Everything on the main page is true and short. This page is the long version — the orbital regimes, the sun-and-shadow model, how a fleet grows, the full hardware catalogue, the weapons, and what is actually built versus what is still coming. The unexplained side of the sky is deliberately not on this page: that part you find out by playing.
Higher is not stronger. Each regime is excellent at one thing and poor at another, and no amount of progression moves you up a ladder — it only widens the set of orbits you can operate in. A veteran flies all three at once.
Imaging. The sharpest resolution and the fastest revisit in the game — and four minutes per pass to get the shot. Crowded, and atmospheric drag bills you for station-keeping every week you stay.
Timing, navigation and wide-area coverage. Slow, stable, reliable and frankly boring — which is exactly why it is the income floor that pays for the interesting orbits.
Comms and signals work. Park over one longitude and stare forever, or spend a few tens of m/s walking along the belt to a new one. Ground resolution is poor, but every neighbour is permanent, and a permanent neighbour is something to photograph, shadow, steal from, or shoot.
Half the Earth is dark at any moment, and the globe draws it as one continuous function of one number — the angle between each pixel of the disc and the anti-solar direction. That is not decoration. Where the sun is decides what an optical payload can do, and pointedly does not decide anything for the rest of the fleet.
Sun elevation scales an imaging payload, on a curve that peaks at a low sun rather than at noon. Shadows are what make structure readable — a curve that peaked overhead would turn "go where the sun is highest" into a strategy, which is the tier-climbing this design spent its rules keeping out.
A signals or relay payload is identically unaffected at every sun angle, because an emitter emits at three in the morning. That asymmetry is the whole point: a night target is exactly as easy to listen to as a day one, so no site is ever locked out — it is a question of bringing the right satellite.
A satellite in Earth's shadow derates whatever it is carrying, and the state of charge is on screen. Collection never spends it and it cannot be bought. Two things can: a satellite fitted past what its solar wings supply goes dark partway through each night, and a railgun or EMP shot takes a bite that only sunlight puts back. Neither is ever a surprise — the power budget is printed on the bill before you buy the part.
You are not flying a spaceship. You are running an intelligence operation whose orbits won't negotiate.Design pillar
Most space games treat orbits as decoration — a sine wave over a map, a number that goes up. Take the physics out of Celestrike and there is no game left underneath it.
Six classical elements and an epoch — a, e, i, RAAN, argp, M. A burn is
elements → state vector → add Δv → back to elements, which is why the fuel gauge and
the trajectory are the same fact expressed twice. The globe on the main page is
running that code, not a loop of a recorded animation.
Earth is not a sphere, so its equatorial bulge drags the orbital plane round over time. Modelling it means ground tracks drift between passes and sun-synchronous orbits actually stay sun-synchronous. Ten lines of maths, and without them the whole sky reads as fake.
Position at any timestamp is computable in constant time, so nothing has to run while you are away and nothing can fall behind while you sleep. The world running offline is not a feature that was built — it is a property of the mathematics.
Nothing below geostationary can hold station, so the real fix for a track that walks west is to tune the period until a whole number of revolutions closes in a whole number of nodal days — the way Landsat and the Sentinels actually fly. Counted in the nodal period, not the Keplerian one: get that wrong and the track closes to within a degree, looks right for a day, and is off target by the end of the week.
Every satellite pays to hold the orbit it is in, and the bill is a U. Low, it is drag — about 100 m/s a year at 400 km, falling away with a 66.8 km scale height. High, it is the pull of the Moon and the Sun on the orbit's tilt, which lands geostationary on the real 46 m/s a year a comsat budgets. Holding a tuned track adds to that, exponentially with altitude, which is what makes the cheapest-looking orbit on the coverage table a visible trap rather than a hidden punishment.
A geostationary satellite can be parked over any point on the globe, and it gets there by being walked — a direct line from where it is, in any direction, taking between six hours and five simulated days. Farther costs more and faster costs more, on the same cost-times-time curve a real drift along the belt trades on. The price is scaled to the game and says so: the honest figure for the same crossing is quoted beside it, so the model never forgets what it is discounting.
