Weekly review, week to 16 August 2026

🎯 Key structural shifts

The jet "strike-interception" race has entered an inversion phase. The key event of the week: Russia's output of jet drones has overtaken its own piston fleet — ~3000 Geran-4/5 versus ~2800 Geran-2 (Shahed-136) per month (HUR data). The Geran-3 (Telefly JT80 engine, up to 370 km/h) has been taken out of production as a transitional model in favor of the Geran-4 on the Telefly LX-WP-160 [TECH] — twice the thrust, speed up to 500 km/h, range up from 600 to 850 km. This has wiped out last year's main achievement — cheap FPV interception: "Pisyuns" and "Stings" physically cannot catch jet targets. Ukraine is responding with jet interceptors >600 km/h (the Kyiv-built Stellar Jet, Brave1 funding), but the question of guidance at closing speeds that bring the price close to a classic surface-to-air missile remains unresolved. Russia is now in an open window of superiority, and that window is finite.

The space domain has become a direct strike target rather than a backdrop. For the first time in the open information space, a massed salvo of FP-5 Flamingo cruise missiles (5–18 missiles, according to various data) was recorded against the Progress Rocket and Space Center in Samara — the only point of serial assembly for the entire Soyuz-2 line, which launches reconnaissance satellites (Liana, Persona, Razdan) and the Rassvet communications constellation (Bureau 1440). A hit, presumably on Building 106A (installation of onboard electronics in clean zones), means not a repair but recertification of the manufacturing process — months of downtime amid a sanctions-driven shortage of electronic components. The Russian drone control loop via its own satellites and the Ukrainian campaign to physically destroy it have become parallel processes.

Logistical and maritime asphyxiation has reached a systemic effect. The massive strike on Novorossiysk coordinated 12+ types of strike assets for the first time (Palianytsia, Peklo, Bars, Long Neptune, MICH, Sargan/Mamai USVs), damaging four Black Sea Fleet warships and halting the NKHP grain terminal (79% of Russia's grain exports). The Russian Navy's Mediterranean group has effectively ceased to exist. In parallel — the blockade of Greater Odesa (loss of ~1/3 of Ukraine's grain exports), strikes on locomotives (20–25 per month) and on refineries 2000+ km away. Both sides have shifted to "total war" against rear-area infrastructure.

📈 Technology dynamics and trends

Jet platforms have scaled up, while piston ones have been downgraded in status. Besides the Geran-4/5, the Russian BM-35 midstrike drone has received a jet engine source (10 kg warhead, >350 km/h, range 200+ km) — previously a carrier of Starlink terminals, it is now out of reach of many current Ukrainian interceptors. The Gerbera has undergone its own "Geranization": the Gerbera-2 source is enlarged, carries a 20–30 kg warhead, a 63 l fuel tank and "effectively half of a Geran-2 engine" — the decoy-imitator is turning into a full-fledged strike drone. Telling is the engineering deconstruction of the British Skyhammer interceptor (Cambridge Aerospace) from wreckage on a beach [TECH]: a production JetCat P250-Pro turbojet (25 kgf, €4000), an FDM-printed skin (~£150), a flight time of only 5–9 minutes, no loitering at all — launch only on a ready track from a ground radar source. On head-on courses the speed is enough against jet Gerans, in a tail chase it is not. This is a fundamental limitation of the entire class of cheap jet interceptors: they are "the executing link of someone else's detection chain."

Satellite-free navigation has become a full-fledged layer of confrontation. The main [TECH] breakdown of the week is the exposure of the Russian ground navigation network SineLink/SNS source, an "inverted GPS" based on multilateration. The drone measures the radio-wave travel time to ground beacon-transponders with known coordinates: range up to 120 km, position update 2 times/sec, accuracy from 60 m, minimum 3 beacons (optimally 4–7), dead zone ~200 m. Signal physics — LoRa 500–1023 MHz, frequency hopping across a 510 MHz band, power up to 2 W, encryption; a narrowband scanner is useless. The key vulnerability: a beacon is silent without a drone overflight, but during an overflight it emits hundreds of short packets in 1–1.5 min — enough for direction finding, and it is better to take bearings at a distance, where the beacon transmits at full power. Service data (coordinates, keys) live only in RAM — a captured unit is valuable "alive." An engineering addendum source: the beacon antenna has pronounced optima rather than covering the whole range — broadband EW can overload the receiver input or burn out the SAW filter. In parallel, optical vision navigation is scaling: the Russian SOVA (matching a camera image to a satellite map, a ~60 MB model on a CPU, integration with ArduPilot/PX4) and KENS, as well as a captured optical module with cryptographic protection via a map pass card (RFID, key held in memory until power-off) source.

