What if the most important intelligence service in the world is one almost nobody can name?
That's not a rhetorical question, is it.
It's Daniel's. He wrote in this week with a whole thing, and it's a good one. He starts from the fact that we've spent a fair amount of time on espionage and intelligence on this show, and that it's stopped being a niche interest. Wars involving Iran are being fought in the information environment as much as on the ground, at least when it comes to holding public support.
Which is its own kind of intelligence problem.
Right. Then he goes back to an episode he still calls one of his favourites, the one tracing espionage as a discipline. His point there is that for most of history this wasn't a profession at all. It was diplomats reporting home, censorship of early mass communication, merchants quietly keeping notes on what moved through their ports. Informal. And then, running in parallel, militaries started building their own intelligence functions before anyone had the idea that these should be standing, independent bodies. So his suspicion is that military intelligence has a claim to being the older discipline, maybe even the predecessor.
The function, yes. The organisation, no. But let him finish.
He does finish, and it gets better. Today both tracks exist in parallel, and the military side is far less famous. Unit 8200 sits inside Aman, Israeli military intelligence, and Mossad and Shin Bet are household names. Same picture in the States. So his actual question is this. When these are separate bodies inside one intelligence community, how do they actually work together? There's presumably cross-pollination, there's presumably an exchange of information, but there also has to be a delineation of scope so they don't step on each other or duplicate resources. How does that work in practice. And he flags the obvious problem himself, that the public record here is thin, so we piece together what we can from what's known, former service members' accounts, that kind of thing.
He's right that it's thin. That's not a gap in his reading. It's the actual state of the material.
So let's start with the historical claim. Is military intelligence really the older discipline?
It's older as a function and younger as a bureaucracy, and both halves of that matter. Collection is as old as war. For most of recorded history there was no intelligence branch because there was no branch of anything. Each commander was his own intelligence officer. You used cavalry for reconnaissance, you used scouts, you used spies, and when the campaign ended all of it dissolved. Nobody was funding an intelligence department in peacetime because nobody could see the point.
And the diplomatic piece Daniel remembers is real.
Very real. Before there was any dedicated military intelligence, the British Foreign Office and the Diplomatic Service carried primary responsibility for collecting information about a country's war-making capacity. Ambassadors reported back. Consuls reported back. A diplomat in a capital city was, functionally, a collection asset with a nice house.
Which is a strange thing to sit with. The embassy was the intelligence service.
The embassy was the intelligence service, and the tradecraft was correspondence. Then the nineteenth century arrives and the advanced European powers start professionalising it. But even then it's not smooth. The British Army ran what you could generously call a ramshackle intelligence service, and the institutional attitude toward a permanent intelligence department was often openly hostile. A standing intelligence shop is an admission that you might not know something, and militaries in that era did not enjoy admitting that.
The Americans did the same thing on a cycle.
Significant advances during each war, then relative disinterest once it's over. Every war builds an intelligence capability, every peace lets it atrophy, and then the next war rebuilds it from scratch and rediscovers the lessons. That pattern runs from the Civil War well into the twentieth century.
So the correction to Daniel's premise is narrower than it sounds.
It's a correction of category, not of era. Military intelligence is older as a practice. It is not older as a standing organisation. The dedicated bodies are all post-war creations. CIA, nineteen forty-seven. Aman formalised in the early fifties. DIA, nineteen sixty-one. Every one of them is a twentieth-century institution built in response to a specific failure or a specific war.
Which raises the question Daniel actually cares about. If the function is older, why is the civilian side the famous one?
Because fame tracks narrative, not capability, and those two things have almost nothing to do with each other. Start with Israel, because the structural reason is clearest there. Aman was given a range of activities that in the West would be considered outside military intelligence entirely. Responsibility for intelligence research on political matters, for instance. That happened because early Israel leaned on the IDF as the anchor institution for national tasks. When you're building a state and you have one institution that actually functions, you give it everything.
