Civilization Resilience Planning

Civilization resilience planning asks a question disaster management never has to: what is worth doing in advance about a catastrophe severe enough that no functioning outside remains to send help. The scenario set is small and specific — nuclear war, a supervolcanic eruption, an asteroid impact, a severe pandemic, an electromagnetic pulse or an engineered biological agent — and it is unified by a single mechanism rather than by a single cause: an abrupt reduction in the sunlight reaching crops, or an abrupt loss of the industrial substrate that moves food from where it grows to where it is eaten.

Established The field has four working programmes and one hole where its foundation should be. The programmes are quantified famine modelling, resilient or alternative foods, germplasm and knowledge preservation, and refuge and recovery analysis. The hole is the restart question. Everything the field does presupposes an answer to whether an industrial society, once stopped, can be started again — and this brief could find no peer-reviewed paper that asks it. That absence is the most interesting thing on this page and it is treated as a finding rather than as an embarrassment.

Scope boundary. The failure, cascade and hardening behaviour of built networks is infrastructure resilience, on which this brief depends and which it does not restate. The institutional machinery for governing low-probability high-consequence risk is existential risk governance. Earth-system limits under normal operation are planetary stewardship. What is here is the deliberate preparation: what is stockpiled, what is preserved, what is modelled, and what has never once been exercised.

How to read the confidence. Frontier A large share of the primary literature behind this brief was reachable only as publisher-deposited abstracts, and a further share only as verified bibliographic records whose text was never obtained. Where a source is record-only it is named and no number is taken from it. That is stated up front because it changes what the page can honestly claim, and because the pattern of what was unreachable is itself informative about how this literature is distributed.

Established The field's single most citable quantity is a famine estimate, not a blast estimate, and it comes from a model chain. Xia and colleagues, in Nature Food in 2022, coupled climate, crop and fishery models across six scenarios of stratospheric soot injection and accounted national calorie availability after stored food is consumed. The deposited abstract states that soot injections larger than 5 Tg would lead to mass food shortages in almost all countries, that livestock and aquatic production would be unable to compensate for reduced crop output, and that more than 2 billion people could die from a nuclear war between India and Pakistan and more than 5 billion from a war between the United States and Russia. Established The abstract also records a negative result that gets quoted far less often than the death tolls: adaptation measures such as food waste reduction would have limited impact on increasing available calories. Frontier The step change relative to the older nuclear-winter literature is resolution — national outcomes rather than global averages — and the price of that resolution is a four-stage model chain whose integrated output has no validation set. The aerosol-climate link in that chain has been validated against volcanic eruptions; the calorie-accounting link at the end has been validated against nothing, because nothing like it has happened.

Frontier The resilient-foods programme is the field's most quantitative limb and it has become quantitative recently. García Martínez, Behr, Pearce and Denkenberger's 2024 review sets out the state of interventions for global catastrophic food shocks, and its abstract is explicit that the scenario set includes not only sunlight reduction but nuclear electromagnetic pulses and extreme pandemics disrupting industry and mass-scale food supply, with the observation that global food storage is limited. Established That framing is the direct bridge from infrastructure failure to civilisational risk, and it is the reason this brief depends on the infrastructure-resilience brief rather than merely resembling it. Frontier Monteiro and colleagues' 2024 land-suitability computation finds approximately 515 million hectares fit for cropland expansion under an abrupt sunlight reduction scenario, with the potential to feed 20 to 39 per cent of the global population — a partial mitigation that states its own shortfall in the same sentence. Speculative Rivers and colleagues' 2022 preprint reports a 95 per cent probability of global food availability between 2,100 and 3,500 kcal per capita daily at a cost given as around $1.73 daily; that figure is a preprint, a model output conditional on successful deployment, and its unit is genuinely ambiguous in the fragment obtainable here — per capita, per day, in which currency year is not disambiguated, and this brief does not resolve it. Speculative Jehn and colleagues propose seaweed as a resilient food on the grounds that it grows quickly across a wide range of environmental conditions, which is a biological argument and not yet a supply-chain one.

Frontier The most politically awkward result in the programme is that maintaining trade mitigates famine. Rivers and colleagues published that finding in Global Food Security in 2024, and the title states it: food system adaptation and maintaining trade could mitigate global famine in abrupt sunlight reduction scenarios. Established This brief verified the record and could not obtain the text — the journal's publisher deposits no abstracts to the bibliographic channel that was available — so the magnitude of the trade effect is not stated here. Frontier What can be said without the text is that the default assumption in collapse scenarios is autarky, that autarky is the pessimistic branch, and that a finding pointing at trade maintenance sits badly with the intuition that resilience means self-sufficiency. Speculative It also sits badly with the degrowth strand of the same field, and that tension is examined below.

