Predictable Ruin

The problem is what comes after. It is always what comes after.

In Isfahan a woman is packing a suitcase and telling herself she will be back in a week. She folds her daughter's school uniform. She adds the math notebook with the good grades in it because her daughter will want to keep up. She wraps a bag of saffron that her mother gave her at Nowruz in a sock and puts it in the corner of the case where it will not get crushed, which is the kind of care you take with something you expect to unpack. She is not packing her jewelry. She is not packing the photo from the mantel. These are things you leave behind when you are coming back in a week.

Her husband is on the roof. He has the car keys in his hand. He has had the car keys in his hand for forty minutes. He can see the glow on the horizon toward Natanz and he has been watching it the way you watch something on television, as though it is not in his country and not forty miles from the room where his son is sleeping. His son is not sleeping. His son is watching a cartoon on a phone plugged into a car battery. The cartoon is in English. The boy does not speak English. He is laughing at something that does not require translation.

The power has been out since the first wave. The neighbors' generator hums behind the wall like a refrigerator in a kitchen that still exists. She zips the suitcase. It is light. She tells herself this is because they will not be gone long.

They do not know where they are going. They know they are coming back in a week. The morning news will say things about corridors and objectives and Phase One and the morning news will be wrong about most of it because the morning news is always wrong in the first 72 hours after the world rearranges itself around a new absence. By then the roads will be full of cars like theirs, packed light, packed for a week that will become a month that will become the rest of their lives.

There are going to be millions of them. The data says so. Here is what the data says.

We built a knowledge graph of every major regime change in the modern record and trained a classifier on the outcomes. Not just the failures. The successes too. Germany. Japan. Spain. South Korea. South Africa. Chile. These are the cases that prove it can be done, that populations can cross from tyranny into lasting democracy. They are also the cases that prove it requires specific, identifiable conditions. A transition plan. International consensus. A path for the security apparatus to survive in the new order. Institutional continuity. A unified opposition. A unifying leader.

Iran has none of them. Not one of six. And no case in the dataset has ever produced a good outcome with this profile. The classifier does not hedge. It assigns Iran a 0.000007% probability of success. We do not trust the decimals. We trust the direction. And the direction is unambiguous.

The Model

We encoded each case as a vector of 23 features: population, ethnic composition, oil dependence, security force characteristics, intervention type, opposition structure, transition planning, institutional continuity, and post-conflict investment. We labeled the six successes and five failures. Then we asked the model two questions. First: which features best separate good outcomes from bad ones? Second: where does Iran fall?

The Features That Matter

The single most predictive feature is not one that gets discussed on cable news. It is whether the old security apparatus was given an economic path to survival in the new order. In Germany, many former officials were absorbed into the new state. In Japan, the emperor and the civilian bureaucracy were preserved. In Spain, the military was given explicit guarantees. In South Africa, the sunset clauses and Truth and Reconciliation Commission gave the old regime a way to participate in the future. In Chile, Pinochet himself stayed on as army commander for eight years after the transition.

This feature perfectly separates the dataset. Present in 6 of 6 successes. Present in 0 of 5 failures. One hundred percent separation.

The logic is straightforward: security forces that face destruction or prosecution have nothing to lose. They fight. They fragment. They reconstitute as insurgencies or criminal networks. But security forces offered a path forward can be negotiated with. They can be transformed rather than destroyed. This is not a moral argument. It is an engineering observation about what happens to institutions under pressure.

Iran 2026 has no such mechanism. Trump told the IRGC to "lay down your arms" and offered "complete immunity." But immunity from whom? There is no occupation force to accept a surrender. There is no transitional government to process an amnesty. There is no institution to guarantee anything. There is a video on Truth Social.

What Iran Is Missing

Zero for six. Not one for six, not three for six. Zero.

No historical case has produced a good outcome while missing all six. No historical case has produced a bad outcome while having all six. The separation is complete.

The Number

The classifier assigns Iran a 0.000007% probability of a good outcome. We do not trust the decimal places. With 11 training cases, the confidence interval is wide enough to drive a convoy through. The point is not the precision. The point is that Iran's feature profile places it unambiguously on the failure side of a clean separation, and that this placement is robust: you would need to change not one or two but all six of the critical features to move it across the boundary.

The model classifies all 11 historical cases correctly. Germany, Japan, Spain, South Korea, South Africa, and Chile are all assigned probabilities above 99.8%. Iraq, Libya, Afghanistan, Egypt, and Syria are all assigned probabilities below 1%. The separation is not subtle.

Model Validation

Egypt is the most ambiguous historical case, and even it sits well below the decision boundary. Germany is the least certain success, and even it sits well above. Iran is not in the ambiguous zone. It is in the deep failure zone, lower than Iraq, lower than Libya, lower than Syria.

What Success Actually Looked Like

The positive cases are not a monolith. They span half a century, three continents, and wildly different political circumstances. But they share a structural logic that the failures do not.

