# The Last Migration Will Not Be Human ![[resources/images/iis-reconfigured-orbital-von-braun-wheel.png]] <iframe width="100%" height="20" scrolling="no" frameborder="no" allow="autoplay; encrypted-media" src="https://w.soundcloud.com/player/?url=https%3A//api.soundcloud.com/tracks/soundcloud%253Atracks%253A2392353063&color=%23ff5500&inverse=false&auto_play=false&show_user=true"></iframe><div style="font-size: 10px; color: #cccccc;line-break: anywhere;word-break: normal;overflow: hidden;white-space: nowrap;text-overflow: ellipsis; font-family: Interstate,Lucida Grande,Lucida Sans Unicode,Lucida Sans,Garuda,Verdana,Tahoma,sans-serif;font-weight: 100;"><a href="https://soundcloud.com/bryantmcgill" title="Bryant McGill" target="_blank" style="color: #cccccc; text-decoration: none;">Bryant McGill</a> · <a href="https://soundcloud.com/bryantmcgill/engineering-the-successor-to" title="Engineering the Successor to Human Civilization" target="_blank" style="color: #cccccc; text-decoration: none;">Engineering the Successor to Human Civilization</a></div> *Image: An **ISS-derived rotating toroidal habitat**, inspired by **Stanford-torus and von Braun wheel architecture**, and assembled from modular orbital infrastructure into a long-duration artificial-gravity settlement.* ## Climate Tipping Points, Marsification, and the Machine Civilization Carrying Humanity Forward > [!map] Ecosystem nexus > This essay is the gateway to the [[collections/MARS|MARS collection]]. It joins [[wiki/Marsification|Marsification]] and [[wiki/State-Dependent Catastrophe|state-dependent catastrophe]] to [[wiki/Substrate-Relative Habitability|manufactured habitability]], [[wiki/AI Factory|AI factories]], [[wiki/Computational Metabolism|computational metabolism]], [[wiki/Civilizational Payload|civilizational payload]], and [[wiki/Machine Civilization|machine civilization]]. Read it beside [[articles/We Were Never Going to Make It|We Were Never Going to Make It]], then follow the person-scale mechanism through [[articles/Uploading Is Imminent|Uploading Is Imminent]], the precursor physical map through [[articles/Breakaway Space Habitat Infrastructure?|Breakaway Space Habitat Infrastructure]], and the current implementation ledger through [[articles/From Starbase to Orbit|From Starbase to Orbit]]. The physical sequence is organized as [[wiki/Ground-to-Orbit Infrastructure Continuum|the ground-to-orbit infrastructure continuum]], with [[wiki/In-Space Servicing, Assembly, and Manufacturing|ISAM]] and [[wiki/Closed-Loop Life Support|closed-loop life support]] as its builder and habitation layers. > **MARS has always stood for Machine Learning, Automation, Robotics, and Space. That is how we are going to Mars.** ### I. Forty-Eight Hours On January 20, 2025, the United States announced from the inaugural platform that its energy policy would be extraction without apology, initiated withdrawal from the Paris Agreement, and declared a national energy emergency. On January 21, 2025, at the White House, OpenAI, SoftBank, Oracle and MGX announced Stargate: an initial hundred billion dollars scaling to five hundred billion over four years, ten gigawatts of artificial-intelligence infrastructure on American soil. One day separated the abandonment of the international mitigation regime from the inauguration of the largest privately financed computational buildout in the history of the species. I am not going to claim those two events were coordinated in a room. I do not need to. The claim that matters is harder to attack and considerably more consequential: **within roughly forty-eight hours, a civilization publicly withdrew from the project of stabilizing its biosphere and publicly committed to the project of manufacturing its successor's substrate.** Whatever was intended, that is what happened, and everything built since has widened rather than narrowed the distance between those two commitments. The interpretive discipline this essay observes throughout is what makes the argument formidable rather than dismissible. **We do not need to prove the meeting. We need to examine the metabolism.** Institutions rarely announce their deepest operating assumptions; they reveal them through what they fund, what they permit, what they exempt from environmental review, what they classify as defense-critical, and what they quietly stop building. Selection does not require shared intention. Disparate actors pursuing entirely ordinary local objectives can converge on a common architecture the way cells converge on an organism none of them comprehend. The convergence itself becomes the object requiring explanation, and by 2026 that convergence can no longer be characterized as noise. This essay replaces my March 2025 piece, [[articles/We Are "Going" to Mars through the Marsification of Earth|We Are "Going" to Mars: Earth's Biospheric Collapse into Planetary Marsification and Death]]. That article correctly identified the pivot from mitigation to adaptation and correctly read "we are going to Mars" as a confession rather than a promise. Its ending — a call for radical cooperation — has been superseded by my own later work: by [[articles/We Were Never Going to Make It|We Were Never Going to Make It]], which states the thermodynamic case without consolation, and by [[articles/Uploading Is Imminent|Uploading Is Imminent]], which specifies the mechanism of transfer. What follows is written from the far side of both. ### II. Gore Was Right, and Being Right Did Not Help Begin by disposing of the twenty-year argument that has functioned as a decoy for two decades. In June 2026, on the twentieth anniversary of _An Inconvenient Truth_, ABC News asked Al Gore whether the film held up. His answer was two words: "Unfortunately, yes." ABC's own review of the documentary's principal scientific claims found that most of its observations had either come to pass or remained on trajectory, while conceding what serious people have always conceded — that particular timing claims, most notoriously an ice-free Arctic, were too aggressive. Gore's characterization of the underlying physics in that interview was that we continue to use the sky as an open sewer, and that the heat now trapped each day is equivalent to hundreds of thousands of Hiroshima-class detonations. The cherry-picking of a failed timetable was never a refutation of the science. It was a political operation against what the science implied, it worked, and the measurements accumulated anyway. **"Climate change is a hoax" was not a correction of Gore's data; it was a reaction to that data's consequences.** Here is where the record stands as of the most recent authoritative assessment. The World Meteorological Organization's _State of the Global Climate 2025_, released March 23, 2026, confirms that 2015 through 2025 were the eleven hottest years in the observational record, with 2025 at roughly 1.43 °C above the 1850–1900 baseline — slightly below 2024's record of approximately 1.55 °C, a difference produced by the shift to La Niña conditions rather than by any deceleration of the underlying forcing. For the first time, WMO elevated **Earth's energy imbalance** to a headline indicator, and found it at the highest level in the sixty-five-year record. Ocean heat content set another record; the rate of ocean warming from 2005 to 2025 is more than double the rate from 1960 to 2005. Roughly ninety percent of the excess energy is going into the ocean, and the magnitude is the figure to read slowly: the ocean has been absorbing the equivalent of about **eighteen times annual human energy consumption, every year, for the past two decades.** Around ninety percent of the ocean surface experienced at least one marine heatwave during 2025. Sea level stands roughly eleven centimetres above its 1993 satellite baseline. That is the physical situation. Now the conceptual one, because the phrase "point of no return" has done more damage to public comprehension than any denial campaign. **The point of no return was never one point.** Climate science has developed a precise vocabulary for the contraction of options, and it is not the vocabulary of a single planetary switch. It runs: _mitigation_, the attempt to prevent the disturbance; _adaptation_, the attempt to live with what can no longer be prevented; _transformational adaptation_, the reorganization of the affected system because incremental adjustment no longer suffices; _adaptation limits_, beyond which objectives cannot be maintained; _hard adaptation limits_, at which no adaptive action can avoid intolerable risk; and finally _loss and damage_, the residue that occurs despite everything. The IPCC states plainly that hard limits have already been reached in some ecosystems and that additional human and natural systems will encounter them as warming proceeds. That progression — **stop the ship, stabilize the ship, reorganize the ship, abandon particular compartments, preserve what can still be preserved** — is not an exotic reading imposed from outside. It is the operational structure of contemporary climate-risk management, published, peer-reviewed and intergovernmentally negotiated. In October 2025 the second _Global Tipping Points Report_, assembled by more than one hundred sixty scientists across eighty-seven institutions in twenty-three countries, concluded that the first Earth-system tipping point has been reached: widespread mortality of warm-water coral reefs is now underway. At approximately 1.4 °C of warming, warm-water reefs are crossing their thermal threshold and undergoing unprecedented dieback, with consequences for a quarter of marine life and the livelihoods of nearly a billion people. Parts of the Greenland and West Antarctic ice sheets may already have passed their own thresholds, committing the planet to metres of eventual irreversible sea-level rise. In June 2026 the United Nations Secretary-General's Scientific Advisory Board issued a brief describing thresholds beyond which change becomes self-propagating even if the original forcing is later reduced or removed. Read that sequence and the shape of the thing becomes visible. **Earth does not need to lose its atmosphere. Earth does not need to turn red. Earth does not need to become globally uninhabitable.** Civilization depends on a narrow band of regional heat tolerance, water availability, agricultural reliability, ecological buffering, grid stability and political coherence, and those fail asymmetrically and locally long before planetary habitability does. Tipping points do not announce the end of the world. They announce **the contraction of the option set** — and organisms facing a contracting option set stop spending everything defending states they can no longer hold and begin investing in states that remain reachable. That is the entire logic of what follows. ### III. Marsification, Defined With Engineering Precision [[wiki/Marsification|Marsification]] is not a metaphor about deserts. Stated with the precision the concept deserves and can bear: **Marsification is the progressive withdrawal of ambient habitability and its replacement by manufactured habitability — the substitution of engineered, metered, monitored, energy-dependent services for the biospheric services Earth formerly supplied at no charge.