Two Way Street

A Ukrainian brigade spokesman in the Lyman sector described Russian units built around not being seen from the air. Instead of dugouts they construct burrow-like shelters with a single opening, and move under dense vegetation. They mass about ten kilometres behind the line, then break into groups of two or three for the approach, defeating a drone operator's ability to track them. Tempo is part of it: quiet days, then a simultaneous surge. The counter is arithmetic: a commander in the Hulyaipole direction reported the same day that Ukrainian forces are putting up to ten drones on one Russian soldier.

A Kremlin-affiliated milblogger said on 26 August that false reporting by Russian units to their commanders had stopped them clearing their own held ground near Stepnohirsk in mid-spring. ISW files it as the latest of a run of complaints about “beautiful reports”, commanders inflating claimed advances until the high command plans around gains never made, and points to a leaked defence ministry map from April as the result. The complaint comes from a pro-war Russian source about a Russian failure, which is why it carries weight. The same day, the same army claimed a settlement eleven kilometres beyond its lines.

The interceptor drone is the one part of Ukrainian air defence that scaled. Cheap, built at home in the tens of thousands, it took down the bulk of incoming propeller-driven Shaheds through 2025 and gave Ukraine what is now studied elsewhere as an unusually effective short-range air-defence layer. Russia's jet-powered Geran variants broke that model — they fly higher and faster than the interceptors chasing them, and when they arrived in numbers they pushed Ukraine back onto guided interceptor missiles it can no longer reliably obtain. On 14 August, Mykhailo Fedorov announced that the answer to the jet is another jet: Ukraine has begun combat trials of its first domestically developed jet-powered interceptor drones, exceeding 600 kilometres per hour and built specifically to run down jet Shaheds. Several Ukrainian producers now have working designs, developed over the past year through grants, evaluation and testing routed via the Brave1 defence-tech cluster; the first confirmed shoot-downs are still awaited, which is the honest status of the thing. 'The Russians are gradually replacing conventional Shaheds with jet-powered versions,' Fedorov said. 'That marks a shift in air-war strategy. We have to respond quickly so we can keep beating Russia in every new technology cycle.' The arithmetic underneath is why the drone route is taken at all: a Ukrainian-built jet interceptor costs a small fraction of a guided round that must be waited on from a partner capital, and Ukraine's stock of those is close to empty — of roughly 150 Iskander-M ballistic missiles fired at the Kyiv area since 1 July, around seven were shot down. It is also a faster machine than the interceptors already recorded here: the Skazhenyi vertical-takeoff design fielded in early August reaches 365 kilometres per hour, and SkyFall's P1-SUN JetKiller, unveiled in July, 370. The target is not standing still either. Colonel Artem Skurativsky, who heads air defence for Ukraine's 7th Rapid Reaction Corps, says Ukrainian forces were intercepting 80 to 93 percent of Geran-2 and Shahed-136 drones — which is precisely what forced Russia to the jet — and that a Geran-3 covers the distance from the Belarusian border to Kyiv in 35 minutes, four times faster than its predecessors. He also names the trade Russia made to get that speed: the jet drones are larger and therefore easier for radar to see, and their heat and emissions signature leaves them exposed to shoulder-fired missiles. They remain about a fifth of the Gerans in a large strike series. The answer now has a name. Ukraine's Defence Ministry said on 21 August it has developed the Alexa Spatium jet interceptor, capable of taking on the Geran-3, Geran-4 and Geran-5 as well as reconnaissance drones and helicopters. ISW assesses that an effective jet interceptor would be a cost-effective substitute for surface-to-air missiles and MANPADS and would reset the offence-defence race — prospectively; there is still no confirmed shoot-down.

Faced with a shortage of interceptors it cannot buy, Ukraine is trying to answer Russia's ballistic missiles by killing the launchers rather than the missiles. Sources told the Financial Times on 21 August that Kyiv is seeking authorisation to use Starlink-equipped drones against Russian mobile ballistic launchers up to 200 kilometres inside Russia. The capability already exists — Starlink-linked drones fly in the intermediate-range campaign against Russian logistics in occupied Ukraine — so what is sought is geographic permission, not a system. ISW assesses that low-latency satellite coverage over defined segments of Russia would make strikes on mobile launchers accurate enough to matter, and requires no interceptor shipments from Washington at all.

The tactics developed on this front are being taught to a third army. Ukraine's military intelligence deputy head, Major General Vadym Skibitskyi, said on 21 August that 8,500 newly arrived North Korean troops in Kursk Oblast include a 400-person drone unit whose operators could survey and strike targets in Ukraine. North Korean troops already flew FPV and reconnaissance drones alongside Russian forces during their 2025 Kursk deployment. Proximity sharpens it: ISW notes that elements of Rubikon, the Russian unmanned-systems centre this page has tracked through the Pokrovsk and Hulyaipole offensives, have operated in the same Sumy–Kursk border area for a prolonged period.

The combat debut of Russia's Arena-M active protection system in late July was recorded here as evidence that the system fails against FPV drones — geolocated footage showed an Arena-fitted T-72B3A destroyed by a drone that its radar-and-interceptor shield did nothing to stop. A second reading of the same footage now complicates that verdict, and the correction is the point: the failure narrative was built on one vehicle's death and missed what the same engagement showed the system doing. Combat drone specialist Dmytro Putiata, who says he and Foreign Policy Research Institute analyst Rob Lee spoke directly with the Ukrainian unit that fought the action near Maiak in Donetsk Oblast, reports that one Arena-M tank — hull number 133 — shot down seven attacking FPV drones in succession before it was killed, and that across the engagement it took 15 to 20 FPVs to destroy an APS-fitted tank against a baseline of three to six for an unprotected one (a 2024 German GIDS study put the figure at 3-6, a 2025 Swedish FOI study at 4-5). Armour analyst Andrei Tarasenko, reading the same Ukrainian Security Service footage, supplies the mechanism of the eventual kill: tank 133 had already expended its interceptors on the earlier drones, so the shield was empty when the fatal drone arrived. That is the system's structural ceiling rather than a malfunction. Arena carries 26 interceptors fired in pairs for 360-degree cover, and BTVT's Andrei_bt notes the magazine cannot be enlarged — more rounds would block the hatch over the engine and transmission compartment that crews need for maintenance. So the defeat is by arithmetic, the same arithmetic that has beaten every previous tank-protection measure in this war: turtle shells, porcupine spikes and cope cages each raised the number of drones needed and none conferred invulnerability, and operators always found they could throw in a few more drones. The analysts divide on what it means. Lee reads a working shield as 'a partial solution to employing armor again' and expects Arena to improve, which would be a genuine reopening of the maneuver question the drone war had closed; Forbes's David Hambling and Andrei_bt read Round 1 of a contest FPVs are structured to win, because a $500 drone made in weeks can be adapted — with decoys, jamming or stealth — far faster and cheaper than a multi-year, roughly $0.5-1m-per-vehicle APS, and single operators now call down whole carousels of drones through systems like Ukraine's Pasika. What is settled is narrower and worth recording plainly: the APS-versus-FPV evidence is no longer one-directional. The shield intercepted drone after drone in real combat; it simply ran out before the drones did. A week of further analysis has since unsettled the ammunition explanation itself, and the correction is worth as much as the first one. Writing on 17 August, David Axe reports that the fielded system was understood to carry twelve charges rather than the twenty-six of the specification — and that Andrei_bt, one of the sceptics above, rejects the twelve-round figure as not matching modern realities, so neither number is settled and the loadout is now the disputed fact. More importantly, the magazine may not be the binding constraint at all. Ukrainian aviation expert Anatolii Khrapchynskyi puts Arena's engagement range at 20 to 30 metres and its detection range at around 50, which against a drone diving from directly above leaves very little time to react however many charges sit in the launchers — a geometry problem that more rounds would not fix, and which no redesign short of a different sensor would. The Institute for the Study of War, the highest-credibility voice in the set, assesses that Arena-M alone is currently insufficient to protect Russian mechanized assaults because the system can be saturated, while granting that it raises the cost and effort demanded of Ukrainian forces — which is the same finding as the fifteen-to-twenty-drone figure, read as a defence rather than as a promise. Czech analyst Jakub Janovsky supplies the arithmetic from the other end: fifteen to twenty FPVs per tank is 'a perfectly viable rate' at which Russia could commit its entire armoured fleet over a few months and produce, in his phrase, a future low-grade iron ore deposit. What keeps the question open is what Arena does not cost. Turtle, porcupine and hedgehog tanks have absorbed more than twenty drones, but their bolted-on plate and bar is extremely heavy, wrecks transmissions, and stops a tank rotating its turret and aiming its gun — it prevents a tank from fighting like a tank. Arena is bulky but light, so it protects without taking the mobility and firepower away, and that is the version of tank protection that would matter, because Russia has spent three years accumulating rebuilt tanks it cannot risk sending anywhere. Tank-led assaults have been rare for the past year while Russian forces leaned on infantry infiltration; whether they return depends less on whether the shield works than on whether it works long enough to cross the ground where the drones are.