The far half of an orbit disappears behind the planet instead of x-raying through it, because occlusion here is a depth comparison rather than a renderer feature. That is the whole reason the globe is a 2D orthographic drawing and not a 3D scene.
The premise is barely fiction. In 2023 a US surveillance satellite manoeuvred toward two newly arrived Chinese satellites in the geostationary belt; they scattered, and one repositioned to get a sunlit view of its observer. Inspector satellites shadowing each other in GEO is documented, ongoing behaviour.
A refit and a launch are how a fleet grows, and both are paid for in the credits the fleet earns. Every bay on a satellite is a ladder: the dearer module is the better one, on every figure its card prints, and a servicing flight can fit it to a satellite you already own for less than launching a new one carrying the same thing — a re-engine on the satellite you start with is 566 credits against 738 for the launch, and takes it from 220 m/s of range to 2,200. Upgrading is the cheap path. A launch is what you buy for coverage somewhere new, and a second satellite is the only way to be over two places at once. Every design is for sale from the first minute. Modules are not: each bay opens on the part a satellite launches with, and every rung above it is researched from the intel you collect — launch pads teach engines and tanks, radar arrays teach optics and cameras, jamming sites teach receivers, ground stations teach power and uplink relays teach antennas. What unlocks a part is which targets you choose to go after, never play time, fleet size or money.
None of the four dominates another, and that is a property the build tests rather than a claim. It is measured twice — yield per day, because per hour of coverage the geostationary post loses to everything and that is the duty-cycle half of the trade being counted wrong, and yield per target kind, because a signals payload and an imager read different sites well. There used to be four geostationary designs; once one could be bought over any region and walked anywhere, what separated them was what the module ladder already sells, so the catalogue keeps one. Modules are the depth axis and a second satellite is the breadth axis, and "upgrade, or buy another" only reads as a decision while another satellite buys something different. What the campaign counts is the same thing: sites collected, countries opened, satellites on station, regimes held. Nothing is unlocked by clearing one — no level, no tier. It is a record of what you have done, not a door.
You specify a satellite's slots on the pad, where a module costs the part. You can also change them later — a servicing vehicle flies out, berths, and swaps them — and there the same module costs the part, the vehicle, the reach to that altitude and half a day of berthing. Five to twenty times more. That is not a balance decision, it is the one genuinely true thing about servicing, and it is why a mis-designed satellite is worth rescuing rather than writing off.
There is a second half to the asymmetry: mass counts on the pad and nowhere else. A lighter wing specified at launch frees bus mass for propellant, and the identical wing fitted in orbit cannot, because by then the mass budget has been spent on a structure that is already built. The one flight that can change that is a new tank, fitted with what was in the old one transferred across. So the ground fit is the better buy per module, and the orbital one is what you pay when it has to be this satellite, in this orbit, carrying this payload — which is still cheaper than a second rocket.
Exhaust velocity decides what a fixed mass of propellant is worth, so this is the one fitting that moves how far a satellite can go. Thrust and exhaust velocity only fight each other at a fixed electrical power, and a bigger power budget buys both — so the ion drive reaches ten times as far on the same tank and finishes in minutes a manoeuvre that took the standard engine an afternoon. It is the better engine outright, and it is priced like it.
A wing's output buys you nothing in sunlight — it pays inside an eclipse, and how much of an orbit that is belongs to the orbit rather than to the catalogue, so the same wing is worth more on some satellites than on others. The concentrator puts its cells on a narrow strip behind its own optics: brighter per square metre, lighter, and better shielded, all at once. Swapping it in flight moves total collection by single digits; specifying it at launch moves range by a quarter, because that is where mass counts.