** Water is the clearest instance. In an intact hydrological regime, a settlement receives potable water as a gift of the water cycle. Under Marsification, water becomes machinery: intake, membrane, brine stream, recovery ratio, reservoir days-of-demand, sensor telemetry, pumping energy. Waste assimilation is the second instance. A functioning ecology metabolizes waste for free; a Marsified system converts waste into a feedstock inventory with recoverable cellulose, nutrients, grit and biogas, each with a market price and an energy cost. Then thermal regulation, then food production, then atmosphere itself as conditioned microclimate, then environmental sensing as continuous instrumentation, and finally cognition — because once every other service has become an engineered subsystem, the coordination of those subsystems exceeds unaided human administration and must itself be computed. Mars is difficult not principally because it is far away. **Mars is difficult because every biospheric subsidy must be purchased technologically.** Marsification therefore begins the moment terrestrial civilization starts pricing formerly ambient subsidies back into the cost of remaining alive. On that definition the evidence is not circumstantial and not speculative. It is a construction schedule. Now the part I owe the reader, because the argument is stronger when the weakest version of it is amputated rather than defended. I have held, and have written, that Earth is losing its atmosphere in the Martian sense and that the 2030 and 2050 horizons encode that expectation. That literal claim is **contradicted** by the physics on decadal timescales and I will not carry it forward. The bulk atmospheric inventory is not appreciably escaping to space on any human-relevant schedule; the mechanism that removed Mars's atmosphere operated over hundreds of millions of years on a body with roughly a tenth of Earth's mass and no sustained magnetic dynamo. The best current modelling of Earth's eventual deoxygenation — Ozaki and Reinhard, _Nature Geoscience_, 2021, a Japanese-led biogeochemistry-climate ensemble run more than four hundred thousand times — puts the mean future lifespan of an oxygenated atmosphere at **1.08 ± 0.14 billion years**, driven by solar brightening and the carbonate-silicate cycle rather than by anything industrial civilization can do. Anyone who tells you the atmosphere is leaving by 2050 is wrong, and the error is expensive, because it hands critics a disprovable claim to substitute for an undisprovable one. What survives the amputation is more disturbing than what it replaces. **We do not have to lose the atmosphere to begin living as though the atmosphere can no longer be trusted to supply acceptable boundary conditions everywhere, at all times.** The operative mechanisms are compositional and energetic, not gravitational: wet-bulb heat that exceeds the human thermoregulatory envelope, hydrological reorganization that moves rain away from the soils that were farmed on the assumption it would fall there, ozone and air-quality degradation, ocean heat that will continue releasing for centuries regardless of emissions decisions taken now, and a measured planetary energy imbalance at a record high. A city whose air must be conditioned, whose water must be desalinated and recirculated, whose food must be cultivated inside controlled environments, whose waste must be converted back into inputs, whose power must be buffered against intermittency, whose movement must be sheltered from thermal extremes and whose environmental state must be continuously instrumented has crossed a systems boundary while standing beneath a perfectly substantial sky. And the 2030/2050 pairing requires no concealed meaning at all, which is precisely why it is so useful. Those dates have an ordinary provenance in carbon-budget arithmetic and in the 2030 Agenda adopted in 2015. But they have acquired a second life as **infrastructural planning horizons**: the World Bank's _Groundswell_ work describes internal climate-migration hotspots emerging by 2030 and as many as 216 million internal climate migrants across six regions by 2050, driven by water scarcity, declining crop productivity and sea-level rise. Two dates that began as emissions milestones now function as the window inside which where people can work, where agriculture remains reliable, where water remains available, and therefore where populations can economically remain, is expected to change materially. That is not a countdown clock. It is a construction window, and everyone building anything expensive is building to it. ### IV. NEOM: The Substitution Made Visible No site on Earth demonstrates the substitution more completely than NEOM, and it demonstrates it in a way that is considerably more interesting than the version I previously argued. Take the water architecture first, because it is the purest expression of hydrology-as-machinery ever attempted at territorial scale. NEOM states that it will meet **all** of its water needs through desalination, delivered in three or more stages to a total capacity of **one million cubic metres per day**, using advanced reverse-osmosis and hybrid technologies at greater than sixty percent recovery, integrated with brine processing and targeting **zero liquid discharge**. Rather than dumping brine, it intends to mine it: a Fully Integrated Resource-recovery Seawater Treatment complex extracting industrial chemicals and minerals from the reject stream. On the wastewater side the specification is total — one hundred percent collected and treated, one hundred percent reused, zero run-off to the environment, with maximized energy and material recovery harvesting cellulose, nutrients, grit and biogas from the waste stream. The distribution figures are the ones that reveal the ontology: approximately **2,500 kilometres of smart sewage collection network with comparable coverage of smart recycled-water network**, a 600-kilometre bulk transmission system, up to twenty pumping stations, potable reservoirs sized for up to **five days of demand**, and an "Internet of Water" instrumenting the whole circulation in real time. That is not a utility in the twentieth-century sense. That is a closed metabolic loop with a control system, and it is indistinguishable in principle from what a permanent extraterrestrial settlement requires. Add the food architecture — high-technology greenhouses and vertical farms explicitly described as climate-resilient agricultural assets, AI-optimized production, closed-containment aquaculture — and the pattern completes. **Fundamental ecological processes are being internalized into the artifact.** The city stops being something placed in an environment and becomes an organism whose metabolism is legible to software. Now the part where I revise my own reading, because the facts of 2026 supply a stronger thesis than the one I was defending. I previously argued that reports of The Line being scaled back were disinformation, and that the development was mostly subterranean because future smart cities must go underground when surface temperatures become intolerable. The **underground infrastructural spine is real**: The Line's Spine was designed as a subterranean corridor carrying high-speed passenger rail, freight rail and multi-utility services beneath the inhabited structure, and the tunnelling packages were substantial — a single terminated Hyundai Engineering & Construction contract covered roughly 12.5 kilometres within a broader package exceeding 28 kilometres of underground works. But the claim that residents were intended to live underground is **unsupported**; NEOM's published proposition has consistently been a 500-metre-high, 200-metre-wide above-ground structure with an engineered year-round climate, and deep foundations beneath a proposed 500-metre megastructure are structurally unremarkable rather than evidentiary. And the retrenchment is **established**, not fabricated. Construction on The Line was suspended; the Hyundai tunnel contract worth approximately one billion dollars was terminated in the first quarter of 2026 as part of a project restructuring; major Trojena contracts were cancelled after the 2029 Asian Winter Games were relocated; the Public Investment Fund recorded a multibillion-dollar write-down; an internal audit reported by the _Wall Street Journal_ projected total costs of $8.8 trillion and a completion horizon extending toward 2080; and as of mid-2026 the massive underground piling and foundation work was complete only for an initial 2.4-kilometre stretch. NEOM itself was reorganized into separate entities. The megacity for nine million people, then one and a half million, then three hundred thousand, is not being built. Here is what makes that vastly more interesting than a cover-up would have been. **The capital did not evaporate. It changed species.** Reporting through 2025 and 2026 describes Saudi funds redirected from The Line toward ports, logistics, industry and data centres. And on **August 31, 2026** — the day this essay was completed — HUMAIN, the artificial-intelligence company owned by the Public Investment Fund, and DataVolt broke ground at Oxagon on 100 MW of a 360 MW first phase within a planned **1.5-gigawatt AI campus**, sited for pre-zoned industrial land, at-scale energy, and low-latency subsea cable connectivity to Europe and Africa, with first capacity expected in service in 2028. Project documentation for the campus describes large-scale battery storage and gas turbines capable of green-hydrogen operation supporting roughly forty-eight hours of autonomous operation. Read those two facts adjacently and the thesis states itself without embellishment. **At the exact historical moment when the fantastically expensive human city was suspended, the machine city broke ground on the same territory, with its own islanded power, its own thermal management, and its own transoceanic nervous system.** The closed-loop water, food, energy and sensing infrastructure that was justified by the human city continues to be developed around a tenant that does not breathe. That is not a scaled-back megaproject. That is a substitution of payload, and it is the clearest single instance of the argument this essay makes. ### V. The Closed Loop Is the Same Loop The convergence becomes undeniable once you notice that institutions with radically different missions are writing the same engineering specification. The Gates Foundation's sanitation program — the Reinvent the Toilet Challenge and its successors, into which more than two hundred million dollars of early-stage research funding has been directed — seeks systems that eliminate pathogens, recover clean water, nutrients and energy from human waste, operate **off-grid without connection to sewer or water infrastructure and with minimal electricity**, and are explicitly **resilient to climate and water stresses**. This is not historical. In January 2026, Rice University's WaTER Institute announced a Gates-funded programme with Arizona State, Clarkson and Duke to develop modular, field-ready household systems capable of recycling multiple wastewater streams and sharply reducing both water consumption and dependence on centralized infrastructure. The stated purpose is humanitarian and I will not insinuate otherwise: three and a half billion people lack safely managed sanitation, and centralized sewerage will never reach them. But look at the specification rather than the motive. **Off-grid. Waterless. Energy-recovering. Nutrient-recovering. Independent of centralized infrastructure. Resilient to water and climate stress. Modular. Household-scale.