# Stilllight · economy and balance

The executable model uses the same `Core/Economy.swift` as the app. Run `swift run Balance`; the saved output is BALANCE_RESULTS.json. This is a deterministic model of explicit choices, not a claim about measured player retention or average human behaviour.

| Companion | Initial cost | Light / second |
|---|---:|---:|
| Pebble | 20 | 0.4 |
| Petal | 180 | 3.8 |
| Tide | 2,200 | 42 |
| Halo | 30,000 | 490 |
| Lantern | 480,000 | 6,000 |
| Archive | 8,000,000 | 78,000 |

Each additional companion costs 1.17× the previous one. Every ten adds 50% of base family output, linearly. Bulk pricing is the exact geometric sum. Light upgrades: touch 25×1.72^level, orbits 250×2.15^level, resonance 1,500×5^level, memory 12,000 once per cycle. Limits are 60 / 35 / 5 / 1. Insight study cost is 1 + rank(rank+1)/2.

Production multiplies cycle Insight (8% each), Renewal growth (60% each), persistent constellation points (3.5% each) and optional Lasting Radiance (2×). Touch and companion study bonuses apply to their own production. Automatic tending adds one ordinary tap per second; it does not simulate manual resonant taps. Temporary multipliers use the highest active value, not their product.

Constellation node costs grow 1.7× within each constellation. Initial costs are 40, 1,400, 18,000, 160,000, 1.5M, 15M, 180M and 3B. Each constellation must be completed before the next opens. Eight sets contain 47 points in total.

## Thirty-day model

One tap per active second, bounded shopping every 30 seconds, no paid upgrades or videos. Long gaps are settled once on return with the real offline cap. All states remain valid throughout all three runs.

| Visit pattern | First constellation | First Reflection | First Renewal | All eight constellations |
|---|---:|---:|---:|---:|
| 15-minute opening; 5 minutes every half hour | 7m | 7m 30s | 2h 30m 30s | 25h |
| 15-minute opening; 3 minutes every four hours | 7m | 7m 30s | 20h 1m 30s | 6 days |
| 10-minute opening; 2 minutes every twelve hours | 7m | 7m 30s | 3d 1m 30s | 15d 1m 30s |

There is no requirement to follow the model's payback-focused purchases. Earlier constellation investment, longer cycles and different Insight studies change these times. The model deliberately limits shopping and does not assume frantic tapping. It exposes meaningful advantages to active visits without withholding the same content from infrequent returns.

Earlier tuning rejected exponential companion milestones and a flat late Renewal threshold because they collapsed long-term progression. The final model uses linear family milestones and rising Renewal requirements. The first Seed costs half the original draft so the opening session includes a completed visible pattern before the first reset.

Offline tests cover eight-hour and extended caps, twenty-four-hour entitlements, boost expiry partway through absence, rollback, duplicate settlements and no production. Save and StoreKit tests cover the same balances across resets and relaunches. See VERIFICATION.md for exact tested environments and remaining limits.