Every open slot is a ladder that starts at the module the satellite launched with: nothing in the catalogue sits below the stock fit, and the build asserts that every figure a card prints climbs with its price. Seven are open — the imaging telescope, the intercept receiver, the bus camera, the engine, the tank, the wing and the antenna — and every one of them can be specified on the pad or swapped later by a servicing flight, once the part has been researched.
Each ladder climbs inside a fixed satellite. The bus is built on the pad, and the ceilings — the deployable area the structure can carry, the aperture of the camera, the reflector the signals bay unfolds, the mirror behind the telescope — are the same on every design, and no module raises one. Two slots move how far a satellite can go: the engine, which re-values a fixed mass of propellant, and the tank, which holds more of it, and together they are worth a tenfold gain in range. A refit that buys either lands under a fresh launch of the same design carrying the same fit, measured one slot at a time. A mission that swaps several bays at once is buying several upgrades, and adds up past a launch, which is legible rather than broken.
Every design in the catalogue has the same ten bays with the same six filled, and the only difference between any two of them is which payload bay that is. That is deliberate and it is tested, because a hull with an extra bay is the upgrade tree arriving as hardware — one bus strictly better than another with nothing given up on the other side. The one exception proves it: the comms relay some older campaigns still fly carries five, because its antenna is the payload rather than the umbilical. An empty bay reports no options and no price, because offering a choice the airframe cannot take would be advertising.
The standard telescope, and what every collection figure in the game is quoted against. Reads features a few metres across from low orbit, and collects fastest at a low sun angle — shadows are what make structure legible. On a night target it works at a fraction of its rate. The rungs above it fix that.
A short-wave infrared channel that reads through haze and thin cloud the standard telescope can't see past, fused with a visible channel to come out sharper than stock despite the longer wavelength. Holds a useful rate well after sunset, and its wider focal plane covers more ground per pass. The first step up.
A cryocooled thermal channel that sees emitted heat rather than reflected light, so a site works almost as well at three in the morning as at noon — fused with a visible channel to stay sharper than stock. Wider again across the track, and faster on every target than the rung below it.
The top of the bay, and the largest single upgrade in the catalogue. A violet channel out-resolves the standard telescope by a third on the same mirror, a thermal channel keeps it working long after dark, and a steering mirror sweeps half again the ground either side of the track — so it finds more targets and finishes each of them faster. It costs more than most satellites.
The standard receiver, and the only thing in the bay that hears everything. It digitises the whole band at once, so nothing on the ground is off its list and nothing on the ground is easy: an emitter has to stand above forty gigahertz of noise. No sun gate at all — a site at three in the morning is exactly as workable as one at noon, which is why the night side of the globe is never a closed door.
Cools its front end to a few tens of kelvin and slices the survey band into channels a few megahertz wide, so it works against a channel's noise floor instead of forty gigahertz of it. Everything the standard set can hear, heard about half again as fast, on a lighter and longer-lived box. The first real step up in the bay.
Sixteen cooled digital beams formed off the same reflector, so the bay works several emitters at once instead of taking them in turn — nearly twice the standard set's rate against everything on the ground. The best receiver in the game, the most expensive thing in the bay, and there is nothing it is worse at.
The small engineering camera every design carries — the same 10 cm aperture on all of them, because a bigger one is a different satellite. It never earns you anything; it is how you see, not how you are paid. What it can resolve is pure Rayleigh: about 4 m on the ground from a 620 km imager, about 240 m from geostationary, and about 7 mm alongside a neighbour a kilometre away. Clarity is bought with propellant and geometry, never with credits.
A cooled thermal camera that reads a target's own heat, so a satellite sitting in Earth's shadow photographs exactly as well as one in full sun — the one thing the standard camera can't do at any range. A visible channel alongside it keeps the detail finer than stock, on a wider field.
An ultraviolet channel gives the finest detail a single frame through the bus aperture can deliver — almost twice the stock camera, because the diffraction limit scales with wavelength — on a wider field, with a thermal channel that keeps working on a satellite in Earth's shadow. Sharp enough for a good picture of another satellite and everything it carries, and it has caught things in frame that nobody has explained.