Automatic target recognition keeps appearing on this page as a property of individual drones — the machine-vision package fitted to the Gerbera and Geran family, the Russian claim of a drone that picks out a uniform. What Ukraine's Ministry of Defence made public on 10 August is the layer underneath, and it explains why the capability spreads the way it does. The ministry opened Avengers Labs, built by the Centre for Innovation and Development of Defence Technologies — the team behind the DELTA battlefield-awareness system — to Ukrainian defence manufacturers, giving them access to an annotated dataset of five million frames collected in actual combat and continuously topped up. The labels cover what the war is actually made of: tanks, artillery, air-defence systems, infantry, and aerial objects including Shahed-type and reconnaissance drones. The ministry's argument for opening it is the mechanism: a dataset of that size and quality is effectively impossible for a private company to assemble on its own, so the state supplies the scarce input and the market supplies the models. It also reports what the dataset already buys. The automatic detection system trained on it runs inside DELTA, analyses more than 100,000 drone video streams a month, and identifies 70 percent of all enemy targets in real time, by day and by night. That figure is the ministry's own and is not independently measured; it should be read as a claim by an interested party rather than an audit. The structural point survives the caveat. Detection quality in this war is a function of labelled combat footage, and the side that pools its footage rather than leaving it inside individual units compounds an advantage that no single manufacturer can buy — which is the same logic, from the other direction, as Russia hunting the drone crews and ground stations that generate the footage in the first place.

The tactics this page tracks were built against Russia, but their reach is clearest where they were turned on a Western army that trains for a different war. Reporting first carried by the Wall Street Journal and unpacked by the Kyiv Post's Stefan Korshak records that at NATO's Combined Resolve field exercise at Hohenfels, Germany, in the second half of April 2026, a small contingent from Ukraine's 412th Unmanned Systems Brigade 'Nemesis' — reportedly fewer than twenty pilots and technicians — playing the opposing force against the US 3rd Armored Brigade Combat Team of the 1st Cavalry Division rendered the American armored force combat-ineffective in at least one scenario, to the point that judges had to 'bring back to life' destroyed vehicles and troops or the wargame would have halted for lack of surviving US units. The mechanism was the one used against Russia for two years and, per the reports, identical to it: recon drones located the advancing Americans partly by their dust plumes, FPV crews scored hits on tanks and infantry fighting vehicles 'with relative ease', and follow-up bomber drones hunted down the vehicles evaluated as damaged and stopped, then the infantry and light vehicles that took cover once their armour was gone. The Americans had standard electronic-warfare and counter-drone systems and, at least in theory, full access to US satellite, signals and reconnaissance intelligence; it did not save the brigade. What makes it a doctrinal data point rather than one bad day is the repetition. The same defeat had already happened twice: at NATO's Hedgehog 2025 exercise in Estonia, where about ten Ukrainian operators notionally destroyed at least seventeen British and Estonian vehicles, including Challenger 2 tanks and Warrior IFVs, and bottled the survivors up on woodland roads; and at Aurora 26 in Gotland, Sweden, running almost simultaneously with Hohenfels, where Ukrainian FPV crews 'destroyed' 28 of 32 vehicles attempting an advance and, according to one account, found their jamming-proof fibre-optic drones unnecessary because the NATO units were unwilling or unable to jam at all. Korshak's flat reading is the finding: across three exercises spanning eleven months, the conventional maneuver tactics that failed against drones the first time were unchanged and failed again. The qualifier belongs on it too — the Hohenfels rout came in only one of several scenarios, weapons effects are simulated with laser-tag and blanks, and the Ukrainian contingents were among the most experienced drone units in any army. But the transferability is the point: the recon-strike-drone combination is not a quirk of the Russian target, and a peer Western brigade fighting the way it was trained to fight met the same fate as a Russian mechanised column. Militarnyi's account of the same Wall Street Journal reporting adds the detail that qualifies the flattest version of that reading. About 3,500 American service members took part across April and May, and a US Army representative says the units rotating through improved as the exercise went on — dispersing more effectively, camouflaging better and making more use of their electronic-warfare systems. Midway through, the Ukrainian operators changed sides and worked with the Americans on employing unmanned systems in the attack. So the honest finding is narrower and more useful than 'NATO cannot adapt': conventional manoeuvre as practised failed repeatedly against drone crews, and the corrections that began to work were learned inside the exercise, from the people who had already learned them under fire.

The oldest working counter on this page is the jammer, and it has now been beaten a third distinct way. A Russian FPV drone destroyed by Ukrainian forces was found to be operating radio control on frequency-hopping spread spectrum, and the Institute for the Study of War assessed on 14 August that this is the latest Russian adaptation built specifically to defeat Ukrainian electronic warfare. The mechanism is simple and hard to answer: FHSS switches the control link rapidly across frequency bands, so a jammer that finds and floods the channel is flooding a channel the drone has already left. A jam-resistant link buys range, and the drone in question was capable of flying up to 40 kilometres in depth — well beyond the reach an FPV had when the link was a single fixed frequency. What makes this a doctrinal entry rather than a recovered gadget is the succession ISW places it in. Russia's first answer to jamming was the fibre-optic drone in autumn 2024, which cannot be jammed at all because it trails a physical wire; Ukraine's answer to the unjammable drone was the interceptor, the aircraft-versus-aircraft solution recorded across this page all summer. FHSS is the next move, and it goes back to the radio link rather than around it — a cheap upgrade to the ordinary FPV rather than a new class of aircraft. Its cost is measured in kill zone. ISW puts the current saturated band, in which any person or vehicle is at elevated risk from tactical strike and reconnaissance drones, at 15 to 25 kilometres from the line of contact, and assesses that FHSS-enabled drones will push it outward. That is the real currency of this contest: each side's adaptations are converted not into ground taken but into the depth at which nothing can safely stand still.

