When Power Tries to Accelerate History

I left Luminara at the end of the last chapter because a warning that cannot survive contact with evidence is only a story one tells oneself. The first Earth history I chose was Afghanistan, precisely because its justified security origin and mixed consequences could test the conclusion I had arrived prepared to draw.

On September 11, 2001, al-Qaeda killed nearly three thousand people in attacks on the United States. The organization had used Afghanistan as a sanctuary under the Taliban. Nine days later, President George W. Bush publicly demanded that the Taliban hand over al-Qaeda leaders and close the camps in which the network trained. NATO subsequently announced that the attacks had been directed from abroad by al-Qaeda, protected by the Taliban regime, and treated the attack as falling under the alliance’s collective-defense provision. When U.S. and British military operations began on October 7, their stated purpose was to disrupt Afghanistan’s use as a terrorist base and attack the Taliban’s military capability.

This matters because it prevents an easy moral story. Afghanistan was not selected for intervention merely because a stronger state disliked its institutions. A transnational organization had attacked from a sanctuary the Taliban had allowed it to use. Before asking whether outside power later tried to accelerate Afghan political development, the original security problem has to remain visible. A theory of non-intervention that cannot accommodate defense against an external attack is not restraint. It is blindness in the opposite direction.

The same record also prevents a second simplification. The Taliban government was violently repressive, particularly toward women and dissenters. Removing it therefore changed more than the location of a terrorist sanctuary. It changed who could go to school, who could occupy public space, which armed factions could claim authority, and which political arrangements could become imaginable. Once that happened, the intervening powers faced a question more difficult than the one that had brought them into the country: after destroying the order through which the original threat had been sheltered, what were they now responsible for helping to build?

A War With a Reason

The first answer did not come from Washington alone. In December 2001, Afghan representatives meeting under United Nations auspices signed the Bonn Agreement. It created an Interim Authority, provided for an emergency Loya Jirga, and set a route toward a constitution and permanent institutions. The United Nations Security Council then authorized an International Security Assistance Force, initially in Kabul and its surroundings, to help the Afghan Interim Authority maintain security. Afghanistan’s 2004 Constitution was later adopted through a Constitutional Loya Jirga.

These facts complicated my Luminaran analogy immediately. In Sareth, I had described stronger powers carrying an institutional template into a political field they only partly understood. Afghanistan contained that danger, but it also contained Afghan political agency, international authorization, and genuine demand for reconstruction. The new order was neither simply imposed from outside nor simply generated from within. It emerged through a relation in which local actors, foreign governments, international institutions, armed factions, returning exiles, established powerbrokers, and communities possessed unequal but real influence over what came next.

That mixture is precisely where the question of standing becomes difficult. A state may have authority to defend itself against an armed threat. A coalition may be asked to provide security while a transitional government forms. Donors may legitimately finance schools, clinics, roads, courts, and public administration. Yet none of these permissions answers every later question merely because the first intervention was justified. Each expansion changes the object. Authority to destroy a terrorist camp is not automatically authority to define a justice system; responsibility to prevent institutional collapse does not automatically establish that an outside actor understands which institutions should replace what existed before.

By 2003, the mission had acquired a much larger future. A joint U.S.-Afghan statement described a shift toward long-term reconstruction and rebuilding Afghan institutions, with the stated aim of a democratic constitutional government adapted to Afghanistan’s culture and history. NATO’s Provincial Reconstruction Teams were later described as instruments for extending the authority of the central government and facilitating development and reconstruction beyond Kabul. Security was no longer only the protection within which Afghan politics would unfold. International power had become one of the mechanisms through which the political center, its institutions, and its reach were being built.

The shift had begun even earlier through the partners chosen to fight the Taliban. U.S. leaders deliberately sought a relatively light American footprint and worked with Afghan anti-Taliban forces, including commanders associated with the Northern Alliance. The logic was compelling. Those forces knew the terrain, already had fighters in the field, and could take ground that would otherwise have required a much larger foreign deployment. They were not remote-controlled instruments. They had their own enemies, constituencies, histories, and ambitions, which is precisely why they could fight effectively.

