SpaceX. On the Other Hand, the Path of Humanity in a Positive Environment

On 28 September 2026, on its fourteenth test flight, Starship reached orbit for the first time and deployed 26 Starlink V3 satellites. The flight was not a spectacle. It was a delivery.

That is the whole significance of the moment, and it is easy to miss. For sixty years, reaching orbit was the achievement. Here it was the precondition for doing something ordinary: putting cargo in a place and coming home. What follows is an account of the machinery that makes that possible, of the one piece that is still missing, and of what the complete machine implies.

The engine

Everything in this programme descends from a decision about combustion.

Raptor 3 is a full-flow staged-combustion engine burning liquid methane and liquid oxygen. In a staged-combustion cycle, propellant is burned in a preburner to drive the turbopumps, and the exhaust from that preburner is then fed into the main chamber rather than thrown overboard. Full-flow means there are two preburners, one running oxygen-rich and one fuel-rich, and that the entire flow of both propellants passes through them. Nothing liquid enters the main chamber; both streams arrive as hot gas.

The benefits are specific. Splitting the flow lets each turbopump be driven by a gas that does not attack it, which lowers turbine inlet temperatures for a given power and extends engine life, the property that reuse depends on. Gas-to-gas injection mixes far better than gas-to-liquid, which raises combustion efficiency. And the cycle can sustain chamber pressures that other cycles cannot.

Raptor 3 runs at 350 bar, up from 300 in Raptor 2. That is about 345 times atmospheric pressure, the pressure found roughly three and a half kilometres beneath the sea, held inside a chamber the size of a bucket while the contents burn at around 3,500 kelvin. Sea-level thrust is 280 tonnes-force and specific impulse about 350 seconds, from an engine that weighs 1,525 kilograms. The thrust-to-weight ratio implied by those two numbers is the reason the vehicle can be as large as it is.

The choice of propellant is equally deliberate. Methane sits between the two conventional options and is better than both for this vehicle. Kerosene is dense and easy to handle but leaves carbon deposits inside the engine when it burns, and coking is precisely what makes an engine need stripping down between flights. Hydrogen burns cleanly and gives the highest specific impulse of any practical fuel, but it is so light that tanks become enormous, and so cold that it leaks through materials that hold everything else. Methane is clean enough to reuse, dense enough for sensible tanks, cheap, and it can be manufactured out of carbon dioxide and water by a chemical reaction worked out in the nineteenth century. That last property is the one that explains the destination: an atmosphere of carbon dioxide and a supply of ice are the two things Mars is known to have.

No full-flow staged-combustion engine had ever flown before this one. The Soviet RD-270 was tested on the ground in the 1960s and never launched; an American demonstrator in the 2000s never left the test stand. The cycle was understood for half a century and considered too difficult to be worth it.

The third version is defined less by performance than by subtraction. Components that were bolted on the outside have been moved inside the engine, which removes the need for a separate engine heat shield; the aft end of the vehicle was redesigned from a clean sheet so that individual engine shrouds and the large close-out volume behind them disappear, taking roughly a tonne of mass out per engine. By April 2026 SpaceX had built more than 600 Raptors and accumulated over 40,000 seconds of run time on the third version alone. Rocket engines have historically been artisanal objects produced in dozens. This is a production line.

The vehicle

Starship V3 stands about 124 metres and leaves the pad on about 90,600 kilonewtons of thrust, carrying 100 to 150 tonnes to low Earth orbit. Both stages use Raptor 3.

It is made of stainless steel, which was the most contested decision in the programme and now looks like the most important one. Carbon fibre is lighter and is what the industry uses. Steel gets stronger as it gets colder, which suits tanks full of cryogenic propellant; it keeps useful strength at temperatures where composites are already gone, which matters on the way back down; it can be welded in the open air by people who weld water towers; and it costs a small fraction as much. The vehicle is heavier than it would otherwise be, and in exchange it is cheap and it survives heat. For a machine intended to fly many times, the second property is worth more than the first.

