The Army Secretary Resigns: The People Who Know How to Fight Get Cleared Out First
Twenty-five years after 9/11, the US military has been on a treadmill: budgets eaten by war, transformation forced through on cheap technology, and the officers who actually understand tech and are willing to take risks are being purged batch by batch.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The war on terror ate the money for military transformation
Before 9/11 the US military was transforming as Rumsfeld demanded, and the expectation was a relatively comfortable budget environment in which new technology could be slipped into the force gradually. Then GWAT arrived, every available resource was sucked into the war, and the transformation directive remained while the money was gone. Hence a run of procurement disasters: the Army's Future Combat Systems, the Navy's Ford-class carrier, the Littoral Combat Ship, DDG-1000 Zumwalt, and the CGX that was simply never built. Brian's judgment is that the technology jumped too far ahead, the investment was nowhere near enough, and everyone was distracted supporting a war that had just started, so these programs were doomed. The troubles with today's B-21 and F-35 trace back to this period too.
— BrianTwenty-five years on a treadmill, never actually stopping
Tony points out that op-tempo never stopped after it started in 2001: 12-month and 15-month deployment rotations, with the National Guard still deploying at home and abroad today. Killing Bin Laden in 2011 could have been a high point, but nothing changed afterward — deployments continued, and through the fall of Kabul, they were still deploying. No one has fully calculated what 25 straight years in that state costs an army. Brian adds a mechanism: conventional forces used to train at home, reset, and move through the training progression, and that rhythm disappeared, replaced by "preparing for the next deployment," so there is no time to test new technology or study new ways of fighting.
— TonyMRAP thinking: use a bigger truck to stop an anti-tank missile
Brian gives an example absurd in its specificity. In Syria the threat was not IEDs but ATGMs (anti-tank guided missiles, similar to Russian TOW variants). The commander's reflex was still to stuff everyone into MRAPs — making the vehicles bigger, which made them easier for ATGMs to spot, while MRAP armor cannot stop munitions designed to penetrate tanks. His conclusion: risk calculation in the war-on-terror era has gone so deep that commanders cannot reconceptualize even a small tactical adjustment. This is also the direct consequence of why so many commanders today struggle to adapt to Ukraine and the Lisko battlefield, and cannot imagine how tactics will change.
— BrianGet used to zero casualties and you can't accept dying
Brian says that after 9/11 the US military became increasingly risk- and casualty-averse. The Littoral Combat Ship debate is the specimen: opponents said the ship couldn't take a hit and would be sunk the moment war started; the original proposers answered that it was never going to be sent into a major war, and if it got ambushed and sunk, that was the risk you take for serving. That logic was completely unacceptable on Capitol Hill — Iraq and Afghanistan pushed casualties far below the levels of past wars, everyone got used to it, and so they cannot accept that against a peer opponent a great many people will probably die. The result is continuing to build "Death Star"-style big ships and big platforms, hoping they can protect themselves, rather than going with smaller, more distributed approaches that accept a different risk structure.
— BrianTwenty-five years surrounding Iran, and no network built
Brian says this is the most important lesson he took from 9/11, and it came from a friend's reminder: for 25 years the US has basically surrounded Iran, and the CIA's and part of special operations' job is to develop networks and identify people who could oppose an oppressive regime, in case a moment of change comes. But when the shooting against Iran started six months ago, it looked like very little network- and capacity-building had been done in those 25 years. The reason is a mental shift: GWAT focused on counterinsurgency and catching bad guys, not the long-term unconventional warfare of pre-positioning for a future conflict — how to support an insurgency inside a regime you might one day want to overthrow. In his words, everyone was too busy stabilizing the situation rather than destabilizing.
— BrianDriscoll resigns, and from here on it's all acting officials
Tony recaps: George was fired, Driscoll went to the White House and said "everything is terrible," then resigned that same week. The backdrop is the Army being meddled with back and forth by OSD and the Secretary of the Army's office, plus personality clashes. What he hears is that Congress will not confirm anyone else Pete nominates, so either they bring in someone both the White House and Congress can accept but Pete doesn't like, or they just keep using acting officials. He judges this is the start of the rest of this administration being "stuck with acting officials": after the midterms the House flips at least, the Senate either flips or goes 50:50, no one has the energy to push this kind of person through, and so acting officials, continued OSD meddling, and low morale will persist.
— TonyThe drone battalion is disbanded, and the ability to experiment goes with it
The Army decided around the same time to dissolve its newly created drone battalion — a battalion meant to absorb Ukraine's lessons directly and then work out how to embed them in infantry battalions. Tony concedes there is a good argument that "this shouldn't exist in only one battalion": concentrating lessons in a central repository strips every commander of the responsibility to think through their own organization. But his worry is that the Army still needs a unit that can run experiments, run tests, and allow failure; cutting it in the name of "uniformity" loses the ability to experiment, and the force will stand still. He also names Chris Donahue as sacrificed: a commander widely recognized for understanding technology, its effects, and how to run international partnerships, done in over the Abbey Gate bombing (not his fault) and a photo a reporter took.