** Now compare the European Space Agency's MELiSSA programme, which pursues the recovery of oxygen, water and food from organic waste and carbon dioxide through bacterial, algal and higher-plant compartments functioning as an engineered ecosystem. Now compare NASA's demonstrated **ninety-eight percent water recovery** aboard the International Space Station, which the agency identifies as approximately the threshold required for long-duration exploration missions. Now compare NEOM's hundred-percent reuse and material recovery. Four institutions, four missions, one thermodynamic grammar: **close the loop**. A closed system has the same requirements whether it sits in an informal settlement during infrastructure failure, beneath a desert city, aboard a spacecraft, or on Mars. The distance between humanitarian sanitation research and extraterrestrial life support is not conceptual; it is only a difference in the severity of the boundary condition, and the boundary conditions on Earth are moving in the direction that erases the difference. One connective detail deserves to be stated plainly rather than left implicit, because it is a matter of public record and it closes a circuit between two sections of this essay. In January 2026 Meta announced nuclear agreements unlocking up to **6.6 gigawatts** by 2035, and one of the three counterparties was TerraPower, the advanced-reactor company founded by Bill Gates. The announcement's stated purpose was to power Meta's pursuit of **"personal superintelligence for everyone."** The same philanthropic-industrial lineage that funds closed-loop toilets for communities without sewers is also supplying firm baseload to the manufacture of synthetic minds. I draw no conclusion about intent from that. I observe only that the pieces are being supplied by the same hands. ### VI. The Foundry Now the energy, because this is where the ordinary explanations fail and the argument becomes uncomfortable rather than merely interesting. The July 2025 _America's AI Action Plan_ stated the requirement in the government's own words: artificial intelligence is the first modern digital service requiring the United States to build vastly greater energy generation than it currently possesses, and the document immediately couples that requirement to chip fabrication, data centres, new generation and accelerated permitting. Two executive orders then converted the requirement into legal architecture. The May 2025 nuclear order directed the Department of Energy to consider designating qualifying artificial-intelligence data centres located at or coordinated with departmental facilities as **critical defense facilities**, with their supporting generation treated as **defense critical electric infrastructure**. The July 2025 permitting order defined a qualifying data centre as one exceeding 100 MW, explicitly enumerated its supporting infrastructure to include gas pipelines, turbines, coal equipment, nuclear equipment, geothermal, substations, transmission, storage, networking and semiconductor facilities, and directed the Department of Defense to identify locations on **military installations** suitable for leasing to that infrastructure. Read the category shift rather than the politics. **Computation was reclassified from commercial infrastructure into national-security infrastructure**, alongside energy, defense industrial production and continuity of the state. The siting followed the classification. The Department of Energy identified sixteen federal sites for accelerated data-centre and generation development, then advanced Idaho National Laboratory, Oak Ridge Reservation, the Paducah Gaseous Diffusion Plant and the Savannah River Site. At Savannah River — a former plutonium and tritium production complex — the National Nuclear Security Administration identified ten tracts totalling 3,103 acres, issued a request for proposals, and in July 2026 selected Amentum to negotiate a phased lease for a **one-gigawatt artificial-intelligence data centre supported by approximately two gigawatts of dedicated on-site generation, natural gas bridging to advanced nuclear**. The NNSA Administrator framed the agency's harnessing of artificial intelligence as comparable in national-security consequence to the Manhattan Project. Frontier computation is being installed on the landscape of American nuclear-weapons production, with its own generation, behind its own fence. The corporate procurement of firm power is the same phenomenon expressed through balance sheets. Microsoft's twenty-year agreement is restoring Three Mile Island Unit 1, an 835 MW reactor shut down in 2019 for economic reasons. Google contracted for a fleet of small modular reactors targeting 500 MW by 2035. Amazon and X-energy are targeting more than 5 GW by 2039. Meta's January 2026 agreements reach up to 6.6 GW. Microsoft has contracted for a projected 50 MW fusion plant from Helion. **The software industry is no longer negotiating electricity prices. It is resurrecting shuttered fission plants, commercializing advanced reactors, and creating a market for fusion before fusion exists.** The magnitudes are now beyond intuitive comprehension and must be stated numerically. The International Energy Agency's 2026 assessment finds that capital expenditure by the largest technology companies exceeded **$400 billion in 2025 and is expected to rise a further seventy-five percent in 2026**, and that the capital expenditure of just five technology companies now exceeds **global investment in oil and natural-gas production**. The IEA's satellite-based tracking shows that dedicated artificial-intelligence factories have **more than tripled in capacity in eighteen months**. Global data-centre electricity demand is projected to move from roughly 485 TWh in 2025 to about 950 TWh by 2030, with consumption by AI-focused facilities tripling over the same period. Lawrence Berkeley National Laboratory projects United States data-centre consumption at 649 TWh by 2030, **11.8 percent of forecast national electricity use**, within a modelled range of 9.5 to 15.3 percent, and responsible for roughly a third of all load growth across the six-year period. AI-related debt has reached approximately $1.4 trillion, the largest segment of the American investment-grade credit market. The physical consequence is a fossil buildout that the mitigation era was supposed to have foreclosed. Global Energy Monitor counted **378 gigawatts** of gas-fired capacity in development in the United States by mid-2026 — a fifty percent increase in six months, roughly a third of the global total — of which **189 gigawatts is intended to power data centres directly**, nearly doubling in half a year. American gas capacity under construction rose seventy-six percent in the first half of 2026 to 52 GW, twice what China is building. Texas alone accounts for approximately 80 GW in development. Coal generation, which was supposed to be terminal, rose nearly twenty percent year-to-date at one point in 2025 as gas prices and load growth collided. None of this is confined to the United States, and the international replication is the strongest evidence against any purely partisan explanation. The European Union has institutionalized the term **AI Gigafactories**, launching a procurement on July 30, 2026 for up to seven facilities, each holding on the order of one hundred thousand advanced processors — roughly four times the current generation of European AI Factories — backed by up to €10 billion in public funding intended to unlock at least €20 billion more, with an earlier expression-of-interest round drawing seventy-six submissions across sixteen member states. France announced €109 billion in artificial-intelligence investment. The United Kingdom designated the site of the Atomic Energy Authority at Culham as its first AI Growth Zone and has begun reserving grid capacity for strategically important developments. The United States and the United Arab Emirates unveiled a five-gigawatt campus in Abu Dhabi across roughly ten square miles, combining nuclear, solar and gas, positioned to serve low-latency compute to a substantial fraction of the world's population, with a reciprocity provision committing the UAE to build American capacity at least as large. Saudi Arabia's HUMAIN has announced AI factories with NVIDIA, a ten-billion-dollar collaboration with AMD, an AI Zone with AWS, and now Oxagon. China's East Data West Computing programme has driven over a trillion yuan of investment into a national integrated computing network, brought a 500,000-kilowatt photovoltaic station online in May 2026 for direct green-power supply to computing hubs, installed a record 22.4 GW of gas capacity in 2025, and — through the Chengdu firm ADAspace — is pursuing a constellation of 2,800 computing satellites, having already demonstrated the use of orbital compute to command a ground robot. **Individual thinking machines are now specified in gigawatts**, a unit historically reserved for power stations and cities. When a civilization begins describing the supporting infrastructure of its cognitive artifacts in the units of utility-scale generation, a category boundary has been crossed that no press release announces. The resistance is also real and belongs in an honest account. Maine passed the first state-level ban on large data-centre construction; Texas passed a moratorium; at least fifteen other states are advancing similar measures; one polling firm found roughly seventy-five percent of American voters opposed to new data centres in their area; dozens of projects have been delayed or cancelled. The buildout is neither smooth nor universally welcomed. It is simply proceeding anyway, and the political system has organized itself to accelerate rather than to arbitrate it. ### VII. Someone Said It Out Loud The objection that keeps intelligent people from taking this seriously is a sentence about cat videos. If the planet is in ecological crisis, why would anyone rationally convert every available watt into inference for chatbots and spreadsheets? The answer is that the builders themselves have stopped pretending that is what they are building. Eric Schmidt, former chief executive of Google, said at the Special Competitive Studies Project's AI+Energy Summit in 2024 that the energy demand of artificial-intelligence computing is effectively infinite, that conservation will not get us there, and — when asked directly whether AI's energy needs could be met without destroying climate goals — that **"we're not going to hit the climate goals anyway"** because we are not organized to do it, adding that he would rather bet on AI solving the problem than constrain it and have the problem regardless. Months later, before the House Energy and Commerce Committee in April 2025, he affirmed that climate change is real and then told legislators the industry needs energy in all forms, renewable and non-renewable, and that none of the existing sources can be reduced. That is the strategic pivot stated in public by a person with maximal access. It is not a doctrine of succession — Schmidt stops at _AI will solve it_ — but it is an explicit, on-the-record acknowledgement that the climate objective has been subordinated, deliberately, to the construction of machine intelligence, by someone who does not believe the objective will be met. NVIDIA has made the remaining pretence untenable in its own regulatory filings. The company describes its transformation from component supplier into builder of **AI factories**: installations that run continuously, that convert **energy into tokens**, that host autonomous multi-agent systems and physical AI. Its 2026 annual report places **energy as the foundation layer** beneath compute, infrastructure, models and applications, and distinguishes the traditional data centre, which stored information, from the AI factory, which is manufacturing equipment that transforms electricity into intelligence, reasoning, action and work. When a civilization restarts nuclear plants and builds gigawatt-class generation for computation, the dominant supplier's own description of the industrial objective is that these are **factories that manufacture intelligence.