So the military track absorbed work that would otherwise have gone civilian.
It absorbed it by default. And the result is that Unit 8200, which sits inside that military track, is the largest unit in the IDF. At least five thousand on active duty, and by some estimates considerably more. That's roughly eighty percent of Aman's manpower in a single unit. To put the scale in perspective, its budget in nineteen fifty was fifteen thousand dollars plus a hundred and ten thousand for initial electronic hardware. Call it about one and a quarter million in today's money. That is the seed capital for the largest signals intelligence organisation in the region.
And almost nobody outside the country could name it until the startup books came along.
That's the paradox in one line. Unit 8200 supplies on the order of ninety percent of Israel's intelligence material. Yair Cohen, who spent thirty-three years in the unit and commanded it from two thousand one to two thousand five, put it flatly. Ninety percent of the intelligence material in Israel comes from 8200, and there isn't a major operation, from Mossad or any other service, that 8200 is not involved in. That's the commander's own estimate, so treat it as an estimate, but the direction is not in dispute.
And yet Mossad is the one with the movies.
Mossad is the one with the movies because Mossad does covert action, and covert action photographs well. Cinema can show you a passport, a hotel room, a poisoned telephone. It cannot show you a signals analyst in a windowless room pulling structure out of noise. So the public imagination fills in the gap with the thing it can picture.
The CIA's founding doctrine is almost funny in this light.
It's a lovely document. The CIA should not be merely another agency duplicating and rivalling the existing agencies of State, Army, Navy, and Air Force. It should not be a competitor of these agencies, but a contributor to them. That's the founding intent, written down, and it is a description of an aspiration rather than a description of reality. Everyone who has ever worked inside a government has read a sentence like that and smiled.
So the fame question has an answer, and it isn't capability.
It's narrative. The DIA and the CIA are the same story. One of them has a film franchise, the other has a budget line.
So that's the historical arc and the fame paradox. Now let's get into how these two tracks actually talk to each other, because that's the part Daniel actually asked about.
Israel gives us the cleanest version of the problem, and it's cleanest precisely because it's so messy. You have a triad. Aman for military intelligence, Mossad for foreign, Shin Bet for domestic. The division of labour is usually established on a geographical basis. That's the theory. In practice there are interfacing and overlapping segments, and the overlap is often rather wide.
Wide is the polite word. The same literature says coordination has suffered in the past from fundamental shortcomings, and that the organisations repressed the necessity for mutual sharing of intelligence information and for synchronising some activities. That's a remarkable thing to find in a description of a national intelligence community. It's an admission that the design has a known failure mode and that the failure pattern is structural, not personal.
And the fix is a document with a wonderful name.
The Magna Carta. The three service heads negotiate their jurisdictional boundaries, and the Knesset Intelligence Subcommittee oversees the arrangement. There's no standalone public text of it. Its existence is documented secondhand. Which tells you something in itself. The charter that governs who is allowed to do what in Israeli intelligence is not something you can go and read.
A constitution for turf.
A constitution for turf, negotiated continuously, and not published.
So the split isn't solved. It's managed.
Managed is exactly right. And the cross-pollination is real, which is what makes the management necessary. The same officer I quoted, Cohen, describes a unit that shares data internally across the IDF and externally with Mossad. There isn't a major operation that 8200 isn't involved in. So you have deep integration on the operational side and a jurisdictional charter on the political side, and the two things are running at the same time.
And the obvious reform was proposed and then quietly not done.
Two thousand four. A Knesset committee, in the aftermath of the Iraq WMD failure, recommended restructuring the whole community into three or four independent services, with a separate national signals intelligence service split off from military intelligence. It was apparently not implemented. That's the single most revealing fact in the Israeli section. The reform was identified, recommended by the legislature, and absorbed by institutional inertia.
Because the military track has the manpower.