Germany and Japan succeeded through overwhelming force combined with equally overwhelming commitment to reconstruction. The intervention was total: ground invasion, unconditional surrender, military occupation, and then the Marshall Plan, the largest peacetime economic commitment in history. The occupying powers did not simply remove the old regime. They physically replaced it, administered the territory, rebuilt the economy, and stayed for decades. The cost was staggering. The commitment was generational. It worked.

Spain, South Korea, South Africa, and Chile succeeded through a different mechanism entirely: negotiated transition. In each case, the old regime participated in its own replacement. The security forces were given guarantees. Institutions were preserved and reformed rather than destroyed. There was a plan, agreed upon by multiple parties, for what came next. The transitions were messy, compromised, and morally unsatisfying. They produced justice deferred, not justice delivered. But they produced stability.

Iran 2026 follows neither model. It has the violence of Germany and Japan without the occupation force, without the Marshall Plan, without the decades-long commitment. It has none of the negotiation of Spain or South Africa, no agreed transition, no institutional continuity, no security force buyout. It occupies a space the dataset has never seen a success emerge from: high violence, no plan, no follow-through.

Why the IRGC Cannot Be Wished Away

The classifier identified security apparatus buyout as the strongest single predictor. It is worth understanding what this means on the ground, in a country where the ground itself is owned by the people you are telling to lay down their arms.

In 1989, after the war with Iraq, Khamenei issued a decree creating Khatam al-Anbiya, the IRGC's construction and engineering headquarters, to rebuild the country. It was supposed to be a temporary reconstruction effort. It became the largest contractor in the Middle East. By the 2000s, Khatam al-Anbiya controlled more than 812 registered companies and had completed over 2,500 infrastructure projects: dams, highways, tunnels, pipelines, metro lines, airports, ports. A quarter of a million people work on its projects. It builds the roads. It builds the buildings on the roads. It builds the pipes that bring the water to the buildings.

Then it bought the phone system. In 2009, a consortium called Etemad-e-Mobin, directly tied to the IRGC, purchased a 51% controlling stake in the Telecommunication Company of Iran for $7.8 billion. The auction lasted one day. There was no competitive process. With that single acquisition, the Guard gained control of the country's telecom infrastructure, which means it controls not just who calls whom but who the government listens to and who the government silences. When the internet went dark during the January protests, it was IRGC infrastructure that killed the signal.

It does not stop there. The IRGC runs banks: Ansar Bank, Mehr Eghtesad Bank, Hekmat Iranian Bank, all formally registered as independent entities, all operated through cooperatives and foundations tied to the Guard. These banks channel public funds to IRGC projects, finance real estate speculation in Tehran and Isfahan and Mashhad, and maintain offshore accounts used to circumvent sanctions. In 2019, under international pressure, five IRGC-affiliated banks were forcibly merged into Bank Sepah, one of Iran's oldest state banks. The merger did not reduce the IRGC's financial footprint. It legalized it.

It controls shipbuilding through SADRA. It controls automobiles through a 45% stake in the Bahman Group. It has moved into agriculture, mining, media, and electronics. The US Treasury has designated Khatam al-Anbiya and its subsidiaries multiple times since 2007. The EU froze its assets in 2008. The UN targeted it in 2010. None of this slowed it down. Sanctions became the justification for deeper economic control: with foreign companies fleeing Iran, the Guard stepped in to fill every vacancy. The Supreme Leader called this the "Resistance Economy." It was a monopoly with a flag.

This is what the model means by "security apparatus economic embeddedness." It is not an abstraction. It is 812 companies and a quarter million workers and the telecom system and the banks and the ports and the metro and the roads and the refineries and the men who run them. These men did not join the IRGC to die for the supreme leader. They joined because the Guard is where the contracts are and the contracts are how you build a house and send your children to university and buy a car and have a life. They have mortgages. They have car payments. They have cousins who work for the subcontractors who work for the subsidiaries who work for Khatam al-Anbiya. You cannot tell these people to lay down their arms because their arms are also their jobs and their pensions and their children's futures and the phone network their wives use to call their mothers.

You can bomb the generals. You cannot bomb an economy out of existence. The mid-level commanders who run the construction contracts and the port authorities and the telecom licenses will still be in their offices when the smoke clears or they will be in their cars on roads they built heading toward provinces they administer and they will still have networks and resources and the phone numbers of every subcontractor in the western half of the country.

In every case in the dataset the security apparatus survived. All eleven. The question was never whether it persisted but in what form. In the successes they were given economic roles and bound by new rules and transformed. In the failures they were left intact and hostile or dissolved on paper and reconstituted as insurgency or fragmented into armed factions that carved the country into territories named after the commanders who held them.

When the Coalition Provisional Authority disbanded the Iraqi army, 400,000 men with training and relationships and weapons and no paycheck did not disappear. They became the insurgency. They became the officer corps of ISIS. The IRGC is larger and deeper than the Iraqi army ever was. Its economic footprint exceeds the Egyptian military's, and the Egyptian military simply took power back. Trump told the IRGC to lay down their arms on a platform that most of them cannot access in a language most of them do not speak and offered them complete immunity from a government that does not exist yet and may never exist.

De-Baathification with a Truth Social account. That is the plan.