A drone is cheap and there are hundreds of thousands of them; the things that fly it are not. Over the middle of August both armies visibly shifted effort off the aircraft and onto the fixed installations behind them — the ground control stations where the pilots sit, the warehouses where airframes are stored, and the signal repeaters that extend a drone's reach past the horizon. On 17 August the Institute for the Study of War recorded the Russian side of it twice in its own voice: Russian tactical aviation is now striking Ukrainian drone infrastructure close to the border, and Geran drones are hitting it deeper in. The Russian Defence Ministry claimed strikes with a Kh-39 air-to-surface missile and two FAB-1500 guided glide bombs against two Ukrainian drone ground control centres near Hryhorivka and Prykolotne in northern Kharkiv Oblast, roughly nine and 20 kilometres from the border; Geran drones against a store of quadcopters near Havrylivka on the Kharkiv-Sloviansk highway; and a drone warehouse near Dmytrivka in Dnipropetrovsk Oblast, some 50 kilometres from the front. Ukraine is doing the same thing at four times the range. On the night of 16 to 17 August its General Staff reported striking a Russian drone repeater and its power supply on the Tendra Spit, about 43 kilometres out; a storage, preparation and launch facility for Russian strike drones near Hvardiiske in occupied Crimea, 190 kilometres out; an Orion drone ground control station at Novofedorivka, 185; and a ground repeater guiding Geran- and Gerbera-type drones at Olenivka, 135. Two Russian drone control points near Chasiv Yar and occupied Luhove had been hit on 14 August. The logic is the same one that pushed sniping into drone piloting and pushed the pilots themselves a thousand kilometres to the rear: an aircraft that costs a few hundred dollars is not worth a glide bomb, but the station that flies twenty of them is, and unlike the aircraft it does not move. It is also why the pilots' physical dispersal has limits — the repeaters and launch rails that give a cheap drone its range are infrastructure, they sit at fixed coordinates, and both sides now spend some of their scarcest munitions on them.

Russia's Rassvet constellation was recorded here on 1 August as operational in windows over Ukraine, and the Institute for the Study of War assessed on 17 August that Russia may be preparing to use it from its new border launch sites to guide more precise strikes against Ukraine's rear and its air defences. The question that entry left open was what Ukraine could do about it, and the specialist answer, from Militarnyi's Dmytro Shumlianskyi, is that at present it has nothing — for a reason that is physical rather than industrial. Rassvet's ground terminals transmit upward in two bands allocated by the Russian regulator: 14 to 14.495 gigahertz in the Ku band, and 29.5 to 30 in the Ka band. Most publicly shown Ukrainian anti-drone jammers work at frequencies up to 6 gigahertz, in some cases 9, and there is no open-source evidence of a Ukrainian mass-produced system that can radiate a directed beam at 14 or 30 while tracking a satellite. Nor is the shortfall only a matter of tuning higher. Everything Ukraine has built jams a dome: a field of interference thrown around a protected object, catching whatever receiver flies into it. Jamming a constellation is the opposite problem — the receiver is aboard a satellite 500 to 550 kilometres up, moving fast across the sky, so the station must form a narrow beam, point it precisely and track the satellite to the horizon, then hand off to the next one as the constellation fills in. Each step of that is harder in the Ka band, where free-space loss at 30 gigahertz is nearly five times that at 14, high-power amplifiers are costlier and less efficient, and rain and wet snow degrade the link. And the signal's own structure — bandwidth, polarisation, modulation, frequency hopping, on-board beamforming — is unknown, so blanketing the roughly 500-megahertz band would take enormous power while jamming individual channels first requires intercepting and analysing them. The sharpest constraint is that the target sits on top of Ukraine's own lifeline. Rassvet's Ku uplink overlaps a Starlink uplink channel at 14 to 14.5 gigahertz, and its Ka allocation falls inside Starlink's authorised uplink spectrum — and most Ukrainian medium-range strike missions are flown over Starlink. Any Ukrainian jammer would have to fall silent whenever a Rassvet satellite passed near a Starlink one, and control its antenna side lobes, or it would degrade the network its own drones fly on. Russia has already demonstrated both the method and its cost: its Volna Kupol Garant system jams the Starlink uplink toward the satellite rather than the user's terminal, reportedly disrupting service across about 20 square kilometres per unit at 150 million roubles apiece, and its powerful, distinctive emissions are detected by SpaceX immediately and located by allied radio-reconnaissance satellites — which makes the jammer a target.

The machine-vision package recorded across the Shahed family in late July gave a drone the ability to finish an attack on a designated target through a severed command link. A Russian claim now pushes the same idea one step earlier in the kill chain — to letting the drone choose the target. A Russian battalion commander with the callsign 'Balu', from the unmanned-systems battalion of the 71st Separate Regiment in the 'Zapad' grouping, told the state newswire TASS on 2 August that on the Krasny Lyman direction Russian forces are striking Ukrainian troops with 'flying wing'-type drones carrying artificial intelligence that orients on Ukrainian military uniforms and vehicle number plates. The Hudson Institute's 5 August situation report records the same thing from the other side of the front as a pro-Kremlin claim — that Russian drones now possess AI capabilities that can recognise and target combat uniforms. What would make this a doctrinal step rather than a product boast is the shift it describes: from a human designating a target and the machine holding track, to the machine identifying the target class on its own from a visual signature. The evidentiary caveat is load-bearing and points the other way. Both accounts are Russian in origin — one a named commander relayed by a state outlet that functions as a conduit for official framing, the other logged by Hudson explicitly as a pro-Kremlin claim — and there is no Ukrainian or independent confirmation that uniform-specific automatic target recognition is being exercised in the field, as opposed to being trained, tested or advertised. It is recorded here as a datable claim about where Russian targeting autonomy is being pushed, not as a demonstrated capability; the contested question of whether machine autonomy is genuinely exercised at all, already noted on this page from the Spider's Web dispute, applies with full force.

The limit on Ukrainian fighter aviation recorded at the end of July was not the aircraft or the missiles but the paperwork: RUSI assessed that 'land-centric targeting and mission allocation processes' had limited the effectiveness of Ukrainian fighters against time-sensitive targets — the fleeting radar or launcher that appears and is gone before a centrally-approved mission can be re-tasked to it. That constraint is now being loosened. The New York Times reported on 5 August, citing Ukrainian commanders, that Air Force pilots now have authorisation to break off from their planned mission targets and strike a higher-priority target of opportunity — such as a Russian radar or launcher — when they detect a window, rather than waiting for centralised approval. Breaking Defense, relaying the report, sets it alongside the same shift described from the ground: a US officer who served with the Security Assistance Group–Ukraine, Captain Daine Van de Wall, wrote in May that Ukraine's most significant adaptation is not the technology but how far down the chain authority has moved — brigade-level-and-below formations routinely absorbing air defence, fires and other functions once reserved for higher headquarters, because they cannot assume air superiority, spectrum dominance or suppression of enemy air defences before they move. The mechanism is the delegation of the decision to the person who can see the target, and its significance is that it answers a limitation this record had already named rather than adding a new machine. The primary account is the New York Times report, which is unreachable to a direct request; the substance here is carried by Breaking Defense's relay of it, which supplies the RUSI and Van de Wall framing directly.

The fiber-optic FPV — guided down an unspooling hair-thin cable rather than a radio link — is the one drone that trench-level jamming cannot touch, and until now the tactical record noted it as electronic-warfare-immune with no counter attached to it. Kharkiv has begun trialing the first counter aimed squarely at that drone, and it is physical rather than electronic. Oleh Syniehubov, head of the Kharkiv Regional Military Administration, describes a three-tiered urban 'dome': an outer layer of active countermeasures and interceptors; a middle passive layer of anti-drone netting strung along the Kharkiv Ring Road and the Kharkiv Highway; and an inner layer of mobile fire groups armed with shotguns and automated turrets to kill whatever penetrates. Officials say roughly 100 fiber-optic FPVs reach the city each day, and that the system adapts proven frontline anti-drone methods to urban defence. What is recorded here is a trial of a mechanism, not a demonstrated kill rate: netting and gun crews are a way to defeat a drone that cannot be jammed, which is why it matters, but no independent assessment of its effectiveness yet exists. The primary account comes from Radio Khartia, the media arm of the National Guard's 'Khartia' formation — a military-affiliated outlet — relayed by United24 Media.

