On May 1, 1960, American pilot Francis Gary Powers took off from Peshawar, Pakistan, bound for Bodø Air Base on the northern tip of Norway. But this was no ordinary flight. Along the way, he would soar over the Soviet Union, secretly photographing enemy missile sites and plutonium processing centers. Powers’ stratospheric U-2 spy plane was specially designed by Lockheed’s top engineers and was believed to fly far beyond the reach of enemy surface-to-air missiles. These experts were mistaken. Catastrophically, Powers was shot down over Sverdlovsk.

Such a brazen incursion into Soviet airspace threatened to turn the Cold War hot. The United States had attempted to prepare for such a disaster, crafting a cover story about a wayward weather plane and instructing Powers to commit suicide if he was shot down. However, this plan was for naught: upon crash landing, the American pilot did not commit suicide and was instead immediately detained by the KGB. Unaware of Powers’ capture and believing he had followed protocol, NASA issued the weather plane story. It was not until days later that Premier Khrushchev announced that Powers had actually been taken alive: “I deliberately did not say that the pilot was alive and well … and now just look how many silly things the Americans have said.”

The U-2 incident dealt a severe blow to President Dwight Eisenhower’s administration, tarnishing America’s reputation on the global stage. What’s more, this event pushed Soviet-American relations to the brink, sparking a diplomatic mess that boiled for two years until Gary Powers was finally repatriated. But although the immediate crisis subsided, the incident left deep geopolitical scars.

The United States found itself in a precarious position. While the need for surveillance remained critical, the risk of another pilot capture was untenable. The solution that emerged was elegantly straightforward: eliminate the need for a human pilot altogether. This realization laid the groundwork for a new era of unmanned systems, reshaping global conflicts and modern military strategy.

The U-2 incident was not surprising to everyone. A year earlier, U.S. Air Force Colonel Lloyd Ryan noticed that Soviet S-75 rockets were rapidly approaching the technical performance needed to reach a U-2 spy plane at cruising altitude. Before satellites, intelligence officials relied heavily on manned aircraft like the U-2 for crucial photographic reconnaissance. To Colonel Ryan, this was a recipe for disaster; the capture of a downed U-2 pilot was inevitable. He preemptively flew to Pasadena to meet with Hycon, the developers of the U-2’s high-altitude cameras. They suggested, for the first time, to attach them to a drone.

Since World War II, San Diego-based Ryan Aeronautical (no relation to Colonel Ryan) had been producing radio-controlled drones for use in weapons testing. Interestingly, it was actually at one of their Radioplane Factories that Marilyn Monroe was “discovered,” featuring her assembling a Ryan Aeronautical drone. Called “Firebees,” these drones were remote-controlled and equipped with a turbojet engine capable of nearly Mach 1.

Alongside Colonel Ryan and Hycon, Ryan Aeronautical developed a new version of the Firebee with a high-altitude camera and a more radar-absorbent airframe, and pitched to the Pentagon as “the least detectable existing aircraft,” as an unmanned photographic reconnaissance system that would allow the Air Force to gather intelligence without any possibility of a captured pilot. Soon, the system—dubbed the “Lightning Bug”—had the green light.

However, the Lightning Bug project would face internal resistance. Despite initial agreement, Assistant Secretary Dr. Courtland Perkins abruptly ordered Colonel Ryan to halt the program, deeming it as wasteful. Edward Teller, of hydrogen bomb fame, later explained that “the Air Force is built around fliers—and unmanned vehicles put fliers out of business.” Some in the Air Force felt threatened by drones, jeopardizing the Lightning Bug’s future.

It was at this time that Colonel Ryan was tragically and suddenly provided evidence for his program’s necessity. Just as he had anticipated, a U-2 pilot, Francis Gary Powers, was shot down over the Soviet Union. Sentiment quickly shifted. “Within a fortnight [of Perkins’ attempt to cancel the program] Francis Gary Powers was in the news,” Ryan recalled. “To us it was sort of an ‘I-told-you-so deal’ because we had started the drone studies in anticipation of just such an event.” Just weeks later, the Air Force officially moved forward with the Lightning Bug program.