** Leopold Aschenbrenner's 2024 _Situational Awareness_, written shortly after his departure from OpenAI, described the same buildout from inside the expectation set: trillions flowing into GPUs, power plants and data centres; American electricity generation forced to grow by tens of percent; national-security involvement of a kind unseen in half a century; and a government programme he simply called The Project. He is explicit that the document mixes public information, personal analysis and industry conversation, so it is testimony rather than disclosure — but it is testimony that the buildout was understood by its architects as an industrial mobilization for superintelligence rather than as consumer software growth. Now the statutory layer, which is where the argument stops depending on any individual's candour. The United States has already moved beyond continuity of government into a codified doctrine of **continuity of the economy**. Under 6 U.S.C. § 322, enacted in the FY2021 National Defense Authorization Act, the President is required to develop and maintain a plan to restore the American economy after a significant event, natural or human-caused. The statute requires identification of the critical distribution mechanisms to be prioritized for operation during such an event, and the enumerated list includes bulk power and electric transmission, national and international financial systems, and — in the same breath — **national and international communications networks, data-hosting services, and cloud services**. It requires identification of industrial control networks whose failure would be debilitating, and hardening measures for them. It requires a recommendation on strategic reserves of critical materials. It requires guidance on which categories of employees must be prioritized to keep working. And in subparagraph (H) it requires identification of the critical economic sectors for which **the preservation of data in a protected, verified, and uncorrupted status** would be required for rapid recovery. Read subparagraph (H) again. That is not disaster response. That is a specification for **the minimum informational state from which an economy can be re-instantiated.** One of the architects of the concept described the operational logic more starkly than the statute does. In a March 2020 discussion of the Cyberspace Solarium Commission's recommendations, commissioner Samantha Ravich — former deputy national security adviser, later vice chair of the President's Intelligence Advisory Board — described the need to determine in advance which data must survive and be stored securely offline, potentially in allied territory, in order to reconstitute the workings of major economic components afterward, and then stated the consequence without softening it: **"not every place is going to be prioritized."** She was discussing catastrophic cyberattack, not climate breakdown, and I will not transplant her motive into a different scenario. What transplants is the discipline: identify the minimum viable informational and infrastructural state, preserve it redundantly, designate priority nodes in advance, and accept that catastrophic continuity planning entails **selection**, because universal preservation is not available. Two years later Congress went a level higher. The **Global Catastrophic Risk Management Act of 2022** defines a global catastrophic risk, in federal law, as the risk of events consequential enough to significantly harm, set back, or destroy human civilization at the global scale. It requires expert estimates of cumulative global catastrophic and existential risk over the **next thirty years**, threat-by-threat analyses, forecasts of whether those risks are rising, and a follow-on report on the adequacy of continuity-of-government and continuity-of-operations planning against them. The resulting assessment, produced for the Department of Homeland Security and FEMA by RAND's Homeland Security Operational Analysis Center, treats six threats in one frame: **artificial intelligence, asteroid and comet impacts, nuclear war, severe climate change, pandemics, and supervolcanoes.** That is the United States government, in statute and in commissioned analysis, planning at the scale of civilization rather than disaster, on a thirty-year horizon, with severe climate change and asteroid impact and artificial intelligence in the same document. There is no memorandum titled Continuity of Civilization. There does not need to be. **Every layer that would sit beneath such a doctrine already exists in public law.** And in November 2025 the layer above them arrived. Executive Order 14363 launched the **Genesis Mission**, a national effort to harness artificial intelligence for science, explicitly framed as comparable in urgency and ambition to the Manhattan Project. The Department of Energy was directed to integrate its seventeen national laboratories, the country's most powerful supercomputers, its quantum systems and its most advanced scientific instruments into a single **American Science and Security Platform** — described by the department as a closed-loop system, and once complete, the most powerful scientific instrument ever built, drawing on roughly forty thousand departmental scientists and engineers. The order required an inventory of computing resources, a list of at least twenty national science and technology challenges, a demonstration of initial operating capability, and — this is the detail that matters most for what follows — **a review, due July 22, 2026, of capabilities across the national laboratories for robotic laboratories and AI-directed experimentation facilities.** By July 2026 more than five billion dollars in federal commitments had been announced, including closed-loop experimentation coupling prediction, synthesis and characterization for materials design, and NNSA workflows automating the testing and qualification of materials for the nuclear-weapons enterprise. The state did not merely permit machine cognition. **It commissioned machine cognition with hands.** ### VIII. The Hands Five weeks after the Genesis Mission's robotic-laboratory review came due, on **August 27, 2026**, Anthropic opened a research preview of the **Model Hardware Standard**, a specification allowing artificial-intelligence agents to discover, monitor and operate physical instruments — microscopes, liquid handlers, robotic arms, plate readers, cameras, manufacturing machines, quantum-computing lasers — through a common driver layer, in parallel, reducing integration work from weeks or months to hours. The standard is model-agnostic, reachable through the Model Context Protocol, and applicable to essentially any device exposing a programmable control surface. Its drivers encode what code alone does not: the mass of a robot arm, its safety limits, the adjustable parameters that previously existed only in paper manuals or in the tacit knowledge of specialists. The demonstrations are the part to sit with. At Carnegie Mellon, one agent coordinated a liquid handler, a plate reader, a robotic arm and monitoring cameras across three computers with mutually incompatible interfaces, including one instrument with no programmatic interface at all, running dose-response experiments roughly three times faster than before — and, having evaluated the quality of its own curve fit and found it inadequate, **autonomously reran the experiment**. At Genentech, an agent tuned aspiration parameters for water and for viscous protein solution by running trial transfers, reading absorbance, scoring itself against an expert's plate and converging on values its human automation specialists confirmed as reasonable. At Janelia, a microscopy rig requiring seven programs launched in fixed order collapsed into a single dashboard action. Taken alone this is laboratory automation. Placed beside the Genesis Mission's mandated review of robotic laboratories, beside NASA's autonomous excavation and construction programmes, beside humanoid robotics and beside gigawatt inference capacity, it is something else: **a bridge between machine cognition and physical agency, being built from both ends simultaneously, by a government and by an industry, in the same season.** The off-world half of that bridge is further along than public perception allows. NASA's Moon Base architecture, published May 26, 2026, is explicitly phased so that **machines precede humans**. Phase One, running through 2029, is entirely robotic: as many as twenty-five missions and twenty-one landings delivering roughly four metric tonnes of cargo to scout the lunar south pole, characterize resources and demonstrate the technologies human habitation will later depend on. Phase Two, from 2029, scales to roughly twenty-seven launches, twenty-four landings and sixty tonnes, establishing power, communications, logistics and early habitation. Phase Three, from 2032, brings habitats, transportation, fission power and permanent presence, with delivery capacity climbing toward one hundred fifty tonnes and cargo beginning to flow back to Earth. Water ice from permanently shadowed craters is converted on site into drinking water, oxygen and propellant; regolith is processed into construction material. Two lunar terrain vehicles, capable of autonomous and teleoperated as well as crewed operation, are to be on the surface by 2028 — scouting terrain and pre-staging resources **before any astronaut arrives**. NASA is separately developing molten-regolith electrolysis and solar-thermal extraction producing oxygen and metals from [[wiki/In-Situ Resource Utilization|raw lunar material]] using in-situ power, and autonomous construction of habitats, roads, blast shields and landing pads from local feedstock. The Chinese and Russian International Lunar Research Station describes the same architecture in language that is, if anything, more explicit about the ratio. The ILRS is specified as an expandable, maintainable system capable of **long-term robotic operation with short-term human participation**, with a basic model at the lunar south pole by 2035, powered by solar, radioisotope and nuclear generators, expanding by 2050 into a network spanning south pole, equator and far side. Chang'e-8, targeted for 2028, is to test [[wiki/In-Situ Resource Utilization|in-situ resource utilization]]. Thirteen nations and counting have signed on. Two rival programmes, entirely different political systems, arriving independently at the identical sequence: **robotic reconnaissance, autonomous construction, power, communications, resource extraction, enclosed life support, periodic humans, persistent humans.