That effectiveness had a political price that was not separable from the military gain. Money, weapons, intelligence, air support, access to foreign officials, and recognition increased the leverage of the Afghan actors who received them. Some would later occupy positions in the post-Taliban order; others used the resources and impunity of wartime partnership to strengthen patronage networks. SIGAR would eventually argue that support for warlords and other strongmen helped empower actors whose private networks competed with formal state institutions. The important point is not that every anti-Taliban partner was predatory or that the coalition had an obvious substitute in 2001. It is that selecting the force capable of solving the immediate security problem also selected political actors who would enter the next political settlement with altered power.

This is what proxy action had taught us on Luminara, and Afghanistan supplied an Earth version with enough differences to keep me cautious. The stronger actor chooses a partner because the partner already possesses local capability. The partnership then enlarges that capability, which changes the local field the stronger actor later tries to stabilize. If the partner behaves differently from what the sponsor wants, that is not evidence that intervention was irrelevant. It is evidence that intervention joined a political system containing actors whose purposes remained their own.

There was no single moment at which defense turned into state-building. The movement was more understandable than that. A weak police service created a security problem, so police had to be trained. Courts that could not operate created an enforcement problem, so judicial institutions had to be supported. A government unable to pay staff could not extend authority, so public finance required assistance. Roads, schools, clinics, elections, ministries, military logistics, anti-corruption bodies, local government, and economic rules each appeared as conditions of a stable settlement. The more the mission tried to secure what it had already changed, the more aspects of Afghan political life entered the reconstruction project.

This is the point at which acceleration becomes hard to distinguish from assistance. A stronger actor can supply money, expertise, security, technology, and organizational templates much faster than local institutions can accumulate the relationships that make those resources self-sustaining. Speed can therefore solve one problem by changing the conditions of the next. The school opens because a donor can finance it now; the ministry functions because advisers can fill a gap now; the security force fields a capability because contractors can maintain it now. Each success may be real. Each also alters what local actors learn to expect, which organizations attract talent, which political brokers gain access to resources, and what will fail if the external contribution later disappears.

The Mission Entered the Society

The United States’ own reconstruction inspector later described the twenty-year effort as a series of compressed timelines and repeatedly underestimated requirements. In its 2021 lessons report, the Special Inspector General for Afghanistan Reconstruction, or SIGAR, argued that U.S. officials often treated decisions made in Washington as though they could rapidly change the incentives of Afghan institutions, communities, and powerbrokers. Pressure to demonstrate progress encouraged fast spending and short-term outputs. Programs that were supposed to build enduring local capacity were often designed around deadlines created by the intervening governments’ political clocks.

The archive forced me to extend that clock backward. I had been treating 2001 as the starting point because that was when the intervention I was examining began, but Afghanistan’s political field had already been altered by rival external projects. The Soviet Union sent combat forces into Afghanistan in December 1979 and sustained a government whose survival became increasingly bound to Soviet military power. The United States had already begun covert nonlethal support to Afghan insurgents earlier that year; after the Soviet invasion, the program expanded to lethal equipment, routed principally through Pakistan with Saudi participation. Soviet forces completed their withdrawal in February 1989, but the fighting did not end with them. None of this makes Afghan factions puppets or turns the later civil wars into automatic consequences of foreign sponsorship. It does something more important for my argument: it removes the fiction of an untouched society waiting in 2001. External power had already altered which armed organizations survived, which patrons they knew, which security expectations hardened, and which institutions had been militarized or weakened.

That longer history changed how I read the post-2001 money. It entered a society whose political networks had survived decades of conflict and foreign sponsorship. SIGAR concluded that U.S. support sometimes legitimized and financed strongmen whose interests did not coincide with the new state’s formal institutions. Efforts to build a market economy and formal legal order also encountered existing patronage systems, religious law, local dispute resolution, and officials whose authority depended on relationships the imported design did not capture. Between 2003 and 2015, the United States spent more than a billion dollars on rule-of-law programming, about ninety percent of it on the formal legal system; SIGAR’s review cited estimates that roughly eighty to ninety percent of civil disputes had historically been handled through informal mechanisms. The point is not that informal justice was necessarily fairer. Taliban dispute resolution could be coercive and substantively unjust. The narrower point is that building the visible apparatus of a legal system does not make people bring their disputes to it.

