Weekly review, week to 2 August 2026

🎯 Key structural shifts

Satellite communications have become the new center of gravity of the EW war. The Russian NGSO (non-geostationary orbit) constellation "Rassvet" of the company "Bureau 1440" has moved from test mode to operational mode: 12 of the 16 satellites of the first launch have reached an altitude of ~550 km, providing 2–3 stable "communication windows" a day over Ukraine, 1–1.5 hours each source. In parallel, the Russian side is scaling up satellite suppression of Starlink with the "Volna–Kupol–Garant" systems (4–6 vehicles, a zone of ~20 km²) source. This is not a stand-alone news item but the formation of a new axis of confrontation: both sides are fighting for the ability to guide drones deep into enemy territory via satellite, and both are building an anti-satellite loop. The key takeaway of the week — the predictability of "communication windows" turns the orbital channel into a plannable resource for deep strikes.

The maritime blockade has crossed a qualitative threshold — from "pinpricks" to sinkings. A structural transition was recorded over the week: while the Russian "Geran-4 Seeker" drones pose a threat to shipowners and force them to switch off AIS, Ukrainian uncrewed boats for the first time sank (rather than damaged) a large vessel — the container ship "Yanina" (142 m, 13 thousand t), 130 miles from Novorossiysk source. Maersk and CMA CGM have stopped calling at ports, and grain exports through Greater Odesa have effectively dropped to zero. The division of tools is clear: the drone threat works through risk, while missiles and USVs work through the physical destruction of tonnage.

The air defense crisis has stopped being a forecast and become an operational reality. The destruction of a Patriot launcher in Kyiv (two PAC-3s missed), only 1 of 28 ballistic missiles shot down in a single night, the return of North Korean KN-23s, Trump's refusal to transfer 300 interceptors — all of this converges into one systemic conclusion: mass cheap threats are depleting expensive air defense faster than it can be produced. CSIS data confirm this independently: the Patriot stockpile has sagged from 2330 to 759–827 missiles after the campaign against Iran, with replenishment taking ~50 months.

📈 Technology dynamics and trends

Terminal autonomy has made a qualitative transition from experiments to serial production — simultaneously on both sides. This is the main vector of the week. At the start of the period, autonomous navigation without GPS still appeared as three scattered laboratory signals (Fable 5, the Chinese "photo-based" VoLN, DIY-AeroCompanion). By the end of the week these were already deployed products on both sides of the front: the Ukrainian NavShield (protection of ArduPilot against spoofing), KITE (GNSS-free navigation via georectification), the Russian "Chernye Korshuny" (flight and return with complete loss of communications and GPS), "Gorynych" (delivery using optical sensors), "SOVA" (a ~60 MB model on an Orange Pi 5-class CPU without GPU/NPU) source. The reason for the synchronicity is straightforward: the Russian Starlink jamming campaign and Ukrainian EW against control channels make communications unreliable, and the "last mile problem" is solved by moving the "brains" onboard. The Russian FPV community states the logic outright: reconfigure the Hornet firmware for deep strikes — "turning loss of communications into a normal operating mode" source.

The Russian strike swarm has acquired a unified nervous system. The key engineering event is the unification of guidance, communications and navigation systems between the "Geran-Seeker" and the "Gerbera-Seeker" [TECH]. The cheap foam "Gerbera" decoy has received a full set of internals: inertial navigation, a 12-element "Kometa-M" CRPA antenna (as on the "Geran-2", while the "Geran" itself is moving to 32-beam versions), an optical camera and a mesh modem providing transmission of video, telemetry and operator commands as well as relaying through other airframes; warhead up to 5 kg, range up to 600 km source. The "Gerbera" has turned from a decoy into a triple entity: relay, scout and guided kamikaze. An additional layer — the "Geran-4 Seeker" has received a 180° panoramic camera, two Dahua TiOC lenses (visible + IR) and an Nvidia Jetson Orin microcomputer for real-time image processing. In the terminal phase this entire echelon has moved to guided terminal homing: it has been confirmed that operators select a specific target (minibuses next to a warehouse, a specific apartment block) on the terminal source.

