Flight

The Reference Drive, the burn, and the fuel that gates everything.

This page owns pillar 3. Every mechanic here exists to make planning and executing a burn the interesting problem.

Changed in v0.2. The previous edition had an Elite-style cruise mode in which a gravity gradient throttled your maximum speed, and the skill was managing a throttle against an overshoot. That was a contrivance — the fiction had to be told to produce the behavior — and throttle-correction is not a compelling verb. It is replaced by honest brachistochrone burns: accelerate, flip, decelerate. Deceleration is still required, because you are going fast and must arrive at rest, which is Newton rather than a rule we invented.

The Reference Drive

The project is named after an inertial reference frame, and its central abstraction is a frame graph in which re-framing provably does not move anything. The fiction is the same object:

A Reference Drive does not push the ship. It holds the ship near a chosen inertial frame while re-anchoring that frame.

Two consequences, and both are load-bearing.

1. Inertial compensation. Because the ship is held near its original frame, the crew experiences only a fraction of the vessel's proper acceleration. This is why a ship can sustain thousands of g and the person in the seat feels one and a half. It is the technology that solves The Expanse's crash-couch problem, and it is why burns can be minutes rather than days.

2. Superluminal transit. At sustained high power the re-anchoring rate outruns the light-speed constraint on the coordinate velocity, while the ship's proper velocity never does. Crossing a star system at 20 c is the same mechanism as a docking nudge, at more power.

One drive. One throttle. One verb. There is no cruise mode to enter and no regime to switch between — only how hard you are burning, which is a continuous quantity from a station-keeping nudge to a system crossing.

🎮 Designer's Note: The rule for writing about the drive: never say "faster than light". Say re-anchor, reference, frame, displacement. The ship is not fast; it is being re-indexed. That framing is what makes both the compensation and the deceleration requirement read as inevitable rather than convenient.

The burn

The most-repeated action in the game, and the micro loop.

   PLOT ──────── BURN ──────── FLIP ──────── BURN ──────── ARRIVE
   the nav        accelerate     ~4 s of      decelerate     at rest,
   computer       toward the     freefall     onto the       relative
   solves for     target;        while the    target;        to the
   a flip point   "down" is      ship turns   "down" is      target
                  aft            180°         forward
      │              │              │             │
      │              │              │             └─ under-burn here and you
      │              │              │                arrive fast. Turn, re-plan,
      │              │              │                lose 40 seconds.
      │              │              └─ silence, weightlessness, and the
      │              │                 world rotating around the canopy.
      │              │                 The game's best four seconds.
      │              └─ thrust gravity. The ship has a floor. Everything
      │                 loose has an opinion about that.
      └─ the decision: <b>how hard.</b> Halving the trip time doubles the fuel.

The three rules

Everything below falls out of these, and every one of them is a real law rather than a design rule.

1. Brachistochrone. Accelerate for half the distance, flip, decelerate for the other half. Trip time and fuel follow directly:

t   = 2 · √(d / a)              trip time
Δv  = a · t = 2 · √(a · d)      total velocity change
fuel = k · M · Δv               k is a drive-efficiency constant

2. Halving the time doubles the fuel. Quartering a doubles t and halves Δv. This is the single most important tradeoff in in-system travel, it is identical in structure to the jump route choice, and it is legible to anyone who has ever driven anywhere.

3. What you carry, you keep. Re-anchoring changes the frame, not your velocity within it. There is no drag, no friction, and no automatic arrival. If you have not spent the second half of the burn shedding what the first half gave you, you arrive at 3 c and you are somewhere else.

Numbers

Class 3 drive · maximum transit acceleration 0.050 c/s (1.5 × 10⁷ m/s²) · Cannon-class at 60 t. Real destinations, real distances.

Two things worth reading off that table. Inner-system travel lands in the 2–7 minute band the micro loop is designed around, without any tuning — it is what the arithmetic gives. And in-system fuel is cheap relative to a jump, by roughly an order of magnitude, which keeps the interstellar economy intact: moving around a system is nearly free, and leaving is the decision.

The Moon at ten seconds is the honest edge case: at full transit power, close targets are trivial. That is correct — nobody should have to sit through a ten-minute trip to a moon — and the player simply throttles down when they want the trip to be a trip.

Drive ratings

Transit is comfortable. Maneuver is not. Compensation is optimized for sustained, straight-line, high-power running; it is much less effective against the short, hard, off-axis accelerations of combat and atmospheric flight. A Class 5 pulling maximum maneuver puts nearly 7 g through the seat, which is survivable for seconds and not for minutes.

That asymmetry is deliberate and it does three jobs at once: it makes crossing a system pleasant, it makes dogfighting physically expensive, and it gives the on-foot layer a reason to care what the pilot is doing.

What makes it a skill

The failure mode of the old design was that the skill was a throttle micro-correction. Here the skill is distributed across five things that are genuinely interesting:

🎮 Designer's Note: The flip is the game's signature image and it should be treated as such. Four seconds of freefall, the engine silent, the ship rotating, and the destination swinging into view — then the drive lights and the floor comes back the other way. Everything unsecured in the ship floats and then falls. It costs almost nothing to build and it is what people will record.

Why there is no ballistic transfer

⛔ Deliberately deferred.

An earlier draft offered a cheap, slow alternative — one impulse, a long coast on a conic, a capture burn — on the grounds that it would make the torch burn's fuel cost mean something.