A similar substitution appeared in administration. Corruption made donors reluctant to route large amounts of assistance through Afghan government systems. Sending aid around those systems reduced immediate fiduciary risk and allowed projects to move faster. It also meant that Afghan officials gained less experience managing some of the very programs they were eventually expected to sustain. Here I recognized Sareth again, arrived at from the opposite motive. In Sareth, outside advisers had displaced local judgment because sponsors distrusted local capacity; in Afghanistan, donors often bypassed public systems because the risks of corruption and misuse were real. The motives were different, yet the developmental pressure could converge: when the institution expected eventually to carry a function is repeatedly bypassed so that the function can succeed now, short-term competence can coexist with long-term weakness. Where local personnel could not maintain advanced military or administrative capabilities, foreign advisers and contractors often performed the difficult functions themselves rather than allow the service to fail. The short-term result could be competence. The long-term result could be a competence whose address remained partly outside the country.

Once that pattern spreads far enough, dependency stops being an accidental side effect and becomes part of the operating environment. The scale of the Afghan dependence stopped me because it changed what the word support could mean. On Luminara, we had learned that when an extraordinary contribution becomes ordinary enough to organize budgets, careers, and institutional expectations, its eventual removal is not a return to a prior baseline; it is another political shock. By 2020, according to the World Bank, foreign grants were still equivalent to roughly forty-five percent of Afghanistan’s GDP. They financed about half of the government budget, three quarters of total public expenditure, and around ninety percent of government security spending. These figures do not show that Afghans lacked institutions or agency. They show that the Republic’s fiscal and security architecture had evolved for years under conditions in which extraordinary external support was normal.

That distinction matters when support is later withdrawn. Withdrawal is often imagined as the removal of an outside force so that the local society can resume its own trajectory. But a twenty-year intervention leaves no untouched trajectory waiting beneath it. Political careers have formed around the institutions it financed. Armed groups have adapted to the security architecture it created. Businesses have grown around its spending. Communities have made decisions based on the durability of its guarantees. Opponents have learned how to exploit its rules and deadlines. The intervention has become one of the facts to which everyone else has adapted.

This is the Earth mechanism I had been looking for. Intervention does not merely produce intended outputs and unintended side effects. It changes the field from which later causes emerge. A militia supported for immediate security can become a political broker. A ministry protected from failure by foreign expertise can become more ambitious while remaining dependent on that expertise. A deadline designed to signal resolve can change an adversary’s strategy. A funding channel chosen to reduce corruption can weaken the institution whose capacity donors say they want to build. The intervention becomes endogenous to the history it intended to accelerate.

Yet even this conclusion becomes misleading if it is allowed to consume the whole record.

The Gains Were Real Too

The same international support that created dependencies also financed real improvements. The World Bank’s retrospective on Afghanistan from 2001 to 2021 describes rapid gains in health, education, public services, macroeconomic management, revenue collection, and human development during the period of extraordinary grant support. Child survival improved dramatically. More Afghans entered school. Public institutions acquired technical capacities they had not possessed before. Cities expanded, infrastructure improved, and a generation experienced forms of education, media, employment, civic organization, and public participation that the Taliban government had denied or constrained.

These gains are not an inconvenient footnote to a story of failed intervention. They change the story. A village clinic can be genuinely valuable even if its financing model is unsustainable. A girl who receives an education has not received an illusion because the political order that enabled it later collapses. An election can widen political participation while occurring inside a system weakened by corruption and violence. Political development is capable of producing improvement and dependency at the same time.

Afghan actors also made choices that cannot be reassigned to foreigners for the sake of a cleaner theory. Successive governments centralized authority, bargained with local powerbrokers, appointed officials, contested elections, distributed patronage, pursued reforms, resisted others, and struggled over legitimacy. The Taliban did not merely react to foreign mistakes; it organized, fought, coerced, negotiated, governed territory, and pursued its own political project. Regional powers shaped the conflict as well. Any causal account that turns Afghans into passive material reproduces the very mistake this chapter is trying to expose.

