Terminus: constellation explorer

Explore the constellation

Everything in the Terminus proposal comes together here: the tidally locked planet with its permanent day and night faces, the twilight band where the civilization lives, the six polar rings of the access wheel, and the MEO anchors that hold the minds. Drag to turn the view, scroll to zoom, right-drag to pan; the Framing buttons jump between the planet, the access wheel, and the whole system — the MEO shell orbits four planet radii out, more than three times the wheel, so it fits in frame only from the last of those. Pole view takes you straight up over either pole, where every ring flattens to a diameter through the planet and the wheel looks the way the duty plate draws it. Change altitude and ring count and watch the coverage numbers move; scrub through a planet rotation to see the orbital planes drift against the fixed terminator (the elegant trap); thin the MEO shell and watch the navigation fix fail.

The sky is not decoration. Everything outside this system is fixed in space while the planet turns beneath it, so the stars sweep around the spin axis once every 11.2 days — the same rate, in the same direction, as the orbital planes drift. The bright pair off to one side is Alpha Centauri A and B, the pair this red dwarf is bound to, 13,000 astronomical units away. From the twilight band they would burn at magnitude −6.7 and −5.3: five times and one and a half times Venus at its best, the brightest things in the sky after the star itself, and still 200 times too faint to cast a shadow. Their true separation is about six arcminutes — a fifth of a full Moon — so they are drawn farther apart than that to keep both visible, in the same spirit as the star's deliberately oversized disc. Skip through Plane drift and watch them go round with the rings: the wheel is pinned to those stars, not to the ground.

Both exaggerations are measured, and you can undo them: switch Star scale to true, and Proxima retreats to its real 1,139 planet radii and real 1.69° — you have to turn the planet's night side toward you to see it at all, and from anywhere else the planet is in the way — while the twins close to their real six arcminutes and merge into a single point. That is what the sky is actually like: mostly empty, with everything interesting either enormous and far away or brilliant and unresolvable. The cinematic default draws Proxima 7.2 times wider than life and 81 times too close, and the twins 24 times too far apart. Everything else in the model — shell altitudes, footprint sizes, the band, the drift, the handover rhythm — is true in either mode.

Nothing in the model waits for you. Ring handovers are 22.4 hours apart, and anchor switches are hours apart. Hence, the console's skip-to-a-moment buttons jump the clock to just before each one rather than winding the speed up: plane drift, coverage, a satellite handover, a ring handover, an anchor switch. The Duty ring strip under the viewport reads out which plane is straddling the terminator right now — two bars contend, they cross at half height, and that crossing is the handover you also see in amber in the scene. The pale column inside each bar is a second, separate quantity: how many of that ring's satellites are lit at this instant. Watch it during a shift, and you will see the duty ring never carrying the band by itself. Coverage is drawn for the duty ring alone by default, so you can see one ring's footprints sitting over the green habitable band; switch it to all rings to see the overlap the whole wheel provides.

The rings are not filed in step with one another. Ring phasing starts on random — each ring's satellites carry an independent, arbitrary offset, because no launch campaign can aim one ring's phase against its neighbors'. That sounds like it should matter, and it does not: switch to aligned, where every ring moves in lockstep, or to half slot, where neighboring rings interleave, and the min visible readout stays at one, and the mean barely stirs from 3.45. Half slot is the instructive failure. Offsetting neighbors by half a slot is how a cellular network tiles a plane with the fewest cells, so the instinct says it should buy satellites — but a polar wheel is not a plane. Every ring pinches to nothing over the poles, each one serves the band at two antipodal longitudes at once, and the wheel closes through a seam where neighboring rings run in opposite directions. No fixed offset survives that, and at this baseline the honeycomb costs coverage instead of buying it. What the fleet gets in exchange is worth more than the satellites it would have saved: it can ignore inter-ring phase entirely.

Turn on Follow a town, and the model draws a whole conversation: the town, the pale link up to whichever access satellite carries it, and the longer line onward to the anchor holding its session. The short link changes every 11 minutes; the long one, at the default Re-anchor margin, roughly every 113.

