Orbit is not one place
It is a stack of floors with radically different physics — from the band where crews sleep to shells that keep their mistakes for centuries. Reaching anything above means crossing every floor below.
The column beside you is to scale. Keep scrolling.
Tap any card's title ▾ to open its story; tap it again to tuck it away.
Tap the sky to park a satellite — it pins the altitude and names the floor you're in.
Watch the 30-kilometer band
6,170 tracked objects · 5,698 Starlink's
Almost too thin to see at true scale — and it holds 6,170 tracked objects, 5,698 of them SpaceX's Starlink. Everything bound for higher orbit crosses it.
a snapshot from the study text
The keys to orbit were never issued by any treaty — no law anywhere requires anyone to coordinate with SpaceX. They were minted by occupation and they are enforced by collision risk: the shells sit across the ascent path, Starlink's maneuvers invalidate everyone else's orbit predictions for days, and ignoring SpaceX prices itself. The keys are structural, not personal — whoever occupied the corridor would hold them — but the corridor is occupied, and by one company.
One launch moved the sky
~4,400 satellites lowered, 550 → 480 km
On 12 December 2025, an uncoordinated satellite passed within 200 meters of a Starlink at 560 km. Within three weeks, the operator announced it would lower ~4,400 satellites from 550 to 480 km — and by this catalog's epoch the move is nearly complete: at the 1-km slot scale, the corridor packs nearly eight times its old peak density; averaged over the full 30-km shell, about twice.
No rule was broken. No venue existed. The episode was litigated entirely by press statement — and the geography of low-Earth orbit changed.
a snapshot from the study text
On 9 December 2025 a Chinese Kinetica-1 launch deployed nine satellites without — SpaceX says — coordination; on 12 December one passed within 200 meters of Starlink-6079 at 560 km. Chinese researchers counter that trajectory data arrived ~14 minutes before closest approach. Within three weeks, SpaceX announced it would lower ~4,400 satellites from 550 to 480 km, explicitly citing safety reconfiguration. The episode is also this study's natural experiment in law: an uncoordinated approach to 200 meters violated no enforceable obligation, triggered no venue, and was litigated entirely by press statement. If the compulsion were legal, that event would have had a legal life. It had none.
A tight view, with illustrative dots — zoomed to 400–640 km here; the full column returns as you scroll on.
A tight view, with illustrative dots: the column zooms to 400–640 km for this beat, then the full stack returns as you scroll on.
A fire below, a ratchet above
the no-dodging gauge: 164 days in 2018 → 2.5 days now
The stress gauge: if every satellite stopped dodging, the expected time to a possible collision was 164 days in 2018 — 2.5 days by May 2026. Our reproduction, run on this catalog: a median wait of about four days. A gauge, not a prediction.
And the danger has two addresses. A first collision almost certainly starts in the thin pulsing band below — the ~480 km internet corridor, where the crowding and that 2.5-day clock both live — a fire that burns hot and burns out. The wreckage that stays gathers in the wide band above, the 700–1,000 km ratchet band, where the air removes nothing on any human timescale — and what settles there keeps falling through every floor beneath, for decades.
a snapshot from the study text
Decomposing the no-maneuver rate by altitude: under either treatment, the probability of a first collision lives overwhelmingly at the 480 shell (92% / 81% of the total rate; essentially all satellite-on-satellite), while the legacy debris band at 700–1,000 km carries ~1% of the trigger rate — but all the persistence. At 465–495 km, a fuel-limited transient cascade: a genuine runaway chain while the intact fleet lasts — fast, self-extinguishing, its fragments flushed by drag within months to years. At 700–1,000 km, where drag removes nothing on policy timescales, crossing the production-exceeds-removal threshold is an effectively irreversible ratchet: reversal requires actively hauling debris out, not merely stopping. Two dangers, two altitudes, and only one of them is forever. Most public discussion blurs them into one word: "Kessler." The no-dodging clocks are the Outer Space Institute's published figures; the altitude decomposition is the seeded engine's.
The calm is manufactured
≈800 dodges a day · 2,113 satellites cannot dodge at all
That clean record is real — and it is in the same ~480 km corridor the 2.5-day clock is about: ≈800 collision dodges a day, one every ≈1.8 minutes around the clock, at a maneuver threshold ten times more conservative than the industry standard. The dodges are what keep that fire from lighting. The strongest reading: not gatekeeping, but SpaceX carrying the commons' safety load.