Established The best-verified preservation datum in this subject is an inventory count, and it is in the right unit. Asdal, writing in Genetic Resources in 2025, describes the Svalbard Global Seed Vault as the largest safety backup of the world's crop diversity, holding 1,331,458 seed samples from 123 institutes across 87 countries since 2008, with European institutions having deposited 178,999 samples. Established The unit is samples — accessions — and each accession contains many seeds; the widely circulated reading of that figure as a seed count is wrong by two to three orders of magnitude in one direction, and the equally common phrasing “over a million seeds” understates the holding while sounding like a boast. Established The same abstract records the vault's one real-world use: it proved vital when ICARDA relocated because of Syria's civil conflict. Frontier That event is genuine evidence that the system works for what it was designed to do, and no evidence at all for the thing it is popularly credited with, because ICARDA was a functioning research institution drawing on its own backup after relocating, not an agriculture being restored after collapse. Frontier Asdal is the vault's coordinator, so this is an authoritative operator statement rather than an independent audit, and it should be read as one.

Frontier How long the deposited seeds remain viable is being measured rather than known. Brodal and Asdal's 2018 record announces an ongoing hundred-year seed storage experiment at the vault; this brief verified the record and did not obtain the text, so no interim result is reported here. Established The honest statement is therefore that the answer to the most obvious question about a seed vault is roughly a decade into a century-long measurement. Frontier A 2025 record notes baobab and acacia among recent African deposits, which signals extension beyond staple crops to tree species — relevant because a large fraction of tree germplasm is recalcitrant and cannot be dried and frozen by the conventional method, though precisely which species are storage-recalcitrant could not be verified in this pass and no list is given here. Speculative A substantial critical social-science literature reads the vault primarily as an object of symbolism and visual culture; von Verschuer's chapters and Craciun's essay on the vault's epic imagery are the located examples, and their central claim — that the symbolic function exceeds and partly displaces the practical one — is a serious minority position rather than a dismissal.

Established Knowledge preservation has a structural problem the field states plainly about itself: it is an operating-cost problem. The standing strategic dichotomy in digital preservation is migration versus emulation, and the Springer volume that formalises both devotes separate chapters to each. Established Both presuppose an unbroken custodial chain — migration because somebody must keep migrating, emulation because somebody must keep the emulator running. Established Chapman's chapter title states the constraint without needing its text: it is money that matters in long-term preservation. Frontier An archive that requires a living institution is not a catastrophe-resilient archive, and every strategy the field has is of that kind. Speculative Jeffrey's A new Digital Dark Age? supplies the appropriate disciplinary voice — archaeology is the field that actually studies what survives — and its title is a question, which is how it should be cited.

Frontier The recovery literature is small, recent, and honest about being underdetermined. Baum and thirteen co-authors formalised long-term trajectories of human civilization as an academic area in foresight in 2019, examining four trajectory types: status quo, catastrophe, technological transformation and astronomical. Speculative Fourteen authors declaring a field is a real institutional event and not the same thing as the field having results, and the paper is best cited for what it is. Frontier Jehn's working note frames recovery time as a three-parameter question — the starting level of complexity, the extent of the loss, and the desired recovery level — which is exactly the right framing and is also a statement that no single recovery-time number is supportable. Frontier Belfield's chapter Collapse, Recovery, and Existential Risk is the most directly on-topic treatment located; its record is verified and its text was not obtained, so nothing is attributed to it here beyond its existence and subject.

Frontier The refuge literature is methodologically respectable and scoped more narrowly than its reputation. Boyd and Wilson's 2021 paper in Risk Analysis optimises island refuges and analyses Australia, New Zealand and Iceland; its title scopes explicitly to catastrophic and existential biological threats. Speculative Turchin and Green's earlier papers argue that the sheer number and diversity of islands raises the chance one provides protection, and separately examine aquatic refuges. Frontier The scoping matters because the scenarios with the largest quantified death tolls are agronomic and atmospheric, and an island is not isolated from the stratosphere. Established Whether refuge rankings survive substitution of an agronomic scenario for a biological one is untested, and the test is arithmetic rather than research.