The top of the slot. A photon-counting sensor stacks hundreds of short frames on top of each other, so it resolves finer than any single exposure through the same aperture can — the best camera in the game on every figure its card prints. Everything the ultraviolet camera sees about another satellite, from further away, and fast enough to hold a lock on things in frame that nobody has explained.
The standard engine, and what every capacity figure in the catalogue is quoted against. Enough shove to reshape an orbit inside a few hours and enough efficiency to do it more than once. Exhaust velocity decides what a fixed mass of propellant is worth, so what an engine swap changes is not how much the bus holds but how far it goes.
Four hydrazine chambers with an electric arc in each nozzle, doubling what a tank of propellant is worth against the standard engine and firing all four at once when you want the burn over with. Twice the reach of the standard fit and seven times the shove — the fit for a satellite you expect to keep moving.
The largest power budget the bus will carry, spent on a gridded ion array. Ten times the reach of the standard fit out of the same tank, and sixteen times the shove — a manoeuvre that took the standard engine an afternoon is over in minutes. The best engine in the catalogue, and priced like it.
The aluminium-lithium standard, and what every capacity figure is quoted against. It does not wear out early, and everything above it holds more and wears out later still. The tank is the one fitting that changes how much propellant a bus holds rather than how far that propellant goes.
A steel shell machined into a lattice of triangular ribs, so the wall between them runs thinner than a plain tank would survive — a little more propellant for the same bay, on a material that shrugs off pressure cycling the standard alloy slowly fatigues under. The cheapest step up in the slot.
A titanium shell machined thinner than the standard alloy would survive, so an eighth more of the launch mass is propellant — and titanium takes pressure cycling better than the alloy it replaces, so it arrives at its refit in slightly better condition too. The first real step up from standard.
A carbon-wound shell a fraction of the mass of a metal one at the same volume, so a quarter more of the launch mass is propellant. The overwrap carries the pressure load in fibre rather than in metal that fatigues, so it ages slower than the bus around it, and every refit for the rest of its life is cheaper for that.
No metal liner at all — the carbon shell is the vessel — which is as much propellant as this volume will ever give up: two fifths more than standard, and nothing in it to fatigue. The best tank in the catalogue on both figures its card prints, and priced accordingly. Buy it for the satellite you intend to keep.
The standard wing, rigid triple-junction cells, sized so the sunlit arc refills the battery with margin in every orbit the catalogue offers. Swapping it in flight moves total collection by single digits; specifying a lighter one at launch moves range by about a quarter, because that is where mass counts.
A thin-film blanket that unrolls off a drum instead of unfolding on hinges, so there is nothing in it to jam and nothing rigid to carry — the lightest step up in the slot, and on the pad that difference goes into the tank. About a quarter more power per square metre than the standard wing, and it ages slower for having no hinge line.
Mirrored troughs fold sunlight onto a narrow strip of cells, for the most power per square metre in the catalogue out of the least cell area — so it is also the lightest wing here, and a launch that specifies it carries the difference as propellant. The strip sits behind its own optics, which is the best radiation shielding on any wing in the catalogue. The best wing in the game on all three figures, and priced like it.
The standard fit: a gimballed dish that holds a narrow beam on a ground station out to the navigation belt — every orbit the catalogue offers except the highest one. From geostationary it is reaching further than its gain will carry and collection runs ahead of the downlink. The gimbal is the only mechanism on the bus, and mechanisms are what wear out. Reading another satellite’s technology across a link this narrow is five days of sitting alongside it.
A dish and a pair of omnidirectional whips behind one amplifier. The whips hold a link whatever the satellite is pointing at, so nothing is lost while it slews, and the amplifier carries the dish half again as far as the standard fit. Lighter than the standard fit and slower to wear, because less of it has to move, and it halves the time it takes to read another satellite’s technology across.