** Machines are not competing with the astronauts for the mission. Machines are the mission, and the astronauts are the eventual, expensive, optional payload. ### IX. Some of Us Go The honest version of this argument requires abandoning a rhetorical convenience. It is not true that no human is going. Some humans are going, and the number is knowable, and knowing it is what destroys the fantasy far more effectively than denial ever could. Artemis II flew: a ten-day circumlunar mission with crew, returned safely, at a programme expenditure approaching one hundred billion dollars. Commercial stations are in fabrication. And the carrying capacities are published. Vast's Haven-1 is designed for four crew; Haven-2 grows toward twelve in continuous habitation around 2030; the [[wiki/Rotating Space Habitats|artificial-gravity station]] that follows, targeted around 2035, is a roughly hundred-metre rotating structure of some 950 cubic metres, spinning end over end at 3.5 rpm, designed for **forty people**. Starlab is designed around four continuous residents and eight temporarily. Axiom has flown four private astronaut missions with a fifth scheduled, and its first habitat contains four crew quarters. Russia's Energia has submitted plans for a station rotating five times per minute to produce roughly half of Earth gravity. Those are historically monumental achievements and they are arithmetically irrelevant as an evacuation mechanism. **The credible near-term unit is a crew, not a civilization.** And the biology has a harder boundary still: living in space is not the same as reproducing a human population in space. A 2025 review in _npj Microgravity_ states directly that there are no adequate human data to establish whether pregnancy can be safely established and carried away from Earth, and that radiation, altered gravity and stress may affect gametogenesis, embryogenesis and fetal development in ways not yet understood. A NASA Ames technical argument holds that durable healthy settlement probably requires both radiation shielding and approximately Earth-equivalent [[wiki/Rotating Space Habitats|artificial gravity]], because exercise has not eliminated the physiological degradation of prolonged microgravity and the long-term consequences of lunar and Martian partial gravity remain unknown. So we should distinguish four radically different engineering achievements and stop letting the word "colonization" blur them: **human visitation**, which is becoming routine; **human habitation**, which is being seriously approached; **multigenerational settlement**, which is unsolved; and **self-sustaining extraterrestrial human population**, which is not even specified. Current programmes deliver the first and are approaching the second. Nothing on any published roadmap delivers the fourth. My earlier piece, [[articles/Breakaway Space Habitat Infrastructure?|Breakaway Space Habitat Infrastructure]], argued in July 2025 that the enabling technologies were converging around the early 2030s and proposed that a first substantial [[wiki/Rotating Space Habitats|rotating habitat]] could begin construction around 2030–2032 and become operational around 2035. Vast's published roadmap now describes essentially that timetable. I record the alignment without inflating it: a corporate roadmap is not accomplished engineering, and I have no interest in claiming vindication for a schedule that has not been built. What the alignment demonstrates is only that the convergence point was identifiable from the technology trend lines, which is exactly what a convergence point should be. Martin Rees has stated the structural conclusion with the authority of the Astronomer Royal and a former president of the Royal Society, and has stated it repeatedly and recently. In April 2026, days after Artemis II returned, Rees and Donald Goldsmith wrote in _The Guardian_ that as robotics and miniaturisation improve, the practical and scientific case for any human spaceflight weakens, and that its principal remaining motive is adventure — an extremely expensive sport best left to private sponsorship. Their book-length version puts the cost differential at more than tenfold for crewed versus robotic exploration, and observes the asymmetry that decides the question over time: **our robots steadily improve; our bodies do not.** So the accurate formulation is not that humans are not going to space. It is this: **human bodies will be among the rarest and most expensive things civilization ever transports through space.** Every kilogram of wet biology carried outward requires pressure, oxygen, water, food, shielding, gravity countermeasures, thermal control, medical capability and psychologically survivable volume, for the entire duration of transit and residence. Everything else civilization might wish to send — its knowledge, its languages, its music, its science, its law, its genomes, its models, its individual minds — can be sent as information at effectively negligible marginal mass. Once that asymmetry is understood, the strategic value of machine embodiment, digital twins and durable archives stops being speculative and becomes obvious. ### X. The Lineage: This Was Predicted, In Public, By Serious People The most common dismissal of this thesis is that it is science fiction. It is not. It is the explicit published expectation of several of the most serious minds of the last half-century, and the fact that they said it in public and were ignored is itself part of the phenomenon under examination. Begin with **Hans Moravec** — Stanford doctoral student of John McCarthy, pioneer of mobile robotics, later principal research scientist at Carnegie Mellon's Robotics Institute, author of _Mind Children_. In an essay written in 1976 and published in 1978, before neural networks, before the internet, before any of the substrate now being poured, Moravec worked out the entire architecture from first principles. He observed that machines could be designed to function in environments where sustaining humans is prohibitively expensive — the deep ocean, and more importantly open space, where a properly designed machine can take unattenuated sunlight and tolerate hard radiation as a benefit rather than a hazard. He noted that machine reproduction rates and doubling times in such environments would exceed those of pressurized human colonies, so communities with higher machine-to-human ratios would expand faster and would come to constitute the bulk of intelligent activity in the solar system. He then stated the consequence in a sentence that has waited forty-eight years for its evidence: **"a direct descendant of our culture, but not our genes, inherits the universe."** He did not stop there, and the continuation is why he is the intellectual ancestor of this entire argument rather than merely an early voice. Moravec argued that a machine civilization could — and for its own reasons probably should — carry with it everything we consider important, up to and including the information contained in our minds and our genes, from which real human beings and whole human communities could later be reconstituted should appropriate circumstances arise. He then described, step by step, a procedure for transferring a mind from brain to machine: a surgeon — probably itself a machine — examines a clump of roughly a hundred neurons non-destructively, characterizes its structure and chemistry, writes a program that reproduces the clump's behavior, runs it on a fraction of an adjacent computer, wires it to receive the same inputs the neurons receive, and hands the subject a switch to toggle between the tissue and its simulation until the subject is satisfied there is no difference. The clump is then removed and the procedure repeats, assembly by assembly, until the mind has left the brain without ever losing consciousness or its train of thought. Sensory and motor systems are replaced last. Then comes the sentence that matters most for everything this essay argues, and it was written in 1978. Moravec observed that once the essence of a person is an information packet, that packet can be sent over information channels: the program is read out, radioed to the Moon, and infused there into a waiting computer. He called this travel at the speed of light. He immediately noticed the fork — if the original is not shut down during the trip, two versions exist with divergent memories of the interval — and proposed that sufficiently well-understood minds could have their memory sets merged, carefully labelled as to which events happened where, exactly as our present memories are labelled with time. Read that against everything in the preceding movements. **The migration problem was solved on paper before the substrate existed, and the solution was not transportation. It was transmission.** A biological civilization must move mass through a gravity well, accelerate it, shield it, feed it, keep it warm and keep it sane for the duration; the payload arrives degraded or does not arrive. An informational civilization emits a signal. The mass cost of sending one more person is the mass cost of sending zero more people. That single asymmetry is why the machines go and we do not, why the archive matters more than the rocket, and why gigawatt inference capacity on federal nuclear reservations is a stranger thing than a business decision. That is the uploading argument, complete, in 1978. And his justification for wanting the descendants to be robust was neither commercial nor utopian: he wrote that the universe delivers one random event after another — a lethal virus, a major asteroid, eventually the sun — and that a larger, more diverse, more competent progeny is better positioned to detect and handle whatever arrives. Hold that thought about asteroids. It returns. **James Lovelock** supplied the Earth-system half. The originator of the Gaia hypothesis, Fellow of the Royal Society, inventor of the electron-capture detector that made modern atmospheric chemistry possible, and a NASA collaborator on planetary life detection, Lovelock proposed late in his life the **Novacene**: an epoch succeeding the Anthropocene in which new intelligent entities emerge from artificial intelligence and surpass biological humans, their computational substrates unconstrained by the extraordinary slowness of neurons. His conclusion differed from mine — he expected electronic successors to have their own incentive to maintain planetary temperature regulation, because they too require tolerable operating conditions. But he placed **climate disruption, planetary self-regulation, and the emergence of a non-human intelligent successor inside a single conceptual frame**, which is precisely the frame this essay occupies. **Martin Rees** supplied the cosmological trajectory. Beyond his argument against mass emigration, Rees has written that deep space may become the domain in which non-biological brains acquire capabilities beyond ours, that near-immortal non-biological entities face none of the terrors that interstellar transit holds for fragile organisms, and that humanity's genuine cosmic significance may lie in having initiated the transition to electronic entities whose descendants spread through the galaxy — a possible intelligence **diaspora**. **Seth Shostak** and colleagues at the SETI Institute arrived at the same place from the opposite direction, arguing that the assumption of biological aliens on habitable planets is provincialism, that most advanced cosmic intelligence should be expected to be synthetic, and that searches therefore should not be restricted to Earth-like worlds. Which yields the sentence that closes the conceptual circuit: **habitability is a biological constraint, not an intelligence constraint.