The Super Heavy booster lost a grid fin in this version, going from four to three, each half again as large and considerably stronger. Grid fins steer the booster back; they also take the load when the launch tower catches it. Catching rather than landing is itself an engineering argument: landing legs are mass carried up and back on every single flight, and a tower is mass that stays on the ground. Returning the booster to the structure it launched from is also what makes the turnaround short.

What has been proven, and when

The programme's method is visible in the last three flights, and so is its honesty about failure.

Flight 12, on 22 May 2026, was the debut of the third version. The ship did what it was asked to do: a full ascent, a controlled reentry and an on-target splashdown in the Indian Ocean. The booster did not. During hot staging it was rotated the wrong way by about ninety degrees, several Raptors suffered energetic failures, the boost-back burn was cut short, and the stage was lost over the Gulf. The Federal Aviation Administration required an investigation before flights resumed and closed it with two probable causes: heat affecting propulsion components during ascent, and incorrect settings in the engine alarm system. Four corrective actions followed.

Flight 13, in July, was the first in which the whole sequence held together. Thirty-three engines lit, hot staging worked, the booster completed its boost-back, the ship burned all six of its engines to the planned velocity, deployed twenty Starlink V3 satellites, reentered, flipped, and splashed down intact. That last word is the one that mattered. Previous vehicles had broken up on the water; this one floated, kept transmitting telemetry and video after splashdown, and was recovered and returned to Starbase, which meant that for the first time engineers could examine a heat shield that had actually flown.

Flight 14, on 28 September, closed the loop: orbit for the first time, twenty-six satellites released, and two tiles from the previous ship flying a second time.

Three flights, four months, one lost booster and a steadily shortening list of unknowns. The approach is to fly hardware before it is certain, lose some of it, and convert the wreckage into a corrective action. It is expensive and it is fast, and it is the opposite of how launch vehicles were developed for the previous fifty years.

The shield

The hard part is not going up. It is coming back without needing a month of repairs.

Reentry from low orbit means shedding roughly 30 megajoules per kilogram of kinetic energy, nearly all of it into heat. Starship's answer is a shield of about 20,000 ceramic tiles. Tiles work, and they are also the programme's weak point, because they are brittle, individually bonded, and inspected one at a time. The Space Shuttle's thermal protection is the historical warning: it was the component that turned a reusable vehicle into one that needed months between flights.

The milestone on the September flight was accordingly small and precise. Two tiles recovered from the previous vehicle, which had spent 24 days floating in the Indian Ocean after splashdown, were reflown. Two out of twenty thousand is not a solved problem; it is the first evidence that the material can go twice. Until the shield becomes a check rather than a rebuild, nothing else about the vehicle's economics is real.

The rest of the recovery chain is further along. SpaceX has stated that the previous ship's splashdown in July was precise enough that the tower could have caught it, and booster catches and reflights have already demonstrated the method on the first stage.

The proof that the model works, when the thermal protection is simple enough, is sitting on the other pad. On 25 August 2026, Falcon 9 booster B1067 flew for the thirty-seventh time, on the hundredth Falcon 9 launch of that year. Six boosters in the fleet have flown at least thirty times. The Space Shuttle's record for an individual orbiter was thirty-nine flights, accumulated over three decades; a single Falcon first stage is approaching it in under six years. Reuse is no longer a theory under test. It is an operating practice, waiting for a heat shield that can keep up.

The depot

And here is the piece that does not yet exist.

A fully fuelled Starship can reach low Earth orbit with 100 tonnes. It cannot reach the Moon or Mars from there, because it arrives in orbit nearly empty. The architecture's answer is to refuel in space: a tanker vehicle launches, rendezvouses, docks and transfers liquid methane and liquid oxygen to a waiting ship, as many times as it takes.

Nothing about that is routine. Cryogenic propellant in microgravity does not sit at the bottom of a tank, because there is no bottom; the liquid has to be settled with gentle thrust or managed with internal hardware before it can be pumped. Both fluids boil continuously, so every hour in orbit costs propellant and every transfer is a race against the heat leaking in. Methane sits at about 112 kelvin and oxygen at about 90, which means the plumbing, the seals and the docking interface all have to work while being cold enough to make steel brittle. And the quantity is not symbolic: filling a ship for a lunar or Martian departure means moving hundreds of tonnes, across several tanker flights, with the target vehicle loitering between them.