The First Drones Flew Over Vietnam

In August 1964, just four years into its development, the Lightning Bug would be put to the test. The USS Maddox had clashed with three North Vietnamese torpedo boats in the Gulf of Tonkin, sparking a flash deployment order for Ryan Aeronautical. The Pentagon knew reconnaissance would be critical in Vietnam, but thick fog that hung low over the jungle made high-altitude surveillance impossible, and low-flying planes extremely susceptible to anti-aircraft fire. Unmanned systems were the only solution. It was up to the Lightning Bug.

Over the next fifteen years of the war, Ryan Aeronautical would iterate on over twenty aircraft designs over the skies of Southeast Asia in the world’s first combat deployment of drone technology. In total, they flew over 1,000 aircraft across 3,435 missions—nearly half of all missions flown during the conflict. Critically, these drones carried cameras and sensors instead of munitions, focusing on intelligence gathering instead of destruction. Lightning Bugs would conduct targeting campaigns, damage assessments, locate prisoners of war, and gather crucial enemy signal intelligence. For nearly four years after President Lyndon B. Johnson’s halt to bombing campaigns in October 1968, Ryan Aeronautical’s drones were the only aircraft flying over North Vietnam.

But this was not an overnight success. When it comes to military hardware, the relationship between cost and survivability is paramount. At $200,000 per unit, the Lightning Bug demanded a high “life expectancy” to justify the program. This was not the case, at first. The drone was initially air-launched from a “mother-plane,” flew a pre-programmed route, then deployed a parachute over a landing zone for recovery. However, these parachutes were routinely blown off course, forcing helicopter crews into hostile jungles for drone collection. This was unacceptable and risked the program’s future. To address this, a cargo plane was employed using a hook to snag the drone’s parachute and carry it to a safe landing location. With this mid-air retrieval system, or MARS, the Lightning Bug’s survivability increased from three missions to nine.

With survivability came expanded use cases. In 1966, the Air Force sought a solution to the Soviet S-75 surface-to-air missiles used by the North Vietnamese (the same system that brought down Gary Powers). With no defense, the Air Force’s only option was to fly at lower altitudes invisible to the S-75 platform’s radar, yet in doing so were now susceptible to standard anti-aircraft fire. Worse, directly targeting these missile sites risked killing Soviet personnel that often assisted in their operation, escalating conflict. Over the course of the war, the U.S. Air Force would lose 110 fighters to such systems. This was a massive problem. Ryan Aeronautical was called into action, and Operation Buffalo Hunter began.

Ryan Aeronautical’s solution was the Model 147D, a Lightning Bug variant specially designed for electronic warfare. When a Soviet S-75 reached its target, a small radar proximity fuse would detect the enemy aircraft and detonate. This was the key to this system’s success; there was no need for the ground radar to precisely track the target, it only needed to get close enough. The Model 147D turned this strength into a weakness.

Notably, the drone’s radar signature was designed to appear much larger than it was, justifying the use of a missile. As it approached the aircraft, the Model 147D would collect the fuse signal, instantly process it, then downlink this data back to a ground receiver stationed far away. Effectively, the drone would fly around until it was targeted, then engage in this suicide mission to collect the valuable data necessary to develop missile warning systems for fighter jets. The Assistant Secretary of Defense for Research and Engineering, Dr. Eugene Fubini—who originally advocated shutting the program down—called the Model 147D “the most significant contribution to electronic reconnaissance in the last twenty years.”

The Lost Decade and Recovery

Though Buffalo Hunter proved successful, in 1979, the Air Force deactivated the program. Despite making critical contributions to reconnaissance, military officials believed Lightning Bugs to be overpriced and the technology still relatively premature. In addition to the annual $100 million in procurement and $250 million in operational costs, each mission required protection by F-4 Phantoms and courier jets for film transport, leading to astronomical hidden costs. Consequently, the Lightning Bug was abandoned.

