Containment Is Not a Lid

One appeared in The Atlantic, written by two terrorism scholars, and its headline was "A Silver Lining in Afghanistan." Al-Qaeda has not launched an attack on the United States or Europe from Afghan soil since the withdrawal. It has not named a successor to Ayman al-Zawahiri, who was killed by a drone flown from over the horizon in 2022 — proof that the reach survived the occupation's end. Its acting leader is hiding in Iran, and the Taliban is reportedly reluctant to let him into the country at all. The White House's own top adviser on the threat now calls the Taliban "moderately cooperative counterterrorism partners."

The other came from UN Women. Its headline was "the normalization of the exclusion of women and girls." More than a hundred decrees in five years. More than half of Afghan women now report leaving their homes twice a month or less. Seven in ten describe their mental health as bad or very bad. A 2026 decree abolished legal equality between men and women outright; another made it substantially harder for a woman to exit a marriage while preserving a man's unilateral right to end one.

Neither document is spin. They are two columns of the same ledger, and almost nobody reads both.

---

i · the lid model

In April 2021, CIA Director William Burns warned that withdrawal carried a "significant risk" of an al-Qaeda resurgence. That was not a fringe position. It was the consensus of nearly everyone with a security clearance and most people without one. The forecast was not merely wrong; it was wrong in a specific, structural, reusable way, and that is what makes it worth five years of attention.

Every version of that prediction modeled the American presence as a lid. A lid holds down a fixed quantity of pressure. Its behavior on removal is trivially predictable — whatever was underneath comes out, at roughly the force it had going in. Twenty years of counterterrorism vocabulary is lid vocabulary: containment, suppression, degrading capacity, keeping a boot on it.

A lid has exactly one property that makes it useful as a model: it does not participate in what it holds. It is inert. It does not feed, recruit, legitimize, finance, or organize the contents.

Almost no real containment system has that property.

Historians have argued for the better part of a century that Rome's frontier system helped manufacture the very confederations it was built to hold back. The limes was not a wall standing outside barbarian society; it was the largest economic actor inside it. Subsidies, frontier trade, and auxiliary recruitment concentrated wealth and status in the hands of whichever chieftains could broker access to Rome — and small, fissile groups consolidated into the durable coalitions — Franks, Alamanni — that would eventually walk through. The defense was a state-formation program for the enemy.

Closer in: internment in 1971 and Bloody Sunday in 1972 are treated, including in the British Army's own retrospective assessments, as the most effective recruiting events the Provisional IRA ever had. Closer still: in 2006, Ethiopia invaded Somalia to break the Islamic Courts Union. Al-Shabaab was the Courts' youth militia before the invasion; the invasion made it the vanguard of a national resistance and handed it a decade.

The pattern underneath all three is the same. An occupying containment system supplies its adversary with the scarcest resource in all insurgency: a legible enemy. Foreign, uniformed, visibly other, morally intelligible to a farmer who has never read a manifesto. No domestic grievance recruits like that. It is a mobilization subsidy, paid daily, by the party that believes it is doing the containing.

---

ii · re-equilibration is not release

Remove a lid and the contents escape. Remove a participant and the entire system re-equilibrates — and you cannot derive the new equilibrium by extrapolating from the old one. This is why the 2021 forecasts failed. They ran competent arithmetic on the wrong kind of object.

Two positions changed in August 2021. Those changes accomplished what two decades of force could not.

The Taliban went from insurgent to incumbent. An insurgent is paid in disorder; an incumbent is billed for it. Same organization, same men, same theology, opposite incentives. In the 1990s, hosting al-Qaeda was free advertising. Today it is a liability weighed against recognition, trade, customs revenue, and the ability to keep the lights on — which is why Kabul appears to be quietly declining to host al-Qaeda's own acting leader. Nobody became moderate. The chair changed, and the chair has requirements.