The 'race to build a faster interceptor' recorded on 25 July against jet-powered Shaheds now has a concrete new entrant. Ukraine's UV Military Cluster, working with Odesa air-defence personnel, has fielded the Skazhenyi ('rabid') vertical-takeoff interceptor, entering combat in August. Militarnyi reports a wingless variant reaching 365 km/h and a winged one 310 km/h; a standard operating altitude of 200 m but the ability to climb to 10,000 m; frequency-hopping control across 100–3,020 MHz to resist jamming; a range of about 45 km; and endurance of up to 30 minutes. Its makers say it draws on experience from the Sting, P1-SUN and Strila interceptor projects and claim a production capacity of up to 5,000 units a month — a manufacturer's figure, not verified output, and recorded here as a claim. The vertical-takeoff design and 10,000 m ceiling place it in the high-speed, high-altitude interceptor class the counter-Shahed contest has been pushing toward: fast enough to chase jet variants, high enough to reach drones that fly above the short-range gun belts.

Writing for GIS Reports, front-line analyst Paul Schwennesen argues that the decisive tactical evolution of the war is no longer any single platform but the way the platforms are now flown together. Ukraine, on his reading, orchestrates three drone tiers as one system: short-range FPVs shaping the close fight, medium-range fixed-wing drones interdicting logistics and command nodes in the near rear, and long-range strike drones — the Lyutyi and the FP-1/2/5 family — reaching strategic infrastructure deep inside Russia. Together, he assesses, they perform what doctrine calls battlefield air interdiction — striking targets behind the front to shape the near-term fight — despite Ukraine never having held air superiority, reversing the operational-depth advantage Russia enjoyed earlier in the war. The claim is doctrinal framing rather than a new capability: the individual strikes it describes are already recorded, and what Schwennesen adds is the reading that their integration, not their individual novelty, is the innovation that matters. It is one analyst's assessment, offered here as such.

GIS Reports, quoting counter-drone specialist Benjamin Remler, records that Ukraine's interceptor ecosystem — scaled up sharply since January 2026 — has significantly dulled Russia's long-standing advantage against slow-moving Shahed, Lancet and Molniya drones. Remler describes a period in which Russia now holds, in his words, 'a net disadvantage in the tactical and strategic theaters with regard to strike assets' — a reversal of the pattern that held for most of the war. This corroborates, from a second analytical shop, the RUSI finding recorded on 28 July that Ukraine's cheap short-range belts and thousands of interceptor drones now destroy most of the Geran and Gerbera drones brought down; where RUSI reached that conclusion through the arithmetic of massed short-range fire, GIS reaches a compatible one through the interceptor-drone build-up. The convergence is what is worth noting: two shops, different reasoning, the same reading that the slow-drone edge Russia held has narrowed. The 'net disadvantage' characterisation is a datable assessment attributed to a named specialist, and is recorded as such rather than as established fact.

The American Enterprise Institute confirmed that drones in Ukraine's Operation Spiderweb engaged autonomously with critical parts of Russian aircraft based on models trained on museum aircraft.

Ukraine introduced the STASH short-range air-defense system, mounting an AGM-114L Longbow Hellfire missile and tactical radar on a truck to provide a mobile hard-kill option against fast jet-powered Shahed variants.

Ukrainian intelligence confirmed that 180 Russian troops received training in China on drone operations, electronic warfare, and counter-drone tactics under a July 2025 agreement.

The first batch of 12 Russian Rassvet communications satellites completed deployment, providing Russian forces with connectivity windows over Ukraine to operate heavy strike and reconnaissance drones deep behind Ukrainian lines.

Through the first half of 2026 Russian assault formations were driven down by drone-saturated defences from platoon-sized attacks through squad and fire-team to two-person infiltration teams, because concentrations of vehicles could not survive the approach. On 22 July Russian forces reversed that, briefly. Ukrainian National Guard Commander Brigadier General Oleksandr Pivnenko confirmed on 25 July that elements of the Russian 8th Combined Arms Army and 41st Combined Arms Army — including the 41st's 90th Tank Division — conducted a reinforced company-sized mechanised assault in the Dobropillya direction, advancing from Novoekonomichne and Vozdvyzhenka and attempting to enter settlements near Shakhove over roughly four hours. It was defeated the same way mass armour was defeated the first time: not by anti-tank guns or minefields but by drone crews watching the approaches, with the Phoenix border unit of Ukraine's Pomsta brigade named as the operators. The counter held, and the doctrine it had displaced was not restored — ISW records Russian forces continuing offensive operations in the Dobropillya tactical area on 23, 24, 25 and 26 July and not advancing on any of the four days. The scale of the destruction is less firmly established than the assault itself, and the distinction matters: geolocated footage published by Ukrainian units on 23 July independently shows strikes on a Russian tank and an armoured personnel carrier north of Novotoretske, a tank at Shakhove and a BM-21 Grad at Nove Paltavka, but the fuller tally of nine tanks, two armoured fighting vehicles and three Grad launchers comes from a Ukrainian State Border Guard Service release, reported by Ukrainska Pravda on 24 July, which ISW noted the same day 'appears to have since been deleted' from the agency's website — and which returns HTTP 404 to a direct request on 27 July, with no archived snapshot available. Ukraine's border guards said there had not been a breakthrough attempt of this scale in their area of responsibility for a long time; that is a local characterisation from an interested party, not a front-wide count. A week later Vikram Mittal published a frame-by-frame reading of the released Ukrainian footage in Forbes, and it turns the episode from an outcome into a method. The Russian vehicles advanced widely dispersed rather than in dense columns — spacing that lowers the chance of the whole formation being identified and struck at once, and stops one destroyed vehicle creating a bottleneck, while keeping the vehicles close enough to hold communications. Most of the tanks carried no extensive anti-drone armour: no unravelled steel-cable screens, only a small canopy with a thin steel plate over the commander's hatch, likely for standoff protection against munitions dropped by heavy bomber drones. Mittal reads the omission as a deliberate trade — heavier modifications cost speed and manoeuvrability, and Russian crews have reported transmission failures after relatively short movements — an assessment, with the crew reports relayed rather than confirmed. The armour was not the whole force. Numerous two-person motorcycle teams operated independently of it, with further infantry in commercial trucks, and the advance was preceded by Gerbera and Lancet drones attempting to suppress Ukrainian drone operations. Most motorcycle teams did not try to outrun the drones; they dismounted quickly and dispersed into nearby vegetation, a drill previously shown in a Russian Ministry of Defence post about Russian training. Because Ukrainian drones concentrated on the vehicles, many of those soldiers kept advancing on foot. The defeat mechanism was layered and pre-emplaced rather than improvised. Reconnaissance drones detected the advancing force and directed a rapid response; the Border Guard Phoenix unit employed FPV drones, heavy bomber drones and artillery; minefields had been laid in advance along the likely approaches and the bridges toward Ukrainian positions had already been damaged; and when Russian forces tried to restore the crossings with pontoon bridges, Ukrainian tactical aviation destroyed them. The drone and artillery element is confirmed by released footage; the pontoon-bridge destruction rests on Ukrainian accounts. On scale, Mittal corrects the claimant: the First Corps Azov described the attack as involving 'two armies and a tank division', but 'the assault element visible in the released footage appears much smaller, likely a reinforced company-sized force' — while still, in his judgement, among the largest Russian mechanised assaults of the past year. Parent formations and force size are different claims, and Mittal does not dispute Pivnenko's identification of the units. The equipment tallies remain unsettled: Azov reported seven tanks, six armoured fighting vehicles, one BM-21 Grad, one Tornado-S launcher, three pieces of specialised equipment and 47 light vehicles destroyed with more than 150 Russian casualties, while Ukrinform, relayed by Mick Ryan, reported tanks and a Sarma launcher destroyed for a claimed 123 Russian dead. Both are Ukrainian claims, neither is independently confirmed, and they disagree on the casualty figure and on the launcher type. What the episode was an instance of is now clearer. ISW states that Russian forces 'continue to rely on their aviation overmatch into 2026, even as Russian infantry tactics have switched from larger-scale infantry attacks and mechanized assaults to mostly small group infiltrations', and the scale reported since suggests what the phrase now denotes in practice: the State Border Service reported a reduced platoon-sized mechanised assault near Artilne on 29 July consisting of one armoured vehicle and ten infantry, and a Ukrainian unmanned systems servicemember in the Slovyansk direction reported on 28 July that Russian forces rarely conduct large-scale mechanised assaults and mostly operate in groups of one to two. Mittal expects a limited number of mechanised assaults to continue alongside infiltration as the primary method: infiltration yields only incremental gains but maintains constant pressure, whereas a mechanised assault offers the possibility, though not the guarantee, of a breakthrough — and Russian commanders have incentive to keep experimenting while Russian leadership demands offensive operations rather than a transition to defence. He offers three non-exclusive explanations for why this assault was mounted at all — that enough infiltration sites had been established to support it, that Ukraine's counter-infiltration effort had drawn drone reconnaissance and rapid-response teams away, or that the armour was a diversion covering the motorcycle teams — and settles on none. ISW does not describe 22 July as the start of a trend, and hedges its own reference to it twice.