The Person in the Chair

We know what the machine does. The question is who sits at its desk.

In January 2026, the United States captured Nicolás Maduro. They had wanted him for years. He was indicted, sanctioned, bounty-raised to fifty million dollars. They got him on a plane to New York. Photographs for the newspapers. A name for the indictments. Victory.

They gave the country to his vice president. Delcy Rodríguez. The woman who ran SEBIN, Venezuela's intelligence service, the one the United Nations found had committed crimes against humanity under her authority. The woman whose father was taken to a building with no windows in 1976 and asked questions until he was dead. The woman who learned from this not that such buildings should not exist but that the important thing was to be on the correct side of the questions.

She gave the Americans what they wanted. One man in chains. A headline. She kept what mattered. The files. The networks. The building that spirals down into the earth. The machine itself.

The Americans told her to hold elections and crack down on drug trafficking and expel the Cubans and stop selling oil to adversaries. She agreed to all of it because agreeing costs nothing. The drug networks are how the military gets paid. The Cubans are her intelligence infrastructure. The oil goes to China, which holds sixty billion in Venezuelan debt. You cannot crack down on your own revenue, expel your own nervous system, or default on your largest creditor and survive. She knows this. The Americans know this. Everyone knows this. The words were said anyway because words are what you say when the cameras are on and the deal is already done.

She did not sell Maduro to hold elections. She sold him to avoid them.

For Venezuelans, the change was real. It was not the change Washington promised. On her second day in office, Delcy appointed Gustavo Enrique Gonzalez Lopez, former director of SEBIN, as the new head of military counterintelligence. Gonzalez Lopez was one of the first Venezuelan officials sanctioned under Obama. He oversaw the policy of torture at El Helicoide, the spiral-shaped building in central Caracas that SEBIN used as its headquarters and prison. He helped run the Liberation of the People operations that used extrajudicial killings to root out dissent in poor neighborhoods, killing hundreds and possibly thousands. His new posting is not a demotion. It is a lateral move from one torture apparatus to another. Diosdado Cabello, Maduro's interior minister, still controls the colectivos, the armed gangs that enforce loyalty in the barrios. Foro Penal, the Venezuelan NGO that documents political detention, counted more than 640 prisoners still held after the amnesty law passed. The amnesty excludes anyone convicted of "terrorism," which is the charge the regime files against activists. Delcy announced she would close El Helicoide and convert it into a sports and cultural center. She did not announce she would close the files inside it, or the methods, or the people who developed the methods. The building changes. The machine does not.

This is the pattern the knowledge graph cannot model because it is not a feature of the intervention. It is a feature of the aftermath. The Americans need a face. They need someone to negotiate with, someone who will say the right things on camera, someone who will provide the appearance of transition while the machine keeps running underneath. They need a person in the chair. They do not care, in the end, who the person is. They care that the person is cooperative and photogenic and willing to say the word elections without laughing.

But the Delcy model requires a precondition the essay has not yet named: a single machine. Venezuela had one power center. Maduro sat on top of SEBIN, the military, the oil revenue, the colectivos. Delcy could sell him because the system was centralized enough that one deal with one person could deliver the whole country. The CIA director flies to Caracas, shakes one hand, and the machinery transfers intact. It is a transaction. A terrible one for the Venezuelan people, but a clean one for Washington.

Iran does not have one machine. It has at least five, and they do not answer to the same switchboard. The IRGC is not a single organization. It is a constellation of semi-autonomous fiefdoms. The Quds Force ran foreign operations independently under Soleimani for decades and its surviving leadership retains those networks. Khatam al-Anbiya has its own revenue streams, its own contracts, its own loyalty structures built project by project across 2,500 infrastructure builds. The Basij answers to the IRGC in theory but operates through local mosques and neighborhood cells that no central commander fully controls. The intelligence arm competes with VAJA, the civilian intelligence ministry, which has its own director and its own files and its own reasons to survive. The navy, the aerospace force, the ground forces each have their own commander with their own budget lines and their own patronage networks. And then there is the clerical establishment, which is a separate power structure entirely: the Assembly of Experts, the Guardian Council, the seminary system in Qom, the bonyads and religious endowments that fund it all independently of the state budget. These are not IRGC subsidiaries. They have their own money, their own legitimacy claims, and their own succession politics.

You cannot buy one person and get Iran. You would need simultaneous deals with the IRGC ground forces commander, the Quds Force leadership, the seminary establishment, the Basij coordination council, and whoever controls the nuclear sites at Isfahan and Fordow. Each of those actors knows that cooperating with the Americans makes them a target for the others. The fragmentation that makes civil war likely is the same fragmentation that makes a clean handover impossible. The Delcy scenario is the best case for American intervention, and in Venezuela it still produced Gonzalez Lopez running a torture directorate and 640 political prisoners in cells. In Iran, they cannot even get to the Delcy scenario. The question is not who will be the Delcy of Iran. The question is whether the concept applies at all.