The study's sharper engine: the stewardship is real and incentive-explained. The fleet itself is the bond — 10,736 satellites of collateral against its own carelessness. The clean record is structure, not virtue — and it cuts both ways: the same structure explains why nothing requires the dodging to continue.
a snapshot from the study text
While the fleet is in the band, SpaceX is what economists call a privileged group — its private benefit alone pays for the public good, so the safety needs no mechanism, no altruism, and no oversight to be supplied. The fleet itself is a bond posted in kind: 10,736 satellites of collateral against its own carelessness. The clean record is structure, not virtue — which is the strongest version of the counter-case, and it cuts both ways: the same structure that explains the record explains why it is unsecured. The arrangement is incentive-compatible exactly as long as the fleet is worth protecting; the shutdown scenario is what the incentive shock looks like. The ≈800/day figure is the SpaceX FCC report's own count — 148,696 maneuvers over Jun–Nov 2025.
The wall every filer has to clear
claims ~110 m² of radiator · needs ≈900 m² at flown density
The deepest zoom on the page — a ≈300 m view of the crewed band; meters are the ruler, and the glow is illustrative, set by implied W/m².
A data center in orbit must radiate away every watt it draws — no air, no water, only infrared glow into vacuum. Radiator area is the textbook bottleneck — and it is exactly the number the biggest proposals leave blank: three of the four big filings say nothing about cooling at all.
What do they claim to run? Orbital Compute filed a 100 kW bus; SpaceX's million-satellite ODC claims 150 kW — revealed only after the public comment window closed; Blue Origin stated no power at all. At 100 kW the implied radiator sheds ≈1,000 W/m² — six times the ISS's flight-proven 166 W/m² — and the 150 kW figures point nearer eight times. And these are per satellite: the SpaceX queue holds up to a million. The sky beside you draws it, to one scale: the ISS itself (109 m end-to-end), a filed compute bus at its own stated span (~50–70 m — no toy), the real ≈420 m² radiator field, the ~110 m² the 150 kW filing claims, and the field it actually needs at flown density. Set the dial — choose the bus's compute power and watch that field draw itself. And IF the business case were real? The study swept every possible world: ~69% stay paper, ~17% build and stay under the physical threshold, ~14% build straight past it — the race either fizzles or overshoots, and nothing in it is calibrated to stop at the line.
a snapshot from the study text
Where a figure does exist it describes a radiator nobody has built or flown: not impossible — a deliberately hot-running chip could get partway there — but unproven at scale, and the study tags it that way throughout. The claim is narrow: not "it can't be done," but that the filings ask to skip the queue before showing they can clear the one wall the physics puts up. The million-satellite filing omitted thermal design entirely; its numbers surfaced months later, in an investor-facing reveal after the comment window had closed. The full deployment-race map — every possible world, the thin middle, the operators' own filings pinned against our bands — lives in the reading edition. Craft sizes are sourced — the ISS from NASA, the compute bus from its own ODAR filing; their placement within the band, and the glow, are illustrative.
The sky already rains — rock, and now metal
today: 0.16× nature's rain · IF the US tier flew: 9.4×
Everything launched eventually returns — and a satellite does not land, it burns, leaving its metal high in the atmosphere. The sky has always rained: nature delivers ≈15,000 tonnes of meteoric rock a year (estimates span 2,000–110,000). Today our entire returning fleet is 2,437 t/yr — 0.16× nature's central delivery.
The sharper — and measured — finding: for aluminum, lithium, copper and lead, reentry mass already exceeds the meteoric source at today's fleet size, and reentry metals now turn up in roughly one in ten stratospheric particles sampled by aircraft. And IF the filing tiers flew (three loud IFs), the bulk rain itself climbs: the first milestone tier reaches 1.9× nature's central; the US 50% tier, 9.4× — its middle estimate above nature's entire range.
a snapshot from the study text
By sheer tonnage, today's whole fleet is a rounding error against nature. But bulk mass hides the composition: the tonnage is small; the mix of metals is not what the upper atmosphere evolved with. The milestone rows stand on three loud IFs — IF every filing were granted, IF each honored its deployment milestone, and IF that stock flew at trigger-band-like shells on 5-year lives. Column arithmetic under the machinery's own rules — not a band forecast, and never a date. Filings are options, not fleets. Nature's rain is sourced (Rojas, Duprat, et al. 2021); the fleet brackets are the seeded engine's (seed 20260717); the metal-excess and one-in-ten findings are measured from aircraft sampling.