Analysis of navigation-jamming geometry has reached a new level of specificity. A Russian teardown of a captured 4-element Kyocera AVX 1002649 CRPA antenna from the Ukrainian Chaklun-B midstrike drone [TECH] source yields a non-obvious conclusion: the CRPA is oriented toward the upper hemisphere, whereas ground-based EW hits from below and from the side — the geometry works against the jamming side, hence the "patchiness" of GNSS suppression. The recommendation is to jam "from the top down."

Smart targeted EW is replacing "dome" logic. The Russian Yastreb/Tishina system (Potapov) source is not a "noisemaker" but a system that simultaneously scans the entire band and works in a targeted way on the frequency, following the UAV through frequency hopping; low weight, consumption ~300 W. It is "a scalpel instead of a chainsaw" — without high-power emission hitting friendly forces. Ukraine's December1, after two years of silence, unveiled the Damba EW system source. A Ukrainian communications expert (Flash) formulated new requirements source: against "Shahed seekers" with target lock-on, the only effective EW is one whose operating range exceeds the optical lock-on range (2–3 km in clear weather), and it must change bands — the adversary has moved off 1300–1500 MHz. "A one-kilometer EW dome became obsolete six months ago."

MESH relaying has carried remote control deep into enemy territory. The "Geran-4 Seeker" is guided "live" via mesh modems; according to Russian data, without additional equipment in Belarus the signal from Russian territory propagates up to 220 km source — enough to hit the outskirts of Kyiv and moving targets (locomotives). In parallel, Starlink Mini is being fitted en masse on ordinary FPV drones source, displacing fiber optics — the link becomes broadband, but the importance of EW against Starlink itself is growing.

The cybersecurity of the wireless layer has been exposed head-on. Three interrelated signals. Ubiquiti AirMAX [TECH]: CVE-2026-21639/21638 — kernel-privilege RCE without authentication, by an over-the-air signal alone, worked from 1.5 and 1.72 km source; the vulnerability in the Linux 2.6.32 kernel has lived for ~10 years, affecting mass-deployed radio bridges on shared masts in the line of contact zone. CATANA [TECH]: a "hostile" SIM uses the RUN AT command to execute code on the modem bypassing the processor — 7 of 8 IoT modules are vulnerable, a direct threat to backup cellular C2 channels of drones source. iFinder: a multi-agent LLM found 84 vulnerabilities in the 4G/5G core — the cost of finding holes in protocol code has dropped sharply for all sides source.

Ground robots have moved out of the experimental phase into standard pairings. The Russian Depesha-3 + aerial reconnaissance ("eyes above, hands on the ground"), the strike TerMIT + a Predator module for intercepting FPV drones on the move, the autonomous Katran PVO system with the Lun radar (AESA) → the Martyshka interceptor drone without an operator. The Ukrainian Hiena (26 kg of explosives, 20 km), Ravlyk (a joint venture in Denmark, >3000 units/year), Fang-1/2. UGVs have been included in the AFU Basic Supply Level — they are already an element of tactics rather than a showcase.

