By Sam Roux
“Gentlemen do not read each other’s mail.”
- Henry L. Stimson (U.S. Secretary of State, 1929, upon learning of the American Black Chamber and ordering its closure.)
Before the first bullet was ever fired in anger, someone had already been reading the enemy's mail.
Signals intelligence didn't announce itself. It crept in during the Second Boer War around 1900, when British forces pressed Marconi Company wireless sets into service and discovered they could pluck Boer transmissions right out of the air. Within a decade, the Japanese and Russian navies were using radio direction finding equipment to locate enemy ships by their radio signatures alone. The electromagnetic spectrum had become a battlefield, and almost no one had noticed.
The First Spectrum Battlefield
The first people who truly noticed were cryptologists. Herbert O. Yardley, a poker player turned codebreaker, stood up the Black Chamber in 1919 and promptly cracked the diplomatic codes of a dozen nations, giving American negotiators at the Washington Naval Conference a significant, secret advantage. His successor in spirit if not directly in lineage, William F. Friedman, transformed cryptology from a gentlemanly art into a rigorous science. Friedman led the Army's Signal Intelligence Service and broke Japan's Purple diplomatic cipher before Pearl Harbor, a feat of analytical patience that proved what a disciplined mind, given enough time and the right tools, could accomplish. These men weren't just codebreakers. They were architects, laying the intellectual foundation that everything since has been built on.
Codebreakers Become System Builders
The Second World War compressed decades of progress into a few frantic years. At Bletchley Park, a sprawling English estate crowded with mathematicians, linguists, and chess champions, the British built Colossus, the world's first programmable electronic computer, specifically to crack the German Lorenz cipher. For the first time in history, machines outpaced human analysts. Colossus could process intercepted traffic at volumes that would have taken rooms full of people weeks to work through. The lesson was unmistakable: if you wanted to win the signals war, you had to automate. The National Security Agency absorbed that lesson and spent the Cold War scaling it, swapping vacuum tubes for digital processors and building listening architectures that could monitor entire communication corridors at once.
From Code Rooms to Computing Systems
By the late 1990s, a different problem had emerged. The Cold War's relatively contained adversaries had given way to a sprawling, borderless internet, and agencies were scrambling to keep up. The Federal Bureau of Investigation deployed Carnivore, a packet-sniffing system that could be dropped at an internet service provider to capture email and web traffic with a few clicks. Early cell-site simulators, the ancestors of today's Stingray, impersonated cell towers to hoover up subscriber data from everyone in range. These weren't elegant solutions, but they were earnest ones: attempts to drag the wiretap paradigm into the networked world before anyone fully understood what that world would become.
The Internet Changes the Collection Problem
Then the towers fell, and everything changed.
September 11, 2001 didn't reveal a failure of collection. The signals were there. What failed was the ability to connect them, to take disparate intercepts held by different agencies and see the pattern they formed together. The Patriot Act expanded legal authorities almost overnight. Fusion centers proliferated. But the more durable response came from an unlikely corner: a pair of entrepreneurs, Peter Thiel and Alex Karp, who founded Palantir Technologies in 2003 with seed money from the Central Intelligence Agency's venture arm, In-Q-Tel. Thiel had just spent years building PayPal's fraud-detection systems, teaching himself to find bad actors hiding inside vast rivers of transaction data. The intelligence problem, he recognized, was structurally the same. Palantir's Gotham platform wasn't just software, it was a new theory of how analysts should work, one where the machine did the correlation and the human did the judgment. It proved itself, and the government bought in.
Correlation Becomes the Real Weapon
The years since have been a story of relentless compression: more signals, faster processing, smaller hardware. Cold War satellite arrays gave way to commercial constellations like HawkEye 360, which can map radio-frequency emissions across the entire globe without a single ground station. Portable IMSI catchers have evolved into 5G-capable systems like L3Harris Technologies' Crossbow, precise enough to isolate a single handset in a crowded city. Machine learning now flags anomalies across metadata streams and telemetry that no human analyst could monitor manually. The volume of signals in the world has grown incomprehensibly large; what's changed is our ability to find meaning inside it faster than events unfold.
The Throughline
What runs through all of it, from Yardley's codebooks to Palantir's AI platforms, is a single recurring insight: the signal is only as valuable as your ability to understand it before it's too late. Every advance in this history was driven by someone who looked at the gap between collection and comprehension and decided to close it. The threats have moved into cyberspace and low-earth orbit. The next chapter will be written by whoever figures out how to follow them there.