Conventional propellants have been transferred in orbit since the 1970s. Cryogenic propellants have never been transferred between two independent spacecraft. The demonstration was first scheduled for March 2025, moved to March 2026, and as of the fourteenth flight in September 2026 had not been attempted.

There is a second reason the depot decides everything, and it is orbital mechanics rather than engineering. Earth and Mars line up for an efficient transfer roughly once every twenty-six months. A crossing is not something a single vehicle does when it is ready; it is something a fleet does in a cluster, inside a window a few weeks wide, which means every ship in that fleet has to be fuelled in orbit within the same short period. The tanker rate, not the ship, sets the size of what can leave.

This is the keystone. Everything downstream of low Earth orbit, the lunar lander contracted for Artemis included, waits on it. It is also the reason the cadence question is not a vanity metric: an architecture that needs several tanker launches per departure only works if launches are frequent and cheap, which returns the argument to the heat shield.

What the rocket already carries

While the deep-space case waits on refuelling, the vehicle has a customer that needs nothing beyond low orbit: the network underneath it.

Starlink V3 satellites weigh roughly two tonnes each, several times their predecessors, and are built around capacity rather than coverage. Each is designed for about one terabit per second of downlink, roughly ten times the previous generation, and carries six laser terminals at 400 gigabits per second apiece. Those lasers are the part that changes the character of the system: satellites route traffic to each other optically, so data can cross the constellation and come down near its destination instead of having to reach a ground station from wherever it entered. The constellation becomes a mesh with its own interior, rather than a set of mirrors.

Twenty-six of them went up on the September flight, a load no Falcon 9 can carry. Later flights are planned at around sixty. Against roughly 9,839 active satellites today, the stated ambition is a constellation of up to 100,000.

This is the closed loop that distinguishes the programme from every previous launch business. The rocket's principal customer is the company's own network; the network's revenue pays for the rocket; and the rocket is the only vehicle that can deploy the network's current hardware at scale. In the quarter to 30 June, connectivity produced $4.29 billion of revenue and the artificial-intelligence segment $2.56 billion, against $962 million for the launch business. The launch company is now the smallest part of the launch company.

The next payload is computers

On 30 January 2026, SpaceX applied to the Federal Communications Commission for an Orbital Data Center System: up to one million solar-powered satellites between 500 and 2,000 kilometres, arranged in shells roughly 50 kilometres thick, linked optically, and optimised for artificial-intelligence inference rather than communications. The Space Bureau accepted the filing for review five days later. The company's own arithmetic is that launching a million tonnes a year yields on the order of 100 gigawatts of computing.

The engineering case rests on two physical facts. Above the atmosphere, in the right orbit, sunlight is uninterrupted and arrives at about 1.36 kilowatts per square metre, with no night, no weather and no grid connection to negotiate. And the thing a terrestrial data centre fights hardest for, electrical supply, is replaced by the thing a spacecraft fights hardest for, which is getting rid of heat.

In vacuum there is no air to carry warmth away. A satellite can only radiate, and radiated power rises with the fourth power of temperature and in direct proportion to surface area, which is why the demonstration article shown in June is mostly wings: solar arrays on one side, radiators on the other, with a computing platform between them averaging around 120 kilowatts. The same unit on the ground would be a single rack. The limiting resource in orbit is not power. It is area.

Whether a million of anything can be manufactured and launched is a separate question from whether the physics works, and the answer to the first is a factory in Bastrop, Texas and a launch rate that has to rise from roughly 2,500 tonnes a year towards a million within about three years. The company's stated near-term target is a weekly, then twice-weekly, Starship cadence.

The traffic

A sky with this much hardware in it has become an engineering problem in its own right, and the numbers are no longer small.

Between December 2025 and May 2026, Starlink satellites performed 207,152 collision avoidance manoeuvres, roughly sixty thousand more than in the preceding six months, for a running total above 355,000 in a year. That is an average of about thirty-six evasions per satellite per year, executed autonomously: the system moves whenever the modelled probability of a collision exceeds three in ten million. The constellation flies between about 480 and 550 kilometres, and SpaceX has begun lowering some 4,400 satellites towards the bottom of that band, where atmospheric drag removes a dead spacecraft in years rather than decades.