That control is the one to play with. Nothing in the sky forces a session to change anchor — a ring can see every anchor at every instant — so how often a mind moves is a policy, and the margin sets it. Pull it to zero, and the session re-anchors constantly, chasing the shortest path; each of those moves drags gigabytes of working memory across the sky, which at a million terminals is more backbone than a hundred-gigabit link can carry. Push it to 25k, and the mind stays put for the whole scrub — on a path so long that the round trip spends 290 of the 300 milliseconds the model was given to answer in. The proposal takes neither end, and why is the backbone's to explain.

Watch the anchor line as you do it. Where a satellite is not the best door out of its own ring, the route steps sideways along the necklace before it climbs — but at the default margin you will not catch it doing so, because a session free to re-anchor always finds an anchor its own satellite can reach. Widen the margin to 20k or 25k, pin the session to a distant anchor, and the sidestep appears: one hop, sometimes two or three. A hop moves exactly one place, because that is where the laser points, so six would cross a ring of twelve.

Not every satellite is radiating. The scene runs the same activation plan the proposal adopts — the duty ring lit as a block, then satellites from other rings switched on one at a time only where the band would otherwise go unserved, then a prune pass switching off whatever the later picks made redundant — so a bright dot is a satellite currently carrying traffic, and a dark one is a satellite coasting between shifts. Typically two-thirds of the seventy-two are dark. Watch the pattern through a ring handover: the incoming ring lights up as a whole while the outgoing one goes dark a few satellites at a time, keeping the towns beneath covered the entire way across.

The Feeder telescope row is the one place the model breaks something. Pooling left each anchor a single telescope into each ring, so Break does not degrade the followed session's link — it severs that (ring, anchor) pair outright, and the Session readout beside the console names what the session is doing about it and what the round trip costs while it does. The other three buttons decide what happens next, and they are not three independent remedies. Plane links are the frozen 37,294-km hops joining an anchor to its neighbors in the same MEO plane: with them the disabled anchor is still reachable through a plane mate, so the conversation goes on being answered; turn them off, and the pair is unreachable, and every session on it goes quiet until it has moved. Spare fits the seventh telescope — cold, steerable, repointed at whichever ring went dark — and Hold-off is the policy rule that stops the anchor from being re-evaluated while it slews. Switch off either one, and the session leaves within a frame, because the detour is worth 6.3 times the re-anchor margin — and the two switches fail differently: without the hold-off, the spare finishes locking onto a ring with nobody left on it, and without the spare, there is no telescope to wait for, so the rule asks the conversation to sit on the detour for nothing. Leave both on, and the session does not move at all. One admitted exaggeration there, in the spirit of the star scale: the acquisition is 5 seconds of spacecraft time, which even at the slowest speed on this dial would pass in 83 milliseconds — about five frames — and inside a single frame at every faster setting, so the scene holds it for about four seconds of wall clock and says so in the log. The backbone argues all four.

The model shows the full architecture at once: the planet's fixed day and night faces and the ±20° twilight band; the six polar rings of 12 satellites each at 2,200 km; and the MEO shell at 20,000 km — six inclined planes of four spacecraft. At the baseline design, every band point sees at least one access satellite (mean ≈ 3.45), each footprint is about 2,500 km in radius, and a satellite is usable for at most 16.6 minutes on an overhead pass — though a town hands over sooner than that, every 11 minutes, when the next satellite in the ring climbs higher than the one it has.

Those 24 MEO spacecraft do double duty. They hold the minds, and the timing fabric that disciplines them is what the RFP's navigation service will be built on — which is why the shell is sized like a navigation constellation rather than like a handful of anchors. Pull the MEO planes and MEO sats/plane sliders down, and the MEO visible readout drops below the RFP's floor of four satellites visible always: a thinner shell still anchors sessions perfectly well, but it can no longer fix a position.

Every number here is computed the same way as the series' reproducible terminus simulator runs.

When the shapes are familiar, the proposal picks up where it left off. Every link drawn here — slate to terminal, terminal to wheel, wheel to anchor — has to survive the glare of a flaring red star, and that is the section that buys the bands to do it.