Military space: intense dynamics along all axes. The Russian Rassvet has reached operating modes — 12 satellites provide fragmentary coverage over Ukraine (windows 2–3 times/day of 15–20 min), with a plan for 292 by 2027 and ~900 by 2035. A Ukrainian expert (Stepanets) clarified the architecture source: Rassvet is closer to OneWeb than to Starlink — "bent-pipe" without full-fledged inter-satellite links, so ground infrastructure is critical; the answer is not to shoot down but to jam (Volna/Kupol/Garant systems, 20 km radius). Two such systems were destroyed near Gelendzhik — Starlink jamming at 14–14.5 GHz on 8 channels, 12 Ku-band transmitters, a 14 kW generator source. Starlink V5 [TECH] source: area −50%, mass 1.1 kg, consumption 35–50 W, wind resistance 265 km/h, but without a GNSS module — simplifies installation on UAVs/USVs and changes geolocation methods. An analysis of the dual-use nature of Musk's constellation source: Starshield (~245 spacecraft, V2 bus with classified payloads — optics, SAR, SIGINT, IR missile early warning); the Space Force has ordered ~480 MILNET satellites for controlling drones/ships. On satellite-free navigation — a Chinese prototype star sensor for hypersonics (recognition accuracy >99% nominally, >80% under aerodynamic interference, attitude <5 arc seconds) and quantum time synchronization of Saab Giraffe 1X radars using AQlock 2.0 cold atoms without GNSS, resistant to spoofing source. US Space Command has officially made "integrated space fires" and counter-pLEO priorities — against constellations of thousands of satellites, striking one spacecraft at a time is pointless; the emphasis is on cyber weapons, directed energy and strikes on ground stations.

🔍 Weak signals (Under the radar)

PWNSAT: a full vulnerable satellite stack opened up at DEF CON 34. "Security through obscurity" in space is over: COTS hardware, open software and CCSDS/SPP standards are ubiquitous, and a single vulnerability is inherited by all missions. Of the 34 findings source, the critical ones: AES-128-ECB with a static hardcoded key, disabling encryption with an open command, GPS L1 spoofing (a geofence on unauthenticated GNSS), unvalidated SPP packet length, a hardcoded auth key 0xC0DEFACE, a default login operator/pwnsat. The authors' key conclusion, confirmed by porting from RP2040 to RV1106: the vulnerability lives in the protocol layer and the ground segment — the same place it will be in combat systems built on the same standards. The source code is open — for the engineering and security community it is a ready-made range for practicing attacks on and defense of the orbital domain.

"Lurker" drones have become a separate class with sensor activation. Russian reconnaissance found ~100 Ukrainian lurkers on a single stretch of road (a "company of copters"). Some are equipped with microphones source — the vehicle hides out of the driver's field of view and registers passing traffic by engine sound. The drones sit passive for hours, are activated by acoustics (tens of meters) or motion, and are invisible to reconnaissance assets until the attack; a "mothership/relay + droppable small lurkers" scheme is possible. Fiber-optic FPV drones emit nothing at all and are not detected by portable RF detectors — a separate C-UAS blind spot. A separate related signal on detection: digital video modules (CADDX Ascent GT PRO 4W, LoRa channel) are not detected by standard analog video intercept equipment, and the video operating frequency has crossed above 6 GHz (intercept at 6470 MHz, range 3.5 km).

Shifts in access to remote sensing data and OSINT. The NRO has expanded contracts for commercial SAR (Capella, Iceye, Umbra) — the state is embedding private players into the intelligence system for round-the-clock all-weather surveillance source. The Open Source Centre disclosed a [TECH] methodology for determining a vessel's load from a satellite image via its shadow (a digital twin + trigonometry in ArcGIS, an Airbus Pléiades Neo image) source — a direct evidentiary base against Russia's "shadow fleet" for sanctions and strikes. Vantor (ex-Maxar) satellites have become a standard tool for real-time strike verification (Novorossiysk, Gelendzhik), and NASA FIRMS records fire hotspots at facilities. An OSINT service, findrassvet.ru, has appeared for tracking the Rassvet constellation using the NORAD catalog (TLE, updated every 3 hours). Playfra Map has integrated FIRMS, terrain and satellite imagery layers into a single geolocation environment.

