament). Effects: lives, gold odds, glide, stumble, scout camp bonus, cruise speed. Applied via ascentSessionMods on both Flight bodies. Ready strip shows WINGS · … and earned toggles.

Stage 3. Conditions (3–5 days) ✅ shipped (v1 Flight)

A data-driven modifier applied to a run or a fight: thin air, fog, crosswind, one life, doubled gold, no scout, hostile crowd, sudden death. Twenty-plus of these is a day of writing plus engine hooks, and it multiplies every existing surface. The purest parametric win available.

Live module: lib/conditions.ts. Thirteen ranked daily skies (UTC day rotation), merged after wing traits via mergeRunMods. Scout practice stays Clear Sky. Ready strip shows TODAY · …; Director can surface the sky as an also-ask. Fight-side Conditions deferred.

Stage 4. Career friction and legacy ✅

Live modules: lib/career-friction.ts, lib/legacy.ts. Form / fatigue / scars stack into Flight after wings + Conditions (applyCareerToMods on Climb + Circuit). Rival chapters escalate via maybeEscalateRival / currentRival. Retire seals a legend + pending heirloom wing for the next claim (retireOwned → Long Vault). Director rotates spent/broken stables and surfaces Face / retire as also-asks.

Stage 5. Expeditions ✅

Live module: lib/expeditions.ts + lib/server/expedition.ts + /api/expedition. UTC-week route: seed (layouts via climbSector(i, seed)), one Condition, 20 sectors, own board per body. Mode expedition on Climb + Circuit. No camps, fractional XP/ Crowns. Ready strip WEEK · …; Director also-ask. Ghost-of-#1 deferred (codec ready in lib/climb-ghost.ts).

Stage 6. Batch content ✅ (dex wave 1 live)

Offline pipeline: content/minds/reviewed/*.json → npm run bake:minds → lib/minds/baked.ts, merged into roster / banter / beats / first-duel / showcase. Wave tools: npm run forge:dex (curated name banks + voice kits) and npm run generate:minds (optional XAI drafts). First hand bake: STILL, KEEL, PRISM, FABLE. Wave 1 forged a Gen-1-scale collectible set on top (see docs/bible/03-champions.md). Phenotype part catalog + species marks at rookie tier so same-Force minds read as different animals. No new GLTF per mind. Further waves = more reviewed JSON + rebake + part-kit expansion. Cadence, Year 1–3 dex targets, and Immortal mint math: champions-supply.md.

7. Explicit non-goals

• A fourth region. It is cheap (a BiomeConfig blob and a nature preset) and that is exactly the trap. Three regions with no job become four regions with no job, and the emptiness gets bigger and better lit. Regions get jobs before they get siblings.
• Anything that displaces the Flight-First feel gates. Retention scaffolding on a core verb that doesn't feel good just makes a legible empty standings. Feel work in flight-first-plan.md stays ahead of everything here.

8. How we'll know

Instrument before Stage 1 lands, or we will be guessing about whether any of this worked.