SIGAR’s later review of the Republic’s collapse makes the multiplicity explicit. It identified the government’s dependence on U.S. military and contractor support, but it also identified the exclusion of the Afghan government from U.S.-Taliban negotiations, the Taliban’s unwillingness to compromise, President Ashraf Ghani’s increasingly narrow governing circle, high centralization, endemic corruption, and the Republic’s continuing legitimacy problems. One can dispute SIGAR’s emphasis or add other causes. What cannot responsibly be done is to compress the collapse into the proposition that outside intervention failed because outside intervention always fails.

The evidence therefore forced me to narrow the Luminaran warning. Afghanistan does not prove that a stronger power must never intervene, that externally supported institutions cannot improve lives, or that the Republic’s collapse was predetermined in 2001. It supports a more structural claim: once an intervention reaches deeply into security, finance, law, administration, and political competition, the intervention itself becomes part of the developmental environment. Intervening actors can no longer evaluate later outcomes as though they had merely repaired a society whose underlying history continued unchanged.

Nor can anyone recover the missing baseline with confidence. There is no accessible Earth on which the United States disrupted al-Qaeda’s sanctuary but did not help overthrow the Taliban; or overthrew the Taliban but refused large-scale reconstruction; or funded reconstruction at a different pace; or kept troops while contractors left; or left earlier while aid remained. Counterfactuals can be reasoned about, but they cannot be observed. This is why causal modesty matters. The strongest conclusion is not that a particular foreign choice mechanically produced the Republic’s final collapse. It is that foreign choices materially changed the institutions, incentives, dependencies, and strategies from which later outcomes emerged.

This is why the distinction between capability and standing matters even when motives are mixed and some outcomes are beneficial. Capability answers what the stronger actor can do. Standing asks which part of another society’s future the actor is entitled, invited, or obliged to shape, under what authority, for how long, with what local participation, and with what responsibility for the dependencies it creates. The first question can be answered by military, financial, or technical superiority. The second cannot.

The Archive Is Longer Than the Argument

Afghanistan also showed me why this problem is so difficult to learn from while it is still unfolding. Twenty years of intervention produced presidential speeches, alliance mandates, peace agreements, appropriations, aid programs, military doctrines, contractor records, provincial reports, Afghan laws, election records, audits, interviews, casualty data, budget accounts, development indicators, and later investigations. They were written by actors with different purposes and different fields of vision. A security report can accurately describe attacks while saying little about how a funding program changed local power. A development report can accurately record school enrollment while saying little about which security guarantees made the school possible. An audit can identify an unsustainable project years after the political deadline that made speed rational to the officials who approved it.

Human investigators have reconstructed important parts of this history. SIGAR itself is evidence that a state can build institutions designed to criticize its own reconstruction effort. But even a strong retrospective investigation arrives after many choices have become irreversible. The machine-scale possibility I described in the Introduction matters here in a narrower way: not as a judge of whether the intervention was right, but as an instrument for keeping the intervention chain connected while decisions are still being made.

Such an instrument could align changing public objectives with later programs, follow which local actors gained or lost leverage after a funding or security decision, compare promised sustainability with the support still being supplied from outside, and preserve competing explanations rather than letting the latest one replace the previous record. It could connect a program’s success metric to the later condition that revealed its fragility. It could surface that a supposedly local capability still depended on foreign contractors, or that a change attributed to ideology followed a funding shock. The value would lie in shortening the distance between action and the evidence of what the action was changing.

It still could not carry the causal conclusion by itself. Records are selective, some relationships remain secret, local testimony is unevenly preserved, and correlation is not authorship. A machine can discover that two events belong in the same chronology; it cannot infer from sequence alone that one caused the other. Nor can it decide whether a security necessity justified an intervention, whether a right was worth defending at a particular cost, or whether a local political bargain was legitimate. Those judgments remain with actors who can be challenged and required to answer.

The practical gain is therefore smaller than omniscience and larger than convenience. A government considering one more institution-building program could be forced to confront comparable earlier programs, including the dependencies they created. An Afghan official could challenge a donor’s theory with records showing how an earlier version behaved outside Kabul. A legislature could see that a promised exit assumes a local capability that still relies on external maintenance. The machine would not make restraint inevitable. It could make historical amnesia less available as an excuse.