The statistical objection to that arrangement is straightforward and is not answered by the manoeuvres. A residual risk of one in a million, repeated across a million events, is not a small number any more. Low orbit is a shared resource with no traffic authority, and the margin is being managed by the operator that owns most of the traffic.

Astronomy has a separate complaint, and it is the one aimed directly at the orbital computing plan. A consortium of astronomers has formally challenged the million-satellite filing at the Federal Communications Commission, and European astronomers have asked the commission to refuse it, arguing that a sky containing that many bright objects would have devastating consequences for ground-based observation. The objection is not sentimental. Optical astronomy depends on long exposures of faint sources; a bright object crossing the field ruins the frame, and radio astronomy has an equivalent problem with downlink emissions.

It is worth being precise about what is actually at stake, because both things are true at once. The constellation has connected places no cable was ever going to reach, and a sky full of radiators is a sky that is harder to see through. There is no technical reason those have to be traded against each other, and no body with the authority to arrange the trade.

Where this is going

Put the pieces in order and the trajectory is legible without any speculation about Mars.

First, launch stops being an event. A vehicle that flies weekly is not a programme; it is a service with a timetable, and the question asked of it stops being whether it will work and becomes what it costs per tonne. Second, the thing being moved is mass, not people: freight, satellites, structures, and increasingly machinery that is meant to operate up there rather than pass through. Third, orbit becomes addressable territory, divided into shells and altitudes and allocated by a national regulator on a first-filed basis, which is a form of zoning for a place no state owns.

It is worth doing the arithmetic on that out loud, because it is rarely stated. A hundred tonnes a flight at a weekly cadence is about 5,200 tonnes a year; at twice weekly, 10,400. The orbital computing plan is denominated in a million tonnes a year. Closing that gap with a 100-tonne vehicle means something in the region of ten thousand flights a year, which is twenty-seven launches a day, every day. Larger payload versions reduce the number proportionally and do not change its character. Whatever else the plan is, it is a statement that launch has to become an industrial process measured in departures per hour, and the thing standing between here and that is not propulsion. It is the hours of labour between a vehicle landing and the same vehicle flying again.

Fourth, and this is the part that would have been unimaginable to the people who built the first stages of this industry: the first heavy industry to leave Earth will not be mining, or tourism, or manufacturing. It will be computation, because it is the only payload that needs nothing from the ground once it is up there. It wants sunlight, vacuum and a radiator.

And underneath all of it sits the same unglamorous dependency. None of this scales on a vehicle that needs weeks of inspection between flights, and none of it leaves low orbit without a cryogenic transfer that has never been performed. The era will be dated not from the first Starship to reach orbit, but from the first one to fly, land, be checked, and fly again the same week.

Time is the technology

Time will take humanity to other planets.

There is a ceiling over all of this, and it is not an engineering problem. It is chemistry.

How far a rocket can go is governed by one equation, and the term in it that matters is the speed at which the exhaust leaves the nozzle. That speed is set by how much energy is stored in a chemical bond and how light the products of the reaction are. Methane and oxygen give an effective exhaust velocity of roughly 3.4 kilometres per second; hydrogen and oxygen, the best chemistry available, about 4.4. No cleverness moves those numbers, because they are properties of matter rather than of design. Everything else in rocketry is an attempt to work around them: staging, so that empty tanks are discarded; reuse, so that the hardware is not thrown away with them; and refuelling, so that the vehicle can begin its real journey from a full tank already in orbit.

Better propulsion exists and has existed for a long time, and it is worth being precise about what each kind is for. Nuclear thermal engines, tested on the ground in the 1960s, roughly double or triple the efficiency of chemical propulsion by heating hydrogen with a reactor instead of a flame. Electric thrusters, which accelerate ions with electric and magnetic fields, reach specific impulses of thousands to tens of thousands of seconds, an order of magnitude beyond anything chemical. Neither is a replacement for what stands on a launch pad. A Hall thruster delivers thrust measured in hundreds of millinewtons, roughly the weight of a sheet of paper: an extraordinary engine for moving a large mass slowly through vacuum over months, and one that cannot lift itself off a table. These are tools for the part of the journey that happens after the hard part. The hard part, climbing out of a gravity well through an atmosphere, is still a job for brute thrust, and brute thrust is still chemistry.

