This is the kind of air defense story that sounds clean and simple on a slide, and then gets messy the second you picture a real night with real drones coming in. A “cheap kinetic interceptor” that can hit small drones at short range? Yes, that’s exactly the direction things need to go. But if we treat this like a magic bullet—“just add AI and a low-cost missile”—we’re going to buy ourselves a comforting demo and a disappointing reality.
Based on what’s been shared publicly, the U.S. tested a compact interceptor called Mk-III/Harpe from Perseus Defense at Yuma Proving Ground during an event labeled Falcon Peak 26.2. The description going around is very specific: about 61 cm long, aimed at small drones up to around 25 kg, with a short reach of 1–2 km. It’s launched from an 8-round container-style launcher that’s described as under 36 kg. The post also says it was integrated with Anduril’s Lattice system, which is positioned as an AI-enabled command-and-control layer that pulls together inputs from cameras, radars, and other detectors to help decide what to shoot, and when.
The price claims are what make people sit up: about $2,500 per interceptor, and about $10,000 for the launcher plus detection and guidance pieces. Whether those numbers hold up outside a controlled environment is a big deal. Still, the direction is the point: low-cost shots against low-cost threats, instead of burning expensive interceptors on something a hobbyist could buy.
From our seat—as a company that builds radar and sensor fusion for air defense—the part that matters most isn’t the interceptor body length or even the container weight. It’s the chain that starts with radar drone detection and ends with a confident decision to fire. That chain is where cheap systems either become the backbone of “small-sky” defense or turn into noisy gear that operators stop trusting.
Because the hard truth is this: at 1–2 km, you don’t get time back. If your detection is late, or your tracking is unstable, or your system can’t sort a real drone from clutter, your “cheap interceptor” becomes a cheap way to miss. And misses aren’t just wasted money. Misses teach the other side what works.
People love the idea that AI glues everything together. Sometimes it does. Sometimes it just makes a messy sensor picture feel more certain than it really is. If a system “fuses” cameras and radar and detectors, great—until you’re dealing with a foggy night, low contrast, birds, reflections, and a drone that’s trying not to look like a drone. Then the question becomes brutally practical: can you hold a stable track, can you classify the target well enough, and can you do it fast enough that the operator doesn’t freeze?
This is why we’re cautious about any story that frames the main barrier as bureaucracy alone. Yes, slow approvals and rules can crush speed. But blaming paperwork is also a convenient way to avoid admitting how hard real deployment is. In the field, you need to integrate sensors, confirm coverage gaps, manage false alarms, train crews, and avoid shooting the wrong thing. The fastest way to get a program stalled isn’t regulation—it’s a highly public “oops.”
The post compares this U.S. approach to an Estonian mini-missile, Frankenburg Technologies Mark I, reportedly around $4,000, up to 2 km range, and able to engage fast targets up to around 600 km/h. It also claims these are already being used in Ukraine, and could even integrate with unmanned maritime platforms. Maybe all of that is true; some of it may be hard to verify from the outside. But the comparison is useful because it highlights what everyone is racing toward: lots of shots, lots of launch points, and short-range coverage that can be moved and replaced quickly.
Here’s the consequence that gets ignored: as soon as defenders can cheaply shoot down simple drones, attackers shift tactics. They add decoys. They swarm. They fly low. They mix real threats with junk. They push you into a “shoot or hold” panic. And suddenly your real scarce resource isn’t interceptors—it’s decision quality. If your radar drone detection and sensor fusion can’t keep up with that kind of chaos, you either waste shots or you hesitate, and either way you lose.
Now picture a few real scenarios.
Imagine you’re protecting a power substation. You get a radar track at the edge of range. The camera can’t see much. The system says “likely drone,” but not confidently. Do you fire a kinetic interceptor over nearby roads? If you wait for certainty and it’s real, you may be too late. If you fire and it’s not, you create a different kind of incident.
Or say you’re guarding a logistics depot. Two small drones appear. Then four more. Then a few “things” that look like drones but don’t behave like drones. A cheap interceptor is great, but only if your system can maintain clean tracks and not collapse into confusion. This is exactly where multi-sensor fusion matters—not as a buzzword, but as a way to keep the picture stable when the sky gets crowded.
And if the post is right that systems like Lattice have been sold abroad for defense against propeller-driven drones, that tells you something else: this isn’t a far-off future problem. It’s already a buying decision, and buyers are choosing between “quick and cheap” and “trustworthy under pressure.”
I do agree with the underlying push: mass, short-range defenses near borders and around energy, transport, telecom, and logistics sites are no longer optional. The cost exchange has to make sense. A $2,500 shot against a cheap drone is a rational conversation starter. But I don’t buy the idea that the interceptor is the center of the story. The center is whether the detection, tracking, and engagement decision can be trusted at speed, day after day, when the other side is actively trying to trick it.
If we’re going to deploy “small-sky” defenses at scale, do we optimize first for the cheapest interceptor, or for the best radar drone detection and sensor fusion that prevents bad shots and missed shots in the first place?