Ukraine’s new Marichka AI targets the one layer of its war stack that Palantir still owns

• July 26, 2026

• Ukraine Central Research Institute has unveiled Marichka, a closed neural network intended to cut brigade-level planning decisions from six to 12 hours down to minutes
• The system calculates force ratios, estimates losses, and models courses of action on an interactive map, but has so far only been trialed on test data, with classified document integration still pending
• Marichka adds another layer to Ukraine’s tech stack for battlefield planning: the country already fields Delta for situational awareness and works with Palantir at multiple levels

Ukraine’s Central Research Institute of the Armed Forces has presented Marichka, a closed neural network designed to compress the operational planning cycle at brigade level and above from hours to minutes.

The system, still in testing, is built to collect and verify battlefield information, then calculate force ratios and estimate required resources and likely losses, presenting commanders with modeled options on an interactive map. Military standards currently allow a regiment or brigade commander six to 12 hours and roughly 20 documents to reach the same point.

It can also alert command posts to changes in the operational situation as new information arrives. Colonel Yurii Kliat, who heads the institute, was explicit about where authority stops, telling the briefing that the final decision remains with the commander.

A battlefield-layer replacement for Palantir

Palantir has been embedded in Ukrainian operations since shortly after Russia’s full-scale invasion in 2022, and it remains a mainstay in that theater, with a scope that dwarfs what Marichka is attempting.

Marichka is still an important step. If it works as described, it would give Ukraine sovereign control over part of the reasoning layer that currently sits with a foreign vendor, even as Palantir’s pipeline remains the more established presence in the country’s software-driven war effort.

According to Colonel Yurii Kliat, head of Ukraine’s Central Research Institute of the Armed Forces, Marichka centrally gathers battlefield information, automatically calculates the balance of forces and combat potentials, and estimates required resources and likely losses.

The intent is to let brigade and regiment commanders reach an informed decision in minutes, measured against a doctrinal standard that currently allows six to 12 hours and roughly 20 documents.

Ukraine already fields the system that most closely answers to the description of a homegrown Palantir: Delta, which has provided situational awareness across the defense forces since 2022. Marichka is a new layer proposed on top of a stack that already includes Delta, Palantir’s platforms for fusion and analysis, and, as of this year, a dedicated Defense AI Center. Whether it will integrate with Delta or run separately has not been disclosed, and that is arguably the most important open question about the program.

Brigadier General Andrii Lebedenko told the Drone Autonomy 2026 conference in May that the Armed Forces had already begun integrating AI agents into command-and-control systems, which places Marichka within an existing effort rather than presenting it as a departure.

This is not Ukraine’s first attempt at an operational-level command system for headquarters. Dzvin was ordered from a contractor in 2016 and frozen in 2021 after the prototypes encountered disputes over cost, timelines, and product ownership.

Delta, developed in parallel rather than as a replacement, succeeded partly because it grew out of a volunteer reconnaissance unit solving an immediate battlefield problem, rather than from a procurement process.

Marichka comes from a research institute, which is a different pedigree again, and the harder test will not be the neural network but the bureaucracy around accreditation for classified data. That hurdle may be reinforced by the argument for the incumbent: Palantir has a proven record in theaters of war, and its early work in Ukraine, much of it provided free, secured a first-mover advantage that a domestic prototype will have to overcome.

Via United24Media

Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.

Ukraine Central Research Institute has unveiled Marichka, a closed neural network intended to cut brigade-level planning decisions from six to 12 hours down to minutes The system calculates force ratios, estimates losses, and models courses of action on an interactive map, but has so far only been trialed on test…