Kremlin Kostiantynivka Assault Force Massacred, Driven Back

Attempting to take a previously unattainable fortress city in Donetsk, the Russians infiltrated a unit of assault troops, then blew up a dam (because who cares about longer term consequences, so why not...?) in order that Ukrainian forces could not reach them. But the Ukrainians were not so easily discouraged, infiltrated their own troops, massacring and ousting the Russians. The lesson, at this point in the war, is that there is little the Russians can do that the Ukrainians cannot undo. JL

Euromaidan Press reports 

On 24 February, the Russian air force lobbed a 3-ton guided bomb at the earthen dam across the Khrushchove Reservoir just west of the fortress city of Kostiantynivka flooding some areas. It was a key supporting effort for a Russian assault on the western edge of the city. The Russian 13th Assault Company infiltrated the Khutir neighborhood in Kostiantynivka. Luckily for the Ukrainians, the Russians were struggling to communicate after Elon Musk's Starlink blocked Russian forces' thousands of terminals. The Ukrainian 100th Mechanized Brigade, garrisoned between the shattered dam and Kostiantynivka, rolled into Khutir and drove out the Russian assault groups. The Ukrainian force counterattack pushed the Russians back. 

Ukraine's Southern Attacks 'Outran Original Scope, Wrecked Russian Plans'

When Ukrainian forces began their initial counterattacks in late January, they were taking advantage of an opportunity caused by the now famous Russian Starlink and Telegram cutoff as well as bad weather that grounded already hobbled Russian drones. 

But the original scope of the attacks was exceeded in a few weeks, permitting the Ukrainians to adapt, and resulting in far greater advances than had been considered possible. Ukraine's attacks, which are continuing at a more deliberate pace, not only gained unexpected new territory, but 'wrecked' Russia's own plans and forced them to react to the Ukrainians, who were then able to dictate both the timing and location of ongoing fighting, contrary to the Kremlin's winter and spring plans. In that regard, it has been an unanticipated double victory. JL

David Axe reports in Trench Art, Yuri Zoria reports in Euromaidan Press:

The Ukrainian push on the Oleksandrivka-Huliaipole axis began in early February, exploiting the collapse of Russian Starlink terminals. By late February, the operation had outrun its original scope and was recapturing villages Russia once firmly held. February was the first month Ukraine recaptured more territory than Russia seized since the 2024 Kursk incursion. Before the operation, Russian forces were actively regrouping and moving in reserves on the Huliaipole axis. Those intentions have been wrecked by the Ukrainian attacks. Russia is now forced to react to what Ukrainian units impose on it, rather than execute its own plans

Don't Trust AI To File Your Taxes, New Research Shows

The more people use AI, the more they learn about what it is really good at - and what it is not so good at. Based on recent research conducted by the New York Times, add doing one's taxes to the latter category.

The issue appears to be, based on these test cases, that AI is good a predicting approximate outcomes. The problem is, as the old saying goes, that 'close enough is only works for (the game of) horseshoes and for hand grenades.' For taxes, precision is crucial. It is fair to assume that AI will get better at this as they absorb more training, probably based on tech companies use, without permission, of your tax filings. JL

Stuart Thompson reports in the New York Times:

The New York Times tested four A.I. - Google’s Gemini, OpenAI’s ChatGPT, Anthropic’s Claude and xAI’s Grok - to see how well they fared with eight tax situations written as training materials by a tax-filing service. The AI miscalculated refunds or amounts owed to the IRS  by an average of $2,000. Even when provided with necessary materials, including all forms to fill out, chatbots whiffed on some calculations. AI do not understand the complex relationships among the information they process - and are exceptionally weak in remembering interconnected information without errors in their responses. Tax software is procedural, following ‘if-then’ logic built for precision.” Large language models, by contrast, are prediction engines that “ fail at tasks that seem simple to humans. With taxes, all those very small details matter."

Data Reveal Putin Is Failing In Ukraine

Putin's intractable negotiating position with the US and Ukraine is based on a bluff and a lie: the bluff is that his faltering forces will 'inevitably' overcome the Ukrainian defenses. The lie is that he will be satisfied with anything less than Ukrainian annihilation. 