Iran has no shortage of candidates. The CIA knew this. According to Reuters, in assessments produced during the final two weeks of February, the agency concluded that even if Khamenei were killed, the regime would not collapse but would reorganize under hardline IRGC commanders. The strikes proceeded anyway. The intelligence community's own analysis confirmed what the model predicts: decapitation produces succession, not liberation.

The names are already circulating. Ali Larijani, secretary of the Supreme National Security Council, is the man Khamenei turned to in January when the protests spread and the system needed stabilizing. He is a former IRGC commander, a former parliament speaker, and the most experienced crisis manager in the regime. He is not a senior cleric, which disqualifies him from the supreme leadership under the constitution, but constitutions bend when the alternative is collapse. Larijani is the Delcy of Iran: not the face of the revolution but the person who knows where every wire connects and can keep the lights on while the brand is updated. Mojtaba Khamenei, the supreme leader's son, has deep ties to the IRGC and the Basij paramilitary. Father-to-son succession is frowned upon in a republic born by overthrowing a monarchy, but Mojtaba controls networks and resources that make him difficult to sideline. Among the clerics, Alireza Arafi heads Iran's seminary system and sits on both the Assembly of Experts and the Guardian Council. He is Khamenei's confidant, fluent in Arabic and English, and has the institutional credibility to hold the clerical establishment together. And then there is Mohammad Bagher Ghalibaf, the Parliament Speaker and former IRGC Air Force commander who reportedly assumed military decision-making during the Twelve-Day War when Khamenei was incapacitated. Ghalibaf ran Tehran as mayor for twelve years, building the patronage networks that function as economic control in a sanctions economy. He is the Sisi model: the military-political operator who already governs before the title catches up.

The extended knowledge graph encodes each of these candidates as instances with structural properties: security network control, economic network control, external interlocutor capacity, clerical rank, constitutional eligibility, depth of IRGC ties. OWL class restrictions then define what a "viable transitional figure" looks like without editorial input: the candidate must control security or economic networks, and must be able to interface with external powers. The reasoner runs. Larijani and Ghalibaf satisfy the class. Delcy Rodriguez, encoded as a prior instance with the same property structure, also satisfies it. The analogy between Larijani and Delcy is not a rhetorical move. It is a computable equivalence: two instances in the same inferred class, separated by 5,000 miles and two months. Mojtaba fails the classification. He controls the networks but cannot sit across from the Americans because he is sanctioned, has no diplomatic history, and cannot say the right things in the right language. The machine might select him but Washington cannot accept him. Arafi fails differently: he has the external capacity and the constitutional eligibility but controls nothing underneath. The dual-structure scenario writes itself: Arafi as supreme leader in name, Larijani or Ghalibaf as the operator who runs the machine from the building with no windows. The reasoner does not assert this. It surfaces the structural conditions that make it the most parsimonious outcome.

The reasoner also fires on the IRGC itself. The ontology defines a class called WillReconstitute: any security apparatus that is economically embedded and has not been offered a survival path in the new order. The IRGC satisfies both conditions. The historical base rate for this class is five out of five. Every economically embedded security apparatus in the dataset that was not offered a buyout reconstituted in some form. SEBIN in Venezuela is the most recent confirmed instance: Gonzalez Lopez moved laterally from one torture directorate to another, and the building got a new name while the files stayed open. The IRGC, with its 812 companies and quarter-million contract workers, will not dissolve because someone in Washington said the word freedom on television. It will put on a suit.

Any of them will say transition and reform and the will of the Iranian people on Western television. Behind them the machine will keep processing because that is what machines do.

The Americans will call one of them gracious. They will call the arrangement progress. They will declare victory and go home because Americans always go home. And in a building somewhere in Tehran that has no windows and no name, the lights will stay on and the questions will continue and the only thing that will have changed is who signs the papers.

The model predicts this as "IRGC reconstitutes." The Venezuelan precedent shows you what reconstitution looks like when it has a face and a desk and a pen and a window overlooking a city that does not know it has been sold.

The puppet has strings too. Some of them go down into the dark. The machine does not care who wins. The machine only cares that it keeps running.

The Consistency Test

The accountability framework from The Weight of Absence applies to the justification for the strikes themselves. The method is the same: encode official claims as structured data, check for internal contradictions.

Test I: Goal Consistency

Within a single day, senior officials offered mutually exclusive descriptions of the operation's purpose. The Vice President says it is not regime change. The President calls on Iranians to overthrow their government. This is a structural contradiction, not an interpretive disagreement. A SHACL constraint fires.

Test II: Behavior-Goal Alignment

If the goal is the nuclear program, you strike nuclear facilities. The June 2025 strikes did this. If the goal is regime change, you strike the supreme leader's residence, the defense minister, the IRGC commander, and the intelligence chief. The February 2026 strikes did this. Target selection is behavior. Behavior reveals intent more reliably than press conferences.

Test III: Negotiation Timeline

Oman's foreign minister reported "significant progress" in U.S.-Iran nuclear talks on February 27. Strikes began hours later. If the goal was the nuclear program and talks were progressing, the strikes are inconsistent with the stated goal. If the goal was regime change, progress in talks was the threat: a deal would have removed the pretext.