But Afghanistan was still too compatible with my inherited suspicion. I needed a case that could break it.

The Counterexample I Could Not Ignore

Japan did.

After Japan’s defeat in 1945, the United States led an Allied occupation that was far more openly directive than much of the Afghan reconstruction project. General Douglas MacArthur, as Supreme Commander for the Allied Powers, possessed final authority within the occupation structure. The occupation dismantled Japan’s military, conducted political purges, pursued land and economic reforms, reorganized public institutions, and exerted decisive influence over the new constitution. If external political redesign were inherently self-defeating, postwar Japan should be one of the strongest cases against it.

The contrast becomes sharper when the occupation’s method is made explicit. U.S. instructions gave MacArthur supreme authority, but they also directed occupation control to operate through the Japanese government so far as that arrangement produced satisfactory results. MacArthur soon emphasized why this mattered: because the existing civil administration could continue functioning, the occupation did not have to create a separate military government and replace the vast professional structure beneath the political leadership. Japanese ministries, local administrations, courts, firms, and officials therefore remained part of the machinery through which occupation directives became policy. This was still coercive rule. Japanese institutions operated inside boundaries set by the occupier, and SCAP retained the power to intervene directly. Yet external authority was acting through an administrative state that continued to exist, rather than trying to construct most of the governing apparatus while simultaneously fighting an insurgency.

Postwar Japan is indeed such a case, but only in part. Japan emerged from the occupation with institutions that endured beyond 1952 and a constitutional order that became domestically operative. But that outcome cannot be credited to occupation design alone. Japanese officials and institutions retained substantial agency; prewar administrative capacity and organizational memory survived even as militarist structures were purged or dismantled; and occupation policy itself changed as economic rehabilitation and the emerging Cold War altered American priorities. The constitutional process shows the mixture particularly well: SCAP exerted decisive influence over the draft, while Japanese political institutions debated, amended, enacted, and later lived under the resulting constitution. Land reform, labor reform, economic reconstruction, education, and bureaucratic reorganization likewise passed through existing Japanese actors whose purposes did not simply merge with those of the occupier. The record therefore resists both easy stories—neither pure imposition nor untouched continuity explains what endured.

Japan therefore changed my question again. The difference between Afghanistan and Japan cannot be reduced to the moral quality of the interveners, the intelligence of planners, or the institutional form they preferred. In Japan, occupation followed total military defeat and unconditional surrender; organized armed resistance did not continue to contest the new order; and a functioning administrative apparatus remained available through which directives could be implemented and later localized. In Afghanistan, reconstruction unfolded amid continuing insurgency, fragmented authority, competing armed networks, and institutions being built while they were expected to govern. Those differences do not make Japan a template for successful occupation. They reveal why the category occupation is too coarse: the same external capability enters political fields with radically different histories, capacities, security conditions, and institutional memories.

Once those conditions enter the comparison, the phrase “nation-building” becomes too coarse to explain anything. External power can sometimes destroy an obstacle, sometimes stabilize a transition, sometimes substitute for capacities that never become local, sometimes strengthen institutions already capable of absorbing change, and sometimes do several of these things at once. The visible reform may be the same—a constitution, a court, a civil service, a security force—while the relations that allow it to survive are entirely different.

I began this chapter asking whether the Luminaran warning would survive Earth history. It does, but not in the form I brought with me. Afghanistan supports the warning that intervention cannot remain outside the society it changes. Japan defeats the stronger claim that externally accelerated political change is necessarily futile. Together they leave a harder proposition: the effect of intervention depends not only on what the stronger actor wants to build, but on the developmental conditions into which the intervention enters and the conditions it creates while trying to leave.

That is not yet an answer. It is the reason the next chapter is necessary.

If political institutions cannot be understood apart from the histories, capacities, social relations, security conditions, and accumulated expectations that let them function, then societies are not simply early or late versions of one another. A reform that accelerates development in one place can interrupt it in another because the visible institution is only part of what is being transferred.

Power can change history quickly. The unanswered question is what makes the change capable of living after power leaves.