Frontier One recent result imports infrastructure-hardening method directly into catastrophe scenarios, and it is the clearest quantified engineering parameter in the subject. Williams and colleagues' 2025 preprint on subterranean water infrastructure under an abrupt sunlight reduction scenario reports peak freeze depths exceeding 30 m, vulnerability affecting over 2 billion people, and 5.4 to 8.8 million kilometres of pipe. Established Those are thermodynamic and inventory quantities, checkable today, requiring no catastrophe to validate. Frontier They are also the model for what the rest of this field's claims could look like if it disaggregated them.

Handwave The load-bearing argument in this entire subject has no literature, and this brief is going to dwell on that rather than write around it. The argument runs: the first industrial revolution was bootstrapped from coal seams reachable with hand tools and drainable with primitive pumps; those seams are gone; a collapsed civilisation would face an energy ladder with its bottom rungs sawn off, and could not climb from muscle power back to an electrified economy. Every remaining hydrocarbon resource requires deep drilling, offshore platforms or unconventional stimulation — that is, requires the industrial base that has just been lost. Handwave The argument is widely made. It is made in popular writing on collapse, in existential-risk discussion, and informally by people who work on preservation and recovery. It is load-bearing: if it is right, the honest civilisational-resilience posture collapses to prevention-only, and preservation programmes become memorials. Established This pass located no peer-reviewed source stating, testing or refuting it. Not a paper that supports it, not a paper that rebuts it, not a review that mentions it as an open question.

Frontier The absence may be retrieval and may be real, and the two possibilities have different implications, so both are stated. The retrieval channel available for this work was strongly publisher-dependent, and one large commercial publisher deposits no abstracts at all to it, which means a substantial slice of relevant engineering and energy-economics literature would have been invisible even if it existed. Could not locate is not does not exist, and this brief has been burned by that distinction before. Speculative Against that: the argument is not obscure, it would be cited by the recovery literature if it existed, and the recovery literature reachable here does not cite it. Belfield's chapter is the one place it would most naturally live, and its text was not obtained — so the honest statement is that the single most likely home for the missing analysis is also the single most conspicuous unread item on this page. Established Either way the situation is remarkable. A claim this load-bearing should have a literature, and if it has one, four search strategies did not find it.

Speculative The case against the restart-impossibility argument is strong enough that its absence from the record cuts both ways. A collapsed civilisation inherits an enormous materials endowment the first industrialisation did not have: above-ground refined metal in the form of scrap, which is dramatically easier to work than ore, and which requires no mining at all. Surface and shallow coal remains in many regions and the claim that it is exhausted everywhere is a geological assertion nobody appears to have evaluated. Solar photovoltaics, hydro and biomass are non-fossil energy ladders that did not exist in 1750. Speculative The crux is narrower than the popular version of the argument: it is whether a non-fossil ladder can bootstrap the metallurgy required to make more of itself, at a scale and quality that closes. That is a concrete engineering question with a determinate answer, and it appears not to have been asked.

Frontier Meanwhile the tractable frontier is quantitative and moving. Famine estimation has gone national-resolution and multi-model. The resilient-foods portfolio has become a set of costed land, biomass and marine pathways rather than a list of ideas. Recovery time has been reframed from a number into three parameters. Refuge analysis has reached a mainstream risk journal. Infrastructure hardening for catastrophe scenarios has produced its first hard parameter. Frontier And the field has begun reaching operational government — a 2025 presentation on food security and global catastrophic risk to a subnational authority in New Zealand is a small datum with a large implication, which is that this work has started to be commissioned by people who have to act on it.

Speculative A newer strand relocates the risk from physical destruction to institutional continuity. Yuan's 2026 working paper distinguishes three nested layers — the biological species, civilisation as an intergenerational operating environment, and philosophy as a reflective kernel — and argues that the risk from artificial intelligence is that it irreversibly weakens the civilisational conditions under which humanity continues as a subject of truth, responsibility, agency and meaning. Handwave That is far from evidence and it is coherent, and it points at the same variable the preservation literature points at from the other direction: what fails is the organisation, not the artefact. Speculative Sandberg's argument that there are distinctively civilisation-level virtues is the normative version of the same move, and it makes explicit that recovery targets are chosen rather than discovered.

Established The target state is a credible civilisational restart capability: a demonstrated, audited capacity to re-establish an industrial society from a severely degraded start, with the pathway specified, the inputs stockpiled or locatable, and the whole thing exercised rather than asserted. It is currently impossible, and not primarily because the physical task is impossible. It is impossible because nobody has established whether the energy-and-materials bootstrap closes; because preservation systems preserve artefacts rather than the institutions that make artefacts usable; and because the capability has never been exercised at any scale, so the word credible has no operational meaning here.

