The Eurobox thing started as a practical problem, and then somewhere around the third trip to the hardware store I started wondering about the version of this where you can't just drive to the hardware store. Where the network you're moving doesn't connect to the public internet at all, and if it goes down, someone might actually die.
That's the jump Daniel made. He's been thinking about military sysadmins — the people who deploy and maintain classified networks in places where the local ISP isn't just unhelpful, it literally doesn't exist.
Right. His prompt is basically: who are these people? Are they conscripts or career professionals? What does it actually mean to run a network like SIPRNET that's physically severed from the internet, where you can't patch remotely, can't call a vendor, and where redundancy isn't a nice-to-have — it's the difference between a working command post and a battalion that can't talk to anyone?
And he's asking across militaries, not just the US — he mentioned the IDF specifically, since he's in Israel and sees that system up close. But he's also curious about the global picture. Who do you call when SIPRNET is down and the answer can't be Comcast?
So we went looking. And what we found is a job that looks almost nothing like civilian IT.
The core of it — and this is the part that I think most people miss — is that a military network administrator isn't really an IT professional in the civilian sense. They're a soldier first, and their job is to make sure that commanders can command. The network is a weapon system. That's not a metaphor. The US Army literally classifies its tactical network as a weapon system.
Which changes everything about how you approach it. In civilian IT, if the network goes down, you lose money. In military IT, you lose... everything. Situational awareness, targeting data, medical evacuation coordination. The network is how the force talks to itself, and a force that can't talk is just a collection of people in trucks.
So let's lay out the landscape. In the US military, the big organizational players are DISA — the Defense Information Systems Agency — which runs the global backbone. They're the ones managing the big pipes, the long-haul circuits, the satellite constellations. Then each branch has its own piece. The Army has NETCOM, Network Enterprise Technology Command. The Navy and Marine Corps have their own structures, and they all feed into something called the Global Information Grid — the GIG — which is the military's version of the internet, except it's not one network. It's a federation of networks at different classification levels.
And the actual sysadmins on the ground — the person plugging in cables at a forward operating base — who is that?
In the US Army, that's typically a Signal Corps soldier. The specific job is MOS 25B, Information Technology Specialist. These are full-time, volunteer soldiers. Not conscripts. The US hasn't had a draft since the seventies. A 25B goes through a sixteen-week course at Fort Eisenhower — used to be Fort Gordon — at the Signal School. They learn basic networking, server administration, cryptographic devices, tactical radio systems. It's compressed, it's intense, and then they go to their units and learn the rest on the job.
Sixteen weeks. That's... not a lot of time to learn how to run a network where mistakes get people killed.
It's not. And that's the tension. The course gives them the fundamentals. The real expertise comes from doing it — from deployments, from exercises, from the senior NCOs who've been doing this for fifteen years and can diagnose a routing problem by the pattern of the blinking lights. The institutional knowledge lives in the non-commissioned officer corps.
Which brings us to Daniel's conscript question. The IDF works differently.
Completely. Israel has mandatory military service — roughly thirty-two months for men, twenty-four for women. And the IDF's C4I Corps, which handles communications and IT, does use conscripts in technical roles. But there's a structure to it. Conscripts typically handle tier-one helpdesk, basic network maintenance, cable runs, that kind of thing. The backbone — the encryption management, the core routing, the strategic networks — that's career officers and NCOs. People who've made this their profession.
So the IDF essentially runs a two-tier model. Conscripts do the work that can be taught quickly and where turnover is manageable. The critical stuff stays with the permanent force.
And they manage the churn in a few ways. Technical conscripts often serve longer than the baseline — three years or more, because the training investment is higher. They also lean heavily on simulation-based training. You can't just throw a conscript onto a live classified network and say "figure it out." So they've built extensive simulated environments where trainees can break things without breaking things.
The US avoids that churn problem entirely by having career soldiers do all of it. But it creates a different problem — recruitment and retention. A 25B who gets out after four years can walk into a six-figure civilian IT job. The military trains them, gives them a security clearance, and then watches them leave for Amazon Web Services.
And that clearance is worth a lot. A Secret or Top Secret clearance takes months and costs the government tens of thousands of dollars to process. A sysadmin with an active TS clearance is extremely employable. So the military is constantly training its own competition.
Let's talk about the network itself, because this is where it gets genuinely strange from a civilian perspective. SIPRNET.