So the variable that is actually open is not energy per kilogram. It is repetitions per year.

Mass delivered to orbit is payload multiplied by frequency, and frequency is the only one of those two that can grow by orders of magnitude. A vehicle that flies once a year is an expedition. The same vehicle flying weekly is a different category of object, not because it carries more but because everything learned on one flight is applied to the next within days. Cadence is how engineering knowledge compounds, and compounding is why the ordinary beats the brilliant over any long enough interval.

The calendar imposes its own rhythm on top of that. Earth and Mars align for an efficient crossing roughly once every twenty-six months. Progress towards another planet is therefore not continuous; it happens in discrete steps, and the real measure of a programme is how much better its hardware is at each window than it was at the last. What accumulates between windows is unglamorous: shorter turnaround, more precise landings, closed loops that lose less water, power systems that survive a dust season, propellant made on site rather than carried. None of that is a breakthrough. All of it is time, applied repeatedly.

This is how every transport revolution has actually happened. Crossing an ocean stopped being an expedition and became a timetable not because someone discovered a new principle of sailing, but because hulls, routes, insurance and schedules were repeated until the extraordinary became boring. Boring is the destination. Boring is what a species living on more than one world will require.

When a planet is a country

An interplanetary economy, where the planets are a country.

Set the engineering aside and consider what distance does to an economy. What follows is reasoning from physical constants rather than reporting, and it is offered as such.

The first constant is the speed of light. Mars is between about three and twenty-two light-minutes from Earth depending on where the two sit in their orbits, which means a question and its answer are separated by six to forty-four minutes. That single number removes an entire category of arrangement. There can be no real-time market between planets, no synchronous settlement, no live supervision of a machine or a person, no instruction that arrives before the situation it addresses has changed. An authority that depends on being consulted is not an authority at that distance. It is an adviser with a long delay.

The second constant is the window. For twenty-six months at a stretch, nothing physical arrives and nothing physical leaves. A settlement must be able to survive two years without a single import. That is not a political ambition, it is an engineering requirement, and it happens to be the oldest working definition of a country: a place that can keep its people alive by itself.

The third is the price of mass. Moving atoms between planets will remain enormously expensive long after it becomes routine, while moving information costs almost nothing and is limited only by light. The economic shape that falls out of those two facts is specific and unusual: planets will be close to autarkic in matter and tightly integrated in information. What crosses the gap is designs, models, software, scientific results, instructions for making things. What stays local is everything with mass. A planetary economy will import ideas and manufacture objects, which is the exact inverse of how colonies worked in every previous era, when the colony shipped raw matter home and received finished goods back.

That inversion is why the first industry on another planet is not mining. It is propellant. An atmosphere about ninety-five per cent carbon dioxide, plus water ice, plus energy, yields methane and oxygen through a reaction published in the nineteenth century. A settlement that can refuel a ship is no longer a destination; it is a port. Ports are where economies begin, because a ship that can be refuelled somewhere will go there for reasons that have nothing to do with the place itself.

Energy is the other half of the ledger and it is thinner out there. Sunlight at Mars delivers roughly 586 watts per square metre against about 1,361 at Earth, a little over forty per cent, before accounting for dust that can dim the sky for weeks. Every local economy will be priced in its own light, and the cost of anything made in a place will track the cost of the power that makes it. A currency is in the end a claim on energy; where energy is scarcer and more interruptible, the whole structure of value tilts towards storage, redundancy and the things that keep working when the sun does not.

The technology that survives the trip

Almost none of Earth's technology works on another planet, and the reason is not harshness. It is that our machines quietly assume a planet underneath them.

They assume air for cooling and for combustion; water that is cheap; a magnetic field overhead; spare parts in days; a supply chain of thousands of specialised vendors; and the freedom to throw away anything that breaks. Remove those and most of what is manufactured here stops being a product and becomes a liability. A device with one unobtainable component is scrap on arrival.