When behavior contradicts the stated goal but aligns with an unstated one, the pattern tells you which goal is real.

Test IV: The Ahmadinejad Strike

On Saturday, strikes hit the Narmak district of northeast Tehran, a middle-class residential neighborhood. The target was the residence of former President Mahmoud Ahmadinejad, who was killed along with his bodyguards. ILNA, the Iranian Labour News Agency, confirmed the death. Two schoolchildren at a nearby school were also killed. Saturday is a weekday in Iran. The school was open.

Ahmadinejad left office in 2013. His influence had waned for over a decade. He held no military command, no nuclear authority, no position in the chain of custody for enriched uranium. His only remaining institutional role was membership in the Expediency Discernment Council, a 48-member advisory body with no operational power. According to Israeli journalist Amit Segal, citing N12News, Ahmadinejad had been under house arrest at the time of the strike, having reportedly attempted a power bid during the January protests. He was already detained by his own government.

If the goal is the nuclear program, killing a former president under house arrest with no access to nuclear facilities achieves nothing. If the goal is regime change, killing a man already removed from power by his own regime achieves nothing. If the goal is decapitation of military command, a member of an advisory council is not in the command structure. No strategic framework produces this target. The only framework that does is a list of names compiled not from operational intelligence but from symbolic meaning: Ahmadinejad is the man who said Israel should be wiped from the map, who denied the Holocaust at the United Nations, who became the face of Iranian hostility in Western media for a decade. He was not a threat. He was a grudge. And two schoolchildren in Narmak paid for it.

Predictions

A framework that cannot be wrong is not a framework. The following predictions are falsifiable. If they fail, the model is wrong and we will say so.

If the Historical Pattern Holds

The prediction that requires the most specificity is the one most people will ask about first: civil war. The model's base rate is 4 of 5 failure cases (80%). Iran's specific profile suggests this is conservative.

Iran is not a homogeneous country. Persians are roughly half the population. The rest are Azerbaijanis, Kurds, Lurs, Baloch, Arabs, Turkmen, and others, distributed across provinces that border countries with direct interests in their ethnic kin. Before tonight's strikes, non-state actors and separatist groups were already asserting local control in Kurdistan, Sistan-Baluchestan, and Khuzestan. The protests that began in December 2025, the largest since the 1979 revolution, had already killed thousands. The regime was cracking before the bombs fell. The bombs removed the central authority that was holding the cracks together.

What makes Iran's fragmentation risk worse than Syria's is a point most analysts have converged on but few have stated plainly: there is no geographic containment logic. Syria's civil war, for all its horror, had a sectarian geography. Assad's Alawites could retreat to Latakia. The Kurds held the northeast. HTS had Turkish-backed Idlib. The factions had territory they could defend and borders they could draw. Iran has no such map. The Islamic Republic is a majority-rule government in a multi-ethnic state. Khamenei himself was Azerbaijani. There are no Alawite mountains to retreat to. If armed conflict erupts, it will be nationwide.

The accelerants are already in position. In the northwest, Iran's Azerbaijani population numbers over fifteen million, more than the entire population of the Republic of Azerbaijan next door, which has spent years building diplomatic and intelligence relationships with Azerbaijani-Iranians and will not waste a power vacuum. In the west, the Kurds are the most organized non-state military force in the region; they have been fighting for autonomy since before the revolution, and Turkey will intervene to prevent Kurdish empowerment along its border, as it has in Syria and Iraq, which means a Kurdish push for autonomy triggers a Turkish military response inside Iranian territory. In the southeast, the Baloch straddle the Pakistan border, where Jaish al-Adl has been conducting armed attacks on Iranian security forces for years and Saudi Arabia has previously funded Baloch separatist groups. In the southwest, Khuzestan's Arab population sits on top of most of Iran's oil reserves, and the province has a history of separatist sentiment that Saddam Hussein exploited and that his death did not extinguish. Each of these groups has an external patron, a territorial claim, and a grievance older than the Islamic Republic. The IRGC's weapons depots are distributed across all of these provinces and each base commander becomes a potential warlord the moment the chain of command breaks. This is not a two-sided civil war. It is a multi-actor fragmentation with external powers backing different factions across a country three times the size of Iraq, with 88 million people and enriched uranium in the basement.

The opposition cannot prevent this because the opposition does not exist as a unified entity. Reza Pahlavi has supporters but Trump himself has expressed doubts about his ability to command support inside Iran. The MEK is reviled by most Iranians. The secular democrats and the ethnic movements and the monarchists and the religious reformers have not agreed on a path, a leader, a constitution, or a flag. The model codes this as "unified opposition: NO" and "unifying leader: NO." The current reporting confirms both at the granular level. There is no Mandela. There is no Walesa. There is no one who can stand in front of a camera and credibly say I speak for Iran and have the country believe it.

The honest probability is 70 to 85 percent for armed fragmentation within 18 months. Not a clean civil war with two sides and a front line. A Libya-after-Gaddafi scenario at four times the population, on terrain that borders six countries, with no geographic logic to contain it.