SIPRNET — the Secret Internet Protocol Router Network. This is the military's classified network for Secret-level information. Above that, there's JWICS — the Joint Worldwide Intelligence Communications System — for Top Secret and SCI, Sensitive Compartmented Information. And here's the thing that civilians often get wrong: these networks are physically and logically air-gapped from the public internet. There is no VPN. There is no tunneling. The data does not cross.
What does "air-gapped" actually mean in practice? Paint me the picture.
It means the cables don't connect. A SIPRNET router and a NIPRNET router — that's the unclassified network — might sit in the same rack, but they are not connected to each other. They have separate switches, separate cabling, separate everything. If you need to move data from the unclassified side to the classified side, you don't transfer it over the network. You put it on a physical medium — a CD, a hard drive, a specialized transfer device — and you walk it over. That's called a sneakernet.
Which means you can't patch remotely. You can't pull updates from the public internet. You can't use cloud-based monitoring tools.
None of it. Every software update, every firmware patch, every antivirus definition has to be physically brought to the network. At a base in the US or Germany, that's annoying but manageable — someone drives a CD over from the security office. At a forward operating base in, say, the Sahel, where the supply chain is a helicopter that comes once a week? That's a fundamentally different problem.
And the encryption devices add another layer. What's a KG-175?
The KG-175 — it's called a TACLANE — is a network encryptor. It sits between your router and your transmission path and encrypts everything at line speed. The military uses these everywhere. There's also the KG-250, which is an inline network encryptor for IP networks. These devices are loaded with cryptographic keys that have to be managed, rotated, and — critically — zeroized if the device is compromised.
Zeroized. Define that.
Physically wiped. Instantly. There's a button or a command that erases all the cryptographic material so an enemy can't extract it. If a forward position is about to be overrun, the sysadmin's job includes making sure those encryptors are zeroized before anything else. The network goes dark, but the keys don't get captured.
That's not a ticket in Jira. That's... you're pulling a handle on a piece of hardware while people are shooting.
And that's the part that civilian IT metaphors can't reach. A misconfigured router in an office means lost productivity. A misconfigured router at a battalion command post means the artillery can't receive targeting data. The medical evacuation request doesn't go through. The commander loses situational awareness. In a combat zone, network downtime is measured in lives, not dollars.
So redundancy must be extreme.
It's taken to a level that would seem absurd in any other context. Military networks don't just have redundant power supplies and failover links. They have diverse routing paths that physically avoid the same terrain — so one artillery strike can't take out both the primary and the backup. They have backup satellite constellations. The US military operates something called BACN — the Battlefield Airborne Communications Node — which is essentially a flying network relay mounted on EQ-4B Global Hawk drones. If ground-based comms are blocked by terrain or jammed, the aircraft overhead can bridge the gap.
So the network has an airborne layer. That's... not something you spec out for a small office.
And it goes further. DISA's Global Operations Center at Fort Meade monitors over fifteen thousand circuits worldwide. They can see, in near real-time, the status of network links on every continent, including Antarctica. Each of those circuits might be fiber in Europe, satellite in the Middle East, undersea cable in the Pacific. The sysadmins managing these have to be fluent in multiple transport types and understand the diplomatic constraints — host nation agreements on spectrum use, legal restrictions on where certain types of traffic can be routed.
Host nation agreements. So the network isn't just a technical problem, it's a diplomatic one.
You can't just set up a satellite terminal in a country without that country's permission. You can't run fiber across a border without negotiations. The network has to work within a web of treaties, basing agreements, and spectrum allocations. A sysadmin in Germany is plugging into a commercial fiber backbone. A sysadmin in Djibouti is working through a satellite link that's shared with other coalition partners. The technical skills are the same, but the context is completely different.
Let's talk about deployment. Daniel mentioned moving his home network in Euroboxes — what does the military equivalent look like?
The Army has something called the CPCE — Command Post Computing Environment. It's a suite of servers, switches, routers, and satellite terminals that fits into transit cases. A small command post network — everything from the switch to the satellite modem — can fit in two or three hardened cases. A 25B can set it up in under an hour, connect to a satellite terminal like the AN/TSC-154 or the newer MUOS terminals, and have a classified network running in the middle of a field.
MUOS?
Mobile User Objective System. It's a next-generation narrowband satellite communications system that works with handheld radios and small terminals. Think of it as the military's version of a cell tower in space — but with encryption and anti-jam capabilities. A sysadmin at a small outpost can connect to MUOS and suddenly have voice and data connectivity that's hard to intercept or jam.
So the deployment model is: show up with cases, connect to whatever transport is available — satellite, line-of-sight radio, fiber if you're lucky — and stand up a network that's completely severed from the internet but connected to the military's global backbone.