Handwave Link 1 — a bootstrap energy-and-materials analysis — is the binding scientific unknown, and it is unattempted. The intermediate result is a specified pathway from a stated degraded endowment to a self-reproducing industrial base, with the energy and materials balance closed at every stage. The measurement is simply whether the balance closes and which stage has the smallest margin. Speculative By the standards of process engineering this is not a hard analysis. It appears never to have been done. If the balance does not close, Links 3 through 7 are ornamental.

Speculative Link 2 — a salvage-endowment inventory — is a data problem and a tractable one. Estimates of above-ground metal stock, its accessibility with low-technology tools, and its degradation rate without maintenance. Material-flow-analysis methods exist and much of the stock data is already published for unrelated purposes. Frontier The half-life of usability is the interesting unknown: scrap that is trivially workable in year one may not be in year fifty.

Frontier Link 3 — an archive that survives institutional discontinuity — is blocked institutionally rather than technically. The medium, the encoding and the index must all be decodable from the artefact itself. The preservation field's own strategies explicitly assume continuity and its economics are operating-cost economics, so a discontinuity-proof archive is a different product and nobody's mandate covers it.

Frontier Link 5 — an exercised food-system pathway — is the binding market failure. There is no commercial demand for a food system that is only needed after a catastrophe. That is a textbook public-good under-provision and it is the reason the resilient-foods numbers remain modelled rather than measured. Established Link 7, an exercised capability scored by an external assessor, is the binding institutional gap and the only thing that would convert credible from a rhetorical term into an evidentiary one. Speculative Links 2, 4 and 6 are all cheap and would each produce a first-of-kind result, which is an unusual state for a field to be in.

Established This brief depends on infrastructure resilience for a specific reason rather than a thematic one. The catastrophic-food literature now includes electromagnetic pulse and pandemic-driven industrial disruption in its scenario set, which makes the failure behaviour of coupled physical networks an input to famine estimation rather than a neighbouring topic. The freeze-depth work on buried water infrastructure is that dependency made concrete: it is infrastructure-resilience method applied to a civilisational scenario.

Frontier The second dependency is a distinction rather than a brief: component claims versus integrative claims. Freeze depths over 30 m across 5.4 to 8.8 million kilometres of pipe, 515 million hectares of expandable cropland, and 1,331,458 vault accessions are ordinary empirical claims checkable now. Total deaths, recovery times and refuge efficacy are integrative outputs of model chains with no validation set. Established The field's credibility depends on keeping those two categories apart, and much of the criticism it receives comes from readers who encounter the second and generalise to the first.

Frontier The third dependency is comparative history that nobody in this literature has assembled. If institutional continuity rather than physical stock is the binding constraint on recovery, then historical recovery times should correlate with institutional survival rather than with the extent of physical destruction. Post-war reconstruction and state collapse are both well documented. The correlation has not been computed. Speculative It would be the field's first result drawn from observed cases rather than from models, which is worth more than its difficulty suggests.

Established And there is a declared evidential dependency: several holdings figures this brief would want are absent and are not filled from memory. The Millennium Seed Bank, the Arctic World Archive, the CGIAR genebank network and the Global Crop Diversity Trust endowment are all unreachable in this pass. No figure for any of them appears on this page.

Established The first required experiment is a paper, and this brief can specify its contents precisely enough that somebody could go and write it. A rigorous energy-and-materials analysis of industrial bootstrap pathways under a salvage endowment would have to contain three things. Speculative First: the energy return on energy invested of accessible surface coal and shallow oil under pre-industrial extraction technology — hand tools, animal haulage, wind or water pumping, no steel casing — evaluated against actual remaining outcrop and shallow-seam geology rather than against total resource. Not whether the coal exists, but whether it returns more energy than it costs to reach with what a degraded society would actually have. Speculative Second: the bootstrapping sequence for metallurgy — the ordered chain from charcoal or salvaged fuel to usable iron, from iron to the machine tools that make better tools, and from those to the precision required for pumps, generators and eventually semiconductors — with the energy, materials and labour balance closed at each step and the step with the smallest margin identified. Speculative Third: the minimum population and knowledge base — how many people, with what division of labour and what surviving documentation, are required to hold the sequence together long enough to complete it, given that each stage must be maintained while the next is developed. Handwave Every one of those three is a determinate calculation. None of them is at the frontier of any discipline. The paper does not exist.

Frontier The second experiment is a blind-decode test, and it would be the preservation field's first empirical result. Give a candidate archive artefact to a team with no prior briefing and no digital infrastructure and measure what they recover: medium, encoding, index, content, in that order. It is cheap and dramatic and nobody has run it.