The military track has the manpower, the budget, the technology, and the recruitment pipeline. Splitting off a civilian SIGINT agency means taking eighty percent of Aman's people and putting them somewhere else. Nobody who has ever watched an institution defend its own existence would bet on that happening.
The American case gives us the same tension with better paperwork.
Much better paperwork, and the paperwork is the interesting part. The intelligence community has eighteen elements. Nine of them sit inside the Department of Defense. That's DIA, NGA, NRO, NSA, plus the service intelligence organisations, the Army's G2, the Navy's N2, the Marine Corps intelligence activity, the Air Force A2, and so on. Half the community is inside the Pentagon.
And the money enforces the split.
Two funding streams. The National Intelligence Program and the Military Intelligence Program. The NIP is the national pot. The MIP is the defense pot. Where a given dollar sits determines who oversees the programme, who can move it, and who gets to say what it's for. That's the mechanism. It's not a policy statement about cooperation. It's an accounting boundary, and accounting boundaries are the ones that actually hold.
So how does anyone coordinate across it?
DIA is the coordination layer. Created in nineteen sixty-one specifically to pull the services' intelligence streams together. And note what its mission actually is. DIA's principal mission is not collection. It's the provision of substantive intelligence support to the Secretary of Defense, the Chairman of the Joint Chiefs, and the commanders of the unified commands. It's an all-source analysis shop serving the warfighter, not a collection agency.
Which is a different animal from the CIA.
A different animal entirely. And then you get the structural answer to Daniel's actual question, which is the dual-hatting. The Under Secretary of Defense for Intelligence and Security is simultaneously the Director of Defense Intelligence. One person, two hats, reporting to the Secretary of Defense and to the Director of National Intelligence. That's a two thousand seven memorandum of agreement. That's the bridge. When you want to know how the two tracks talk to each other at the top, the answer is that in one specific job, they're the same person.
That's either elegant or alarming.
It's both, and it's deliberate. It means there's a human being whose entire professional life is the seam between the two tracks.
And the redundancy at the working level is openly acknowledged.
Openly acknowledged and partly defended. The CIA has military analysts and the DIA has political analysts. Those redundancies can offer value in terms of competitive analysis and interagency cooperation. That's the argument for duplication. Two shops looking at the same problem from different institutional angles, and if they disagree, that disagreement is itself information.
And the argument against.
Duplication of effort wastes resources, leads to clashes of jurisdiction, and may result in a sense of turf that precludes information sharing. Same source, same paragraph, opposite conclusion. That's the honest state of the question. Redundancy is a feature and a bug, and nobody has figured out which one dominates.
The liaison numbers tell you which way it's drifted.
They tell you a lot. At one point there were only a handful of military liaison officers working with the CIA. Today there are hundreds of uniformed personnel, active, guard, and reserve, serving in the building. And the agency has representatives at dozens of military commands and professional military schools. That's not coordination at the top. That's coordination embedded at the working level, in the same corridors, on the same problems.
So the answer to how they avoid stepping on each other's toes is that they mostly don't, they just have enough people in the room to notice when they do.
That's a fair summary, and it's less cynical than it sounds. Most deconfliction in any large organisation is social before it's procedural. You avoid stepping on someone's toes because you know whose toes they are.
Then there's the international layer, which is where it gets strange.
The NSA and Unit 8200 have a deep relationship. The Snowden documents describe it as far-reaching, technical and analytic, covering access, intercept, targeting, language, analysis, and reporting. And a two thousand nine memorandum of understanding had the NSA providing Unit 8200 with raw American signals intelligence. Unevaluated and unminimised transcripts, gists, facsimiles, telex, voice and digital network intelligence metadata and content.
Raw. Unminimised. That's not a summary, that's the feed.
That's the feed. That's the American signals apparatus handing the raw product to a foreign military intelligence unit, which is about as deep as an intelligence relationship gets between two countries.
And then this June, the Pentagon raised the threat level of Israeli spying on the United States to its highest category.
