Weekly review, week to 5 July 2026

Key structural shifts

1. The death of the radio-electronic defense paradigm. The key event of the week is not a single strike but the recording, by a whole series of independent sources, of the transition of autonomous strike drones from the experimental segment into serial mass use. The Russian "Molniya" (developed by "Kronstadt") and "Molniya-2" have gone in en masse near Zaporizhzhia with the same internals: a camera + an onboard computer, without a single control antenna. Previously, autonomy had been seen only on the expensive reconnaissance-strike V2U and the "Lancet". Bringing neural-network guidance down to an expendable carrier means the industrialization of a threat against which the first link of classical defense (a radio-emission detector + wideband EW [electronic warfare]) is nullified — there is nothing to jam. Both camps confirm this: both the Ukrainian electronic reconnaissance (ELINT) specialist "Flash" and Russian developers.

2. The fuel-and-logistics war has become a symmetrical strategy of attrition. Over the week, strikes on oil refineries, fuel and lubricant depots and filling stations turned from tactical episodes into a systemic loop on both sides. Ukraine claimed hits on 11 refineries and 7 fuel logistics facilities in June, at ranges of >2000 km (Omsk for the first time, ~2500 km); [TECH] the "Geran", equipped with objective monitoring systems, has moved to a "detect → strike → verify → finish off" scheme. Russia is knocking out >150–300 filling stations in the 150-km zone of the LOC (line of contact). The result is a fuel crisis in Russia with rationing, electronic queues and gasoline imports from India/Kazakhstan. The war has shifted from the plane of "destroy the vehicle" to the plane of "make the adversary's supply unprofitable".

3. The institutionalization of military AI at the NATO level and the reframing of air defense as an economic problem. Palantir Maven has reached full operational capability in the Alliance; the British ASGARD/DECIDE with an update cycle of 8–12 weeks; a revision of the Pentagon's targeting doctrine toward an "AI initiates — human controls" model. In parallel, the PDF framework RAMPART (West Point) has formalized cost asymmetry as a weapon.

Technology dynamics and trends

The vector of moving away from the radio link — along two roads. Besides full autonomy (machine vision + inertial navigation + optical flow), a second line of escape from suppression has been recorded: [TECH] according to "Flash", on the night of July 2 almost all "Shaheds"/"Gerans" over Kyiv operated at 3900–4100 MHz — one floor above the 900/2400/5800 MHz familiar to EW. The third channel — [TECH] the "Geran-4 Seeker" with automatic target lock-on on a Raspberry Pi and terminal guidance via Starlink, which gives immunity to jamming even while controllability is retained. Low-power EW on the target (a locomotive) no longer saves it — hence the appearance of disruptive ("broken") AI camouflage on diesel locomotives on both sides as a countermeasure against machine vision.

The formation of an interceptor drone class and the transition of C-UAS to an automatic loop. Over the week the segment took final shape: the Ukrainian ZIRKA (NOCTIS/Vyriy, >340 km/h, ceiling 6 km, radius 30 km, ~$2000), "Talion", and the returned Bumblebee V2; the Russian "Molniya-P" ([TECH] MSL-20045 radar, FPV up to 8 km), "Rita", "Varvar". A qualitative transition: [TECH] "Katran PVO" ("Lun" radar with AESA [active electronically scanned array] + the "Martyshka" interceptor) operates in a fully automatic mode from detection to engagement. The logic has shifted from an "interceptor drone" to a distributed sensor-and-algorithm short-range air defense system (acoustics + optics + RF sensors + radar + machine vision during the final approach).

Layered defense economics as a consensus framework. [TECH] The RAMPART PDF formalized the "least-cost-first" principle: from the outside in — EW (5–8 km) → directed energy (2–5 km) → interceptor drones (0.3–2 km) → kinetics (reserve). Combat data are given: Iron Beam (100 kW, 4.3 km, $2–3/shot, first combat use, Israel 2025), the British DragonFire (~50 kW, <£10/shot, on the Type 45 by the end of 2027). The same thesis was independently voiced by A. Rogozin (a drone is cheap at the item level and expensive at the system level). The multi-bullet "Mnogotochie" cartridges (STs-226/STs-228, a splitting bullet, up to 300 m, standard AK) have entered serial production — a cheap lower link of the echelon.