Frontier The third is a germplasm-to-field exercise. Withdraw an accession, multiply it, and grow it to harvest under constrained conditions, without the depositing institution. The vault holds 1,331,458 samples and already runs a century-long viability experiment, so the culture of long-horizon experiment exists — it simply stops at germination instead of continuing to a crop. Established No source located here describes such a pathway, which means the step from a vault deposit to food is at present asserted rather than demonstrated.

Speculative The fourth is arithmetic and is the cheapest high-value analysis identified in this subject. Cross the national-resolution calorie outputs from the famine modelling with the refuge criteria from the island-optimisation work, and see whether refuge rankings reorder between biological and agronomic scenarios. Both datasets exist. Nobody has crossed them. Frontier A fifth, harder and more valuable, is a pilot deployment of at least one resilient-food pathway at a scale that tests the supply chain rather than the biology, reporting achieved calories per day at achieved cost against the modelled band.

Frontier Catastrophe hardening has exactly one well-specified engineering parameter and it arrived in 2025. Peak freeze depths exceeding 30 m under an abrupt sunlight reduction scenario is a design condition, not a warning: it says what burial depth a water main would need, and it says that the existing 5.4 to 8.8 million kilometres of buried pipe were not designed against it. Established Whether hardening that inventory is affordable is a separate question this brief does not have a costing for, and does not invent one.

Speculative Resilient foods are an industrial engineering problem disguised as a biology problem. Seaweed cultivation at scale, cellulose and stored-biomass conversion, and single-cell protein from methane are all processes whose biology is understood and whose supply chains do not exist at the required scale. The founding monograph of the programme devotes chapters to stored biomass and fossil-fuel conversion to food and to fibre supply for conversion to food; those chapter records are verified here and their contents were not read, so no conversion efficiency or yield is stated. Frontier The binding engineering question is ramp rate: how fast a pathway can go from zero to feeding a population, given that the scenario that requires it also destroys the industrial capacity to build it.

Frontier A discontinuity-proof archive is a specifiable artefact and nobody is building one. It needs a medium with a documented physical lifetime, an encoding that a reader can infer from the artefact without external reference, and an index that is part of the object. Every existing digital-preservation strategy fails at least one of those by design, because every existing strategy assumes a custodian. Handwave The engineering is not hard. The mandate does not exist.

Speculative The restart question, if it were ever answered, would generate the largest engineering specification in this subject. A staged bootstrap pathway is a process-engineering artefact: a sequence of unit operations, each with an input inventory, an energy budget, a labour requirement and a tolerance on the precision of the previous stage. Handwave Nothing about writing that specification requires a new technology. It requires somebody to decide that the question is worth a year of a competent process engineer's time.

Established The nearest neighbour is infrastructure resilience, and the relationship is directional. Cascade behaviour, restoration capacity and interdependency modelling are inputs here; this brief supplies the scenario severity that those methods are rarely run against.

Frontier Existential risk governance is the institutional counterpart and the two are frequently conflated. Governance asks who decides and under what mandate; this brief asks what would physically have to be in place. A governance regime with nothing stockpiled and nothing exercised is a committee.

Frontier Future agriculture and this brief share a literature and disagree about the default. Agricultural research optimises yield under expected conditions; resilient foods optimise calories under conditions in which yield optimisation has failed. Speculative The interesting question is whether the two portfolios are complementary or competing for the same land, capital and attention, and no allocation analysis was located here.

Speculative Energy corridors and economic resilience both bear on the trade finding. If maintaining trade is a first-order famine mitigation, then the physical and financial substrate of trade is a resilience asset, and the case for interconnection is a catastrophe case rather than only an efficiency case. Handwave That argument has not been made in either literature as far as this pass could establish.

Speculative Climate migration planning and civilizational planning share the refuge problem from opposite ends: one plans movements of populations under gradual pressure, the other plans destinations under abrupt pressure, and the destinations in both cases are chosen by geography and closed by politics.

Frontier The least obvious adjacency is materials-flow accounting, which is not usually thought of as a resilience discipline at all. The salvage-endowment question — how much refined metal sits above ground, in what accessibility class, degrading at what rate — is answerable with methods built for entirely different purposes, and the stock data has largely been published for those purposes already. Speculative Repurposing an industrial-ecology instrument to answer a catastrophe question is the cheapest transfer available in this subject and nobody has attempted it.