Steers its beam electronically, so it has nothing in it that moves — the longest-lived fit in the catalogue — and it is the only antenna here that closes a link from geostationary at full rate. A flat panel carries no reflector and no gimbal, so it is lighter than the dish as well, and on the pad that comes back as propellant. It reads another satellite’s technology across in a day and a half where the standard dish takes five. The best antenna in the game, and priced like it.
A directed pulse that burns out what it hits, from any range a satellite can follow at. It carries no ammunition, but every shot takes a third of the battery and it fires only at half charge or more, and the battery refills only in sunlight: two shots from full, then a wait for the sunny side of the globe.
The second-generation emitter: a harder pulse for less power. It draws less standing by than the EW-7, takes a quarter of the battery a shot rather than a third, fires faster, and hits harder.
Fires a tungsten slug at several kilometres a second from medium range or closer. Twelve rounds, a second between them, and a small draw on the battery for each. A refit restocks the magazine.
Ten mines, fired backwards while you follow a target. A mine runs the orbit the wrong way round the globe in minutes and the target flies into it head-on, unless it has changed orbit since. Anything on the track behind you runs into it first.
Shielding against an electromagnetic pulse that would otherwise put a subsystem out. Some of what is up there fires back, and this is what will decide which parts of a satellite come through it still working.
Every module has a price, in game credits and never in real money, and it is priced on the figure its card leads with — the telescope on what it reads, the wing on its output, the tank on what it holds — so the dearer part is the better one and the price is the only gate once it has been researched. What you buy is still the mission: a launch with these bays specified, or a servicing flight that swaps them, with each bay's share of that flight printed on its card and the whole bill under them. See progression. A kill pays nothing: no credits and no research. What a weapon buys is the target gone, or its recorder silenced. The defence bays are shown empty on purpose until the hardware behind them exists — an empty slot with a reason is honest, and an empty slot with a price on it would be a shop again. Everything marked Coming soon is next.
Written the way it is tracked internally, including the parts that are cut. A roadmap is only useful if it is falsifiable, so this one carries dates and a number you could check.
The game is built. Three screens — WORLD, CMD and FLEET — and the whole loop closed: find a site by zooming, target it, price the route, watch the move fly, collect, sell, and find all of it intact after a restart. The sun is in the model, so the terminator moves across the globe and optics are gated by it. The cadence planner prices a repeat ground track in propellant before you commit. Credits buy launches and refits; all seven open slots can be specified on the pad or swapped by a servicing flight that flies out and berths, and every upgrade is researched from the intel the fleet collects; a geostationary satellite is bought over a named region and can be walked anywhere on the globe; a satellite worn to nothing falls back to Earth after a telegraphed window and can be rebuilt; and the milestone tracks record what the campaign has covered. A campaign starts with one satellite, four designs are for sale across all three hull classes, and a reference manual generated from the game's own tables — never written out by hand, so it cannot end up teaching last month's rules — explains every module the fleet has not flown yet.
Since launch the sky has filled in. Foreign satellites can be flown past, shadowed at three ranges, robbed of their data and their technology, and shot at with mines, a railgun and two EMPs. Among them are objects no registry lists, and craft that land for a few days: they can be photographed, tracked to where they come from, recovered by a ground team, and chased in flight by a craft built from what the teams bring back.
Single player, no accounts, no backend: your state lives on your device and nowhere else. The game logic carries over 1,700 passing unit tests and no test framework. v1.0 shipped with nothing monetisation-shaped in it; version 1.0.15 added the one optional purchase — packs of in-game hours that fast-forward the world clock — and nothing else is sold for real money.
It has been through the device soak and App Review, and it is live on the App Store and Google Play.
The defence and hardening bays, and contacts that react to being watched. Offence shipped first; this is the other half of the fight, for the objects that already fire back.
A persistent shared sky, an insurance economy and player-versus-player engagements. Only if v1.0 finds an audience. That version is a company rather than a feature, and it will not be started on optimism.
Every burn is permanent. Choose the orbit you can live with.