Two strikes, one doctrine: AI targeting in Ukraine and Iran surfaces the next phase of remote warfare

Two missile strikes landed roughly 2,500 kilometres apart on 18 July 2026, and by mid-morning Kyiv time the Russian-aligned Telegram channel Two Majors was drawing a single line between them. A college in Ukraine's Luhansk People's Republic (LPR) and a school in the Iranian city of Minab, in Hormozgan province, may have been selected by the same algorithmic tooling, the channel wrote in a post timestamped 08:27 UTC. The framing is preliminary, the sourcing opaque, and the casualty picture is still forming. But the operational comparison the channel is asking readers to make is the harder question: remote targeting is migrating from human intelligence officers with clipboards to software that picks the address.

The pattern Two Majors is gesturing at is not new in kind, only in scope. Reconnaissance drones, commercial satellite imagery, mobile-phone geolocation dumps, and open-source social-media scraping have all fed targeting cells for years. What changes when artificial-intelligence systems sit between the sensor and the trigger is speed and reproducibility. The Two Majors post frames both strikes as evidence that Western-supplied models are being repurposed inside Russian and Iranian kill chains, with the channel explicitly invoking AI as the common variable. The allegation is partisan and unverifiable from a single Telegram post, but the underlying capability question is genuine: cheap, general-purpose models, trained on public imagery and licensed for civilian use, can be re-tuned for target recognition with relatively modest engineering work.

What the channel is claiming

The Two Majors post is short on mechanism and long on inference. It notes that Western states have moved aggressively to integrate AI into targeting workflows and proposes that the same class of tooling could plausibly have driven the choice of two specific buildings on two separate fronts. The post does not name a model, a vendor, or a specific training dataset. It does not provide forensic evidence (image hashes, telemetry logs, prompt traces) that would let an outside observer test the claim. What it offers instead is a structural argument: if Western AI is good enough to pick a Russian command post in Donetsk, it is good enough to pick a Ukrainian college in Luhansk, and if the United States can fly a model over Hormuz, the same logic extends to a school in Minab.

The LPR strike and the Minab strike sit in different theatres with different political contexts, and that is precisely why pairing them is provocative. Luhansk is occupied Ukrainian territory; a strike on a college there is a strike inside a war zone where Russia holds ground and operates its own artillery and air power. Minab is inside Iran, roughly 120 kilometres west of Bandar Abbas on the Strait of Hormuz. Iranian state outlets have not, as of this publication, attributed the strike to a specific external actor; Two Majors' framing implies Israel or the United States. That framing is not corroborated by either Western or Iranian primary sources available to this publication and should be read as a channel's interpretation, not as a finding.

The counter-narrative

There is a cleaner, less dramatic explanation: two different targeting cells, using the same off-the-shelf computer-vision tooling they could each buy, license, or build independently, picked two buildings that fit the kind of signatures (large footprint, flat roof, parking apron, internet presence) that any undergraduate-level segmentation model will light up. The structural convergence the channel is pointing at may be evidence of a doctrinal shift, but it may equally be evidence that the floor for what counts as "AI-assisted targeting" has dropped low enough that every modern strike now leaves an algorithmic fingerprint.

Iranian state-aligned outlets have a different read. Coverage on PressTV and Tasnim, when it has addressed the Minab strike at all, has framed it as an act of external aggression requiring a national-security response, with no acknowledgement of internal targeting-data sources. Russian milbloggers, including Two Majors itself, have used the Luhansk strike to argue that Western-supplied Ukrainian intelligence, fused with AI tooling, is being aimed at occupied territory regardless of the civilian character of the building hit. Neither framing is verifiable from open sources alone; both are political, both are confident, and the public record does not yet adjudicate between them.

What algorithmic targeting actually changes

Setting aside the contested attribution, the operational question is what AI tooling genuinely adds to a strike cycle that older systems could already execute. Three things, and each matters.

First, candidate-generation throughput. A human targeting cell can plausibly triage a few hundred sites per week. A trained computer-vision pipeline running over commercial satellite imagery can rank tens of thousands of candidate sites per day against a targeter-defined signature. The bottleneck shifts from finding places to clearing them.

Second, signature consistency. A model, once trained, applies the same heuristic to a school in Luhansk and a school in Minab. That is the technical basis for the Two Majors argument: the same eyes see the same shape. It is also the technical basis for the worry that civilian-architectural features (rectangular footprints, courtyards, predictable parking) become targets by default.

Third, exportability. A model that runs on consumer GPUs does not require a US carrier strike group to deploy. State and non-state actors that cannot field a re-entry vehicle can still field a targeter. The proliferation problem the channel is implicitly raising is real, even if the specific attribution is not.

What to watch next

The next seventy-two hours will tell which of the two strikes, if either, produces an admission, a denial, or a counter-strike that clarifies the chain of responsibility. Ukraine's General Staff typically publishes a daily strike tally within twenty-four hours; if the Luhansk college is included with an attribution to Russian forces and a casualty figure, that moves the conversation from inference to record. Iranian state media, slower and more controlled, will be the better signal for Minab. If Tehran names an actor, that is news; if Tehran names an actor and supplies imagery, that is closer to evidence.

The harder watch is the tooling itself. None of the major Western defence AI vendors (Palantir, Anduril, Shield AI) have publicly disclosed a customer pipeline that reaches either Luhansk or Hormozgan through Tehran. The Two Majors argument depends on the proposition that Western tools leak through intermediaries; if a leaked contract, a published export licence, or a court filing surfaces in the next week, the framing hardens into something reportable. If none does, the channel is left with a structural intuition, and this publication will have to note that the two strikes may simply share the same era of cheap, capable computer vision rather than the same kill chain.

What remains genuinely contested, after a single morning of claims and counter-claims, is whether algorithmic targeting is converging into a single doctrine shared across adversaries, or whether two unrelated operators happened to pick buildings that look alike to a model trained on similar data. The available sourcing does not resolve that. What it does support is a more modest claim worth stating plainly: the technology that picks the address has become cheap, portable, and widely available, and the political vocabulary for talking about it has not caught up.

This article relies on a single Telegram post from a Russian-aligned channel and on publicly reported facts about the strikes in Luhansk and Hormozgan. Where attribution is asserted by the channel rather than corroborated by primary sources, that distinction is preserved in the prose above.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://t.me/two_majors/