Established Nobody owns the restart capability, and that is the institutional fact that explains most of the others. Seed vaults are owned by genebank consortia and answer to depositors. Digital preservation is owned by libraries and archives and answers to funders on annual cycles. Famine modelling is owned by universities and a small number of philanthropically funded research organisations. Refuge analysis is owned by two or three academic groups. No institution's mandate is the whole chain, and the chain is the product.

Established Preservation fails on operating cost, which means it fails on institutional survival. The preservation literature says this about itself in a chapter title. An archive funded from an annual budget line is exactly as durable as the organisation holding that budget line, and the scenarios this brief is about are scenarios in which organisations stop.

Frontier The field is beginning to reach operational government, at subnational scale first. A 2025 presentation on food security and global catastrophic risk delivered to a New Zealand regional council is the located example; the record is verified and the content was not obtained, so nothing is claimed about what it recommended. Speculative The direction of travel matters more than the instance: catastrophe-scale food planning being commissioned by an authority that has statutory duties is a different mode of existence for this work than being published in a foresight journal.

Frontier The institution that does not exist is an external assessor. Every claim in this field is currently self-graded. Nothing has been exercised, scored and published by anyone other than its proponents. Established That is not a criticism of the proponents, who are candid about it; it is a description of a field at a stage that other risk disciplines — aviation safety, nuclear licensing, financial stress testing — passed through and left behind by adopting adversarial assessment.

Frontier Several institutional holdings this brief would want are simply unreachable here and are declared rather than estimated. The Global Crop Diversity Trust's endowment, the CGIAR network's aggregate holdings, and the Arctic World Archive's contents would all sharpen the picture of what is actually preserved and by whom. None appears on this page.

Established This literature is unsentimental about its own subject in a way readers should be warned about. The founding monograph of the alternative-foods programme contains a chapter titled Food Storage, Food Conservation, and Cannibalism. That is not sensationalism; it is a research programme taking seriously that a starving population does things, and choosing to analyse rather than avoid them. Frontier A field that will not name its worst branches cannot cost them.

Frontier The refuge concept has an unresolved allocation problem that its literature does not foreground. Island refuge optimisation names Australia, New Zealand and Iceland. Those are sovereign states with immigration policies, and the analysis that ranks them does not address who is admitted, on what basis, or by what authority. Speculative A refuge that works is a refuge that excludes, and the ethics of that exclusion are load-bearing for whether the concept is a public good or a private one.

Speculative The single most-quoted number in the field is doing political work it was not built for. More than 5 billion could die is a conditional output for one of six scenarios, from famine following soot injection, in a specific bilateral exchange. Quoted bare, it functions as an argument-ender. Established The paper's authors gave the range, the mechanism and the scenario; the compression is done downstream, and it is the compression rather than the modelling that deserves the scepticism it attracts.

Frontier Deep Adaptation is the collapse-acceptance strand and it must not be presented without its critique, which this brief could not obtain. Bendell and Carr's peer-reviewed 2021 paper describes an agenda and framework for people who believe in the probable, inevitable or unfolding collapse of industrial consumer societies. Established Note carefully what that paper is: its subject is the facilitation of gatherings, and it takes collapse as a premise held by its participants rather than defending it as a forecast. Frontier The substantive academic critiques of the original 2018 argument are a separate literature that this pass could not locate, and presenting either without the other would be a misrepresentation. This brief presents the paper as what it is and declares the missing half.

Established The civilizational implication is unusual because the subject is the implication. What this field is for is the question of whether the current arrangement is a one-shot event or a repeatable one, and no work programme anywhere is presently constituted to answer it.

Handwave If the restart argument is right, everything downstream changes character. Preservation becomes memorial rather than instrumental; refuges become lifeboats rather than seed crystals; the correct allocation of effort moves almost entirely to prevention, because recovery has no pathway. Speculative If it is wrong — and the salvage-endowment counter-argument is strong — then restart capability is an engineering programme that could be specified, costed and staged, and the current absence of one is a straightforward failure of institutional imagination rather than a physical constraint.

Frontier Nobody knows which, and the cost of finding out is a paper. That is the sentence this brief exists to deliver. Civilisational restart is not known to be impossible; it is not known at all, because the balance has never been computed and the capability has never been exercised.

Speculative The secondary implication concerns what a resilience posture says about a society that adopts it. Sandberg's argument that there are distinctively civilisation-level virtues makes the normative content explicit: choosing a recovery target is choosing what is worth recovering, and that is a political act performed in advance by whoever writes the specification. Frontier No current preservation programme has published such a specification, which means the choice is being made implicitly by what happens to get deposited. Speculative A vault whose contents reflect which genebanks had functioning administrations in the 2010s is not a neutral backup of world agriculture, and nobody has audited it as one.