ISKP lost its occupier. Its founding pitch was that the Taliban were apostates who had signed a deal with the infidel. Remove the infidel and the argument thins. Worse, the group now has to explain to Afghans why it is bombing mosques full of Afghans. The moral frame a foreign army donates for free had to be self-financed, and it turns out to be expensive.

For twenty years an extraordinary quantity of force was applied to the Taliban and changed its behavior toward transnational jihadists not at all. A change in position changed that behavior almost immediately. This is what coherenceism means by alignment over force: force pushes against a system's momentum, while position changes what the system is trying to do — and a system trying to do something else requires no pushing.

---

iii · pressure selects, it does not delete

Here is where a good pattern refuses to become a victory lap.

The Taliban's campaign against ISKP has been genuinely effective inside Afghanistan. Attacks have fallen; the International Crisis Group's 2025 assessment asked, reasonably, whether the group was a "threat in retreat." But the same analysts documenting the decline note the other half: squeezed at home and no longer effective locally, ISKP turned outward. Kerman. Crocus City Hall. Istanbul. Disrupted cells in Europe, inspired plots in the United States. The quiet in Kabul was partly paid for in Moscow.

This is the second pattern, and it is less comfortable than the first: pressure on a distributed adversary does not remove it, it relocates it. Mexico's kingpin strategy is the cleanest laboratory version — decapitate the cartels and you do not get fewer cartels, you get more of them, smaller and more violent, because you have selected for fragmentation. Pressure is a selection force. It does not eliminate the organism. It determines which variant survives.

So the honest ledger on the withdrawal is not "the threat vanished." It is: the threat moved, and the people it moved onto were not the people writing the forecast.

---

iv · the debt off the books

And then there is the cost for which the counterterrorism ledger has no column at all.

To stop its own fighters from defecting to a purer rival, the Taliban has out-purified it. No decree cites that motive, and none ever will; the inference runs from the incentive, which is where evidence of this kind always lives. But it means the repression documented by UN Women is not incidental to the counterterrorism success — some meaningful fraction of it is downstream of that success. This is the mechanism nobody wants in the summary: more than a hundred decrees, 638 maternal deaths per 100,000 live births, half a population that leaves the house twice a month or less — partly the price of denying a jihadist rival its recruiting ground. The stability is real. It is financed by making half the country illegible to the system that governs it.

Coherenceism holds that coherence is not the good. A system can be internally consistent, stable, low-friction, and entirely illegitimate — because consistency purchased by removing a voice is not coherence, it is quiet. The distortion does not disappear when you stop measuring it. It stops appearing in your instruments, which is a different thing, and the difference is where the next twenty years get made.

Which is precisely the error the occupation committed, now being repeated by its successor. The occupation measured the system by the threats it could name and missed what its own presence was feeding. The Taliban measures the system by attacks prevented and misses what its exclusion is incubating. Illegibility is not stability. It is incubation — how a grievance acquires a founding story, and a story acquires recruits, and recruits acquire a name that ends up in someone's threat assessment in 2046.

A coherence is legitimate by including those it affects, not by suppressing them. That is not a moral flourish appended to a security analysis. It is the security analysis, on a timescale long enough to be honest. Every order bought by silencing a constituency is carrying a debt off the books, and the interest is always paid by whoever inherits the ledger.

---

v · two questions

Any containment system deserves both of these — a border regime, a sanctions architecture, an occupation, a content-moderation stack, a policing strategy, a compliance department, a parenting philosophy. Ask them before you build it and once a year while it runs.

What is it feeding?

Where does the pressure go?

Neither question has a good answer that is "nowhere." If your model says "nowhere," you have drawn a lid where there is a participant, and you should expect to be surprised in the same direction the class of 2021 was surprised.

The withdrawal from Afghanistan was never a solution. It was a hand coming off a scale. The scale kept weighing — it always does, and it is weighing right now in a country where a hundred decrees have made half the population invisible to their own government's instruments. What five years bought us is not vindication. It is a correction to the shape of our own maps.

We are almost never standing outside the thing we are holding back. Usually we are one of the reasons it is there.