Stagnation and degradation: classical post-Soviet air defense is structurally exhausted — with the AFU reaching ~10,000 long-range drones/month, even Moscow has become vulnerable; the economic asymmetry of "cheap drone versus expensive SAM" cannot be solved by adding SAM systems. The AFU's shortage of Patriot missiles has been exposed: the interception rate of ballistic missiles on the night of July 2 fell to 16.7% (4 of 24) versus 44% on June 15 — a direct harbinger of the growing effectiveness of Russian ballistic missiles. The Russian drone industry has exposed a systemic dependence: [TECH] the Ministry of Industry and Trade's points system encourages a "Russian copter on Chinese props" (an STM32H743 does the actual work, while a domestic Baikal-U blinks next to it "for show").

Weak signals (Under the radar)

1. Drones as signals intelligence platforms. [TECH] "Flash" recorded "Gerberas" with an additional fuel tank, without radio modems or cameras, which hold a stable altitude, fly to the target and return without triggering alerts. The hypothesis is operation in RTR (radio-technical reconnaissance) mode: collecting the EW picture/relaying. This is a shift of the expendable kamikaze into a reconnaissance platform; it requires confirmation from wreckage, but it sets a new type of threat.

2. Bypassing EW via the stratosphere and the infrastructure layer. [TECH] The Ukrainian DART — a missile launched from a stratospheric balloon (12–18 km): after descending to ~6 km it switches off navigation, ignites a solid-propellant motor and follows its program — immunity to near-ground jamming. Separately — "BakhmutTelecom": a covert military 3G/4G operator of the AFU (~2500 towers, fiber optics + radio relay links), supporting Zvook/SkyFortress acoustic sensors and control of FPVs/UGVs (unmanned ground robotic systems) at up to 70 km. This is the "nervous system" of the drone war — a priority but almost undiscussed target.

3. Documented Russian resource superiority in AI recognition. [TECH] A comparison of AlissNet-4 (Russia) vs. WUSTCA (China): at F1=0.98, the Russian algorithm requires 256 kB of memory versus 345 MB for the Chinese one (~1000 times less), classifies the entire 300–10,000 MHz band in 5–7 s and works against previously unknown drones. A rare case of documented technological leadership in military AI.

Vector projection and implications

Confirmed dynamics (high confidence):

• A shift in the weight of C-UAS from RF means to sensor-kinetic ones. Against silent autonomous drones, investment will go into electro-optical detection, acoustics (with AI filtering of wind/rain noise), thermal imagers and kinetic interception. Radars remain the most effective all-weather means, but their shortage, high cost and self-exposure through emissions will become the bottleneck of the coming weeks.

• The race to make directed energy cheaper (Iron Beam/DragonFire lasers) as the only answer to cost asymmetry against mass "Shaheds" — at a cost per shot of single-digit dollars, it is the only echelon that breaks the economics of a drone attack.

• The fall wave of Ukrainian drones has been paid for by a targeted EU tranche of €3.9 billion specifically for UAVs — short-range air defense and interceptors need to be prepared for the quantitative jump in attacks right now.

• Growth in the effectiveness of Russian ballistic missiles against the background of the AFU running out of Patriot missiles is almost inevitable without emergency deliveries.

Analytical hypothesis (moderate confidence):

• Autonomy devalues not only EW but also the method of reconnaissance by radio emission — the adversary will have to build sensor fields (acoustics + optics + radar in a single loop with a common time reference), where the bottleneck becomes not the "hardware" but a real-time network architecture without a single point of failure. It is precisely this weakness that the Ukrainian side is methodically exploiting (operation "Polifem": 6 people knocked out ~20 small SKVP radars, opening a corridor for strategic strikes) — the countermeasure lies at the level of architecture and "exchange discipline", not procurement.

• Carriers are losing their tie to infrastructure: Chinese mobile EMALS (electromagnetic catapults) on truck chassis (up to 2.2 t, a plan of 2000 systems/year), USVs (uncrewed boats) as multi-domain hub-carriers of interceptors (Guardian-1) and robotic relays. The next frontier is moving the drone war underwater (the "transparent ocean", Sea Trident), where a "Lancet" will no longer work and autonomous underwater hunters will be needed.