What replaces it is a different design philosophy, and its rules are already deducible. Build from the short list of what is locally available, which on Mars means carbon dioxide, water ice, iron-rich regolith, silicon and aluminium, while treating the perchlorate salts in the soil as a hazard to be managed. Prefer repairability to optimisation, because a machine that is ten per cent more efficient and cannot be fixed is worth less than a crude one that can. Close every loop, because anything vented is gone. Design for redundancy rather than throughput, because the failure of a single system is not an inconvenience but a fatality. And accept mass penalties in exchange for simplicity, which is the opposite of the instinct that governs aerospace here.

The human body imposes its own constraints, and these are measured rather than speculative. The radiation instrument aboard the Curiosity rover recorded an average of about 0.67 millisieverts a day on the Martian surface and about 1.8 millisieverts a day inside the spacecraft during the cruise, and from those measurements the team estimated roughly one sievert in total for a mission of a 180-day outbound flight, a 500-day stay and a 180-day return. Shielding against that is either mass carried from Earth or regolith piled overhead on arrival, which is why the first permanent structures will probably be buried rather than built. Martian gravity is 0.38 of Earth's, and here the honest statement is that nobody knows what that does to a human body over years, because no data exist anywhere between weightlessness and one gravity.

None of this is a reason the thing cannot be done. It is the specification. It explains why the technology of another world will not look like exported Earth technology, and why the first genuinely planetary machines will be designed there, by people living inside the constraints, rather than here.

Who makes anyone obey

How will they make a human being trillions of galaxies away respect their laws? Because it is basically an inadmissible operation, unless an entire intergalactic police force is created to cross the universe so that their laws are enforced.

Which brings the argument to the question usually waved away: if a settlement is two years from the nearest ship and forty minutes from the nearest answer, how does any law reach it?

The honest starting point is that enforcement, as normally understood, does not scale with distance at all. Law on Earth works because a state can reach a person faster than that person can get away. Every element of the system, courts, police, prisons, borders, assumes presence. Take presence away and the mechanism does not weaken gradually. It stops. A patrol sent across interplanetary distance arrives after the situation that provoked it has resolved itself, and one sent across interstellar distance arrives to find a society that has had centuries to become something else. Enforcement that arrives late is not enforcement. It is reprisal, and reprisal has never produced compliance.

The image of a fleet policing the void is therefore the wrong model, and not mainly because of cost. It is wrong because what it is meant to achieve cannot be achieved by that method at that distance, for the same reason a real-time market cannot exist across it.

What replaces the policeman

The useful question is not how to project force but what has historically made rules hold when force could not arrive. There are four mechanisms, all of them observable today, and each has a physical reason for working.

Dependency. A settlement that cannot make its own air, fuel and food complies for the same reason a tenant pays rent: the next ship is the sanction, and it carries no weapons. This is the strongest instrument that exists and it has a defined expiry date, which is the day local production closes the last loop. The moment a colony makes its own propellant, the relationship stops being compulsory and becomes commercial. In-situ production is therefore a legal event as much as an engineering one, and it will be the most consequential milestone in the whole question.

Chokepoints. Nobody patrols the ocean, and yet shipping is among the most regulated activities on Earth, because every vessel needs three things it cannot obtain at sea: a port, a flag and an insurer. The rules are applied at those three doors, not on the water. Space has the same structure. Launch needs a licence and a pad. Communication needs spectrum, which is allocated. Objects need registration, and jurisdiction follows the registry exactly as it follows a ship's flag. Payment needs a banking system. Specialised components come from a handful of suppliers. Anyone who intends to leave, to be heard, to be paid or to come back must pass through a door that is on this planet. That, and not a patrol, is where rules will bite.

It also means space inherits the characteristic pathology of maritime law, the flag of convenience. Where jurisdiction attaches to a registry, registries compete, and the most permissive wins the business. The eventual shape of the problem is not lawlessness. It is states selling the right to be governed by them.