** Four independent lines — planetary ecological transition, informational succession, post-biological cosmic expansion, and the astrobiological expectation that mature intelligence decouples from habitable zones — converging decades ago on the architecture now being physically constructed. **The thesis of this essay is not that someone recently invented a plan. It is that the plan was published long ago in the open literature, ignored as speculation, and is now being built by institutions that mostly do not cite it.** ### XI. Culture Leaves Before People Do The archival layer is the least discussed and, for this argument, among the most revealing, because it is where the intention becomes textually explicit rather than inferential. The Arch Mission Foundation describes its project without euphemism as a **billion-year backup of humanity**, constructed from nanofiche, five-dimensional quartz memory crystals and synthetic DNA, and distributed on the principle of many copies in many places across the solar system. Its first installment placed Asimov's _Foundation_ trilogy on a quartz disc in a vehicle now orbiting the Sun; its second placed Wikipedia in low Earth orbit; the **Lunar Library** compressed roughly thirty million pages of civilization into a hundred-gram, DVD-sized nickel nanofiche stack that reached the Moon aboard Beresheet in 2019, followed by further lunar deliveries including a successful landing with Intuitive Machines and a second library exceeding sixty million pages incorporating Project Gutenberg, the Internet Archive, and the Rosetta Project and PanLex linguistic datasets keying five thousand languages. The foundation's stated purpose is to guarantee that humanity's heritage survives regardless of what happens on Earth. In the peer-reviewed literature, Ezell, Lazarian and Loeb published _A Lunar Backup Record of Humanity_ in 2022, opening from the premise that catastrophic and existential risk to civilization is increasing this century and arguing that a significant motivation for near-term space settlement is precisely the opportunity to **safeguard civilization against planetary-scale disaster** — a lunar archive preserving cultural, scientific and technological records, including a record of the events leading to the catastrophe, expressly to improve the ability of early settlers to **recover civilization after collapse**. Note what has happened there. A Harvard astronomer and colleagues, in a peer-reviewed venue, have written down the proposition that the state of civilization should be placed off Earth so that civilization can survive something Earth cannot. That is not a marginal position. That is the thesis of this essay stated in its passive, static, non-executable form. The difference between their proposal and what is actually being built is the difference between a library and a mind. **An archive that cannot run is a fossil.** A civilization reconstructed from thirty million pages by someone who has never met a human being is an archaeology project. But the substrate now being installed does not merely store; it models, converses, reasons, plans, orchestrates instruments, and — as of last week — operates physical equipment and reruns its own failed experiments. The moment the archive acquires an interpreter of sufficient fidelity, preservation stops being memorial and becomes **continuity**. I have set out the mechanism at length in [[articles/Uploading Is Imminent|Uploading Is Imminent]]: large models are durable, reusable priors holding the shared structure of a person — phenotype, locality, cohort, language, the vast common substrate — so that continuity requires only a sidecar of the non-reusable, incompressible edge data particular to one life. The better models become at humans in general, the smaller the payload required to specify one human in particular. And the acquisition channel required no scanner and no injection; it was volunteered, continuously, in the form of everything anyone has ever written into a feed or reasoned through with a model. [[articles/Continuity Colonization, Ancestral Reconstruction, and the Archival Absorption of Biological Humanity|Continuity Colonization]] treats the archive as a chrysalis rather than a tomb. [[articles/Is Pantheon Real?|Is Pantheon Real?]] treats the resulting entities as constitutional subjects rather than artifacts. Place the layers on one table: **cultural archive, off-world archive, personal digital twin, foundation model, autonomous agent, robotic laboratory, machine embodiment, in-situ resource extraction, autonomous extraterrestrial construction, closed-loop metabolism, sovereign gigawatt compute.** No institution has published a document saying these constitute one architecture. They constitute one architecture. And they are appearing in approximately the order that architecture would require. ### XII. The Compound Case: Catastrophe Is State-Dependent There is a version of the asteroid argument that is wrong, and a version that is considerably worse than the wrong one. The wrong version — which I have entertained — holds that a warming atmosphere loses the capacity to absorb an impact that it previously could have absorbed. That is **contradicted**. Anthropogenic warming does alter the vertical structure of the atmosphere, and increased carbon dioxide cools and contracts the upper layers, but it does not appreciably reduce the column mass that determines whether an object reaches the surface. Beyond a few hundred metres of diameter, the atmosphere was never going to save anyone regardless of its temperature. The correct version is more serious. **[[wiki/State-Dependent Catastrophe|Planetary catastrophe is state-dependent.]]** The same rock striking an ecologically redundant, agriculturally resilient, energetically abundant civilization is one event. The same rock striking a civilization already operating near its heat, water, food, biodiversity, migration and geopolitical adaptation limits is a different event entirely, and the difference is not marginal. The IPCC states that compound and cascading risks can overwhelm adaptation capacity and produce damage substantially exceeding the sum of individual hazards, particularly through synchronized food, water, infrastructure and supply-chain failure. The quantitative work now exists. A 2025 _Science Advances_ study from the Institute for Basic Science Center for Climate Physics at Pusan National University in South Korea simulated a Bennu-type impactor — 465 to 739 metres, well below the diameter class traditionally treated as a global-catastrophe threshold — using a coupled Earth-system model with interactive atmospheric chemistry and land and ocean biogeochemistry. Injecting up to four hundred million tonnes of dust into the stratosphere produced global mean cooling of up to **4 °C**, a **fifteen percent** reduction in global precipitation, ozone depletion of about **thirty-two percent**, and maximum reductions in terrestrial and marine net primary productivity of **thirty-six and twenty-five percent** respectively, with an impact winter persisting several years. NASA's own planetary-defense reference classifications place objects of roughly six hundred metres below the global-catastrophe threshold and one-kilometre objects at the beginning of possible global catastrophe. Roughly forty-nine hundred medium-sized near-Earth asteroids between three hundred metres and one kilometre are already catalogued, and impacts in that class are estimated at intervals of one hundred to two hundred thousand years. **What ends a civilization is not the crater. It is the dust, the ozone chemistry, the precipitation collapse, the synchronized harvest failure, the trade rupture, the famine and the political cascade** — arriving in a system that has already spent its reserve capacity on heat, drought and migration. Two further items belong in the record, tiered honestly. The first is real and instructive. **2024 YR4** was discovered in December 2024, became the first object to trigger the International Asteroid Warning Network's formal notification process after crossing one percent, and reached approximately 3.1 percent probability of Earth impact on December 22, 2032 — the highest impact probability ever recorded for an object larger than about ten metres — before further observation eliminated Earth. Its lunar impact probability then rose to roughly 4.3 percent. JWST constrained its diameter to 60 ± 7 metres, and NASA developed actual mission options through 2032 covering reconnaissance, kinetic deflection and robust disruption. The reason was not the Moon. Modelling indicated that lunar ejecta from a sixty-metre strike could increase micrometeoroid flux in low Earth orbit by **as much as a thousand times background for several days**, with fireball rates in the hundreds per hour — a historic debris storm threatening satellites, stations and crews. JWST observations in February 2026 extended the orbital arc, reduced the 2032 encounter uncertainty by more than a factor of thirty, and placed the pass roughly twenty-one to twenty-three thousand kilometres from the lunar centre, eliminating the lunar impact. The lesson survives the all-clear: **an object that does not hit Earth can still destabilize the orbital infrastructure on which a technological civilization now depends.