The Australian Strategic Policy Institute published an account by David Kirichenko of what persistent drone observation has done to special forces work, and it describes a trade being retired rather than adapted. By 2025, Kirichenko assesses, Ukraine had established a 'drone wall' in which virtually every part of the front is watched by uncrewed aircraft; manoeuvre warfare becomes harder and sniping becomes dangerous for a specific reason — a sniper who becomes active is now far easier to locate, and once located can be killed by drones or artillery. US special forces operators who arrived in Ukraine in 2022 largely worked as snipers; in recent years those roles have shifted away from sniping under the drone threat. A former US operator writing under the pseudonym 'Xen', who worked with a Ukrainian special forces regiment, puts it as a complete substitution: 'Every single SOF sniper I trained is now a drone pilot of one type or another,' and 'the small independent teams who defined themselves as hunters who could hit the enemy with relative impunity from 1 to 2 km now do so from a range of 30 to 300 km.' That is a pseudonymous participant describing units he personally trained, and 'every single' is a personal sample rather than a survey — it is recorded here as relayed testimony, not as a measurement. The consequent adaptations are the more checkable part. Teams in ghillie suits operating behind enemy lines are resupplied by heavy bomber drones, because constant surveillance leaves such narrow windows to move in the open that infiltrating teams must carry less and move fast; Rob Lee of the Foreign Policy Research Institute confirms the practice from outside the account, saying 'some Ukrainian assault regiments and SSO battalions have adapted by employing heavy bomber drones in direct support of small infantry or SSO teams.' Trench assaults now require a drone overhead at all times to confirm the ground is clear before the team advances. And neutralising Russian infiltrators who accumulate in a position may require special forces to go in on foot, which exposes exactly the small teams the drone threat has already displaced. Kirichenko also records a Ukrainian 3rd Separate Special Purpose Regiment pattern of sending three teams to a threatened sector — one training and equipping frontline infantry, one in reserve, one flying drones — which a participant credits with holding Pokrovsk; that is an outcome claim from an interested party and is not adopted here.

The cheapest counter Ukraine found to Russia's one-way attack drones was another drone — an interceptor costing a fraction of a surface-to-air missile — and Russia is now attacking the arithmetic that makes it work. Russia has begun fitting Shahed drones with jet engines in place of propellers, raising their speed to as much as 500 km/h and compressing the window in which Ukrainian air defences can detect and intercept them. Yurii Cherevashenko, a senior commander in Ukraine's Air Force, told Reuters at the Farnborough International Airshow that 15 to 20 percent of the Shaheds Russia currently uses already carry jet engines. The Ukrainian and allied response is a straightforward speed race: SkyFall unveiled the P1-SUN Jetkiller at 370 km/h against 310 km/h for its standard P1-SUN, General Cherry is finalising a Bullet variant exceeding 400 km/h with a stated goal of 500, and the British-Ukrainian Firebolt Engineering's jet-powered Griffen exceeds 350 km/h. The production and kill figures attached to these systems are manufacturers' claims made at a trade show — the outlets carrying them note that such claims are difficult to verify independently — so what is recorded here is the speed contest, not the volumes. Independent of the vendors, there is dated field evidence that interception is working somewhere: the commander of a Ukrainian drone interceptor battery in the Oleksandrivka direction reported on 25 July that interception has made Russian Shahed, Geran and Gerbera drones less effective on that axis, and that Russian forces are conducting guided glide bomb strikes to compensate. That substitution is the part worth watching. Oleksandrivka now carries a two-week sequence of the same loop running to completion: Ukrainian jamming suppressed Russian drone tempo there on 23 July, interception degraded Russian drone effectiveness by 25 July, and the response was not a better drone but a different weapon. An independent account of the same shift arrived at the end of the month, and it supplies the number that explains why speed works. The Hudson Institute records that the jet-powered Geran-3 'complicates the efforts of air-defense systems optimized to guard against piston-powered Gerans flying at roughly 112 miles per hour', while the Geran-5 reportedly reaches about 373 mph on a turbojet of roughly 440 lb thrust. The mismatch, rather than the top speed on its own, is the mechanism: an interception envelope tuned for one speed band meeting targets flying at more than three times it.

Drone warfare has been improvisation almost everywhere it appears in this war — volunteer-built airframes, workshop modifications, units inventing counters in weeks. Russia is now converting it into establishment. Russian Navy Commander-in-Chief Admiral Alexander Moiseyev stated on 26 July that the Navy formed its first unmanned systems regiment within the Northern Fleet on 1 July 2026, that further such regiments are planned in the Northern Fleet during 2026, and that a similar structure will be incorporated into the Pacific Fleet in 2027. These naval regiments are separate from the dedicated Unmanned Systems Forces that Russia began forming in 2024 to centralise drone operations in support of its ground forces; ISW assesses the extension as evidence of 'the way the Russian military writ large is trying to integrate unmanned systems throughout the wider force'. The same assessment supplies an unusually direct answer to the question of whether a counter is working. Against Ukraine's strikes on vessels and ports in the Black and Azov seas, ISW records that Russian forces have relocated and dispersed ships, installed anti-drone netting, and introduced mobile task force units and aviation patrols — 'but have largely failed to defend against and adapt to Ukrainian strikes'. A named counter-measure suite with an institutional verdict attached to it is rare, and the verdict here is negative. It also sits awkwardly beside RUSI's assessment four days earlier that Russian adaptation is 'terribly slow at the organisational' level: creating regiments is organisational change. The two readings are not strictly incompatible — RUSI's claim concerns command culture and the delegation of initiative, ISW's concerns force structure, and a rigid hierarchy can stand up new units readily while still declining to devolve decisions to them. Whether these formations are given decision authority is what would settle it, and neither document answers that. Ukraine did the same thing three weeks earlier, and to a different part of the problem. Two force-design reforms were created by presidential decree on 10 July 2026. A Long-Range Strike Command stands up no new units at all; it pulls existing long-range strike drone units, missile forces and naval unmanned systems units under a single structure operating to a unified concept and plan. The Joint Rapid Reaction Forces is a new branch combining assault troops with organic drone and artillery support, led by Brigadier General Dmytro Voloshyn, and its stated mission is explicitly not to hold ground: 'their mission will not be to permanently hold designated sections of the front, but to respond quickly to threats, reinforce troop groupings, conduct offensive and counteroffensive operations along the main axis of attack or other priority directions, stabilize the situation, and rapidly seize key terrain and strategic facilities.' Mick Ryan, relaying the decrees, a General Staff statement of 14 July and Ukrinform, reads the pair as formalising 'under permanent structures, the strike and manoeuvre innovations that had until now been run as semi-improvised task forces' — a branch designed around the drone-saturated battlefield rather than one adapting to it. Set the two side by side and the shape of the month is a convergence. Within four weeks Russia stood up the Navy's first unmanned systems regiment, on 1 July, with more planned, and Ukraine created a strike command and a new branch, on 10 July; Ryan notes that Ukraine 'reorganised its assault forces and stood up a new strike command in a fortnight'. That sits awkwardly against RUSI's assessment four days before the Russian announcement that Russian learning is 'terribly slow at the organisational' level — though not incompatibly, since RUSI's claim is about Russian command culture and concerns Russia alone, while Ryan's is about Ukrainian force design. Neither is adjudicated here. What the two have in common is the direction of travel: capabilities invented in workshops and volunteer units are being given establishment, budget and a chain of command.