The reality is that data reveal early to mid 2025 was Russia's high water mark in Ukraine. Since then, his depleted forces have not only failed to gain any ground of significance, but are now giving up territory to advancing Ukrainian troops. The only question is how long Putin can keep this up before China and North Korea, the suppliers on whom he depends, force him to stop wasting their resources. JL

Frederick and Kimberly Kagan report in the Washington Post:

Since 2022, Russia has seized only 1.5% more Ukrainian land while suffering over 1 million casualties: it has needed 3.5 years to seize 9,318 square kilometers, an area smaller than Los Angeles. Ukraine's forces have not only held Russian advances to a slow crawl, they have begun to push Russian forces back in local counterattacks since last autumn. Ukraine has also embarked on a long-range strike campaign, using both drones and a new domestically produced cruise missile. They are beginning to deploy midrange weapons to disrupt Russian attacks before they reach the line of contact. And they have been able to take advantage of Russian overextensions and missteps to retake ground. Military leaders we spoke to during a recent trip to Ukraine are increasingly confident they can continue to keep the Russians at bay. The Russian negotiating position is based on a bluff and a lie.

Ukrainian Forces Show First Net Gain In Land Retaken Since 2023

Not only are Ukrainian forces advancing, but they have attained a net gain in territory - eg, subtracting Russian gains from Ukrainian, the result is a net gain for the Ukrainians. 

The significance is less in the total amount of territory than in the fact that the Russians no longer appear capable of major strategic breakthroughs. JL

The Kyiv Post reports:

Ukrainian forces liberated more territory than they lost during the last two weeks of February 2026, marking the first net territorial gains since Ukraine’s summer counteroffensive in 2023. Ukrainian forces achieved a net gain of nearly 33 square kilometers between Feb. 14 and Feb. 20, followed by another 57 square kilometers between Feb. 21 and Feb. 27.

In 1 year, Ukraine's $880 MM Drone Line Has Made the Front A Russian Death Trap

A conceptual talking point has become a military reality. Every third Russian soldier now faces the possibility of a drone attack. 30,000 a month become casualties, with Ukraine's goal being to raise that to 50,000. 

This is what executing military strategy looks like. And apparently the US military in the Middle East is now asking for Ukraine's assistance. JL

Olena Mukhina reports in Euromaidan Press:

A year ago, Ukraine launched the "Drone Line" to scale up its UAV units. The initiative allocated $880 million allowing Ukraine to acquire FPV drones, ISR systems, and unmanned aerial bombers on a large scale. Now, the project operates more than 1,000 drone crews who systematically strike Russian targets related to its war. In 2025, Ukraine carried out 371 strikes on Russian territory, inflicting $13 billion in losses on Russia’s oil and gas sector. The project’s goal creates zones where Russian troops cannot move without taking losses, as Ukrainian units detect and destroy targets before they approach Ukrainian positions. At the start of 2026, every third Russian soldier on the front faced UAV attacks.

Anthropic's Claude AI Has Been Central To US Attack On Iran Despite Pentagon "Ban"

The harsh reality for the Pentagon and for a White House attempting to bend the world to its will is that Anthropic has developed a superior product already deeply intertwined with the US military and security services' needs and uses. Replacing it may be possible, but not soon - and certainly not quickly enough for current operations. That probably explains some of the Administration's harsh urgency. 

But as in so many other realms, once the order to attack was given, necessity took priority over theory. Anthropic has already proven its worth. Whether it will ever really be 'banned' is hard to know, but unlikely, except as a piece of political rhetoric. JL

Anna Zhadan reports in Cyber News, Marcus Weisgerber and colleagues report in the Wall Street Journal:

Within hours of declaring that the federal government will end its use of artificial-intelligence tools made by tech company Anthropic, President Trump launched a major air attack in Iran with the help of those very same tools. Commands around the world, including U.S. Central Command in the Middle East, use Anthropic’s Claude AI tool, people familiar with the matter confirmed. Centcom declined to comment about specific systems being used in its ongoing operation against Iran. Claude’s use in high-profile missions, such as the Iran attack and the U.S. military operation that captured Venezuelan President Nicolás Maduro, shows why the administration said it would take six months to phase out the technology