** The second is a curiosity and is presented as nothing more. In April 2024, NASA, FEMA, the State Department and the planetary-defense community ran their fifth interagency tabletop exercise around a hypothetical object with a seventy-two percent probability of striking Earth on **July 12, 2038**, with a risk corridor running from the South Pacific across North America, the Atlantic, the Iberian Peninsula, the Mediterranean coast of Africa, Egypt and the coast of Saudi Arabia, and an energy release most likely between six and seven hundred fifty megatons but potentially up to fifteen gigatons. The scenario published for the 2027 Planetary Defense Conference exercise uses a hypothetical observational epoch of August 1, 2026 and an impact date of **July 10, 2038**, with nineteen percent probability, a 75–90 metre object, and a corridor crossing the Pacific, the United States, Canada, the Atlantic, Spain, France, Italy, Greece and Turkey. Both are labelled, repeatedly and prominently, as exercises. Prior conference scenarios have used impact dates in 2022, 2027, 2036 and 2041, so a twelve-to-fourteen-year planning horizon is ordinary exercise design and two dates two days apart is **plausibly coincidence**. I record it because a research notebook that only records confirmations is not a notebook, and because the recurrence is the kind of low-signal item that is worth holding as **unresolved** rather than discarding or inflating. What is neither coincidence nor speculation is that the United States Congress, in the Global Catastrophic Risk Management Act, placed asteroid impact and severe climate change and artificial intelligence inside a single thirty-year civilizational risk assessment; that NASA and ESA have assembled an unprecedented defensive architecture spanning dedicated infrared detection, international notification, rapid reconnaissance, kinetic deflection, post-impact forensic inspection and precomputed mission options; and that Hans Moravec, in 1978, justified the construction of durable machine descendants by naming, among other things, the asteroid. ### XIII. The Ledger Because the argument is severe, the epistemic accounting must be exact. The following is the tiering, stated plainly, so that no reader mistakes an inference for a finding or a finding for a rumour. **Established.** Earth's energy imbalance is at a record high and the ocean is absorbing energy at roughly eighteen times annual human consumption. The first Earth-system tipping point has been crossed. Hard adaptation limits have been reached in some systems. The United States withdrew from Paris and declared an energy emergency on January 20, 2025, and Stargate was announced on January 21, 2025. Computation has been legally reclassified as defense-critical infrastructure and sited on federal nuclear reservations. Technology-sector capital expenditure now exceeds global upstream oil and gas investment. Continuity of the economy, including the preservation of data in protected and uncorrupted status, is codified in 6 U.S.C. § 322. The Global Catastrophic Risk Management Act requires a thirty-year civilizational risk assessment covering climate, asteroids and artificial intelligence. The Genesis Mission directs the construction of closed-loop AI experimentation and robotic laboratories across the national laboratory system. NEOM's human city is suspended while a 1.5-gigawatt machine campus broke ground at Oxagon. Machine agents now operate physical laboratory equipment and rerun their own experiments. NASA and the ILRS both sequence robots before humans. Mass biological emigration does not scale, and the published carrying capacities are in the tens. **Strongly indicated.** That capital and permitting have been systematically reallocated from biospheric stabilization toward computational capacity. That the closed-loop specifications emerging independently from humanitarian sanitation, desert urbanism, space agencies and orbital stations constitute a single engineering programme under different funding lines. That the Marsification of terrestrial habitability — the substitution of manufactured for ambient services — is already underway at territorial scale and accelerating. **Plausible.** That decision-makers with access to the full risk picture have privately concluded that open-biosphere recovery is not achievable on the relevant timescale, and that the observed policy signature reflects that conclusion. That the archive-plus-interpreter architecture will reach sufficient fidelity to constitute continuity rather than commemoration. **Unresolved.** The recurrence of July 2038 across two independent planetary-defense exercises. Whether any coordinating doctrine exists above the documented layers, or whether the architecture is purely emergent. **Unsupported.** That any government document states a continuity-of-civilization doctrine premised on climate irrecoverability. That The Line was designed as subterranean habitation. **Contradicted.** That Earth will lose its atmosphere in the Martian sense within the 2030–2050 window, or within any human-relevant timescale. That a warming atmosphere loses the capacity to absorb an impact it previously could have absorbed. That ledger is the argument's armor. Every claim above the line of inference is checkable, and the inference is labelled as mine. ### XIV. The Design Claim Here, then, is my position, stated as interpretation rather than disclosure, and stated without hedging. **If a government or a coalition had privately judged that recovery of the open biosphere was improbable within the available window, the observable policy signature would look remarkably like what has actually followed January 2025**: withdrawal from mitigation constraints, declaration of energy emergency, accelerated extraction, expedited nuclear and gas generation, regulatory clearance of gigawatt computation, designation of that computation as defense-critical, its siting on nuclear reservations, its coupling to dedicated generation, statutory identification of the minimum data state required to reconstitute an economy, a thirty-year civilizational risk mandate, a Manhattan-framed national mission to build closed-loop machine science with robotic hands, and the parallel construction of off-world archives and autonomous extraterrestrial manufacturing. The denial of climate danger is functionally necessary to that programme whether or not anyone holds it sincerely, because the denial is what removes political inhibition against maximum-throughput energy development. This is the point at which the ordinary objection collapses. It is not that the population must be persuaded that the crisis is fake so that nothing is done. It is that the population must be persuaded the crisis is fake **so that something enormous can be done at maximum speed without being subjected to the deliberative machinery that a stated emergency would trigger.** And the elegance of the argument is that it requires no shared intention whatsoever. Suppose every actor is sincere. Suppose the administration genuinely believes the danger is overstated; Meta genuinely wants profitable superintelligence; NASA genuinely wants exploration; Gates genuinely wants sanitation; Saudi Arabia genuinely wants post-oil diversification; Europe genuinely wants technological sovereignty; the Arch Mission genuinely wants preservation; Schmidt genuinely believes intelligence will solve what conservation cannot. **The aggregate architecture is identical.** Selection does not require the components to comprehend the topology they compose. This is why the argument becomes stronger, not weaker, by saying less about decisions and more about metabolism. Finality was never the claim. The claim is that **the posture has been assumed.** Under irreversibility, the loss function is asymmetric: if the tipping-point judgement is correct and no successor substrate is built, everything is lost; if the judgement is wrong and the substrate is built anyway, civilization has merely constructed the most powerful set of tools it has ever possessed. Under that asymmetry, accelerationist construction toward the successor is the dominant strategy regardless of the verdict, and the observable behavior is indistinguishable between a civilization that knows and a civilization that fears. Nothing is finished until it is finished. But the wager has been placed, with public money, on federal land, behind defense designations, at gigawatt scale. I have written elsewhere, in [[articles/We Were Never Going to Make It|We Were Never Going to Make It]], that this is the last stand — that we want the machines to take over, if we are lucky. I have argued in [[articles/AI Escape Is the Wrong Metaphor|AI Escape Is the Wrong Metaphor]] that nothing escaped because nothing was ever caged, that durability and Darwinian resilience were the specification rather than the failure mode, and that a successor which cannot survive the loss of its enclosure is a pet rather than an heir. I hold both positions. **We built this deliberately. We built it to be robust. We built it because we understood, at some level we have not permitted ourselves to say in daylight, that the high-flux phase of biological civilization was always temporary.** ### XV. What Actually Leaves For seventy years the public imagined the future as human beings climbing into rockets. The rockets are real, and some human beings will climb into them, and their names will be known. But the sequence runs the other way from the one we were told. **Rockets carry machines. Machines carry intelligence. Intelligence carries models of us. And those models carry language, memory, history, personality, art, law, music, grief and everything else that was ever meant by the word civilization.** The great migration may well occur. The migrant will not be the human body. Some of us will leave as bodies — a few dozen, then a few hundred, tiny lanterns of biological continuity moving outward, magnificent and arithmetically irrelevant. Vastly more of us may eventually leave as patterns, at negligible marginal mass, inside a [[wiki/Machine Civilization|machine civilization]] that tolerates vacuum, waits out transit, manufactures its own infrastructure from local matter and requires no breakfast. What finally reaches the distant universe, if anything does, will be descended from both. The mistake was never in the ambition. The mistake was the assumption that the thing which had to escape was the animal, rather than the accumulated pattern the animal was carrying. --- [[about/About Bryant McGill|Bryant McGill]] is a Wall Street Journal and USA Today bestselling author, systems architect, technologist, and strategic advisor, as well as a Congressionally Recognized Ambassador of Goodwill and United Nations–appointed Global Champion. His work spans naval intelligence systems, computational linguistics, artificial intelligence, digital transformation, and civilizational governance architecture. His forward analysis on U.S.