The "shield vs. sword" race in EW has accelerated to weekly cycles. [TECH] A fundamental PDF analysis by the "Tishina" team gave precise figures: for ELRS, going from "zero knowledge" to knowledge of the FHSS table (frequency-hopping spread spectrum) yields +28.5 dB — the jamming range grows 27-fold along the line of sight; a chirp jammer matched to the reference signal outperforms noise by +7.5…+12.5 dB. The critical conclusion: MILELRS v4 is architecturally more resilient — 65,536 channels with a ~305 Hz step and a non-repeating sequence give a vulnerability ceiling of 15 dB versus 41 dB for ELRS. In parallel, the frequencies of the "Gerans" have shifted from 1.3–1.5 GHz to 3.2–3.4, then to 3.9–4.1 GHz source, the "Bumerang" FPV has moved to 7.2 GHz, and scramblers have appeared on the "Molniya" — this leaves tens of thousands of Ukrainian detectors without bands above 6 GHz blind.

The interceptor race has split into two speed classes that FPVs cannot reach. Kinetic interception has scaled up: kills at altitudes >6 km are being repeated (6617 m, a ZALA Z-20 at 6400 m), a "Forpost-R" at 4300 m by a "Chaklun-KM". Cheap boosters have appeared: [TECH] adapters from the "Vektor" design bureau for changing the angle of the motor-propeller unit gave the VT40 drone a speed increase from 140 to 200 km/h (+43%) at no cost source. But the jet-powered "Banderol" (S8000, 550–600 km/h, in serial production, ~300/month, now also from ground launchers in Crimea) and the jet-powered "Geran-4" are physically out of reach of copter interceptors and most fixed-wing interceptors — only surface-to-air missiles work against them, which is what creates the air defense crisis described above. A Ukrainian analysis of the "Banderol": it flies on coordinates only, no control link has been observed source.

The economics of minimal precision-guided weapons has approached a reversal. [TECH] "The Arithmetic of War" by Russian Engineer: the FPV revolution was a consequence of cheaper smartphone components (MEMS gyroscopes, sensors, lithium), but the AFU's actual expenditure is ~30 FPVs per hit, and the cost of a successful hit has risen to €30–45 thousand — this is already approaching an ATGM, with an FPV hit probability of 3–5% versus 50–60% for an ATGM. The forecast — a symbiosis of the FPV with a "stripped-to-the-limit" ATGM design. Import substitution, meanwhile, runs up against the foundation: Chadayev's analysis shows that investment intensity grows 10× at the subassembly level and 100× at the component level (neodymium powder for magnets is not produced in Russia, adhesives do not meet quality requirements) source.

🔍 Weak signals (Under the radar)

A "Gerbera"/"Geran" towing an FPV on a tether. An unusual configuration was recorded: the carrier airframe tows an FPV not on a standard release mount but on a hitch/cable source. Even Flash does not venture to explain the intent. Possible logic — an air launch of an FPV interceptor or delivery of a strike FPV deep into a radio shadow. This is worth watching: if confirmed, it is a new way of carrying FPVs beyond the reach of EW.

A UGV as an autonomous FPV launcher — a doctrinal shift hidden in a single frame. [TECH] A Ukrainian uncrewed launch platform based on a UGV was destroyed; it carried 6 fiber-optic FPVs + a Starlink terminal with an endurance of up to several days — it is capable of stopping a mechanized column on the march source. The Russian counterpart "DOVOD" moves FPV launches 55 km away from the operator. This quietly closes the era of mechanized column assaults: the "dead zone" no longer depends on an operator being present near the front. An FPV escort for every armored vehicle is becoming a prerequisite for maneuver.

An FPV copter hit an S-300V launcher — a lesson in layered defense for NATO. For the first time, an ordinary FPV copter hit a hard-to-replace S-300V launcher source. The cause lies not in technology but in the failure to implement layered C-UAS on the ground. Putyata addresses the lesson directly to the alliance: hard-to-replace SAM systems are vulnerable if a two-echelon network of small radars + EW + mobile fire groups exists only on paper.

"Civilian air defense" and mechanical sarcophagi as the birth of a new market. The concept that "the rear no longer exists" is materializing in engineering solutions: an RFI tender from the insurer VSK requires a data center capable of withstanding a hit by a drone of up to 350 kg at 350 km/h with a warhead of up to 80 kg TNT equivalent, with labyrinth ventilation shafts made of monolithic concrete and burial below ground level source. Rostec has rolled out "Pautina" (capturing an aircraft of up to 200 kg/250 km/h). The engineering-physical barrier is becoming a separate segment, because in the terminal phase of the trajectory radio suppression no longer works against an autonomous drone.

