A fifth branch the framework is missing
The four-scenario framework on this site varies speed, access, ownership, and institutional control: managed diffusion, concentrated acceleration, open proliferation, and capability shock with institutional lag. Those four variables do real work. But notice what every cell quietly assumes: that the institutions doing the managing, accelerating, proliferating, and lagging survive the transition intact. There is a fifth variable the framework leaves out — does institutional continuity itself hold through the arrival of AGI?
This page is scenario analysis, not prediction. Nothing here requires a specific arrival date, a specific failure mechanism, or a belief that collapse is likely. It simply takes seriously the possibility that transformative AI capability arrives during — or triggers — a partial breakdown of states, supply chains, and shared information systems, and asks what that world looks like from inside. The extremes are covered elsewhere: full civilizational collapse on one side, prolonged crisis and recovery on the other. This page is about the muddy middle, which history suggests is where disasters usually land.
What partial collapse actually looks like
Real disasters rarely produce either the apocalypse or a quick return to normal. They produce patchworks. A famine, a war, a pandemic, a grid failure — the standard outcome is a geography where some functions hold and others fail: local food systems working while import chains seize, one grid segment stable while the neighboring region goes dark, a state fragment keeping courts and payroll running while another dissolves into rule by whoever controls the fuel depot.
The disaster-sociology evidence is unusually consistent on the human response. Decades of field research find that most communities self-organize effectively in the first phase of a disaster — the predicted panic, looting, and social breakdown mostly do not occur. In their comparative study of emergency survivors, Drury, Cocking, and Reicher found solidarity and coordinated helping, not each-for-themselves behavior (Drury, Cocking & Reicher, “Everyone for Themselves? A Comparative Study of Crowd Solidarity Among Emergency Survivors,” British Journal of Social Psychology, 2009); the same research program distilled the practical implications into twelve recommendations for building collective psychosocial resilience in emergencies (Drury et al., 2019, Frontiers in Public Health), and the British Psychological Society’s research digest summarizes the broader finding bluntly: emergencies inspire crowd cooperation, not panic (BPS Research Digest). Rebecca Solnit’s field reporting in A Paradise Built in Hell (2009) reaches the same conclusion from a different direction: ordinary people are often at their best in catastrophe, and the disasters come from institutions that arrive late, misread the situation, or treat survivors as threats.
The second phase is harder. Mutual aid fills days and weeks; it does not rebuild transmission lines, settle property disputes, or restore a currency. That is where patchworks either re-consolidate or decay, and where the patchwork of a post-AGI transition would be decided.
Life inside the patchwork
Inside a patchwork, the defining feature is unevenness. Some of the consequences are familiar from any fractured society; others are specific to arriving there during an AI transition.
Access to AI benefits becomes geographic. Regions with intact data centers, power, and connectivity keep automated logistics, medical triage support, and agricultural optimization; regions without them do not. The gap is not just economic — it compounds, because the regions with working AI recover their infrastructure faster than the regions without it.
Mobility becomes the key privilege. The person who can cross a border, physically or digitally, can shop among fragments for safety, work, and law. The person who cannot is subject to whichever fragment controls their street. This is why the manual treats community and local networks as infrastructure rather than sentiment: in a patchwork, the local network is often the only network.
Information becomes a terrain feature rather than a utility. Information deserts sit next to information abundance: one valley with functioning connectivity and machine translation, the next with rumors and a single radio. Legitimacy flows to whoever keeps the lights on, whatever their politics — a dangerous but predictable dynamic that rewards whoever controls energy and comms hardware, not whoever governs best.
AI as recovery asset or extraction tool
The same capability that intensified the crisis becomes the scarcest recovery resource. An intact frontier model with working inference is, in a collapsed economy, roughly what a shipping fleet was in 1945: the bottleneck through which everything else passes. That makes the control question decisive, and it is the question where the ownership analysis of the transition scenarios page bites hardest in a crisis.
In the benign version, AI is a reconstruction accelerator: it designs replacements for broken equipment, optimizes scarce fuel and fertilizer, translates across fragments, models epidemic spread, and lets a region with almost no surviving specialists act as if it had them. Recovery timelines compress from decades to years.
In the malign version, AI is an extraction multiplier. Whoever controls the surviving capability uses it to price, predict, and police a population that has no countervailing leverage — the pattern Acemoglu and Johnson document across the history of technology in Power and Progress (2023), where new tools defaulted to enriching whoever already held power until institutions forced a different distribution (Penguin Random House). A crisis does not suspend that history. It accelerates it, because the checks that normally slow extraction — courts, press, elections, exit — are exactly the institutions a patchwork lacks. The scenario to plan against is not the dark region. It is the well-lit, orderly region whose residents discover, too late, that their stability was rented from whoever runs the models.
The short, local window when the patchwork’s future gets decided
Patchworks are not stable equilibria. They consolidate, federate, decay, or get conquered — usually within years, sometimes months, and almost never through the preferences of the people living in them. The choices that matter are made early, by whoever is organized at the moment of maximum fluidity: what gets rebuilt first, who owns the recovery infrastructure, whether fragments merge on whose terms, and whether the region that kept its AI capability shares, rents, or weaponizes it.
This is why the response-and-leverage material in this manual — grassroots organizing, local reserves, redundant communications, community institutions — is not a hobbyist appendix to the main argument. In a partial collapse scenario it is the mechanism by which the people who will live in the patchwork get any say in what it becomes. The difference between a fragment that consolidates into something livable and one that becomes an extractive enclave is usually decided by whether ordinary people were organized before the fluidity closed.
The adjacent pages trace the boundaries of this one: civilizational collapse scenarios describes what lies past the far edge, and protracted crisis vs. recovery examines the two directions a patchwork can go once it stabilizes. Partial collapse is not the likeliest future and not the worst. It is the one where preparation matters most relative to its probability — because in the others, either nothing you do changes the outcome or everything continues roughly on schedule.