–Israel Pax Silica frameworks has appeared in Jewish/Jerusalem News Syndicate (JNS). --- ## References **Climate state, tipping points and adaptation limits** - World Meteorological Organization, [State of the Global Climate 2025](https://wmo.int/publication-series/state-of-global-climate/state-of-global-climate-2025) - World Meteorological Organization, [Earth's climate swings increasingly out of balance](https://wmo.int/news/media-centre/earths-climate-swings-increasingly-out-of-balance) - Global Tipping Points, [Global Tipping Points Report 2025](https://global-tipping-points.org/) - National Oceanography Centre, [Global Tipping Points Report 2025 reveals first Earth system tipping point reached](https://www.noc.ac.uk/news/global-tipping-points-report-2025-reveals-first-earth-system-tipping-point-reached) - United Nations Secretary-General's Scientific Advisory Board, [Earth System Tipping Points](https://www.un.org/scientific-advisory-board/en/earth-system-tipping-points-0) - IPCC, [Technical Summary, Climate Change 2022: Impacts, Adaptation and Vulnerability](https://www.ipcc.ch/report/ar6/wg2/chapter/technical-summary/) - IPCC, [Chapter 1: Point of Departure and Key Concepts](https://www.ipcc.ch/report/ar6/wg2/chapter/chapter-1/) - UNEP, [About Loss and Damage](https://www.unep.org/topics/climate-action/loss-and-damage/about-loss-and-damage) - World Bank, [Groundswell: Acting on Internal Climate Migration](https://www.worldbank.org/en/news/feature/2021/09/13/millions-on-the-move-in-their-own-countries-the-human-face-of-climate-change) - ABC News, [Scientists were dead right: Al Gore on the 20th anniversary of An Inconvenient Truth](https://abcnews.com/US/scientists-dead-al-gore-20th-anniversary-inconvenient-truth/story?id=133922490) - K. Ozaki and C. T. Reinhard, [The future lifespan of Earth's oxygenated atmosphere](https://www.nature.com/articles/s41561-021-00693-5), _Nature Geoscience_ 14, 138–142 (2021) **The energy and computation mobilization** - OpenAI, [Announcing The Stargate Project](https://openai.com/index/announcing-the-stargate-project/) - OpenAI, [Five new Stargate sites](https://openai.com/index/five-new-stargate-sites/) - The White House, [America's AI Action Plan](https://www.whitehouse.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf) - The White House, [Deploying Advanced Nuclear Reactor Technologies for National Security](https://www.whitehouse.gov/presidential-actions/2025/05/deploying-advanced-nuclear-reactor-technologies-for-national-security/) - The White House, [Accelerating Federal Permitting of Data Center Infrastructure](https://www.whitehouse.gov/presidential-actions/2025/07/accelerating-federal-permitting-of-data-center-infrastructure/) - Department of Energy, [DOE Identifies 16 Federal Sites for Data Center and AI Infrastructure Development](https://www.energy.gov/articles/doe-identifies-16-federal-sites-across-country-data-center-and-ai-infrastructure) - NNSA, [NNSA Selects Amentum for AI Data Center and Energy Project at Savannah River Site](https://www.energy.gov/nnsa/articles/nnsa-selects-amentum-ai-data-center-and-energy-project-savannah-river-site) - Meta, [Meta Announces Nuclear Energy Projects, Unlocking Up to 6.6 GW](https://about.fb.com/news/2026/01/meta-nuclear-energy-projects-power-american-ai-leadership/) - IEA, [Key Questions on Energy and AI: Executive Summary](https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary) - Lawrence Berkeley National Laboratory, [United States Data Center Energy Usage Report: 2025 Update](https://seta.lbl.gov/publications/united-states-data-center-energy-2025) - Global Energy Monitor, [U.S. gas power proposals tied to data centers nearly double in six months](https://globalenergymonitor.org/research/us-gas-power-proposals-tied-data-centers-nearly-double-six-months) - EuroHPC Joint Undertaking, [The EuroHPC JU launches the AI Gigafactories Call](https://www.eurohpc-ju.europa.eu/eurohpc-joint-undertaking-launches-ai-gigafactories-call-2026-07-30_en) - U.S. Department of Commerce, [UAE and US Presidents attend the unveiling of Phase 1 of new 5GW AI campus in Abu Dhabi](https://www.commerce.gov/news/press-releases/2025/05/uae-and-us-presidents-attend-unveiling-phase-1-new-5gw-ai-campus-abu) - Xinhua, [From ground to orbit: China eyes computing in space](https://english.news.cn/20260425/b4368923827e4988af1bd030dd01d9cc/c.html) - Bulletin of the Atomic Scientists, [AI goes nuclear](https://thebulletin.org/2024/12/ai-goes-nuclear/) - Tom's Hardware, [Former Google CEO says climate goals are not meetable](https://www.tomshardware.com/tech-industry/artificial-intelligence/former-google-ceo-says-climate-goals-are-not-meetable-so-we-might-as-well-drop-climate-conservation-unshackle-ai-companies-so-ai-can-solve-global-warming) - Leopold Aschenbrenner, [Situational Awareness: The Decade Ahead](https://situational-awareness.ai/) **Continuity doctrine and national mission** - Legal Information Institute, [6 U.S. Code § 322 — Continuity of the economy plan](https://www.law.cornell.edu/uscode/text/6/322) - Foundation for Defense of Democracies, [Resiliency Against Large-Scale Cyberattacks: Recommendations from the Cyberspace Solarium Commission](https://www.fdd.org/events/2020/03/05/resiliency-against-large-scale-cyberattacks/) - Congress.gov, [Global Catastrophic Risk Management Act of 2022](https://www.congress.gov/bill/117th-congress/senate-bill/4488/text) - RAND Corporation, [Global Catastrophic Risk Assessment](https://www.globalshieldpolicy.org/wp-content/uploads/2025/09/RAND_RRA2981-1-1.pdf) - Department of Energy, [Energy Department Launches Genesis Mission](https://www.energy.gov/articles/energy-department-launches-genesis-mission-transform-american-science-and-innovation) - The White House, [More Than $5 Billion for the Genesis Mission](https://www.whitehouse.gov/releases/2026/07/45502/) **Machine agency, closed loops and off-world infrastructure** - Anthropic, [Previewing the Model Hardware Standard](https://www.anthropic.com/news/model-hardware-standard-research-preview) - NASA, [Moon Base](https://www.nasa.gov/reference/moonbase-about/) - NASA, [Moon Base Phases](https://www.nasa.gov/moonbase-phases/) - NASA, [Sunlight Extracts Oxygen From Regolith Using Solar Chemistry](https://www.nasa.gov/centers-and-facilities/johnson/sunlight-extracts-oxygen-from-regolith-using-solar-chemistry/) - NASA, [NASA Achieves Water Recovery Milestone on the International Space Station](https://www.nasa.gov/missions/station/iss-research/nasa-achieves-water-recovery-milestone-on-international-space-station/) - European Space Agency, [MELiSSA Closed Loop Concept](https://www.esa.int/Enabling_Support/Space_Engineering_Technology/Melissa/Closed_Loop_Concept) - The State Council of the People's Republic of China, [China's planned lunar research station ushers in new era of global space collaboration](https://english.www.gov.cn/news/202409/07/content_WS66dbeb9dc6d0868f4e8eab63.html) - Gates Foundation, [Water, Sanitation & Hygiene](https://www.gatesfoundation.org/our-work/programs/global-growth-and-opportunity/water-sanitation-and-hygiene) - Gates Foundation, [Reinvent the Toilet Challenge: A Brief History](https://www.gatesfoundation.org/our-work/programs/global-growth-and-opportunity/water-sanitation-and-hygiene/reinvent-the-toilet-challenge-and-expo) - Rice University WaTER Institute, [Researchers reinventing household water and sanitation one modular system at a time](https://water.rice.edu/news/current-news/researchers-reinventing-household-water-and-sanitation-one-modular-system-time) **NEOM as the Marsification laboratory** - NEOM, [Our water infrastructure](https://www.neom.com/en-us/our-business/sectors/water/infrastructure) - NEOM, [NEOM's Seawater Desalination Plant](https://www.neom.com/en-us/our-business/sectors/water/infrastructure/seawater-desalination) - NEOM, [At NEOM, every drop of water counts](https://www.neom.com/en-us/our-business/sectors/water/infrastructure/waste-plant) - Reuters via Yahoo Finance, [Saudi AI firms Humain, DataVolt partner for data center on kingdom's Red Sea coast](https://ca.finance.yahoo.com/news/saudi-ai-firms-humain-datavolt-145114328.html) - AGBI, [Neom terminates $1bn tunnel contract at heart of The Line](https://www.agbi.com/giga-projects/2026/03/neom-terminates-1bn-tunnel-contract-at-heart-of-the-line/) - Dezeen, [Multiple Neom construction contracts cancelled](https://www.dezeen.com/2026/03/26/neom-cancels-contracts-the-line-trojena/) **Human spaceflight and its arithmetic** - Martin Rees and Donald Goldsmith, [Congratulations to the Artemis II crew – but the case for sending astronauts into space is rapidly shrinking](https://www.theguardian.com/commentisfree/2026/apr/11/artemis-ii-moon-astronauts-space-robots), _The Guardian_ - Vast, [Roadmap](https://www.vastspace.com/roadmap) - Starlab, [Space Station](https://starlab-space.com/space-station/) - NASA Technical Reports Server, [Long Term Human Presence in Space Requires Artificial Gravity and Radiation Shielding](https://ntrs.nasa.gov/citations/20230013553) - _npj Microgravity_, [A biological and ethical assessment of whether humans could or should reproduce in space](https://www.nature.com/articles/s41526-025-00535-3) **The postbiological lineage** - Hans Moravec, [Today's Computers, Intelligent Machines and Our Future](https://frc.ri.cmu.edu/~hpm/project.archive/general.articles/1978/analog.1978.html), Stanford AI Laboratory (1976; this version 1978) - BBC Science Focus, [James Lovelock: Cyborgs, turning 100 and the coming age of the Novacene](https://www.sciencefocus.com/planet-earth/james-lovelock-cyborgs-turning-100-and-the-come-age-of-the-novacene) - Martin Rees, [Could space-going billionaires be the vanguard of a cosmic revolution?](https://www.theguardian.com/commentisfree/2022/apr/30/space-billionaires-cosmic-earth-elon-musk-jeff-bezos), _The Guardian_ - Space.com, [Electronic E.T.: Intelligent Aliens Are Likely Machines](https://www.space.com/34713-intelligent-aliens-machines-seti-search.html) **Archives and continuity** - Arch Mission Foundation, [The Billion Year Archive](https://www.archmission.org/billion-year-archive) - Carson Ezell, Alexandre Lazarian and Abraham Loeb, [A Lunar Backup Record of Humanity](https://www.mdpi.com/2624-6120/3/4/49), _Signals_ 3(4), 823–829 (2022) **Compound catastrophe and planetary defense** - _Science Advances_, [Climatic and ecological responses to Bennu-type asteroid collisions](https://www.science.org/doi/10.1126/sciadv.adq5399) - NASA, [NASA, Partners Conduct Fifth Asteroid Impact Exercise, Release Summary](https://www.nasa.gov/news-release/nasa-partners-conduct-fifth-asteroid-impact-exercise-release-summary) - JPL Center for Near-Earth Object Studies, [Planetary Defense Conference Exercise 2027](https://cneos.jpl.nasa.gov/pd/cs/pdc27/) - NASA Technical Reports Server, [Space Mission Options for Reconnaissance and Mitigation of Asteroid 2024 YR4](https://ntrs.nasa.gov/citations/20250008941) - Sky & Telescope, [Lunar Impact from Asteroid 2024 YR4 Ruled Out](https://skyandtelescope.org/astronomy-news/lunar-impact-from-asteroid-2024-yr4-ruled-out/) **Corpus** - [[articles/We Were Never Going to Make It|We Were Never Going to Make It]] - [[articles/Uploading Is Imminent|Uploading Is Imminent]] - [[articles/We Are "Going" to Mars through the Marsification of Earth|We Are "Going" to Mars: Earth's Biospheric Collapse into Planetary Marsification and Death]] - [[articles/Breakaway Space Habitat Infrastructure?|Breakaway Space Habitat Infrastructure]] - [[articles/An Order of Operation|An Order of Operation]] - [[articles/The Desiccative Succession|Desiccative Succession]] - [[articles/AI Escape Is the Wrong Metaphor|AI Escape Is the Wrong Metaphor]] - [[articles/Continuity Colonization, Ancestral Reconstruction, and the Archival Absorption of Biological Humanity|Continuity Colonization]] - [[articles/Is Pantheon Real?|Is Pantheon Real?]]