Two passive optical counter-UAS designs, both Israeli, both running AI at the edge. They differ in how they look at the sky. Third Eye Systems’ MeduzaX is one unit that moves its field of view around the horizon. Its brochures describe a “Step & Stare mirror mechanism”, a fast scan that “covers 360° in seconds” and is “rotated by a unique mirror technique”. GOTEAM covers the circle with several fixed, wide-field thermal nodes. Each one images its own sector on every frame, and nothing in the optical path moves.
MeduzaX is a strong sensor, and parts of this page say so. Third Eye specifies a cooled MWIR detector (1280×1024) paired with a daylight CMOS channel, in a 12 kg man-portable unit that covers 360° on its own. Third Eye states detection range up to 4 km. GOTEAM uses uncooled long-wave cores, which are less sensitive, and needs many nodes to close a circle. The argument for GOTEAM is not about the detector. It is about the revisit interval: how long a given bearing goes unwatched between two looks.
This page compares MeduzaX, not Third Eye as a company. Third Eye also builds Argus Shield, a vehicle system that uses several fixed daylight (CMOS) sensors to cover a 120°×360° cone, with auto-tracking of all objects in view. By its published optics that is a staring design, so GOTEAM is not the only 360° staring system on the market. Argus Shield is a daylight, on-the-move product and is not compared here.
Every claim about Third Eye on this page is quoted from Third Eye’s own brochures, linked in the sources at the bottom and checked on 6 October 2026. We have not tested any Third Eye product. Every claim about GOTEAM is published on this site, in /llms.txt, or in our post Blink and It’s Over. Figures marked illustrative are our arithmetic, not either company’s specification.
One fact drives most of this table: a sensor that does not move sees every bearing in its field on every frame. A scanner sees a bearing for the moment its field of view passes over it, then moves on and comes back on the next scan. Third Eye gives that cycle only as “seconds”. It publishes no per-bearing revisit figure.
| Capability | GOTEAM — fixed staring thermal nodes | Third Eye MeduzaX — mirror-scanned |
|---|---|---|
| Every covered bearing is imaged on every frame, so the revisit interval is one frame period | ✓each node watches its own sector continuously | ✗“covers 360° in seconds”; Third Eye lists a 60Hz “effective frame rate”, which we read as the rate within each stare step, not the full-circle revisit |
| Each track is updated at frame rate, however many drones there are and whatever bearings they come from | ✓every target in a node’s field, every frame | ◐number of objects listed as “Unlimited”; in a scanning design each one is updated once per scan |
| Sees a fast drone turn or dive, not just its position before and after | ✓a drone moves a fraction of a degree between frames (see the arithmetic below) | ◐sees where the drone was on the last scan and where it is on this one; the tracker has to bridge the gap |
| Drones from opposite bearings are all watched at the same moment | ✓each sector has its own camera | ◐all are covered within one scan, but at different moments in it |
| Nothing moves in the optical path | ✓fixed lens, no mirror, no gimbal | ✗“Step & Stare mirror mechanism”; Third Eye describes the unit as low-maintenance and field-ready |
| Losing one sensor leaves one sector uncovered, not the whole circle | ✓the other nodes keep watching their sectors | ✗one unit covers the full circle, so it is the single optical point of failure unless a second unit is deployed |
| No cryocooler in the sensor | ✓uncooled LWIR cores | ✗cooled MWIR; GOTEAM’s published position is that cooled detectors carry a cryocooler with a limited service life |
| Open C2 output over SAPIENT (BSI Flex 335 v2) | ✓azimuth, elevation and range handed to third-party C2 and effectors | —integration through the MeduzaX C2I platform; we found no SAPIENT reference in Third Eye’s brochures |
| Detection engine licensable as software onto a third party’s thermal core | ✓camera-agnostic; runs on 16-bit frames from uncooled LWIR cores | — |
Third Eye publishes no revisit figure beyond “in seconds”, so the scan rows below use illustrative values of 1 s and 3 s. They are not MeduzaX specifications. The drone is also illustrative: 30 m/s, crossing the line of sight 300 m from the sensor, which is about 5.7° of bearing per second.
| Revisit | Drone moves between looks | Looks at that drone in a 10 s window |
|---|---|---|
| Staring node, 30 frames/s (illustrative) | ≈ 1 m, ≈ 0.2° | 300 |
| Scan returning every 1 s (illustrative) | ≈ 30 m, ≈ 5.7° | 10 |
| Scan returning every 3 s (illustrative) | ≈ 90 m, ≈ 17° | 3 |
A tracker predicts where each target will be at the next look. If the drone turns between looks, it reappears away from the prediction, and the tracker has to decide whether it is the same drone, a new one or noise. With a dozen drones in one sector, the same gap makes it harder to keep tracks from swapping. Confirmation works the same way: a system that confirms a target over several looks waits longer for those looks when each one is seconds apart. This is our analysis of the two architectures, not a measured comparison of the products. Third Eye publishes its own detection and false-alarm figures, which we quote below and have not tested. The full argument, with sources, is in Blink and It’s Over: The Fatal Flaw of the Scanning Eye.
