Ships Are Assembled, Not Authored

The Three-File Model

Craft hardware lives in three JSON files, and the split matters: two of them are definitions (what a part is), one is an instance (what a particular ship has).

Definitions never carry state, and instances never carry stats. A tank.lh2-20t knows it holds 20,000 kg of liquid hydrogen and weighs 800 kg empty; only the installation on a specific ship knows it's currently full. An engine definition knows its thrust curve; only the installation knows it's throttled to 100%.

Worked Example: Hermes

Hermes — the fleet's crewed fast-transit ship — is five sections. Reading her build file top to bottom is reading the ship:

Drive Bay        (engine-bay.fusion-16x7, 16 m × 7 m tube, 4,000 kg structure)
  ├─ Direct Fusion Drive    1 GW fusion power → 13.66 kN at 6,000 s Isp
  ├─ DFD Power Tap          2,500 kW electrical, bled off the drive plasma
  └─ Droplet Radiator ×2    5 MW heat rejection each — the drive's 10 MW
                            of waste heat has to go somewhere

Cryo Tanks       (tank-module.cryo-14x8, 14 m × 8 m tube, 2,600 kg structure)
  ├─ LH2 Reaction Mass      20,000 kg — the drive's propellant
  ├─ LOX Tank               4,000 kg — oxidizer for the descent engines
  └─ Flywheel Bank          250 kWh buffered for burn transients

Command          (command.capsule-5x4, 5 m × 4 m capsule)
  └─ Flight Deck            5 crew stations

Habitat Ring     (habitat.ring-10, 10 m ring, 5,000 kg structure)
  ├─ Habitat Fit-out        berths for 5
  ├─ ECLSS                  life support rated for 5 crew, 7.5 kW draw
  └─ CMG Cluster            4,000 N·m attitude authority

Landing Truss    (landing.truss-9x6, 9 m × 6 m open truss)
  ├─ LH2/LOX Descent Cluster  85 kN at 450 s — the drive can't land her
  ├─ Cold-Gas RCS
  └─ Cargo Hold             5,000 kg loaded

Nothing above says what Hermes weighs. But sum the structural masses, the component dry masses, the tank fills and the cargo, and her mass budget falls out — and from that, with the drive's Isp, her Δv. Change one line (halve the LH2 fill, swap the cargo) and every downstream number moves with it: the engineering panel, the mission compiler's burn solutions, the collision scan. There is no second copy of the truth to drift.

The design also encodes real engineering tension. The fusion drive is 11 tonnes of engine producing all of 13.66 kN — a thrust-to-weight ratio that would embarrass a lawnmower, attached to an Isp seventeen times better than the best chemical engine ever built. So Hermes carries a second propulsion system, an 85 kN chemical descent cluster, purely because a high-Isp low-thrust drive cannot fight a landing. Two engines, two fuels, two jobs: that trade is the whole game in miniature.

Slots: Why You Can't Build a Monster

Sections constrain components. An engine-bay.fusion-16x7 offers two large engine mounts aft at 3 and 9 o'clock, two medium power bays inside, four medium radiator mounts round the rim and a small utility mount — 42 points in nine slots, plumbed for fuel, oxidizer, power and heat — so you cannot rack six fusion drives into it, and a part that needs a line the slot does not carry does not go there either. The little tank-module.tube-5x3 takes three storage parts and one thermal, and that's the ship's fuel margin decided at the drawing board. Slot budgets are what make hull choice a strategic decision rather than a texture swap: the envelope you launch is the option space you live with.

The Tech Tree: Production, Not Propulsion

The systems below are the design direction currently being built toward; the shipped catalog covers the hardware side (engines, tanks, hulls), with the production economy following it in.

Three commodities

• Helium-3 — fusion fuel, and the economy's bottleneck. Effectively unobtainable on Earth; present at parts-per-billion in lunar regolith; abundant only in the outer system.
• Deuterium — the other half of the fusion reaction, extracted from water, or harvested where nature has pre-enriched it.
• Water — reaction mass, life support, electrolyzed LH2/LOX, and the feedstock deuterium comes from. The most broadly useful and the easiest to get.

The sourcing ladder

Each rung is gated by ISRU and robotics maturity — not by a date or a research checkbox:

• Earth (start) — everything the fleet burns is made here and lifted out of a 9 km/s gravity well. The tax on every mission.
• The Moon (early) — the first place fuel can be made instead of lifted. He-3 at whisper concentrations, water ice at the poles — and a real-world anchor: at ~20 ppb regolith concentration, a miner drawing ~350 kW harvests roughly 66 kg of He-3 a year. That number is why the early game is a grind worth winning, and why He-3 mining co-produces water and methane for free.
• Mid-system (Venus, the asteroids) — specialty sources. Venus's atmosphere is 120× enriched in deuterium but almost bone-dry: the boutique refinery, not the tank farm.
• The outer system (late) — everything, abundantly, at the bottom of enormous gravity wells and weeks of light-lag. The prize, and the trap: the obvious giant may not be the profitable one.

ISRU: upgrade the bottleneck, not the base

In-situ resource utilization follows the real lunar process — excavate → beneficiate → heat → separate — and each stage is a separately installable, separately upgradeable component, the same catalog schema as an engine or a tank (mass, power draw, waste heat, throughput, yield). A mining site is just a craft that doesn't move: sections, slots, components, and a derived number at the end — kilograms per day. First-generation gear is heavy, power-hungry, and inefficient by design. "Unlocked" is not "supplied"; the site's daily output is the real constraint on fleet operations, and upgrading the bottleneck stage beats replacing a monolithic base every time.

Robots: three families, one loop

Mining robots gate the surface. Processing robots run the plant. Assembly robots build in orbit what no one wants to launch whole. All three start slow and unreliable, and their progression stats — autonomy, throughput, reliability — feed the game's central compounding loop:

better robots → higher ISRU throughput → more fuel
     ↑                                        ↓
  capital  ←  more sites  ←  more ships  ←────┘

Reliability is the sleeper stat: a robot fleet that breaks down is a production shortfall, a production shortfall is a depot that didn't fill, and a depot that didn't fill is a ship stranded somewhere rescue can't reach. Failure is permanent here — which is also why a deliberately boring, never-upgraded backup thruster earns its mass on the one flight where the main drive quits.

Tiers with headroom

Every threshold technology ships good enough to matter, not good enough to trivialize. The catalog's Baseline tier is game-start hardware; the Future tier is a generation out. The fusion drive gets cheaper and harder-burning; shielding gets lighter; ISRU yield climbs; robots stop needing supervision. The one deliberate optimism — the Direct Fusion Drive performs several times better than the published physics says it should — is written down as a setting premise, because a game about logistics needs its engines to be merely expensive, not fictional-forever.

See It in the Data