Behind the Killing Glass

In December 2024, near Lyptsi, north of Kharkiv, Ukrainian units staged a battle using dozens of robotic ground platforms and flying strike craft without deploying a solitary foot soldier. A captured Russian fighter later described the assault, explaining that on that ground, there was not a single soldier, only drones until he finally laid down his arms.

A listener wrote to ask what combat becomes if computers and autonomous machines fight one another, and what's really at stake when violence moves behind a screen. People often picture a bloodless duel of machines, with soldiers staying safely on the sidelines. Yet as defense analyst Paul Scharre cautions, robots are not going to wage neat contests. In Ukraine, uncrewed systems cause roughly 70 to 80 percent of troop casualties, with some frontline estimates climbing close to 90 percent.

Understanding this shift starts with dividing combat technology into three tiers. First come teleoperated platforms, which a remote operator must guide by hand every second. In between are semi-autonomous weapons that take over specific steps, such as following a moving target. At the far edge sit lethal autonomous weapons: systems that use onboard sensors and programmed algorithms to find and attack targets all on their own, with no person authorizing the blow.

Most aircraft above current trenches are not thinking machines. They are first-person-view quadcopters: cheap civilian craft flown by human pilots looking through video headsets. During the 1991 Gulf War, the U.S. military flew only a single operational drone type, the Pioneer, whose camera famously prompted frightened Iraqi troops to surrender by waving white bedsheets. Today, Ukrainian factories produce roughly 4.5 million small craft annually, while troops on the line burn through over 200,000 each month. If human hands still steer these frontline craft, why are armed forces racing so urgently to give algorithms lethal control?

Militaries aren't chasing artificial intelligence for sci-fi robots. They're turning to it because the radio airwaves are blocked. Electronic warfare, where forces blast frequencies with static, severs the invisible link between pilots and their drones. Most uncrewed craft need a constant connection to navigate and attack. When jamming cuts that feed, a drone drifts off or drops straight to the dirt.

To bypass that static, developers turn to terminal guidance: the machine runs on an optical autopilot during the last twenty seconds before impact. A pilot tags a target while the signal holds, and if jamming severs the link, onboard visual software takes over the final dash. That was how a Ukrainian drone struck what appeared to be a Russian tank once the controls died. According to the Center for Strategic and International Studies, a defense think tank, letting software finish the run lifts strike success under heavy interference from around 10 or 20 percent to roughly 70 or 80 percent.

Guiding that final sprint quickly slides toward letting machines choose what to kill. In early 2025, Russian units reportedly deployed the V2U, a drone Ukrainian intelligence believes was engineered to select targets independently using onboard software. Then in 2026, the artificial intelligence company Anthropic revealed that independent Russian coders used its Claude tool to build software for an autonomous swarm, generating code aimed repeatedly at targets in Ukraine's Donetsk region.

That slide toward unguided lethality has set off alarms at the highest levels. In August 2026, leaders from the United Nations and the Red Cross issued a joint appeal: "We are now dangerously close to crossing a moral red line, the autonomous targeting of humans by machines." At the Red Cross, Laurent Gisel warned that delegating decisions about using force to machines creates a fundamental problem under international law. Sensors cannot evaluate human intent, recognize surrender, or weigh proportionality. Yet even if an automated tracker flies that final dash, the deeper shift is happening behind the glass display, where living operators watch the killing.

Behind the monitor, drone combat might look like a video game, but keeping one small craft in the air takes a full team. Launching a single first-person-view strike drone is no solitary task; it typically ties up 3 to 6 people to manage piloting, payloads, signals, and targeting. As tactical reports point out, executing even one strike against an enemy trench demands a surprisingly heavy human footprint.

Looking through glass changes what killing feels like. When monitors shrink human lives into clusters of pixels, the emotional weight begins to slip away. An Army chaplain, D. Keith Shurtleff, warned that seeing the enemy as blips on a screen risks losing the natural restraint that the horror of war provides. An Air Force lieutenant colonel who directed Predator missions noticed how easily crews acted like "gods from afar," casually choosing targets while forgetting real humans stood on both ends.

For soldiers underneath those lenses, the reality is pure dread. Constant surveillance overhead causes sanctuary deprivation, stripping away any chance to rest, hide, or recover in cover away from enemy eyes. Troops endure chronic anticipatory anxiety as bunkers and earthworks turn into glaring targets. As one tactical assessment concluded, under loitering munitions—exploding drones that hover overhead waiting to strike—"the trench becomes a mass grave."

Whether this distance protects operators is still an open question. Clinical studies haven't established whether screen-based combat spares operators from trauma or produces distinct rates of moral injury and clinical PTSD. Yet there is no mystery about who absorbs the blow. Drones cause an estimated 70 to 90 percent of front-line casualties in Ukraine. As Paul Scharre argued, uninhabited systems do not pull humans from the fight; they funnel lethal power onto troops exposed in the dirt. What happens, though, when that screen is no longer a human pilot's window, but an algorithm deciding independently whom to kill?

For military planners, handing over the trigger offers an obvious edge: software never panics. Major Kenneth Rose of the U.S. Army observed that machines don't tire, blink, or hide under trees, knowing no fear when human attention fades. A 2018 U.S. white paper argued automated targeting reduces civilian casualties, and scholar Nathan Wood suggested autonomous weapons can sometimes prove superior, morally and legally, to terrified soldiers.