🔮 Vector projection and implications

The jet race (high confidence). The vector is unambiguous: both sides are switching to jet strike drones, and the coming months will be defined by the race of jet interceptors. The key technical problem is guidance at closing speeds: the Skyhammer teardown shows that a cheap interceptor requires external targeting from a ground radar (the radar itself becomes a critical node). What will be needed: X-band AESA radars to form tracks (such as Lun/Stena), terminal-phase machine-vision guidance (an Nvidia Jetson Orin is already fitted on the S-71M Monokhrom, V2U, Geran-2), and — out of necessity — a return to classic surface-to-air missiles and aircraft cannons (interception of a Geran-4 by a MiG-29's 30 mm cannon and by an Osa SAM system has already been recorded). This will bring back the issue of the anti-aircraft missile shortage: FT/Guardian acknowledge that reliable coverage of Ukraine requires 3–4 times more Patriot missiles than the entire world produces in a year.

EW countermeasures (high confidence). Flash's requirement — EW with a range greater than the optical lock-on range and mandatory band switching — will become a de facto standard. Against MESH relaying at 220 km and Starlink on FPV drones, targeted systems such as Yastreb/Tishina and "top-down" jamming that takes CRPA geometry into account are needed. Dome-type broadband EW is becoming a thing of the past.

The space domain (split confidence). Confirmed dynamics: the strike on Progress will slow the deployment of Rassvet by months, which gives Ukraine a window but does not cancel the goal. Analytical hypothesis: the next round will be targeted EW against Rassvet (by analogy with Volna/Kupol/Garant against Starlink) and strikes on the published "Bureau 1440 map" (Moscow, Dubna, Tomsk's Mikran, Smolensk, Minsk). Against proliferated constellations kinetics is pointless — the vector is shifting toward cyberattacks (the PWNSAT lessons are directly applicable) and directed energy. GNSS spoofing and alternative navigation (SNS, vision, star sensors, atomic clocks) will continue to scale on both sides — a stable long-term trend.

The cyber domain (high confidence). The AirMAX, CATANA and iFinder vulnerabilities show that the wireless and protocol layer of C2 is a new front line. German intelligence is already proposing preemptive cyberstrikes on Russian UAV plants. The cost of finding vulnerabilities through LLMs is collapsing — expect a rise in attacks on backup cellular/radio-bridge drone control channels.

🌐 External analysis (Western sources)

• A Taiwanese state audit exposed the failure of the island's counter-drone defense: a $111 million fixed system does not detect the 5.1 GHz frequency of the newest Chinese UAVs, does not cover BeiDou-3 and is defenseless against fiber-optic and AI-controlled drones; the army's procurement of 26 systems was canceled after three failed acceptance tests source. This independently confirms the Russian-Ukrainian information space's thesis about the death of "dome" EW against autonomous and non-radio-controlled vehicles.

• IISS turns NATO's calculation in the Baltics upside down: Russia does not need to physically seize the Suwałki Corridor — it is enough to keep it under constant strikes by fiber-optic FPV drones (range ~40 km, EW-resistant) plus mines and sabotage; ~56% of the Baltic states lie within 100 km of Russia/Belarus source. A direct transfer of the Ukrainian kill-zone experience to a future NATO theater — something the tactical Telegram information space does not formulate strategically.

• At the Lanternfish 26 exercise, ThayerMahan demonstrated a distributed passive hydroacoustic network (ShorePicket/Outpost) that detects underwater drones of all classes — from 15 cm vehicles to XLUUVs, with deployment in under 24 hours source. This extends the picture beyond the land theater: the "drone-counter-drone" race has moved into the underwater environment, where the Russian-Ukrainian information space provides almost no data.

• Firefly Aerospace received a DIU/SDA contract for the Elytra orbital vehicle for rendezvous, inspection and controlled deorbiting of satellites in all three regimes (LEO/MEO/GEO), based on the technology of the Blue Ghost lunar lander source. This is dual-use anti-satellite infrastructure: a vehicle capable of approaching and deorbiting "its own" satellite is technically capable of more — a direct complement to Space Command's declarations on counter-pLEO.

• At exercises in Hohenfels/Combined Resolve, fewer than 20 Ukrainian drone operators notionally routed an American armored brigade; after two weeks of repetitions, the Americans mastered dispersal, camouflage and EW — but the initial gap between updated doctrine and practical skills remains a critical vulnerability of the US Army source. Telling is the coincidence with the cancellation of Russia's Dronnitsa-2026 due to the threat of a strike — both armies are learning drone warfare under pressure from a real adversary.