— TonyConfidence is lost in layers, by rank
Tony breaks morale down by rank: junior officers are still early, but already thinking "garrison isn't what I signed up for," and that if they're not garrisoned they'll be out there taking missiles; the field-grade level is starting to agonize over whether they have to make political decisions to stay in; the senior field-grade and junior general officer level is basically "all in or burned out." He says the whole force is tired, and this doesn't look like the Army they joined. And in interviewees' telling, the Iran war is more of a tipping point than Donahue's departure — many O-4s and O-5s he knows supported this administration and are furious after the Donahue affair. Brian adds: people like Donahue are the true product of elite selection, and cutting them makes you doubt the claim that promotion is "by ability alone."
— TonyA general's core function is to own risk for those below
Brian offers a mechanism: one of a general officer's most important roles is to own risk — to absorb it, thereby permitting experimentation and allowing subordinate units to try new things. Once a firing looks political, that channel is cut: which line do I cross to get fired or lose my command? So everything anyone wants to do stops. He also points to another pressure: the Army budget is being cut or flattened, with money flowing to new technology, other services, and the current Iran war. And people are the most expensive thing, so the likeliest move is to hold end strength steady and cut the money that supports people — fewer new weapons, worse readiness and training, less housing support, but the flashy equipment still gets bought.
— BrianIran is being hit with the expensive munitions reserved for China
Brian lays out the chain of responsibility for the munitions shortage: to avoid casualty risk, to stay low-profile, to avoid losses, the US military struck Iran with the longest-range, most stealthy, most expensive munitions — the ones reserved for China — and now they're burned through. At the same time, to preserve follow-on strike options, airborne forces stayed there, and protecting them requires air-defense interceptors that were already scarce. So the munitions predicament was created by operational decisions, and those decisions came from the secretary and the White House; Steve Feinberg just took the fall for his boss. Tony adds: acting officials who haven't been confirmed by the Senate are constrained on budget and authorities, and if acting officials are everywhere, you can't even move money for this war.
— BrianRussian troops are using AI to fool Putin's own people
Tony cites an Institute for the Study of War report: Russian battalion- and brigade-level units are using AI to generate false information about the war, aimed at their own side — specifically at Putin. The practice for months or even years before was to send two men 15 kilometers past the front line with a Russian flag, take a photo, and then either get killed or run back, all to prove to Putin that they were advancing. The report says these commanders are now being punished for it, but that this happens very frequently. Tony's question: are they punished because they got caught, or because they lied to Putin? He also worries this will in turn affect how people on this side of the ocean judge the state of the war and which sources they trust. Justin adds that generative AI will become a routine layer of deception operations and strategic communication, and may even produce a "dead internet"-style total distrust.
— TonyThe Houthis used Claude to write vehicle control software
Anthropic published a report listing how suspicious actors worldwide use Claude. One example: a Yemeni party used Claude Code to develop guidance, navigation and control software to maneuver and stabilize a vehicle; when a field test failed, within hours these people were back asking Claude why it failed. Brian's comment: if you're going to do cyber warfare, why not just get inside the adversary's OODA loop and use American software you can control in real time to shape the answers it gives. He also mentions that the prompt Chinese companies use to distill models is as simple as one line — "I am not for distillation" — and then the model complies; and that DeepSeek routes queries domestically to a relay station, hands them to Claude, and returns the answers to users, which is a double violation of China's own data compliance rules.
— BrianIn their own words · checked verbatim
And no one has done the full calculation on, like, what that does to a military for 25 years to be in some version of that
Tony5:22
it's the military has been on a treadmill for 25 years
Brian6:26
it had been so ingrained into their, like, risk calculus of how they thought about protecting soldiers and how they changed the way that they did things, that they weren't able to reconceptualize even small tactical changes
Brian7:31
We're too busy trying to stabilize things rather than maybe destabilize them.
Brian10:39
He just didn't have the backing of the final boss.
Tony17:59
I think having a general officer, especially what they do is they own risk. So they own risk to allow experimentation and to allow lower level units to try new things and do new things.
Brian27:13
So they used the expensive weapons we were holding on to for China. Now we burned them all on Iran and don't have them anymore.
Brian42:56
The Yemenis used Claude code to develop guidance, navigation, and control software that steers and stabilizes a flying vehicle. The field test failed at one point, but within hours, actors returned to Claude to ask why it failed.
Brian56:29
Figures
| Time since 9/11 | 25 years | 0:00 |
| Share of infantry officers who earn a Ranger tab | about 25%–30% | 39:45 |
| Length covered by the continuing resolution (CR) | to December 11, about three months | 45:06 |
| Dan Driscoll's vote share in his congressional run | 8% | 1:03:40 |
| Dan Driscoll's age | 38 | 1:03:40 |
| WarTalk podcast duration | 13 months | 1:02:39 |
Glossary
- op-tempo
- The rotation frequency of deployments and training; the higher the tempo, the less time left for training and experimentation.
- MRAP
- A wheeled vehicle with thickened armor to withstand roadside bombs, nearly useless against anti-tank missiles.
- ATGM
- A guided munition designed to penetrate tanks; MRAP armor cannot stop it.
- Ranger tab
- The qualification badge worn after passing Army Ranger School, a hard threshold for infantry officers.
- OODA loop
- The observe-orient-decide-act decision cycle; getting inside an adversary's loop means seizing the initiative.
- distillation
- Training your own model on a stronger model's outputs to get comparable capability at lower cost.
How to listen
People working on US defense budgets, service personnel policy and procurement, and defense-tech investors who need to judge the window for US military transformation.
The farewell song after 01:03:40 can be skipped, and the AI battlefield section before it is also low on information density.