One collision, three directions
crewed-band load, the event's own share: ×2.7 year 1 → ×1.02 year 10
If the corridor's cascade ever runs, a first collision throws its wreckage three ways at once. Drag the clock and watch it. Sideways: the fragments smear along the shell into a closed ring within days. Down: two-thirds of them dip through the crewed band every orbit from the first hour — but the air there digests the inheritance fast.
Up: about a third arc into the 700–1,000 km ratchet band — and this is the part everyone gets wrong. They visit; they do not stay. Every fragment's orbit still passes through the collision point, so its low point stays in the corridor's air, which pulls it back down over months to years. With ~99% probability a corridor event leaves nothing permanent above 700 km. The ring never even completes its torus at this altitude: the air drains it as fast as it fans.
a snapshot from the study text
The danger has two clocks and the beat keeps them apart. The crewed band's inheritance is the fast, self-limiting one: paired against the no-event world draw for draw, the lethal-debris load runs about 2.7× the counterfactual in year 1 and washes to 1.02× by year 10 — the corridor's air digests it. (Shown against today, the trigger world reads 6.2× / 7.2× / 9.4× at years 1 / 5 / 10, but its no-event baseline is already 2.2× / 6.5× / 9.2× — most of that ladder is ordinary growth, not the event.) The ratchet band's is the slow, effectively irreversible one — a century clock — but this event almost never reaches it. Two directions, two fates, and only one of them is forever. Years here are time elapsed after a hypothetical event, never a date. The measured proof that source altitude sets that clock: 3 Cosmos-1408 fragments still carry orbits ~5 years after a 485-km breakup, versus 1,912 Fengyun-1C fragments 19 years after an 865-km one (July 2026 catalog, MEASURED). The fragment dots are a Keplerian impression — right in shape, not tracked orbits.
Play the fifty years
walk away: ≈73,000 in the corridor at year 5 · steer it down: ≈2,400
Every fleet eventually stops — commerce, sanction, or time. The engine plays both endings of the same shutdown, on its own published draws. Walk away, and by year five the corridor holds ≈73,000 tracked fragments. Steer the fleet down on the way out, and the same shutdown leaves ≈2,400 — thirty times less, from tidiness alone. Step through the years and switch endings.
Then watch the wide band above. The corridor's fire is fuel-limited — even walked-away wreckage mostly washes out by year 50, because the corridor sits in air. The 700–1,000 km ratchet band is the opposite: it only gathers, under either ending — what settles there is the part no ending gives back. That fork, priced, is why the next beat's exit-door levers exist.
a snapshot from the study text
The study prices the fork rather than forecasting it: paired on the same random numbers, walking away leaves 73,113 tracked trigger-band fragments at year five against 2,407 for a responsible wind-down — thirty times the debris, from institutional tidiness alone. The view here is the seeded engine's five reported years (0 / 5 / 10 / 25 / 50; kessler_qb_engine.py v1.2, seed 20260716, N=100k), drawn as the median with the [p5–p95] band in the readout. Disposal compliance is ASSUMED 0.85–0.97; magnitudes are f_imp-conditional; years are time elapsed after a hypothetical shutdown — never a date. Dots draw the median at 1 dot ≈ 100 objects, capped for legibility (the cap is disclosed in the readout when active); no individual collision is ever depicted — the engine models populations, not events.
Eyes and manners — no ledger, no will
in force: monitoring and norms · billing, bonding, continuity: zero anywhere
Every fix has an address on the column you just climbed. A cap for the corridor you watched pack to eight times its old density. A disposal bond at the exit the continuity gap named — decisive in the model: disposal jumps to nearly 100% the moment the bond matches the ≈$71k it costs to deorbit a satellite. Removal for the ratchet band the wreckage only ever visits. Tap a marker in the sky for its source.
Now look at what is actually in force: monitoring and norms. Eyes and manners. Everything that would price the crowding, bond the exit, or guarantee a failed operator's fleet still gets flown down is modeled only — zero implementations anywhere, and no debris object ever removed. The fixes aren't fictional; several are cheap, one is decisive on paper. Each just needs someone with the authority to require it — and the treaty architecture gave that job to no one.
a snapshot from the study text
The study ran each lever through the model, and they turned out not to overlap. The deployment fee moves where satellites are allowed to fly — and nothing else. The holding price moves how much filed paper ever becomes hardware — and nothing else. The disposal bond moves the end-of-life outcome completely, once it is sized right — but barely touches deployment. No single instrument reaches both games at once. That is why the honest answer is not one clever rule but several levers, each at its own address — which is exactly what the coverage matrix shows is required.