Horizons here track the production of results and the exercise of capabilities, not the arrival of catastrophes, which are not on a schedule.

• 10 yr: Frontier The tractable analyses land or they do not: a bootstrap energy-and-materials paper, a salvage-endowment inventory, and the refuge-ranking cross-analysis are all achievable within a single grant cycle by a small group, and none of them needs an instrument that does not exist. Frontier The hundred-year seed viability experiment passes roughly its second decade and should have publishable interim results, which would convert the field's most-asked question from an assumption into a measurement. Speculative A first blind-decode test of a candidate archive is plausible if anybody funds it, and it would be cheap enough to run twice. Frontier On present trends the famine-modelling limb continues to disaggregate by nation and by crop, and the resilient-foods limb continues to publish costed pathways that nobody deploys.
• 25 yr: Speculative A germplasm-to-field exercise executed without the depositing institution, and a first pilot of one resilient-food pathway at supply-chain scale rather than laboratory scale — the two results that would move this field from modelled to measured. Speculative An external assessor for catastrophe-resilience claims either exists by then or the field has settled into permanent self-grading, and the second outcome is currently the more likely of the two. Frontier Also within this horizon: the comparative-history test of whether recovery correlates with institutional survival rather than physical destruction, which requires no new data at all.
• 50 yr: Speculative A discontinuity-proof archive as a deployed artefact rather than a proposal, with its medium, encoding and index all decodable from the object without a custodian. Handwave A costed and staged restart specification — the thing that would make the phrase restart capability mean something — is a fifty-year institutional achievement rather than a technical one, because the technical work is decades shorter than the institutional work and the funding case is the hard part.
• 100 / 250+ yr: Handwave An exercised, externally scored, repeatedly demonstrated civilisational restart capability, maintained across generations by an institution constituted for the purpose. Handwave Nothing in the physics forbids it; nothing in the present institutional landscape produces it; and the honest statement about this horizon is that it is contingent on a decision nobody has yet been asked to make.

• Depends on Cascade, interdependency and restoration behaviour of coupled physical networks, supplied by infrastructure resilience and used here as the input to catastrophe-scale food and water modelling rather than re-derived.
• Enables A specified, costed and eventually exercised restart capability; catastrophe-scenario design conditions for buried and surface infrastructure; and a preservation product whose readability does not depend on a surviving custodian.
• Adjacent Global catastrophic and existential risk analysis, germplasm conservation and genebank practice, digital and archival preservation, food-system modelling, comparative history of state collapse and reconstruction.

Established “Nuclear war would kill five billion people.” The figure is more than 5 billion in the United States–Russia scenario, one of six stratospheric-soot-injection scenarios, arising from famine following climate disruption — not from blast, and not from nuclear war in general. The India–Pakistan scenario gives more than 2 billion. Quoting the single number collapses a six-scenario range into a point and discards the mechanism, which is the part that determines what could be done about it.

Established The same paper is routinely cited under a title it does not have. It circulates as Global food insecurity and famine from regional and global nuclear conflict. Its published title is Global food insecurity and famine from reduced crop, marine fishery and livestock production due to climate disruption from nuclear war soot injection. The short form resembles the 2021 preprint title Global Famine after Nuclear War, which is the probable origin of the drift. Frontier Citing the short title against the journal DOI is a citation error propagating from a preprint, and it is now widespread enough to be worth naming.

Established “The Svalbard vault holds about a million seeds.” This one is enthusiast-side and it understates. It holds 1,331,458 seed samples — accessions, each containing many seeds — from 123 institutes across 87 countries. Sample is not seed, and the difference is two to three orders of magnitude.

Frontier “The seed vault guarantees we can restart agriculture.” This is the most important misconception in the subject, and it is also enthusiast-side. What the vault provides is a safety backup of deposited accessions for depositing genebanks. It does not hold every crop, because storage-recalcitrant species cannot be conserved by the conventional dry-and-freeze method. It does not hold soil biota, locally adapted management knowledge, or seed-multiplication capacity. Its one real-world use was a withdrawal by an institution that still existed and had relocated. And how long the seeds remain viable is a question currently being measured over a hundred years, which is to say not yet answered. Established A vault preserves germplasm; agriculture is germplasm plus everything else, and no source located here describes a pathway from a vault deposit to a harvested crop without the depositing institution.