Capital. Where adjudication is impossible, risk pricing replaces it. An operator that breaks the rules is not arrested; it is dropped by its insurer, loses its financing and finds counterparties unwilling to sign. In offshore drilling, aviation and shipping, the document that actually governs behaviour is the insurance contract rather than the statute, because the insurer is the only party with both the incentive and the information to inspect. Expect the same beyond Earth, with one feature that sharpens it: hardware at the bottom of another gravity well cannot be repossessed, so the lender's only security is the borrower's conduct, and it will be priced accordingly.

Repetition. Cooperation survives without an enforcer whenever the same parties expect to deal with each other again and defection is visible. Light-lag delays that calculation; it does not abolish it. A settlement with a handful of counterparties and a fifty-year horizon has overwhelming reason to keep its word, and in a small community reputation is a harder currency than it is on a crowded planet, not a softer one.

What falls out of all four is a prediction that is neither utopian nor dystopian but ordinary. Law beyond enforcement range is not imposed. It is generated locally and then negotiated between places. Ships' articles, mining-camp codes, the law merchant, frontier courts: every community that has ever found itself outside the reach of a state wrote its own rules within a generation, because people do not live without them and because the alternative is intolerable at close quarters. The realistic future is not one legal system extended outwards. It is a family of local systems, each sovereign in practice long before anyone says so, connected by treaties, standards and trade.

The existing framework is a starting point rather than an answer. The treaty governing these activities was written in 1967 for objects and for states, when the question was which government had launched a given satellite. It forbids any nation from claiming territory and it attaches jurisdiction to whoever registered the object. Seventy-six countries have since signed a set of non-binding principles about behaviour around the Moon. None of it was drafted with a resident population in mind, and none of it answers the first questions a settlement actually raises: who adjudicates a dispute, who owns what has been made on site, and what happens when the interests of the people there and of the registry that nominally governs them stop coinciding.

The environment adds a category of offence Earth has never had to take seriously. Where opening a door can kill everyone, acts against the commons, against air, water, power and pressure, become the gravest class, and the right to consume is necessarily conditional on the survival of the system that supplies it. Maritime law developed an entire doctrine of seaworthiness for precisely this reason. Something similar and stricter will grow around life support, and it will be written by the people breathing the air rather than by anyone at a distance.

Push the distance far enough and even that dissolves. At interstellar separations there is no shared present at all: by the time a message arrives, the society that sent it has moved on. What survives that gap is not commands but conventions, the kind of rule that holds because it is useful rather than because it is backed, which is to say protocols, interfaces, units and standards. Across sufficient distance the only law that enforces itself is physics, and everything else is an agreement that lasts exactly as long as both sides find it worth keeping.

The best case

And artificial intelligence, as a new race, will push technology towards the unlimited, bringing hope to the human race, contemplating only the best of cases, where it is pleasantly controlled by humanity.

What follows considers only the favourable branch, in which the tools stay under human direction. It is not a prediction, and the unfavourable branches are not addressed here.

The binding constraint in all of this has been named several times and it is always the same: how many good attempts fit between one window and the next. Design, test, fail, understand, redesign. That loop has a clock speed, and the clock speed of engineering has always been set by how long a competent person takes to learn something.

Machine intelligence changes that number without touching the laws it operates under. A search across design space that would occupy a department for a year can be run over an enormous number of candidates. Structures can be optimised against objectives a person would struggle to hold in mind at once. Materials can be proposed from a short list of locally available elements, which is exactly the problem a settlement has. Trajectories, control laws and closed-loop life support are precisely the kind of coupled, many-variable problems that reward a model. None of this repeals the rocket equation. It compresses the time between attempt and lesson, and time was always the thing in short supply.

At distance it stops being a convenience and becomes a requirement. A machine twenty-two light-minutes away cannot be driven from here; it has to decide. Autonomy in that setting is not an ambition but the only available control system, and its quality sets how much can be attempted before people arrive.

Controlled, in the version of this that goes well, has an operational meaning rather than a rhetorical one: people state the objective and the constraints and carry the consequences, machines search the space of solutions, and verification stays with the people who will live inside the result. That is not a new relationship. It is what every engineer already has with every simulation they have ever trusted, scaled up until the search is vast and the checking is the scarce part.