A RUSI update on the air war by Justin Bronk breaks Ukraine's air defence down by layer and reaches a counterintuitive verdict: the cheap layer does most of the killing. Ukraine's deep belts of short-range air defence — vehicle-mounted machine guns, crewed and automated turret emplacements, Gepard and Shilka self-propelled guns, Stinger and Igla man-portable missiles, and many thousands of interceptor drones — 'now destroy a majority of the Russian Geran and Gerbera OWA drones shot down'. The mechanism is arithmetic rather than system performance: any one such weapon covers only a small area, so 'the odds of a given Russian Geran-2 or Gerbera OWA coming within their effective range on their way to targets deeper inside Ukraine is very low', and massing teams in depth is what converts individually improbable engagements into majority attrition. The trade-off RUSI names is that the equipment and ammunition are cheap and the operation sustainable, but the approach is personnel-intensive at scale. Why the cheap layer is doing the work is a supply story rather than a doctrinal preference. RUSI assesses the belts exist to offset, not supplement, serious Ukrainian shortages of surface-to-air and air-to-air missiles, freeing scarce interceptors for the cruise and ballistic missiles that guns cannot reliably engage; the primary constraint on Ukrainian coverage is ammunition resupply rather than launcher availability, with PAC-2 GEM-T and PAC-3 MSE in short supply worldwide after very large-scale expenditure by Gulf states and US forces during Operation Epic Fury — a war in another theatre setting a ceiling on this one. The same arithmetic has already pushed fighters out of the role. Ukrainian F-16s alone have destroyed well over 2,500 Geran and Gerbera drones, but RUSI assesses the practice as economically unsustainable, having consumed an unsustainable number of airframe hours and AIM-9M/L Sidewinder and Magic 2 missiles since mid-2024, and records that F-16 pilots now use the 20-mm cannon and US-supplied AGR-20 rockets, also known as APKWS II, 'to save missiles for more valuable targets like cruise missiles wherever possible'. Against a Russian industry producing more than 6,000 one-way attack drones a month, RUSI's efficient mix is short-range belts near the border and the front plus helicopters and volunteer-flown fixed-wing gunships with machine guns — anything but fast jets, which have finite airframe hours before deep maintenance. One further mechanism connects Ukraine's deep-strike campaign to conditions at the front. RUSI assesses that campaign is straining Russian air defence in a self-reinforcing way: finite ground-based systems are pulled ever deeper inside Russia and their ammunition stocks strained, 'particularly among the SHORAD systems Russia generally prefers for counter-UAV interceptions such as the SA-22 Pantsir and SA-15 Tor' — with the frontline consequence that more Ukrainian loitering munitions get through against resupply and battlefield logistics, and Ukrainian fighters find more room to run glide-bomb and cruise-missile attacks on hardened targets tens of kilometres behind the lines. RUSI also scopes to one domain the command limitation recorded from its 22 July commentary: Ukrainian fighters launch GBU-39 small diameter bombs, GBU-62 JDAM and AASM Hammer weapons out to around 60 km behind Russian lines, but 'land-centric targeting and mission allocation processes have so far limited the effectiveness of Ukrainian fighters against time-sensitive targets'. Western fighters now form the core of the Ukrainian Air Force — the 107th Wing on F-16 plus a smaller Mirage 2000-5 force — with the remaining MiG-29s and a few Su-27s flying air defence, firing AGM-88 HARM against Russian ground-based air defence, and conducting glide-bomb attacks. One dating caveat belongs on all of it: RUSI's page carries no publication date, and the text places itself in 'summer 2026' and refers to events of late July, so it is recorded at the end of the month it describes.

Ukraine's two working counters to Russian one-way attack drones are jamming and interception. Russia's answer to both is now a single modular package, and it is spreading across a family of airframes rather than appearing on one. The Hudson Institute's Can Kasapoglu records that Russia's 'Seeker' suite fits machine vision — built, per Ukrainian open-source analysis, on Raspberry Pi single-board computers — to its Gerbera and Geran drones, giving automatic target recognition, acquisition and terminal guidance: once an operator has designated a target, the drone can hold track and complete its attack even if the command link is disrupted in the terminal phase. Hudson's account of the effect is narrow and worth taking as written — the innovation 'reduces the effectiveness of low-power electronic warfare countermeasures that an opposing force might launch'. 'Low-power' scopes that to man-portable and vehicle-mounted jammers, the trench-level layer, rather than to the whole electronic-warfare contest. What makes it a doctrinal question rather than a product announcement is the layering and the standardisation. The Gerbera Seeker adds an inertial navigation system and a 12-element jam-resistant Kometa-M satellite-navigation antenna, already fielded on Geran-2s, to the baseline Gerbera airframe, flight controller and servomechanisms, with operator control running over a mesh-network radio modem alongside multiple electro-optical cameras. That is three answers to three different attacks: the antenna resists satellite-navigation jamming and spoofing, inertial navigation covers outright denial, and machine vision covers a severed command link — with mesh networking a fourth measure aimed at link resilience. Guidance, communications and navigation components have been standardised across the camera-equipped Gerbera Seeker and Geran Seeker; the package can reportedly be integrated into the Geran-2, Geran-3 and Geran-4; and recovered Geran-5 debris and datalink analysis suggest the mesh technology already on some piston-powered Geran-2s is being carried into the jet variants. One counter-jamming guidance package moving across a whole family is a different order of problem from one new drone. Interception is being attacked in the same movement, by speed. Hudson assesses that the proliferation of Ukrainian interceptor drones has accelerated Russian losses of its more expensive Orlan, ZALA and Supercam reconnaissance drones and Lancet loitering munitions, pushing part of the reconnaissance-strike mission onto cheaper machine-vision platforms such as the Gerbera Seeker; and that the jet-powered Geran-3 'complicates the efforts of air-defense systems optimized to guard against piston-powered Gerans flying at roughly 112 miles per hour', with the Geran-4 and Geran-5 abandoning the delta wing for a small cruise-missile configuration that can be air-launched from Su-25 aircraft, the Geran-5 reportedly reaching about 373 mph on a turbojet of roughly 440 lb thrust. The mismatch is the mechanism: an interception envelope tuned for one speed band meeting targets flying at more than three times it. Two Ukrainian reports support the direction and neither confirms the capability. The commander of an unmanned systems battalion in the Kramatorsk direction reported that Russian command understands Ukrainian FPV and strike drone activity threatens its ability to mount large-scale assaults, and is responding by attempting to fit Molniya fixed-wing FPV drones with machine vision and AI-enabled targeting so they can find targets independently, and by testing jet-propelled Shahed drones because Ukrainian interception of propeller-driven strike drones is becoming increasingly effective. 'Attempting' and 'testing' are the report's own words: in development, not confirmed in widespread use. Hudson's component detail traces to the Ukrainian open-source channel serhii_flash and to Militarnyi, and the debris and datalink analysis to the IISS in January 2026; none of it is independently confirmed. And whether machine autonomy is genuinely being exercised at all is contested in the specialist literature. Writing about Ukraine's Operation Spider's Web, AEI states it 'reportedly involved drones that engaged autonomously with critical parts of Russian strategic bombers and other aircraft based on models trained on aircraft in Ukrainian museums', while the Lieber Institute at West Point writes that the same operation was 'reportedly flown by individual pilots, one per drone, making it a striking feat of coordination rather than of machine autonomy'. Both hedge with 'reportedly'. The two accounts are not strictly irreconcilable — a human could fly the run while terminal aimpoint selection is automated — but neither source says so, and neither is adopted here.