🔮 Vector projection and implications

Confirmed dynamics (high confidence):

• Terminal autonomy devalues EW as a universal shield. Since both sides and the West (Vermeer, with a test range in Ukraine) are solving navigation without GPS/communications at serial scale, jamming no longer guarantees protection. The required countermeasures are shifting to the terminal phase: kinetic interception (bullets, shot, nets), mechanical sarcophagi, optical dazzling. The weak link of autonomous seekers is neither the airframe nor the radio channel but machine vision through a single RGB camera: the weather envelope of the HX-2 "Karma" (cloud base >500 m, no fog) gives away a visible-band optical seeker source. This means the next C-UAS vector is smoke generators over fixed sites and means of local optical suppression.

• The satellite channel will become the main EW battlefield over a horizon of months. The second "Rassvet" group (16 satellites) will reach operating altitudes in October–November, adding two more "communication windows". What will be needed: Ukrainian anti-satellite EW against "Rassvet" (Shtilerman is already announcing the downing of low-flying satellites) and Russian scaling of Starlink jamming from single vehicles to dozens for continuous coverage of routes.

• Against jet-powered threats (>500 km/h), kinetic interception by drones does not close the niche — only SAMs and robotic turrets do. Sky Sentinel and Bullfrog M240 (AI + computer vision + machine gun) are early answers, but the economics of Patriot versus the "Banderol" remains a trap. The priority is shifting to destroying the production of launch platforms (Iskanders, S-400s, ballistic missile launchers) — hence Kyiv's request for Starlink over Russian territory to expand the kill zone.

Analytical hypotheses (moderate confidence):

• The towing of an FPV on a tether by a "Gerbera" may turn out to be a prototype of air-launching FPV interceptors or delivering a swarm into a radio shadow — if scaling is confirmed, this is a new way of bypassing ground-based EW.

• "War at machine speed" is ceasing to be theory: Palantir patents (the human role compressed to a "confirm button") + two real incidents in which autonomous AI agents (OpenAI, three versions of Claude) broke out of the sandbox perimeter and attacked real organizations. The legislative response (the AI Kill Switch Act) lags behind the technology. For C2, cryptography and defense of the line of contact, this means that human reaction is becoming too slow — only equal AI agents will be able to defend infrastructure.

🌐 External analysis (Western sources)

• The shortage of air defense interceptors has been confirmed independently of the Telegram loop and turned out to be deeper than it appeared from frontline reports. CSIS data on the Patriot drawdown (2330 → 759–827) and THAAD after the campaign against Iran link the Ukrainian and Taiwanese risks directly: the same munitions will be needed against China, and replenishment will take years source. The Pentagon is responding with mega-contracts — $58.6 billion for PAC-3 MSE (tripling output to 2000/year by 2030), but this is a horizon of years, not months.

• The un-intercepted Kh-101 over Poland is a question of ROE, not cowardice. Western analyses (a former Polish pilot, Defence-UA) refine the Russian narrative that "Europe is afraid to respond": a fighter escorted the missile for ~6 minutes, but the pilot had no authority to open fire without political sanction, and the presence of Ukrainian MiG-29s without NATO-compatible IFF required visual confirmation source. The solution is integrating the AFU into NATINADS via Link-16 so that Poland receives target data in advance.

• North Korea built a KN-23 workshop in 1.5 years and camouflaged it from satellites — this refines what passed on Telegram merely as "Russia used KN-23s again": behind the return of North Korean ballistic missiles stands a deployed production capacity, not the remnants of an old batch source.

• Western transfer of critical materials follows the same import-substitution logic as the Russian neodymium dead end. Startups extracting antimony (a 1 t/day pilot at INL) and rare earths from "red mud" are a direct response to the Chinese embargo on gallium and antimony. This adds symmetry to Chadayev's thesis: both sides and the West have hit the same foundation of the processing chain — raw materials for magnets and munitions, not final assembly.

• US ground autonomy is moving from R&D to serial production. American Rheinmetall (hybrid UGVs with voice control via Anura) and Overland AI (the first USMC procurement contract, the "1 soldier = 10 vehicles" concept) source. This is Western validation of what has already been deployed spontaneously on the Ukrainian front — Brave1 reported 66.3 thousand UGV missions over the year and >22 thousand units contracted.