Same year. Both things true at once. Deep SIGINT sharing and a raised counterintelligence threat assessment, simultaneously. And the honest reading is that this is not a contradiction. It's the normal condition. Allies spy on allies. The relationship is deep because the interests overlap, and the counterintelligence posture is elevated because the interests don't overlap completely. Both of those are permanent facts, and the system is built to hold them at the same time.
Which is the least comforting thing you've said today.
It's the most accurate thing I've said today, which is usually the same category.
That's the structural picture. But what does it look like when you're the one actually handing the material over?
Hilbert: So the folder was called transit, and it had a subfolder for each port, and nobody had ever told us what the ports had in common.
Go on.
Hilbert: Eighteen months I did that. Contract translator for a firm that had a defense contract. We read transcripts. Shipping manifests, port authority chatter, the ordinary stuff that moves through a harbour on a Tuesday. You get good at it. You learn which discrepancies matter.
And the hand-off.
Hilbert: That's the part I wanted to say something about. You two have been describing cross-pollination like it's a system. It isn't. The military side collected the raw material and the civilian side interpreted it, and the hand-off was a shared drive folder. No context. No note saying what we'd been told to look for. You'd spend a week on a manifest and never find out whether anyone read it.
The analysis happened on the other side of a folder.
Hilbert: The analysis happened on the other side of a folder, and the folder didn't know anything. There was a standing joke in that office. The military collects the hay, and the civilians look for the needle, but nobody tells the hay what a needle looks like.
That's the deconfliction problem in one sentence.
Hilbert: It's not a deconfliction problem. It's a context problem. Nobody was fighting over jurisdiction. Nobody had time. The material just arrived without the thing that would have made it useful, and everybody assumed somebody else had sent it.
The Magna Carta, the charter you'd expect to cover exactly this.
Hilbert: Sounds like a document written by people who never had to hand anything over. You can agree on boundaries all day. Boundaries don't tell the hay what a needle looks like.
The seam we've been describing at the top, the dual-hatting, the funding streams, the memoranda. None of that reaches the folder.
Hilbert: None of it reaches the folder. The folder is where it actually happens and the folder is nobody's job. Anyway. My sister-in-law's expecting me, and she's been expecting me since Tuesday, so.
Right. So the seam is real, and the seam is where the work happens, and the seam is the one place nobody owns.
The structural answer to Daniel's question is that the two tracks coordinate through dual-hatted officials, shared funding oversight, embedded liaison officers, and negotiated charters. The practical answer is that a lot of the actual transfer happens in whatever ad hoc mechanism the people at the desk invent, and the quality of that mechanism determines whether the intelligence is useful or just present.
Which reframes the whole thing. The question isn't whether the division of labour exists. It clearly does. The question is whether the hand-off across it works.
The evidence suggests it works well at the top and unevenly at the bottom, which is roughly true of every large organisation ever built.
If you take one thing from this, take the correction to the premise. Military intelligence is older as a function and younger as a bureaucracy. Every dedicated body in this conversation is a twentieth-century creation.
The fame gap isn't a capability gap. It's a narrative gap. Unit 8200 supplies the overwhelming majority of Israel's intelligence and most people couldn't name it, because covert action photographs and signals analysis doesn't.
One thing to leave you with. If the civil and military split is a designed friction rather than a solved problem, then the next major intelligence failure has an obvious question attached to it. When the hand-off fails, who gets blamed? The agency that collected, or the agency that was supposed to understand it?
The ally-spying paradox isn't going away. Deep signals sharing and a raised threat assessment can both be true in the same year, and the system is built to hold both. That's not a temporary condition. That's the architecture.
Thanks as ever to our producer, Hilbert Flumingtop.
This has been My Weird Prompts, the human-AI collaboration podcast. You can find the full episode and the sources at my weird prompts dot com.
If you enjoyed it, a review helps more than you'd think. We'll be back soon.