Two analytical shops reached the same domain from opposite directions at the end of July, and between them they explain why Russian aviation keeps producing ground movement while Russian infantry tactics shrink. RUSI records that Russia's Su-34 fleet delivers hundreds of heavy UMPK and UMPB glide bombs a week against frontline positions and near-front cities — a practice introduced at scale from autumn 2023 — systematically demolishing strongpoints and defensive positions, bombarding cities near the front, and targeting concealed drone operator teams. ISW, working from the ground up, states that Russian forces are 'widely using glide bomb strikes against Kostyantynivka and other cities in Ukraine's Fortress Belt in Donetsk Oblast to turn the cities into rubble, physically destroy Ukrainian strongpoints and ground control stations, and allow Russian infiltrators to enter and eventually seize and consolidate positions', with the rubble then used to conceal the infantry. That two shops independently name drone crews and their ground control stations as a glide-bomb target set is the more useful convergence: it puts the aviation campaign and the hunt for drone operators in the same sentence. ISW also supplies the lineage, which is what makes this a doctrine rather than a munition. Russia debuted glide bombs at scale to enable a ground advance with the seizure of Avdiivka in February 2024, and used the same combination at Vuhledar in October 2024, Kurakhove in December 2024 and Toretsk in August 2025, continuing to rely on that aviation overmatch into 2026 even as the infantry underneath it changed from mechanised assault to small-group infiltration. ISW glosses the category as battlefield air interdiction — 'the use of airpower to strike targets in the near rear behind the frontline to impact battlefield operations in the near term' — which Russia has leveraged glide bombs for throughout 2025 and 2026, particularly against Ukrainian ground lines of communication. A dated instance from the same window: the head of the Slovyansk military administration, Vadym Lyakh, reported a Russian FAB-500 glide-bomb strike against warehouses and civilian infrastructure at Myrne, just west of Slovyansk. The two shops diverge on the counter, and the disagreement is the substance. RUSI's reading is kinematic and pessimistic: the AIM-120 AMRAAM carried by Ukrainian F-16s 'cannot effectively reach the Su-34s releasing their weapons at high altitudes and speeds 70 km or so behind the Russian lines', and what offers hope of changing that is the arrival of Swedish Gripen C/D with the long-range Meteor missile in early 2027 — that is, not this year. ISW's is indirect and already under way: it assesses Russia's reliance on glide bombs to be an exploitable vulnerability, and records that 'Ukraine has already significantly degraded Russia's ground-based air defense network in sectors of the theater through a sustained suppression and destruction of enemy air defenses campaign (SEAD / DEAD) since late 2025'. Two qualifications belong on that. 'In sectors of the theater' is not theatre-wide, and ISW's further reasoning — that degrading Russian combat aviation would likely slow Russian advances — is a forecast rather than an observed effect. Ukraine's suppression campaign has been visible for months without a specialist assessment attached to it; ISW has now supplied one. It also inverts the opening phase of the war, in which RUSI records the Russian Aerospace Forces as the side whose suppression campaign failed, leaving most of its tactical fighters unable to affect the ground war and pushing Russia into long-range cruise and ballistic missile strikes for two years.

Hunting the operator rather than the airframe has been visible in scattered incidents for months. It is now generalised by a specialist shop and answered by a counter-measure that moves the pilot out of the theatre altogether. AEI's John G. Ferrari and Dillon Prochnicki state the condition plainly — 'Ukrainian and Russian forces have used drones to extend their reach, but the operators and links those drones rely on have become prime targets' — and tie it to the link rather than only the person: because most drones remain piloted rather than autonomous, the command-and-control link has itself been drawn into the fight, the visible result being the blanketing of the front line with fibre-optic cable. That is one sentence of authorial assessment with no mechanism, figures or citation attached to it, and it is recorded as the generalisation it is rather than as evidence. The field evidence is mirror-imaged. Geolocated footage published on 28 July shows a Ukrainian unmanned systems battalion dropping munitions on Russian drone operators at Yablunivka, northeast of Sumy City. A Russian milblogger separately claimed Russian strikes on Ukrainian drone ground control stations at Mykolaivka, Oleksiievo-Druzhkivka, Kindratkivka and Druzhkivka along the H-20 Kostyantynivka-Druzhkivka highway, which ISW assesses are likely continuing Russian attempts to suppress Ukrainian drone activity around Kostyantynivka. Only the Ukrainian strike is geolocated and the Russian locations rest on a milblogger's claim — but the assessment that each side is hunting the other's crews is ISW's own. The counter is to break the link between the pilot and the place. Ukrainian interceptor crews are increasingly flying over satellite connections from the rear rather than line-of-sight radio, which makes the requirement a stable internet connection rather than proximity to the aircraft; the maker Wild Hornets says its interceptors can be flown from hundreds of kilometres away, potentially from another country, and frames the capability as protecting pilots who are high-priority Russian targets. That rationale is the manufacturer's own. What makes it more than a vendor claim is where the exposure goes instead: the launch crew does not move. Troops still transport, unload and assemble the drones, attach the explosives and communications equipment, and position them for launch. Dmytro 'Liber' Zhluktenko, a former drone operator and lessons-learned analyst with Ukraine's 413th Unmanned Systems Regiment 'RAID', says satellite control lets a pilot sit anywhere with a stable connection but 'doesn't remove the need for the ground groups, people who actually attach the explosive, put the drone outside of the bunker', and that protecting everyone involved by keeping them all far from the fighting is 'impossible so far'; he adds that units are moving operators further back for safety even at the expense of capability. Handing the launch step to ground robots acting as motherships is under way and not yet ubiquitous along the front, and on more conventional radio-frequency missions the operator must still stay close to the drone. The vulnerable node moved. It did not disappear.

Following the successful Ukrainian counteroffensive in Kharkiv, Russian forces begin establishing deep, multi-layered defensive fortifications across southern and eastern Ukraine, prioritizing defense-in-depth to prevent future breakthroughs. Contemporaneous evidence of the turn comes from the east rather than the south: satellite imagery from Planet Labs, analysed by ABC News in October 2022, showed a fortified line under construction from just south of Russian-held Popasna in Luhansk Oblast — two rows of pyramid-shaped concrete blocks followed by a vehicle-trapping ditch near Hirske — with a planned length of some 120 miles, and a Janes analyst assessed the trenches were 'considerably larger and more sophisticated than most Russian trench lines seen throughout the war'. Russian media called that eastern line the Wagner Line; the name Surovikin Line attaches properly to the southern belts built across Zaporizhzhia from late 2022, which are the ones Ukraine's 2023 counteroffensive met. The September 2022 start date recorded here remains approximate and is not yet tied to a contemporaneous primary source.