The academic literature on civil war termination makes this worse, not better. Fearon and Laitin's foundational work at Stanford, analyzing every civil war from 1945 to 1999, found that the factors predicting civil war are not ethnic or religious diversity per se but the conditions that favor insurgency: poverty, political instability, rough terrain, and weak or collapsing state capacity. Iran after tonight's strikes satisfies all four. Fearon's subsequent research on duration found that civil wars involving multiple armed groups with external patrons and access to natural resources or smuggling routes last dramatically longer than two-party conflicts. The median duration for what he calls "sons of the soil" conflicts, where ethnic groups fight over territorial homelands with outside backing, is 29 years. Iran has at least four such groups, each with its own patron and its own territorial claim, stacked on top of an IRGC that will fragment into regional commands rather than surrender as a unit.

The research on how civil wars end is equally grim. Licklider's study of 91 post-1945 civil wars found that half of all negotiated settlements broke down into renewed fighting, compared to 15 percent of military victories. But Downes refined this: in ethnic civil wars specifically, two-thirds of negotiated settlements collapsed. Every single case of a civil war reigniting after a settlement in Walter's dataset occurred in an ethnic conflict. Power-sharing agreements in Lebanon, Angola, Chad, Sierra Leone, and Rwanda all failed. The alternative, military victory, is more stable but Licklider found it is also more likely to be followed by acts of genocide. This is the dilemma the literature describes and Iran will face: negotiated settlements between ethnic factions fail at catastrophic rates, and military victories produce mass atrocities. There is no clean path through a multi-ethnic civil war. There is only the question of which form the suffering takes.

Beyond the model's predictions, three risks demand naming. First, and most dangerous: Iran's nuclear material. As of the last verified IAEA count, Iran possessed 441 kilograms of uranium enriched to 60 percent purity, enough for roughly nine nuclear weapons if further enriched to weapons-grade. Most of this material is stored in an underground tunnel complex at Isfahan whose entrances were damaged in the June 2025 strikes but whose interior survived largely intact. The IAEA has not had access to any of Iran's enrichment facilities since June. It cannot verify the size, composition, or location of the stockpile. It cannot confirm whether enrichment has resumed. Satellite imagery shows regular vehicular activity around the Isfahan tunnel entrances, but no inspector has been inside. Under centralized control, however hostile the regime's intentions, there was a chain of command responsible for this material. Under fragmentation, there is not. The question is not whether Iran will build a bomb. The question is whether 441 kilograms of near-weapons-grade uranium will remain in one place, under one authority, when that authority ceases to exist. The IAEA's own director has called verification "long overdue." After tonight, overdue does not begin to describe it.

Second: the Strait of Hormuz. Twenty-one percent of global oil transit. The IRGC Navy, which was not destroyed tonight, has the capacity to mine it, harass tankers, or simply create enough ambiguity to spike insurance rates and reroute global shipping. A fragmented IRGC with nothing to lose is more dangerous to the strait than a centralized IRGC that could be deterred. Third: ISIS or its successor. The historical rate of terror group emergence from regime collapse is 3 of 5 in the failure cases. The conditions that produced ISIS in the Iraqi vacuum, a disbanded security apparatus, sectarian fracture, ungoverned territory, weapons proliferation, are reproducible in western Iran within 18 months.

These are not prophecies. They are the modal outcomes across 11 cases with feature profiles matching Iran's. The burden of argument falls on those who claim this time will be different, not on those who observe that it has never been different before.

If the Model Is Wrong

Iran 2026 would be the first case in the dataset where air-only intervention, without ground troops, without a transition plan, without an occupation force, without international consensus, without a security apparatus buyout, without institutional continuity, and without a unified opposition or unifying leader, produced a stable democratic outcome. Every feature that has historically been required would be shown to be optional. Every feature that has historically predicted failure would be shown to be survivable. The model would require not adjustment but abandonment.

We would welcome this outcome. We do not expect it.

The Soviet Comparison Is Wrong

Western commentators have already begun describing the strikes as Iran's "fall of the Berlin Wall" moment, the collapse of a sclerotic regime under the weight of its own contradictions. The comparison is not just optimistic. It is historically illiterate. The Soviet Union collapsed from internal rot over years, with a reformer at the top who chose not to use force against his own population, who negotiated the transfer of nuclear arsenals through arms control infrastructure built over decades, and who had counterparts in the West willing to manage the transition patiently. Iran has no Gorbachev. The reformists were arrested during the January crackdown. The arms control infrastructure does not exist. The IAEA has not been inside an Iranian enrichment facility since June 2025. And the Soviet collapse, even under these comparatively favorable conditions, still produced wars in Chechnya, Georgia, Transnistria, Nagorno-Karabakh, and Tajikistan that killed hundreds of thousands and are still burning three decades later. If that is the optimistic template, the optimists have not read their own example.

