Ukraine has spent this war building some of the densest electronic warfare cover on earth. Along much of the front line, GPS and GLONASS simply do not work. A drone that depends on satellite navigation arrives over the target having forgotten where it is.

Russia’s answer, on at least some of its cheap Molniya attack drones, was to screw a magnetic compass to the airframe where the onboard camera can tilt down and look at it.

Five screws. About two dollars. The operator reads the needle through the video feed and corrects the heading by hand.

What they actually did

The photograph that started it showed a small dial fixed to a panel inside the airframe with five Phillips screws, positioned in the camera’s field of view when the gimbal points down. Nothing is wired to it. It is not connected to the autopilot. It does not talk to anything.

That is the entire system. The human in the loop becomes the navigation computer. He watches the video feed, glances at the needle, works out that he is drifting south of where he wants to be, and feeds in a correction. It is exactly how a pilot flew in 1925, except the pilot is two hundred kilometres away looking at a screen.

Why satellite navigation is so easy to kill

One of the sharpest questions under the original post was this: if the GPS signal can be jammed, why does the control and video link keep working?

The answer is distance and power, and it is the single most useful thing to understand about this war.

A GPS satellite sits about 20,200 kilometres up and transmits with the power of a domestic light bulb. By the time that signal reaches the ground it is astonishingly faint, below the background radio noise, and a receiver only extracts it by knowing exactly what code to look for. A jammer sitting a few kilometres away, running a few watts, arrives at the drone’s antenna orders of magnitude stronger. It does not need to be clever. It only needs to shout.

The control link is the opposite geometry. That transmitter is on the same battlefield, tens of kilometres away rather than tens of thousands, so it reaches the drone as a comparatively enormous signal. Jamming it means out-shouting a shout, from further away, on a frequency that hops. Harder, and often not worth the effort.

So the satellite link dies first, every time. Which is why the interesting question in drone warfare is no longer how to navigate, but how to navigate when the sky has been taken away from you.

But drones already have magnetometers

This is the other objection worth taking seriously, and it came up repeatedly in the comments. Every hobby quadcopter has an electronic magnetometer. It costs pennies, it is soldered to the flight controller, and it feeds heading into the autopilot automatically. Why bolt on a dial and look at it through a camera?

Because a magnetometer is a sensor in a system, and a system has an attack surface. It can be disturbed by local magnetic fields, it can be fed into an autopilot that has been confused about its own position, and a heading source that the software distrusts gets ignored. Above all, it is one more component in a navigation stack that Ukrainian electronic warfare has spent three years learning to break.

A painted dial with a magnetised needle is not part of any stack. There is no firmware, no bus, no sensor fusion, nothing to spoof and nothing to disagree with. The only processing is a human eye. You cannot jam a man looking at a compass, and that is a genuinely different problem for an electronic warfare officer than jamming a chip.

It is also, obviously, a step backwards. That is rather the point.

The fourth fix in six months

The compass is not an isolated piece of improvisation. It is the fourth answer Russia has fielded to the same question on the same airframe since the start of the year.

January 2026: Starlink. Terminals fitted to Molniyas to route control traffic through commercial satellite internet rather than jammed radio. Reported results were mixed, with around one in three Starlink-guided aircraft reaching the target.

Fibre optic. A physical spool of glass paying out behind the drone. Completely immune to jamming, because it is a wire, and completely limited by the length and weight of wire you can carry.

1 July 2026: autonomy. A FlyCore single-board computer allowing terminal navigation and target selection with no operator link at all, which sidesteps jamming by having nothing to jam.

July 2026: the compass. Two dollars.

Read as a sequence, it is more interesting than any one of the fixes. This is not a weapons programme. It is a workshop iterating against a defence in something close to real time, and the cost asymmetry runs the wrong way for the defender: each new countermeasure has to be developed, funded and fielded, and the counter-countermeasure arrives by post from a hardware shop.

What a compass cannot do

Here is where the viral version oversells it.

A compass gives you heading. It does not give you position. Knowing you are pointing at 047 degrees tells you nothing about where you are along that line, and errors accumulate: wind drifts you sideways, airspeed is estimated rather than measured, and there is no fix to correct against. This is dead reckoning, the oldest navigation method there is, and its failure mode is well known. You arrive confidently in the wrong place.

Against a city, a refinery or a power station, that may be enough. Against a specific vehicle in a treeline, it is not. The compass keeps the drone flying in roughly the right direction long enough for the operator to recognise something in the video feed, and recognition does the rest of the navigating.

Ukrainian analysts have been careful about this. The modification is documented; its battlefield effectiveness is not established. What it does prove is that the jamming is working well enough to be worth defeating.

One note on timing

This is doing the rounds again this week, and it is worth knowing it is not new. The photograph surfaced on 17 July 2026 and was reported within days. What has changed since is only that the clip found a larger audience.

The underlying question the post asks is the right one, though. Is a plywood drone with a two-dollar compass proof that improvisation beats investment, or a stopgap that better jamming eventually erases? Probably neither. It is evidence that the contest has moved to a place where the cheapest available answer is often the most robust one, and that is an uncomfortable finding for anyone whose procurement cycle is measured in years.

Sources: Serhii “Flash” Beskrestnov; United24 Media; Tom’s Hardware; 19FortyFive; UAS Vision; TechRadar; Defence Blog; theaviationaddict on Instagram; Afterburner reporting