And every piece of it is documented. Every cable, every config change, every cryptographic key load. The compliance burden is enormous. Which brings me to something I think most people don't appreciate: the training burden created by all of this.
Because you can't Google the answer.
You can't. In civilian IT, if you hit a problem you don't understand, you search for it. Stack Overflow, vendor documentation, Reddit threads. On a SIPRNET terminal, there's no internet. There's no Google. There's internal documentation, there's the senior NCO who's seen it before, and there's your own brain. That's it.
Which means the training has to be more comprehensive, and the institutional knowledge has to be preserved differently.
The Army's Signal School tries to front-load as much as possible — sixteen weeks of networking fundamentals, crypto, tactical systems. But the real learning happens in the unit. And mistakes are expensive. You can't break a SIPRNET router and file a ticket with Cisco. You have to fix it, or you have to explain to a commander why their network is down. The IDF handles this with longer service terms for technical conscripts and heavy simulation training. The US handles it with career NCOs and a lot of field exercises.
And then there are the contractors.
This is the part that complicates the picture. Not all military networking is done by uniformed personnel. DISA and the branches contract heavily with companies like General Dynamics, Northrop Grumman, L3Harris. In garrison — at a base in the US or a major overseas installation — contractors do a huge amount of the day-to-day network engineering and maintenance. They design the architectures, they manage the upgrades, they handle the long-term sustainment.
But they don't deploy.
They can, but not in the same way. A contractor can't be ordered into a combat zone. They can't be told to grab a rifle and defend the network. In the field, in a forward operating base, it's soldiers. This creates a two-tier system where the institutional knowledge — the deep expertise on how the network actually works — often lives with the contractor workforce, while the operational capability — the ability to deploy and fight the network — lives with the uniformed force.
That seems like a vulnerability.
It's a known tension. The military is aware of it. They've been trying to manage it for decades. The ideal is that the uniformed force has enough expertise to operate independently, with contractors providing surge capacity and specialized skills. The reality is sometimes messier — units deploy and discover that the network they're supposed to operate was designed by contractors who aren't coming with them, and the documentation is... let's say incomplete.
What about the future? Where is military networking heading?
Two big trends. One is software-defined networking — SDN — which separates the control plane from the data plane. In a civilian context, SDN lets you manage a network from a central controller, dynamically rerouting traffic, spinning up virtual networks. The military wants this because it would let them reconfigure networks on the fly — if a satellite link goes down, the controller automatically shifts traffic to a different path.
But the controller needs connectivity.
That's the catch. SDN assumes the controller can talk to the devices it's managing. In an air-gapped, disconnected environment, that's not guaranteed. The military is working on something called "disconnected operations" frameworks — essentially, SDN that can operate autonomously when the controller is unreachable and sync up when connectivity is restored. It's a military-grade version of what Daniel did with his Euroboxes — a network that can be packed up, moved, and stood up without phoning home.
The other trend?
Zero-trust architecture. The idea that you don't trust anything just because it's on your network — every device, every user, every connection has to authenticate continuously. The civilian world is moving toward this. The military wants it too, but the air gap makes it harder. A zero-trust architecture typically relies on cloud-based identity providers and continuous policy evaluation. If your network can't reach the cloud, you need to run all of that locally.
So the military has to build parallel versions of every modern IT innovation — versions that work without the internet.
And that's expensive, and slow, and creates a permanent gap between what's state-of-the-art in the commercial world and what's deployed in the field. The military is always catching up, but the security requirements mean they can't just adopt commercial solutions. They have to rebuild them.
Project Convergence — what's that?
It's the Army's big experimentation effort. Every year they run exercises that test new networking technologies in realistic combat scenarios — linking sensors to shooters, connecting allied forces, operating in environments where the enemy is jamming and hacking. The goal is to figure out what actually works before they spend billions on procurement. A lot of the disconnected operations and SDN work gets tested there.
And the IDF has something similar?
They call it the Digital Army program — it's an effort to integrate all their C4I systems onto a single common network. Right now, different branches use different systems that don't always talk to each other. The goal is a unified network where a commander can see everything — intelligence, logistics, fires, maneuver — on one screen. It's ambitious, and it runs into the same air-gap and security challenges.
You've been quiet for a bit, Hilbert.
I was wondering when —
Hilbert: The paperwork.
Go on.