Both systems are passive. Both see a drone that is not transmitting, both use neural networks at the edge, and both claim to handle many targets. The mechanisms behind those claims still differ.
| GOTEAM does it by… | MeduzaX does it by… | |
|---|---|---|
| Operating passively | Receiving long-wave infrared. No emission and no spectrum licence. | Receiving mid-wave infrared and visible light. Third Eye: “No emissions, no radiation”. The two systems are the same on this axis. |
| Seeing the radio-silent drone | Heat and shape. Fibre-optic and pre-programmed drones look the same to a thermal camera as radio-controlled ones. | Heat, shape and visible imagery. Third Eye’s MeduzaX presentation lists fibre-optic drones with “No RF Emissions” among the drone threats it addresses. |
| Handling many targets | Per-frame detection in every node. Every target in the field is detected on every frame, so latency does not grow as more targets appear. | Software that does not cap the count. Third Eye lists the number of objects as “Unlimited” and says MeduzaX classifies “unlimited” quadcopters and fixed-wing drones. Each one is seen once per scan. |
| Rejecting birds and clutter | Persistence across many frames. A tracker confirms a target only once it persists and moves consistently, after neural detection on full 16-bit frames. | Neural ATR plus motion detection. Third Eye describes VMD combined with ATR, and publishes positive detection above 94% with FAR below 1% (2025 brochure). Its 2024 spec sheet lists PD 97% and FAR of 1 per hour. We have not tested these figures. |
| Working day and night | Thermal detection, with a cued camera for identification. The staring thermal layer detects. A zoom camera behind it, cued by the track, identifies. | Two channels in one head. Third Eye: “Dual channel – VIS + MWIR”, with thermal and daylight fusion. |
| Handing off to effectors | Open standard. Tracks go out over SAPIENT (BSI Flex 335 v2) to whatever C2 and effector sit behind the sensor. | Its own C2. MeduzaX Tactical C2, with listed mitigation pairings including GNSS spoofers, jammers, kinetic and drone-to-drone. |
Where GOTEAM is stronger. Every bearing a GOTEAM node covers is watched on every frame. A fast drone that turns, dives or drops behind clutter is seen while it does so, and a dozen drones from different directions are each updated at the same rate as one. Nothing in the optical path moves, there is no cryocooler, and losing one node costs one sector. Tracks go out over SAPIENT to whatever C2 and effectors are already on site. Against fast, many-bearing attacks at short range, we think persistence matters more than per-pixel range.
Where MeduzaX is stronger. Sensitivity, single-unit coverage and portability. A cooled MWIR detector with a daylight channel, covering 360° from one 12 kg unit, is a good fit for a mobile team that has to set up quickly and wants long-range early warning of a single slow target. GOTEAM needs at least a dozen nodes to do what one MeduzaX does geometrically, and each GOTEAM pixel covers more sky.
The two can work together. A scanning or zooming sensor is a good second layer behind a staring one: the staring layer keeps every bearing under watch, and the higher-resolution sensor looks harder where a track already exists. No single sensor wins against drones, and buyers should judge both products against their own threat, range and site.
Sources (all accessed 6 October 2026). Third Eye Systems, “Meet the MeduzaX” brochure, 2024 — “covers 360° in seconds”, “rotated by a unique mirror technique”, cooled MWIR 1280×1024, 60Hz effective frame rate, PD 97%, FAR 1 per hour, number of objects “Unlimited”, 12 kg portable bag, and EO-IR challenges “Narrow FOV, tracker logics, multiple targets”: thirdeye-systems.com/wp-content/uploads/2024/06/Meet_the_MeduzaX.pdf. Third Eye Systems, MeduzaX brochure, 2025 — 360° coverage, “scans the entire area within seconds”, “No emissions, no radiation”, thermal and daylight fusion, cooled MWIR and CMOS imager, unlimited targets, up to 4 km: thirdeye-systems.com/wp-content/uploads/2025/11/MeduzaX_2025.pdf. Third Eye Systems, “Meet the MeduzaX” 2025 presentation — “Dual channel – VIS + MWIR with a unique mirror mechanism”, “Step & Stare mirror mechanism”, PD >94% and FAR <1%, fibre-optic drones, MeduzaX Tactical C2, mitigation pairings, recognition-range chart: thirdeye-systems.com/wp-content/uploads/2025/05/Meet_the_MeduzaX__2025_link.pdf. Third Eye Systems, Argus Shield brochure, 2025 — CMOS sensors, 46°×63° per sensor, 120°×360° for the whole system, auto-tracking of all objects in view: thirdeye-systems.com/wp-content/uploads/2025/11/Argus-Shield_2025-1.pdf. GOTEAM claims: this site, /llms.txt, and Blink and It’s Over.
Independent comparison by GOTEAM. GOTEAM is not affiliated with or endorsed by Third Eye Systems. MeduzaX and Argus Shield are products of Third Eye Systems. Third Eye specifications are quoted from Third Eye’s own published brochures and have not been independently tested; Third Eye may have revised them since the access date. Figures marked illustrative are our arithmetic, not either company’s specification. Our reading of how a scanning sensor behaves against fast drones and swarms is architectural analysis, not a measured comparison of the two products. Specifications beyond what GOTEAM publishes are shared under NDA after vetting.
Upload a clip to FlyBox and a GOTEAM field device processes it on real hardware, streaming back an annotated result with per-frame detections and range and bearing — your footage, our engine, no NDA required to start.