Pentagon pizza traffic surged at 2am. Four hours later, Maduro was captured.
How Google's location data turned a Virginia pizzeria into an accidental intelligence indicator
Pizzato Pizza sits two miles from the Pentagon in Arlington, Virginia. It stays open late. At 2:04am on Saturday, according to Google Maps, the restaurant was "busier than usual." By dawn, Venezuelan President Nicolás Maduro was in American custody, extracted from Caracas by Delta Force in an operation involving 150 aircraft.
A Twitter account with a quarter-million followers had noticed the pizza traffic and posted it. The account exists for exactly this purpose.
The Pentagon Pizza Report scrapes Google's "Popular Times" data for restaurants near American government buildings and alerts its followers when something looks unusual. The premise sounds absurd until you learn that the phenomenon it tracks has been documented for over forty years, that it has preceded nearly every major American military operation since Grenada, and that Soviet intelligence once monitored the same signal. They called it Pizzint.
The franchise owner who noticed first
Frank Meeks owned forty-three Domino's locations around Washington during the Reagan years. He kept meticulous delivery logs. And he noticed that on certain nights—always unexpectedly, always late—orders to government buildings spiked.
The night before American forces invaded Grenada in October 1983, Pentagon deliveries doubled. The pattern repeated before Panama in December 1989. But August 1, 1990 made Meeks briefly famous: the CIA ordered twenty-one pizzas that night, a record for Langley. The next morning, Iraq invaded Kuwait.
When Desert Storm began five months later, Meeks told the Los Angeles Times that late-night Pentagon deliveries had surged from three to one hundred and one. Wolf Blitzer, then CNN's Pentagon correspondent, offered advice that reads differently three decades later: "Bottom line for journalists: Always monitor the pizzas."
The Washington Post documented another spike in December 1998, when Clinton faced impeachment while bombing Iraq. White House orders ran thirty-two per cent above normal. Extra cheese was in high demand. The pizza index had entered Washington folklore—not quite serious, not quite forgotten.
The algorithm that remembers everything
Meeks needed delivery slips. The modern version needs only Google.
When you open Maps and check a business, Google shows you when it's typically busy—a bar chart compiled from the aggregated location data of every Android phone and opted-in iPhone that has ever passed through. The Pentagon Pizza Report monitors these charts for establishments near government buildings. When the graphs spike at odd hours, the account posts.
The methodology has obvious flaws. Google measures bodies in restaurants, not delivery orders dispatched to secure facilities. A Friday surge might mean a Nationals game ended early. The account tracks Freddie's Beach Bar, the closest gay bar to the Pentagon, for the inverse signal: when it empties on a night it should be packed, that suggests Pentagon employees are stuck at work. Clever, but crude.
And yet the correlations persist. April 2024: unusual traffic at a Papa John's before Iran launched drones at Israel. June 2025: a "huge surge" at Pentagon-area pizzerias at 6:59pm, hours before Israel struck Iranian targets. Saturday morning: Pizzato Pizza busier than usual at 2am, Maduro captured by dawn.
The critics are not wrong
Marcel Plichta spent years as an analyst at the Department of Defence. He has written extensively about why the pizza index fails under scrutiny.
His objections cut deep. Which restaurants count? The accounts monitoring the index never agree. The core assumption—that Pentagon staff respond to crises by driving past the building's own food court to eat at nearby restaurants—strains belief. Google's data captures foot traffic, not phone calls to delivery drivers. If staff are ordering in, the signal wouldn't register at all.
Zenobia Homan, a senior researcher at King's College London, raises the statistical problem that haunts all pattern-matching: confirmation bias. Spikes that precede events get publicised. Spikes that precede nothing get forgotten. "I'm not saying the theory is wrong," she told reporters, "but I want to see way more data. When else do spikes occur? How often do they have absolutely nothing to do with geopolitics?"
The Pentagon itself denies any connection.
These criticisms land. The pizza index generates false positives constantly. It cannot distinguish a Venezuela planning session from a budget deadline. As prediction, it is worthless.
But prediction may not be the point.
What the jogging paths revealed
In January 2018, an Australian university student named Nathan Ruser was browsing Strava's Global Heatmap—a visualisation of everywhere the fitness app's users had ever exercised. In the deserts of Syria and Afghanistan, jogging paths glowed in strange rectangular patterns. Ruser realised he was looking at the perimeters of secret American military bases, traced by soldiers who had tracked their runs.
The internal road layouts were visible. Patrol routes leading in and out. Guard posts. All rendered in the accumulated GPS data of personnel who wanted to log their workouts.
The discovery cascaded. In 2020, investigators identified fourteen members of Britain's SAS through Strava profiles at their supposedly secret Hereford base. In 2024, French journalists found that bodyguards protecting Emmanuel Macron had been tracking their jogs. By analysing their routes, researchers could identify which hotels the French president would stay at before his arrival was announced.
Hedge funds noticed similar patterns in corporate aviation. Flight-tracking services now sell data on executive jet movements to investors hunting merger rumours. When Occidental Petroleum's Gulfstream appeared at an Omaha airport in April 2019, analysts correctly guessed executives were meeting Warren Buffett. Two days later, Buffett announced a ten-billion-dollar investment.
The vulnerability is identical across domains: human beings generate data through ordinary activities, and that data aggregates into intelligence. Classify your documents, encrypt your communications, compartmentalise your operations. You cannot prevent your staff from wanting dinner, going for a run, or flying to meetings.
The confession no one asked for
During the Cold War, monitoring Pentagon pizza deliveries required assets in Washington. Human sources. Careful observation. Today, a single person with Google Maps can replicate that work from anywhere on earth.
This is what has actually changed. The signal Meeks noticed in 1983 still exists. But detecting it no longer requires resources. It requires only curiosity and a free app.
The Pentagon could theoretically neutralise this specific vulnerability. Install better kitchens. Order from randomised locations. Generate noise. But other signals would remain. Uber knows when ride requests spike near government buildings. Power companies track consumption. Even waste management data carries information—full bins mean overtime.
The surveillance economy produces an endless exhaust of behavioural data. Some of that exhaust will always reveal what institutions prefer to conceal. The game has become whack-a-mole, and the moles multiply faster than anyone can swing.
When pizza traffic spiked early Saturday morning, it did not predict that Maduro would be captured, or that 150 aircraft would launch from twenty bases across the Western Hemisphere, or that American forces would be operating inside a sovereign nation's capital within hours. It simply registered that something was happening. The Pentagon was awake and hungry at 2am. That information, trivial in isolation, became meaningful the moment the bombs began falling.
Frank Meeks noticed a pattern in his delivery logs in 1983. Four decades later, his observation has evolved into a crowdsourced surveillance network the Pentagon cannot fully suppress. The pizza index may be methodologically flawed, statistically noisy, and useless for prediction.
But it demonstrates something that matters: in a world saturated with data, our ordinary behaviours constitute an involuntary confession. The patterns are there for anyone who thinks to look. The data is free. The lesson is not.