Three Scenarios

Best realistic case: the IRGC state. The Assembly of Experts convenes quickly. Reports already indicate Ayatollah Arafi has been named interim leader. If the IRGC command structure holds together under a Larijani or Ghalibaf figure, and if the clerical establishment rallies around Arafi as supreme leader, there is a narrow path to reconsolidation. The regime hardens. It becomes more militaristic, more insular, more hostile. This is exactly what the CIA's pre-strike assessment predicted. The IRGC state is worse for Iranian civil liberties, worse for regional stability, worse for nuclear nonproliferation than the status quo of February 27. But it maintains territorial integrity and centralized control over 441 kilograms of enriched uranium. The irony is brutal: the best thing that can realistically happen now is the thing the strikes were designed to prevent.

Most likely case: center holds, periphery fragments. Tehran, Isfahan, Shiraz, and the major Persian-majority cities come under a reconsolidated IRGC-dominated government within months. But the periphery fractures. Kurdistan, Sistan-Baluchestan, and parts of Khuzestan enter varying degrees of insurgency and separatist conflict. Turkey conducts military operations inside Iran's western border. The Baloch insurgency intensifies along the Pakistani frontier. The new government fights a multi-front counterinsurgency while trying to maintain nuclear custody and keep the oil flowing through Khuzestan, which is the province most likely to be contested. This is the Egypt-after-Mubarak meets early-stage-Syria scenario: not full state collapse but not functional governance either. A decade of low-grade civil conflict, refugee flows into Turkey, Iraq, and Pakistan, and regional destabilization that makes the current Middle East look stable by comparison. Fearon's research on duration says the median for this type of conflict is measured in decades, not years.

Worst case: full fragmentation. The IRGC fractures along regional command lines. No successor consolidates authority. Each provincial commander becomes a warlord with access to heavy weapons and, in the case of Isfahan, proximity to the nuclear stockpile. Kurdish groups declare autonomy, triggering Turkish intervention. Khuzestani Arabs move on oil infrastructure with Gulf state backing. Baloch and Azerbaijani separatists activate simultaneously. The nuclear material at Isfahan, Fordow, and Natanz falls under the custody of local commanders who answer to no one and use it as leverage, sell it, or lose control of it entirely. The Strait of Hormuz gets mined or harassed, oil prices spike past $200, triggering a global recession. ISIS or a successor organization establishes a foothold in western Iran's ungoverned Sunni-majority border regions. Refugee flows reach the millions. Iran becomes a failed state at four times Libya's population, with nuclear material unaccounted for, bordering six countries, three of which (Iraq, Afghanistan, Pakistan) are themselves unstable. This is not a tail risk. It is a scenario the model assigns 70 to 85 percent probability, and it is the scenario for which no one in Washington has articulated a plan.

The Berlin Wall fell because East Germans walked through it and nobody shot them. That is not what is happening in Iran. What is happening in Iran is that the central authority that held a multi-ethnic, multi-factional, nuclear-armed state together has been destroyed from the air, and the people who will decide what comes next are not the Iranian people in any collective sense but the individual commanders, clerics, warlords, and ethnic leaders who control territory, weapons, and the loyalty of men willing to use them.

What This Piece Is Not

This is not an argument that the Iranian regime deserved to survive. It was a theocracy that murdered tens of thousands of its own citizens. It funded proxy wars across the region. It pursued nuclear ambiguity as a strategic weapon. The people in the streets demanding change were right. The Woman, Life, Freedom movement was the most significant democratic uprising in the Middle East in a generation.

This is not an argument that regime change can never work. Germany, Japan, Spain, South Korea, South Africa, and Chile prove that it can. The data shows what success requires. It requires things that are hard: transition plans, institutional continuity, security force buyouts, international coalitions, sustained investment. It sometimes requires things that are morally uncomfortable, like letting former regime members participate in the new order. It always requires a plan for the day after.

The administration has not provided one.

The model has not found one.

What the model has found, what the historical record confirms, what the CIA's own pre-strike assessment concluded, and what the structural analysis of Iran's five competing power centers makes nearly certain, is armed fragmentation. Not as a worst case. As the central tendency. The convergence is unusually tight: every analytical lens pointed at this intervention produces the same shape. The Naive Bayes classifier gives it a 99.99999% probability of bad outcome. The historical base rate for civil war in matching cases is 80%. The ethnic fragmentation analysis identifies four groups with external patrons and active armed movements, more than any case in the dataset. The IRGC reconstitution pattern has a 100% base rate. The CIA told the administration this before the bombs fell. The bombs fell anyway.

The shape of what is coming, stated plainly: within three to twelve months, armed groups will control territory in Kurdistan, Sistan-Baluchestan, and Khuzestan. Turkey will conduct military operations inside Iran's western border to prevent Kurdish consolidation, as it has in Syria and Iraq. The IRGC will fracture along regional command lines, with each base commander becoming a de facto warlord with access to heavy weapons, and in the case of Isfahan, proximity to 441 kilograms of enriched uranium that no international body can currently verify or secure. Refugee flows will begin toward Turkey, Iraq, Pakistan, and Azerbaijan, and they will not be in the thousands. Iran has 88 million people. When the infrastructure fails, when the hospitals run out of supplies because the ports are contested and the roads are cut, when the power grid collapses because the technicians have fled and the gas lines have been bombed, the displacement will be in the millions. The UN system is not resourced for this. UNHCR's entire global budget would not cover the first year of an Iranian refugee crisis at the scale the model predicts.