Cut, deliberately. One travel verb is this design's strength, and a second one earns its place only if in-system fuel is ever genuinely tight. At 0.05–0.15 t for an inner-system burn against a 16 t tank, it is not. It also drags in time compression and a determinism question about analytic skips, for a mode players would rarely choose. If in-system fuel pressure ever becomes a design goal, this is the first thing to reconsider.

Maneuver and landing

✅ Built and proven.

Below transit power, this is ordinary Newtonian 6-DoF flight and it already exists — packages/simulation/src/flight.ts, with patched-conic gravity, atmospheric drag, thruster resolution and surface attachment, verified against free fall to within 0.03%.

Flight assist

A toggle, defaulting on, that adds rotational damping and a velocity-matching cap. Assist off removes both, and the ship becomes a rigid body in space with thrusters.

Rationale. Assist-off is the most-loved advanced mechanic in the genre and it costs nothing to implement, because it is the absence of a feature — the honest physics is already what flight.ts computes. Making the honest mode the advanced mode is correct for pillar 3, and it means the skill ceiling is set by physics rather than by a designer.

Landing

🟡 Partially built.

Contact under 3 m/s vertical and 12° off local vertical attaches the ship to the surface frame. Outside those bounds it is a crash: integrity loss proportional to the square of impact speed, destruction above 25 m/s.

Landing works today. Slope response, landing gear as a discrete module, and contact at anything other than a single ground point are ⬜.

Jump

⬜ Designed, not built.

Discrete re-anchoring to a distant system's frame. The interstellar mechanic, and the one that makes fuel matter.

[ select target ]→[ align, 15° ]→[ charge 12 s ]→[ TUNNEL 6 s ]→[ arrive at 40 c ]

The six-second tunnel is the one place the game leaves the world. It is not a loading screen — the destination is already generated and the simulation is already running — and its length is set by how long the visual is worth watching. Resolved: hold-to-skip becomes available after the first dozen jumps. The skip still costs ~2 s, because the tunnel is where shader pipelines pre-warm and there is nowhere else in the game with six spare seconds.

Range

d_max = R_drive · (M_opt / M_total)^0.6

The range spread across the whole game is roughly 7.7×, comparable to Elite Dangerous, and this is a deliberate revision from v0.1's much flatter 2.2×.

Why the reversal. v0.1 argued for a flat curve on the grounds that with no retention metric to defend, the first ship should be nearly as good as the last. That reasoning was about fairness and it missed the thing that actually matters: a large range spread changes the character of the game rather than merely its speed. At 8 ly you thread a dense web and every leg is forced; at 55 ly you leap between chosen anchors and the sparse outer regions open up for the first time. That is a qualitative transformation, and wanting a faster ship is a legitimate and well-earned desire rather than a retention hook.

What is kept from the old argument: the early curve is steep, so the cage phase is short. Elite's problem is not its top end, it is the hours spent at the bottom. See progression.

Fuel cost

F = C_drive · (M_total / 100) · d²      tonnes

For a 60 t Cannon on a Class 4A (C = 0.0025, 40 ly rated) with a 16 t tank:

Short hops are eight times more fuel-efficient per light-year than maximum jumps. That single fact generates the entire route-planning game: the fastest route and the cheapest route are different routes. See galaxy.

Fuel

⬜ Designed, not built.

One resource: hydrogen, in tonnes. It powers jumps, burns, and the reactor everything else draws from. No second fuel, no consumables, no ammunition.

Scooping

Fuel is taken from a star's outer atmosphere. Which stars are scoopable is real data — the spectral class from the catalog, not a flag we assign.

The Milky Way's most common star by a wide margin is class M [Source: standard stellar population statistics; see e.g. Chabrier, Galactic Stellar and Substellar Initial Mass Function, 2003], so the frontier is mostly slow refuels and a route that threads G and K stars is genuinely better. Nothing about that was designed; it fell out of the data.

The scoop

Rate = S_module · classMultiplier · (1 − (r − r_min) / (r_max − r_min))²   t/s

Closer is faster, and closer is hotter. Heat accumulates from the star's irradiance and is shed by radiators, which makes refuelling a skill with a risk gradient rather than a wait — and ties fuel into the same heat system that governs burns, weapons and stealth. One resource system, four uses.

Running dry

The one genuinely unrecoverable state, and it must be visible long before it is reached. Three safeguards: the router refuses to plot a route it cannot complete and says which leg fails; a 0.5 t reserve is held back and is not spendable on a jump, so a stranded ship can always reach a star in-system; and Resolved: in solo modes, self-scuttle returns you to the last station with unbanked data lost — clean and immediate, with no waiting for help that cannot come. In the persistent universe, a distress beacon other players can answer, with the rescue unrewarded and visible. Degrades gracefully with population, which is the property that matters.

🎮 Designer's Note: Resist adding a second fuel type, a repair consumable, or ammunition. One resource that is simultaneously the travel budget, the combat budget and the survival budget is what makes every decision in the game commensurable — "is this detour worth it?" always has an answer in the same units.

The one sanctioned exception is the relay beacon, and it is sanctioned because it does not break the rule: a beacon is manufactured from banked survey data, so the thing being spent is still the one thing the game meters. It is data committed in advance against a risk, not a new currency.

Related

• ships — the modules that set every coefficient here
• loops — the burn as the player experiences it
• galaxy — jump range and fuel as a puzzle
• onfoot — what thrust gravity does to a person
• docs/concepts/frames.md — the frame graph the drive fiction describes