Nevertheless, the concept of unmanned systems would persist. But, after Vietnam, drone programs were routinely delayed and often obsolete against Soviet equipment by the time they were ready for deployment. In 1962, Lockheed’s Skunk Works was contracted to build an unmanned variant of the SR-71 Blackbird, but, after numerous accidents, the program was canceled. In June 1966, Ryan Aeronautical would build a more sophisticated high-altitude surveillance drone, but ran three years over its projected timeline, making it quickly obsolete against advancing Soviet technology. In 1971, the D-21 was raised from the ashes and flew four missions over China—all of which ended in failures and the ultimate cancellation of the program. Boeing, too, would build an unmanned system that would never fly. The American defense industry understood the advantages of drones, but were unable to effectively deploy them for nearly a decade. These were dark days for drones.

Things began to change in 1988. While the defense “primes” struggled, Abraham Karem—an Israeli engineer that had built target drones for the Yom Kippur War—diligently worked in his garage in Los Angeles. With DARPA funding, Karem and his team aimed to build a long-duration “unmanned aerial vehicle,” or UAV, soon constructing and flying the “Albatross” with fiberglass and carbon epoxy. In 1985, DARPA contracted Karem to develop a larger version, the Amber, which flew reliably for 38 hours at 5,500 feet. However, in 1988, Congress cut UAV funding and canceled the Amber, a move Karem later said was to protect incumbents. Karem’s company was then acquired by General Atomics, where he shifted focus to exporting the runway-launched Gnat 750, a modified Amber.

By then, the U.S. had largely abandoned reconnaissance drones. However, the introduction of the Soviet SS-25 road-mobile intercontinental ballistic missile renewed interest in long-duration high-altitude surveillance. In the 1990s, civil unrest in Yugoslavia escalated into war, capturing U.S. attention. It wouldn’t be Congress or the Department of Defense that changed focus. Karem was approached by the CIA, and the Gnat 750 was deployed over Bosnia within months. By July 1994, Karem had upgraded the design with satellite links, rebranding it as the Predator.

Buffalo Hunter had growing pains. But the flagship deployment of the Predator was a mature and permanent success. From 1994 to 1997, the RQ-1 Predator was used in Bosnia and Herzegovina to support NATO operations in the Bosnian War. Contemporary reports on the program are glowing. Congress drastically increased its investment, dedicating a full 30% of DARPA investment in 1997 for unmanned reconnaissance, with 46% of that budget going towards long-duration surveillance systems, and 29% to the Predator program alone.

Congress declared that UAVs had finally moved “from words to deeds.” A report by the Office of the Undersecretary of Defense credits the first Predator deployment in Gjader, Albania, from July to November of 1995, with influencing the overall outcome of the conflict. By using persistent video surveillance of roads for treaty verification, NATO found that opposing forces had not removed field weapons from a contested area as they had promised to do, and was able to confirm a bombing campaign that led to the Dayton Peace Accords of December 1995. Critically, the Predator also provided evidence of civilian massacres by monitoring fields that became mass grave sites outside Sarajevo in 1996.

The Predator’s successful deployment in Bosnia marked the end of decades of clandestine drone reconnaissance without which this generation of the technology couldn’t have matured and the first moment that drone technology was integrated into joint operations in the Department of Defense, with the full and lasting support of Congress.

The Commercial Roots of the Modern Drone

A decade ago, drones were still the exclusive domain of great power militaries. Behemoth fifty-foot-wingspan systems like Karem’s, costing hundreds of thousands to millions of dollars. But in 2010, the first consumer drone came out. Suddenly, anyone with an iPhone and $299.99 could easily fly a simple camera drone. The first consumer drone, from French electronics manufacturer Parrot, was essentially a toy, little different than a model airplane. But as hardware reliability and camera quality improved, off-the-shelf drones evolved from toys, to “selfie cameras,” to simple commercial tools for things like roof inspection.