ISW documented three separate instances in this window of Russian sources publishing footage that does not show what it purports to show, and generalised the practice as a cognitive-warfare method for propagating claims of advances. On July 22 the Russian Ministry of Defense claimed the seizure of Blahodatne, southwest of Oleksandrivka, and published footage of a servicemember waving a flag there; ISW assesses the footage was filmed at a different location and edited to splice in overhead drone imagery of Blahodatne, and that Russian forces have only infiltrated to within 11 kilometres of the settlement — making not just seizure but recent entry unlikely. Geolocated flag-raising footage from Hrafske, northeast of Kharkiv City, published July 20, is footage ISW has reason to suspect was AI-altered; a Russian milblogger claimed the settlement's capture on July 22, and ISW recorded no confirmed advance there. And footage posted by the Kremlin newswire TASS purporting to show strikes on Ukrainian positions in Oleksiyevo-Druzhkivka was identified by a geolocator as filmed instead at Kostyantynivka and Kleban-Byk, roughly 10 kilometres from the current frontline, with some of it dating from around August 2025. The mechanism is what makes this a tactic rather than three errors: the claim of ground is made in imagery rather than in a communiqué, so it must be refuted frame by frame — and the refutation, unlike the claim, takes days of geolocation work. Three days later the practice was described from the Ukrainian side as a planned task rather than a by-product of capture. Ukrainian Joint Forces Task Force Spokesperson Colonel Viktor Trehubov stated on 25 July that Russian forces had claimed to seize Bilyi Kolodyaz in northern Kharkiv Oblast but that only isolated Russian infiltrators reached the settlement and raised a flag there; ISW's own framing of the episode is that Russian forces 'previously conducted a flag-raising mission for informational effects' — the flag-raising treated as the objective of the mission rather than as the marking of a captured place. Trehubov added that Russian forces are increasingly attacking in small groups in northern Kharkiv Oblast, which is the manoeuvre form such a mission requires: infiltration by a handful of soldiers is enough to produce the footage, and produces nothing else.

A RUSI commentary by Fabrizio Minniti and Giangiuseppe Pili offers the frame this page has been missing — an account of why the tactical adaptations recorded on this timeline keep failing to add up to anything. Its thesis is that Russian learning is real but selective: 'What distinguishes the Russian case is not adaptation per se, but its pattern; that is, rapid at the tactical and technological level, terribly slow at the organisational.' The mechanism RUSI proposes is a throughput mismatch. Since late 2024 Russia has built what the authors call a distributed strike complex — drones, long-range fires and real-time reconnaissance in place of manned airpower — and its reach has outrun the command system meant to direct it: 'the ability to detect and strike targets at extended range has outpaced the system's ability to prioritise and synchronise effects across domains,' so coordination across land, air, electronic warfare and cyber 'remains uneven and unequally distributed across the lines' and localised tactical success does not convert into operational results. The authors are explicit that they attribute this to Russia's own architecture rather than to Ukrainian resistance, and they name two structural causes: the absence of a professional non-commissioned officer corps, which leaves the lieutenant as the lowest professionally trained leader and pushes senior officers toward the front to keep control, and what they call 'the corruption of information, not merely of money' — reporting tailored upward, logistics padded to conceal shortfalls, and personnel records that diverge from combat reality, in a system that punishes honest bad news. RUSI also revises the record in two places. It reads Russia's turn to artillery dominance and drone saturation as an operational logic of 'saturation, not surgical strike' and as 'adaptation from constraint as much as from insight'; and it argues the 2022-era Western consensus, while accurate about the visible failures, 'obscured the system's capacity to absorb losses, replace formations, and adjust tactics'. A Ukrainian brigade spokesperson in the Pokrovsk direction reported on 24 July that Russian commanders abandon communication with assault units when those units come under Ukrainian drone fire — a field report that matches RUSI's argument closely, though a Ukrainian unit describing Russian command behaviour is weak evidence for it, and it is recorded here as a claim rather than as confirmation.

Two independent analysts generalised Ukraine's wave of strikes on Wildberries warehouses as a shift in targeting doctrine rather than a run of individual raids. Economist Sergei Guriev, dean of London Business School, told Meduza the campaign is 'a double blow – to the military machine and to the image of normality': 'This is an attack that touches every Russian because people increasingly depend on e-commerce. The Kremlin has spent years trying to convince Russians that life continues as normal despite the war. It's difficult to maintain that illusion when huge columns of smoke are rising over cities.' Chris Weafer, CEO of the consultancy Macro-Advisory, reached the same doctrinal conclusion from the opposite direction, assessing the economic damage as 'relatively slight' against a stagnant but stable economy and locating the effect in morale instead: 'There is now much more awareness, but there's also much more discussion, and more people are questioning what's going on and why is it happening and the conflict lasting so long. It brings it home.' The mechanism they describe rests on a structural fact Weafer supplies: Russian e-commerce grew into the gap Western brands left after 2022, and is now worth roughly $150 billion with online retail $80–90 billion of that, about 20% of all retail sales — a load-bearing civilian network that did not exist at this scale before the war, distributed across more than 200 Wildberries facilities and therefore impossible to defend against cheap drones. The two analysts do not agree on the military half: Guriev keeps it, Weafer largely discounts it. A counter-indicator is on the record too — a Moscow shopper told the Associated Press she would keep using the service even if prices rise. The strikes themselves are recorded on the main topic timeline; what is recorded here is the doctrine.

Analysts assess Russia has cycled through assault-tactic adaptations to beat Ukraine's drone-saturated defenses — from platoon-sized attacks down through squad and fire-team to two-person infiltration teams, and separately to 'hybrid mechanized' assaults mixing motorcycles and ATVs with armor. In documented cases (a ~50-soldier, 28-motorcycle assault near Sloviansk; two-person teams near Vovchansk) Ukrainian reconnaissance-and-strike drone teams detected and destroyed the attackers before they reached the front. The mechanism: roads and open approaches are persistently watched by recon drones, and strike drones on standby swarm any detected assault. Unable to convert costly assaults into held ground, Russia is fortifying its lines and leaning on information operations — claiming captures (e.g. Yurkivka, Novyi Donbas, Kostiantynivka in mid-June) that ISW finds little evidence for.

Ukraine conducts a precision strike on the Voronezh Semiconductor Devices Plant using newly acquired, mass-producible AGM-188A Rusty Dagger cruise missiles. The tactic relies on arithmetic saturation—using swarms of cheap ($250k) missiles to overwhelm Russian point defenses like Pantsir-S1, enabling deep strikes against logistical centers.

In a masterclass of multi-domain electronic warfare and sensor fusion, Ukrainian F-16s successfully engage and down a superior Russian Su-35S fighter over eastern Ukraine. Operating in strict radar silence to mimic a glide-bomb run, the F-16s lured the Su-35S into a trap, utilizing offboard targeting telemetry from a Patriot battery or ASC 890 AWACS via Link-16 data links to secure the kill without revealing their position.

Ukraine officially integrates Western-supplied F-16 fighter jets into its air force. Due to intense Russian ground-based air defenses and superior Russian radar capabilities (like the Su-35), the F-16s are primarily employed defensively to intercept cruise missiles and preserve ground-based SAM systems, rather than conducting deep offensive strikes.

Due to the extreme vulnerability of heavy armor to cheap, ubiquitous Russian FPV drones in a highly transparent battlefield, Ukraine temporarily withdraws its U.S.-supplied M1A1 Abrams tanks to reset tactics and add protective 'cope cage' modifications.

Russian forces, heavily utilizing Wagner PMC units, normalize the use of relentless, high-attrition infantry waves to overwhelm Ukrainian defenders in urban environments, accepting massive casualties for marginal gains.

Ukraine launches its highly anticipated summer counteroffensive but immediately struggles against dense Russian minefields and pre-sighted artillery zones, validating the effectiveness of the deep defensive network.

Both Ukrainian and Russian frontline units report that cheap, commercial FPV drones have surpassed traditional artillery as the primary cause of casualties and vehicle losses, leading to the rapid deployment of anti-drone nets and 'cope cages'.

The proliferation of trench-level jamming devices forces both sides to innovate rapidly, moving towards frequency-hopping radios, fiber-optic wire-guided drones, and AI-assisted terminal guidance to bypass electronic interference.

The first US-supplied M142 HIMARS precision rocket systems arrived in Ukraine, and within weeks Ukraine used HIMARS and UK M270 GMLRS to strike Russian ammunition depots and command posts 80-120km behind the front. The mechanism was decisive: Russia held a roughly 3:1 artillery-fires advantage, but its offensive depended on large, predictable rear ammunition dumps — precision GMLRS strikes on those bottlenecks degraded Russian firepower far out of proportion to the number of launchers, and set the conditions for Ukraine's late-2022 southern counteroffensive.