That speed is real. In Army experiments with Project Maven, algorithms compressed targeting timelines from 724 minutes to about 20 minutes. Across Ukraine, the Defense Ministry reported its system analyzes over 100,000 drone video streams monthly, detecting about 70 percent of targets in real time.

Spotting a shape, though, is not valuing human life. Algorithms struggle with the principle of distinction—the requirement under international law to separate fighters from civilians. Former UN rapporteur Christof Heyns warned machines lack the qualitative judgment human life demands, unable to weigh intent or surrender. Roboticist Amanda Sharkey observed that machines cannot feel pain or extend empathy, while philosopher Peter Asaro noted only mortal humans who experience loss grasp life's worth. Moral agency, the Red Cross insists, cannot be delegated to code.

Surrendering that choice, even during the last twenty seconds of terminal flight, opens an accountability gap: the legal void when an autonomous weapon kills unlawfully, yet no individual can be prosecuted. Human Rights Watch warns that deep learning relies on opaque black boxes. Software cannot be tried for war crimes, making it almost impossible to assign criminal liability to commanders, programmers, or manufacturers.

That fog is already here. On July 6, an AI-enabled Russian drone struck a Zaporizhzhia gas station, killing Tetiana Bubynets, amid conflicting accounts over whether software or a pilot approved the hit. Other reports cited three civilians killed by an AI-guided drone, though a UN official declined to verify the strike for lack of information. In that rubble, untangling who chose to fire proved impossible. Does erasing friendly casualties tempt governments to start wars more readily?

Protecting soldiers from lethal fire is simple military logic. In 2011, Lieutenant General Richard Lynch testified: "As I think about what’s happening on the battlefield today... I contend there are things we could do to improve the survivability of our service members." On front lines today, analyst Can Kasapoğlu notes Ukrainian forces deploy robotic ground vehicles for evacuation, engineering, and fire support. Ukraine's General Staff estimates fielding robotic platforms can cut human casualties by up to 30 percent.

The danger begins when leaders realize fewer soldiers come home in coffins. A study led by Riley Simmons-Edler argued that swapping soldiers for autonomous systems reduces the political friction of launching an attack, softening domestic blowback. Boaz Golany warns this lets governments sustain operations without paying the political price of mounting casualties. In October 2016, Barack Obama cautioned that remote capabilities could tempt leaders to "carry on perpetual wars all over the world, and a lot of them covert, without any accountability or democratic debate."

Between peer rivals, combat can accelerate into hyperwar—where artificial intelligence and autonomous systems compress decisions to machine speed. The Red Cross warns this runaway velocity risks escalating conflicts unpredictably, while researchers caution minor clashes could quickly widen into broader warfare. With tensions high, the U.S. Strategic Futures Group warns leaders might feel pressured to strike first out of fear that an opponent's salvo will wipe out their advanced weapons.

Yet machine speed brings no clean finish. That earlier memory of Iraqi troops waving white bedsheets offered a rare fantasy of swift surrender. For the next 20 years, intelligence forecasts warn future conflicts will be extraordinarily violent yet protracted and inconclusive. That devastation repeatedly spills onto non-combatants. On September 17, 2026, prosecutors reported a Russian drone killed a civilian near a public transit stop in Kotivka. As Boaz Golany observed, "More importantly, modern wars increasingly blur the distinction between combatants and civilians." Where does that leave combat when the drone batteries die, leaving only the cold dirt of the battlefield?

Once batteries run out and propellers halt, machines cannot hold ground. That truth rests on combined arms, the military doctrine proving no single technology wins without infantry, armor, and artillery moving as one. An uncrewed vehicle can rake a trench line, but it cannot sit in the mud and hold it. As military analysts put it, "The soldier on the ground remains the decisive element of combat power." Troops take and secure territory, enduring where hardware fails.

That dynamic decided the 2020 conflict across Nagorno-Karabakh. Drone feeds led the news, but taking the mountain fortress of Shusha was done on foot. In heavy fog that blinded airborne sensors, commandos climbed steep rock walls and cleared city blocks room by room. Drones softened the target, but tactical proficiency and an infantry assault finished the fight.

That truth explains what happened to the captured Russian soldier near Lyptsi, who remembered that on the battlefield he did not see a single Ukrainian soldier, only drones. It felt like a clash of robots, but behind every device sat a human crew managing sensors, hauling batteries, and aiming fire. Robotic craft do not displace soldiers; they augment them. As Boaz Golany pointed out, the real question is how humans and smart algorithms fight together, preserving contextual judgment and accountability where lives hang in the balance.

Future combat will never be an empty arena where machines duel machines. Warfare behind monitors does not spare blood; it directs lethal force straight into human flesh while pushing decisions to machine speed. As defense researchers Kateryna Bondar and Matt Mande observed, "The locus of lethal decisionmaking is not the thing that flies; instead, it is the software that commands it." And as Pope Francis cautioned, waging war without direct human hands invites "a lessened perception of the devastation" and dulls our sense of moral responsibility. Looking through glass never erases the human toll—it just alters how far back we stand.

If you want to understand how combat is shifting, look into FPV drones in Ukraine. And follow autonomous weapons; we have more stories on them so you will be ready when modern warfare turns. Thanks for listening. Until next time, this has been Angle.