By the end of the decade, every major outdoor industry—agriculture, mining, public safety, construction, energy, and utilities—was deploying off-the-shelf drones for more sophisticated commercial uses. Today, drones flying LiDAR sensors conduct land surveys, and drones equipped with optical gas imaging sensors detect methane leaks on offshore oil rigs. If you happen to be robbed in Santa Monica, where this author lives, a $13,000 drone will take off from the roof of the police department, fly within three minutes to the scene of the crime, and help responding officers on the ground chase down the fleeing perpetrator.

But the significance of the off-the-shelf drone was more than commercial. In the 1970s, when computers went from six-figure mammoths whose servers took up entire rooms of universities to $700 desktops, the invention of “personal computing” was also the invention of the “computer nerd.” Drones were no different. The 2013 release of the Chinese DJI Phantom, the first truly airworthy consumer camera drone, created a Cambrian explosion of drone tinkerers, drone racing pilots, and hobbyists. An incredibly generative ecosystem of skilled drone pilots, cheap drone component suppliers, and open-source drone software platforms quickly emerged to serve them.

Just over twelve months after the DJI Phantom hit the market, in February 2014, Russian President Vladimir Putin annexed the Crimean Peninsula. Ukraine still relied on Soviet-era drones, like the Tupolev Tu-143, for aerial reconnaissance (think Lightning Bugs—they had to be launched from another vehicle, shot stills on physical film, and deployed a parachute to land). These were woefully inadequate—but foreign-made military drones were prohibitively expensive at over $500,000 in unit cost, heavily restricted to import, and could not be purchased from governments wary of provoking Russia.

The Ukrainian military had few resources. But, with four aerospace universities which graduated 10,000 IT specialists per year, Ukraine was rich in technical talent, which, by this time, included drone pilots and hobbyists. The “nerd ecosystem” of component supply chains, skilled pilots, and tinkerers which sprang up around the off-the-shelf drone allowed Ukraine to leverage its technical talent to create an entirely new domestic weapons capability despite having little capital. The Ukrainian government took full advantage, even modifying its commercial code to facilitate the import of drone sensors and components.

That spring, Ukrainian volunteers founded the People’s Project to raise funds to supply the Ukrainian army, and sourced two Phantom 2s (~$600), one Skywalker X8 (~$300), and one octocopter, all commercial off-the-shelf drones. Ukrainian battalion commander Natan Chazin founded the Aerorozvidka Project, which by 2015 grew to a twenty-man tactical unit of repurposed vans launching modified hobby drones. In June of 2014, Artem Vyunnik, who had been founding a consumer drone company before the invasion, modified his consumer drone prototype into the A1-CM “Fury.” With a seven-foot wingspan, two-hour flight time, and high-resolution zoom camera, the Fury series cost between $10,000 and $22,000.

By 2016, Vyunnik supplied over forty aircraft to the Ukrainian Armed Forces, National Guard, and volunteer organizations, which were used to discover hundreds of enemy targets for artillery reconnaissance. Companies rapidly iterated on combat drone models in conjunction with the armed forces, and produced around thirty different models of drone that were officially used by the Ukrainian army in the first three years of the war. In 2018, a U.S. Army National Guardsman advising the Ukrainian Delta Center, told The Smithsonian that “In the last two years . . [The Ukrainians] have rapidly advanced from using dirigibles or balloons to do reconnaissance to building their own UAV systems . . . from zero.”

Reaping the much-needed rewards of its investment circa 2014, Ukraine entered the present conflict with a three-to-one advantage over Moscow in small tactical drones. The deployment of drones in Ukraine is widely considered to have tipped the balance in the successful resistance to the initial Russian offensive, and broadly defined the battlefield of the future. This is due not to multimillion dollar military-grade systems like the Predator, but to a consistent bend towards smaller, more attritable systems. Ukraine has only increased its investment in unmanned systems.