Frontier “Alternative foods could feed everyone.” Also enthusiast-side. The modelled result is a 95 per cent probability of 2,100 to 3,500 kcal per capita daily at around $1.73 daily — from a preprint, conditional on successful deployment, with an ambiguous cost unit this brief does not resolve. The same literature reports that food waste reduction would have limited impact, and that cropland expansion covers 20 to 39 per cent of the population. Established These are partial mitigations that state their own shortfalls. They are not a solution and their authors do not present them as one.

Speculative “Existential-risk research is unfalsifiable, so it is not a research programme.” This is the sceptic's misconception and it deserves a real answer rather than a dismissal. The steelman is strong: the field has no observed instances of its dependent variable, its predictions are unfalsifiable within any assessment lifetime, and scenario construction generates unlimited scenarios with no selection principle. Established But the objection treats an epistemically heterogeneous chain as homogeneous. Freeze depths over 30 m across 5.4 to 8.8 million kilometres of pipe is a thermodynamics-and-inventory calculation. 515 million hectares of expandable cropland is a land-suitability computation. 1,331,458 accessions from 123 institutes is an inventory count. None of those awaits a catastrophe. Frontier The aerosol-climate response at the head of the famine chain is validated against volcanic eruptions. The ICARDA withdrawal is the preservation system's dependent variable observed, at small scale, in the world. And the field demonstrably updates on evidence: the trade-maintenance finding and the negative result on food waste reduction both changed advice. Established The defensible residue of the objection is real and this brief endorses it: the field's integrative claims — total deaths, recovery times, refuge efficacy — carry uncertainty that is not propagated and cannot be validated, while its component claims are ordinary science. Be sceptical of every headline integrative number, including five billion. Do not extend that scepticism to an inventory count.

Handwave “We cannot restart without accessible fossil energy.” The correct status of this claim is not true, not false, but unstudied — and that is a stranger situation than either. It is asserted confidently in popular and semi-technical writing about collapse. It is load-bearing for the whole subject. This pass found no peer-reviewed treatment of it in either direction. Speculative The absence may be a retrieval artefact, and the retrieval conditions here were genuinely poor in a way that would systematically hide engineering and energy-economics literature. It may also be real. A reader who knows of such a paper should regard this paragraph as an invitation. Established What is not in doubt is that a claim doing this much work in a field's argument ought to have a literature, and that specifying the missing paper — the energy return of accessible surface coal and shallow oil under pre-industrial extraction technology, the bootstrapping sequence for metallurgy, and the minimum population and knowledge base — is more useful than restating the claim more forcefully.

Speculative “Islands are refuges.” The refuge literature scopes explicitly to biological threats requiring isolation, and an island has no stratospheric border. Frontier Islands may still rank well under agronomic scenarios for a different reason — maritime thermal inertia and fisheries access — but whether the rankings survive scenario substitution is untested, and cheap to test.

Established “Digital archives will preserve our knowledge.” Migration and emulation both presuppose an unbroken institutional chain, and the field's own literature says the binding constraint is money. An archive that requires a living institution is not a catastrophe-resilient archive.

Established “We know roughly how long recovery would take.” Recovery time is stated in the field's own working literature as depending on three unspecified parameters: starting complexity, extent of loss, and desired recovery level. Any single recovery-time number is unsupported, including the reassuring ones.

Speculative “Managed degrowth is the resilience strategy.” The resilient-foods literature is, on its face, a refutation of this, and the tension is unremarked in both literatures. The degrowth argument is that reducing throughput reduces the complexity that must be maintained, and complexity maintenance is what fails first. Frontier But the interventions that actually mitigate modelled famine are industrial — seaweed cultivation at scale, biomass conversion, single-cell protein — and the finding that maintaining trade is a first-order mitigation points at a complexity-increasing institution. Established Nobody has defined the complexity metric that would adjudicate this, let alone measured it, so the disagreement is currently between two intuitions wearing different citations.

Frontier Finally, a misconception about citations themselves, which this subject is unusually exposed to. The bibliographic instruments used to assemble evidence at this scale fail silently. During the research behind this brief the fetch layer began returning stale cached results to fresh, distinct queries — three consecutive different requests returning the same digital-preservation result set — and the correct response was to stop fetching rather than keep going. Established Related failures have been recorded elsewhere in this corpus: fuzzy matchers returning unrelated papers for a well-formed query, and a publisher URL serving a different article than the DOI addressed. None produces an error message. Frontier That matters here because civilisational-risk assessment is exactly the kind of work assembled by machine from other people's citations — and this field's own founding monograph is double-deposited under two ISBNs, every chapter carrying two DOIs, silently splitting its citation counts.