None of this is inevitable in the way that gravity is inevitable. A transition plan could still be produced. An international coalition could still be assembled. The security apparatus could still be offered a path that gives its members something to protect other than themselves. Every one of the six missing features could, in theory, be supplied after the fact. Germany's transition plan was written after the war ended, not before. But Germany had 2.5 million occupation troops on the ground and the will to stay for a decade. The current administration has ruled out ground forces, has no coalition, has no plan, and has told the Iranian people to "take over your government" as though this were advice rather than abandonment.

The window for preventing the worst outcomes is measured in weeks, not months. After the regional commanders consolidate their positions and the ethnic militias establish territorial control, the cost of reversing fragmentation increases by an order of magnitude. This is not a call for more intervention. It is a statement of what the data shows: the intervention that occurred, without the structures that make intervention survivable, has set in motion a sequence that has never once, in the modern record, resolved itself without catastrophic human cost. The people who will pay that cost are not the people who ordered the strikes. They are the family in Isfahan with the suitcase by the door.

The Iranian people deserved better than the regime that killed them. They also deserve better than an intervention modeled on every failure in the modern record and none of the successes.

In Isfahan the suitcase is by the door. It is light because they are coming back in a week. The saffron is wrapped in a sock in the corner where it will not get crushed. The car keys are in her husband's hand. The cartoon is still playing and the boy is still laughing at something in a language he does not understand and the glow on the horizon toward Natanz has not dimmed.

They will leave before sunrise. They will join a column of cars on the highway south and the column will grow and it will not stop growing. She will tell her daughter they are visiting relatives. Her daughter is eleven. Her daughter will not ask which relatives. Children know when not to ask questions. It is something they learn without being taught, in countries where the buildings disappear overnight and the adults stop explaining why.

The saffron will make it to Turkey. Her daughter will outgrow the school uniform before they ever see Isfahan again.

Framework and Data

The Regime Change Outcomes Knowledge Graph is implemented in RDF/OWL (Turtle serialization). The extended ontology contains approximately 928 triples modeling 13 cases (including Venezuela 2026) across 23 boolean features and 2 continuous features, plus successor candidate profiles, ethnic fragmentation risk modeling, security apparatus survival patterns, nuclear stockpile tracking, intelligence assessments, and goal consistency constraints. OWL class restrictions define four inferred categories: PredictedFailure (cases missing all six critical features), ViableTransitionalFigure (successor candidates who control security or economic networks and can interface with external powers), WillReconstitute (security apparatuses that are economically embedded and not offered a survival path), and HighFragmentationRisk (ethnic groups with an external patron, active armed group, and separatist history). The reasoner identifies which candidates, apparatuses, and groups satisfy each class without editorial intervention. Classification uses Naive Bayes with Laplace smoothing, acknowledging conditional independence violation as a known limitation. All feature separations, predicted probabilities, and historical rates are computed from the data rather than set by authorial weight. The ontology, data, and classification code are available for inspection, replication, and critique.

Sources

Population, territory, and regime duration: CIA World Factbook, World Bank Development Indicators.

Security force estimates: IISS Military Balance 2024-2025; RAND Corporation assessments.

IRGC economic embeddedness: Brookings Institution, "The IRGC's Economic Empire" (2023); RAND, "Iran's Political, Demographic, and Economic Vulnerabilities" (2024).

Positive case transition features: Linz and Stepan, Problems of Democratic Transition and Consolidation (1996); Huntington, The Third Wave (1991); O'Donnell and Schmitter, Transitions from Authoritarian Rule (1986).

Civil war onset and duration: Fearon and Laitin, "Ethnicity, Insurgency, and Civil War," American Political Science Review 97:1 (2003); Fearon, "Why Do Some Civil Wars Last So Much Longer Than Others?" Journal of Peace Research 41:3 (2004); Fearon and Laitin, "Sons of the Soil, Migrants, and Civil War," World Development 39:2 (2011).

Civil war termination and negotiated settlements: Licklider, "The Consequences of Negotiated Settlements in Civil Wars, 1945-1993," American Political Science Review 89:3 (1995); Downes, "The Problem with Negotiated Settlements to Ethnic Civil Wars," Security Studies 13:4 (2004); Walter, Committing to Peace (2002); Hartzell and Hoddie, Crafting Peace: Power-Sharing Institutions and the Negotiated Settlement of Civil Wars (2007); Fearon and Laitin, "Civil War Termination" (2007).

Outcome assessments: Iraq Body Count; Syrian Observatory for Human Rights; UNHCR displacement data; Libya Analysis (Chatham House); Freedom House country reports.

Intervention details: Congressional Research Service; UN Security Council resolutions; DOD press briefings.

© 2026 August Holloway / Dead Letter Publishing. All rights reserved. No part of this work may be reproduced, distributed, or transmitted in any form without prior written permission.