The country invested over $1 billion in drone technology in 2023. In February 2024, President Zelensky established the Unmanned Systems Forces as a separate branch of the armed forces, and set a target of producing one million drones domestically this year. There are now around 200 companies producing drones in Ukraine, and in 2023, the Ukrainian government increased its drone purchases from domestic companies by over 100 times the previous year. The acute need for unmanned systems, combined with enthusiastic support from the government of Ukraine, has given rise to a dense startup ecosystem, in which novel drone technologies receive tight feedback from the battlefield and are able to rapidly iterate.

The drone war in Ukraine rages on two fronts. The first is a strategic war, in which adversaries exchange “deep strikes” hundreds of miles into enemy territory at targets designed to impede the enemy’s ability to supply the front, like munitions factories, military bases, or energy infrastructure. Strategic drones, like Russia’s Shaheds, typically cost $20-30,000 and are able to fly 600-1,500 miles to crash and explode “kamikaze”-style, into a target. In this war, strategic strikes have been especially focused on drawing limited air defenses away from an increasingly deadlocked front across a geographically vast rear.

The second drone war takes place across its 600-odd-mile front line. Here, most famously, first-person view racing drones equipped with a few pounds of munitions fly suicide missions into targets, and “Baba Yaga” bomber hexacopters drop larger payloads on targets below. But reconnaissance drones play a critical role. Small, commercial quadcopters equipped with high-resolution visual or infrared cameras fly across the contact line to locate enemy targets a mile away or more, then relay the locations to brigade headquarters who may call in a strike from heavier-duty ground artillery like mortars. Commercial drones have become so ubiquitous and so effective at locating targets across the front lines that the war has been called a “transparent battlefield” in which it is impossible to achieve the element of surprise. Both Ukrainian and Russian militaries find that they cannot mass forces or conduct offensive operations without creating a target.

AI at the Edge

UAVs had lost decades where little development occurred, and it took individual technologists as well as functional institutions such as DARPA to see the potential that larger military bureaucracy and elected government neglected. In the history of military technology there are often decades where nothing happens; and there are weeks where decades happen. In that sense, decades have passed since the Russian invasion of Ukraine in 2022 began. That the tactical quadcopter has transformed the modern battlefield over the last 24 months is consensus. But even these cutting-edge systems are flown predominantly by hand. The AI revolution of the same period has yet to make its impact on rapidly changing warfare. What does unmanned reconnaissance look like in the age of AI?

Recall the Lightning Bugs. In deploying a drone, the most important calculation one makes is cost per mission. A $2,000 drone which flies one kamikaze mission essentially costs the same as a $20,000 drone which can be expected to survive 10 missions. Artificial intelligence changes this cost calculation. Put a GPU on a drone, you add a few hundred dollars to the cost, but suddenly, you can expect it to come back on its own.

The drone operates on an electronic battlefield. To defend against a reconnaissance drone, the adversary must jam the radio frequency connecting the aircraft to its ground station controller, so it can no longer receive manual commands, or jam the frequency connecting the drone to GPS satellites, so it can no longer navigate. Introducing AI to the drone allows it to use computer vision to navigate like a human would by looking at a map, without the need for any satellites, and make decisions about where to fly without constant input from a human operator. “You cannot jam such a drone, because there is nothing to jam,” one Ukrainian drone pilot from the 92nd Brigade told Reuters in March.

Electronic warfare and counter-drone systems are rapidly growing more sophisticated. To achieve decisive superiority on the battlefield of the future, drones will need to be capable of operating, processing data, and making decisions independently of a human operator. Against a technologically capable adversary such as the U.S. might face in China, new capabilities such as recognizing objects and obstacles and autonomous action will very quickly become necessary. The next generation of unmanned aerial systems will not be achieved with attritable systems alone. The winner of the next great power conflict will be the one that invests in intelligence at the edge.

History can take the same lesson from the successes of Ryan Aeronautical in Vietnam, of Abraham Karem in his garage, and of the grassroots drone industry in Ukraine. Key breakthroughs in unmanned systems are made when startups are able to receive tight feedback from performance on the battlefield, rapidly iterating in conjunction with the warfighter. As the Department of Defense seeks to reshore America’s drone fleet, it should prioritize the startup companies that create war-winning innovations.