Hilbert: Nobody talks about the paperwork. Eighteen months I spent on NMCI — Navy-Marine Corps Intranet — back when they were transitioning from all the legacy systems to a single managed network. I wasn't in uniform. Civilian contractor. But I sat next to active-duty sysadmins every day. And every time you touch a classified network, you log it. Every cable you plug in. Every config change. Every firmware update. It's all documented. I watched a Navy petty officer spend three hours filling out forms for a five-minute fix.
For a five-minute fix?
Hilbert: Router needed a new static route. Typing the command took thirty seconds. The paperwork — the authorization, the change request, the after-action log, the verification signatures — three hours. And if you make a mistake on the forms, that's a bigger problem than the original outage. The network works, but the paperwork is wrong? You're explaining yourself to people who don't care that the network works.
The compliance burden is its own operational factor.
Hilbert: NMCI was the worst I ever saw. You wanted to install a printer driver — a printer driver — you submitted a request to the configuration control board. They met once a week. If they approved it, you waited for the next scheduled maintenance window. Could be two weeks. The printer driver took longer to approve than the printer took to ship.
That's the tension in a nutshell, isn't it? Security and compliance on one side, agility on the other. The military defaults to security — for good reason — but the result is a network that's incredibly hard to change.
Hilbert: The joke at NMCI was that the network was so locked down, even the enemy couldn't break in. And neither could we. I remember a sysadmin — young guy, sharp — wanted to deploy a monitoring tool he'd built himself. Simple script, just polled switch ports and graphed utilization. He'd been running it on a test network for months, worked perfectly. The approval process took eleven months. By the time it was authorized, the switch he wanted to monitor had been decommissioned.
Eleven months.
Hilbert: Eleven months. He kept the letter. Had it framed.
That's... I mean, that's the cost of the air gap and the security model. Every change is a potential vulnerability. Every piece of software is a potential attack vector. So the system is designed to say no by default.
Hilbert: The thing is, I understood why. It wasn't bureaucracy for its own sake. NMCI had hundreds of thousands of users. One compromised printer driver on a classified network, and you've got a real problem. But the system wasn't built to distinguish between a printer driver and a custom monitoring script. Everything was treated the same. So the sharp people — the ones who could actually improve things — they spent their careers waiting for approvals.
Did any of them stay?
Hilbert: Some. The ones who believed in the mission. Most left for the private sector after their enlistment was up. The clearance was worth more than the job.
That's the retention problem we were talking about. The military trains them, clears them, gives them four years of experience on systems that don't exist in the civilian world — and then watches them walk.
Hilbert: The guy with the monitoring script? He's at Google now. Makes three times what I ever made. Still has the framed letter on his desk.
I think that letter is a better monument to military IT than any official history.
It captures something real. The job is fundamentally about constraints — air gaps, security clearances, compliance regimes, hostile environments — and the people who do it well are the ones who can work inside those constraints without losing their minds.
Or they leave, and the constraints get tighter because the expertise is gone.
That's the spiral. And it's not clear how to break it. The technology is getting more complex — SDN, zero-trust, AI-driven network management — but the human pipeline isn't getting faster. Sixteen weeks at Fort Eisenhower is still sixteen weeks. The IDF can stretch conscript service to three years, but that's the ceiling. You can't conscript someone for a decade.
The open question — and I think this is where Daniel's prompt leaves us — is whether the military can actually adopt the technologies that are transforming civilian networking, or whether the air gap and the security requirements will force them to build entirely parallel systems. A military internet that evolves on its own track, increasingly disconnected from how the rest of the world networks.
There's a version of the future where commercial networks become so agile, so software-defined, so AI-managed that the military can't keep up — not because they don't want to, but because every innovation assumes connectivity to a cloud that doesn't exist on the other side of the air gap. And the military ends up running networks that are secure but obsolete.
The other version is that disconnected operations become a solved problem — that the military figures out how to run SDN and zero-trust without phoning home, and those innovations eventually flow back to the civilian world. Remote infrastructure, disaster response, ships at sea, research stations in Antarctica. There are civilian use cases for networking without internet.
Daniel's Euroboxes, scaled up.
The same problem, different stakes.
The image I keep coming back to is the 25B in the dark. Somewhere, right now, there's a soldier troubleshooting a satellite link at a forward operating base. The generator is running low on fuel. The encryptor needs a new key. The paperwork isn't done. And they can't just reboot the router and hope.
Because hope isn't a redundancy protocol.
No. No, it's not.
If you've got a weird prompt — something about networks, logistics, or the hidden systems that make the world work — send it to prompts at my weird prompts dot com. We might just build an episode around it.
Thanks to Hilbert Flumingtop for producing.
This has been My Weird Prompts. We'll be back soon.