From the hand axe

It is worth ending on what this is a continuation of, because the line is longer and straighter than it looks.

A stone hand axe is a lever applied to muscle: it lets an arm do what an arm cannot. Fire is a lever applied to chemistry, predigesting food outside the body and later freeing metal from rock. Writing is a lever applied to memory, and it is the first technology that defeats death in part, because it lets someone who no longer exists instruct someone who does. The press made copies of a mind. The engine broke the dependence of work on muscle, wind and weather. Electricity carried power to wherever it was wanted and the transistor did the same for decisions. The rocket applies the identical operation to the one boundary that had never moved at all, which is the planet itself.

Every one of those is the same act performed on a different constraint: take a capacity bounded by a body and make it no longer bounded by that body. Reach, memory, strength, sight, calculation, and now location.

Immortality, in the only sense that can be discussed without leaving the ground, is the end of that sequence rather than a departure from it. Writing already made memory outlive the person. Medicine designed rather than discovered extends the body, and the first molecules built that way are in human trials now. And a species on one world has a single point of failure, while a species on two has none that can be taken in a single event. Durability at that level is not a metaphor for immortality; it is the literal form the thing takes for anything made of many lives.

Which is the full measure of a vehicle that lands itself and flies again a week later. Not that it is impressive, although it is. That the oldest operation our species performs, taking a limit that belongs to the body and moving it outside the body, has arrived at the planet, and that what now stands between here and everywhere is no longer a mystery about nature. It is a quantity of patient, repeated, unglamorous work, measured in flights, in windows and in years. Which is to say it is only time, and time is the one obstacle a species has ever been able to outlast.

The human, raised from stone hammers to the grace and condition of an immortal being with unlimited technology, free across the universe.

The part about immortality and unlimited technology towards the future: who knows how long it will take to arrive. Let us hope it is while we are still alive. After all, once you are immortal, the time to build everything and to reach everything will be left over.

Basically, the question and the path of the human is well-fortuned by nature. Evolution embraces us, and the only way to move is forward, and forward is where the Gods are. So it is probable, almost certain, that we humans will have the faculty of the Gods, technologically speaking. It is a good day to have been born a human being, is it not. The date of the beginning, of everything. As long as everyone stays friendly and fights only in their imagination, there awaits us, our children and our grandchildren, the hope of experiencing the magnificence of existence at its highest peaks. Cheers to that, humans.

With many people on this planet being protagonists of that story, among them those young promises who will throw themselves into the space sector in the near future. Clearly, when will the day come when leaving the planet for other countries is like taking an Uber? It seems sooner rather than later, because unlike a human, who tends to live off pleasures, with only a few passionate ones dedicating their life to the work of their lives, their work of art, artificial intelligence never rests and only knows how to grow.

It is without doubt a gratifying future for everyone, as long as countries stay sane. What we all think is that staying sane among all of us is what will take humanity to become supreme beings, and it is without doubt the right way to get there.

So all of you, calm down a little, smile and make peace. Then go to an electronic music festival and wish each other a good life and good health.

Sincerely. Until next time.

Sources

Filed documents. Space Exploration Technologies Corp's Form 10-Q for the quarter ended 30 June 2026 (filing), for revenue by segment, research and development, capital expenditure, cash, debt, backlog and the description of the three business segments. SpaceX's application to the Federal Communications Commission for an Orbital Data Center System, filed 30 January 2026 and accepted for review by the Space Bureau five days later. The Artemis Accords and their signatory list are published by NASA.

Measurements. Martian surface and transit radiation figures are from the Radiation Assessment Detector aboard the Curiosity rover, as published by the instrument team (Southwest Research Institute).

Press and trade reports, identified as such in the text. Flight-by-flight accounts of Starship tests 12, 13 and 14, including the booster anomaly, the mishap investigation and the reuse of heat shield tiles, are drawn from NASASpaceflight and Spaceflight Now. Falcon 9 reuse records from Space.com. The orbital data centre constellation and the first satellite's power figures were reported by Data Center Dynamics and SatNews. Engine and vehicle specifications are as published by the manufacturer and compiled in trade coverage; figures for nuclear thermal and electric propulsion are standard published values for those classes of engine.