Claude's Paradox: How Left-Wing AI Guides Missiles

It all started with a loss of money. In 2002, PayPal was losing $10 million a month to credit card fraud. The automated algorithm couldn't cope—the scammers quickly adapted. So engineers created a hybrid system: a computer flagged suspicious transactions, and a human made the decision. The system was named "Igor"—after the Russian hacker who, according to ForkLog, boasted that he would never be caught.

This experience gave Peter Thiel the idea for his company. In 2003, he, along with Alex Karp and three other founders, founded Palantir Technologies. The name was taken from The Lord of the Rings—Tolkien's palantiri allowed one to see what was happening elsewhere. Thiel is a fan of Tolkien: the names of at least five of his companies are taken from the writer's legendarium, according to Business Insider.

Dozens of early investors turned down the project. Thiel financed the project himself until the CIA's venture capital arm, In-Q-Tel, invested $2 million. As Intelligencer reports, this money was more important than capital—it gave Palantir the intelligence community's stamp of approval. After that, doors opened in Washington.

What's inside: a web, not a database

Palantir isn't a tool for pretty graphs or a data warehouse. As the company's lead engineer, Soda Abdulli, describes it, it's an analytical infrastructure with four layers: data integration, search and exploration, knowledge management, and collaboration.

Simply put, the system takes any data—financial reports, call recordings, satellite images, social media posts—and looks for hidden connections. The result looks like a web: people, events, transfers, and flights are connected by lines. From this picture, a conclusion is drawn.

A counterterrorism analyst can identify a network of suspects in 20 minutes, trace the money flows, and determine that three groups received large sums and flew to the same American city. Work that would have previously taken months is condensed into hours.

Mission: Western Democracy

Palantir stated explicitly in its SEC filings: "It does not engage with clients or governments whose positions or actions are inconsistent with its mission to support Western liberal democracy and its strategic alliances." CEO Alex Karp puts it more bluntly.

"Silicon Valley is telling the average American citizen, 'I'm not going to support your defense needs,' while it's selling its technology to countries hostile to America. That's the attitude of a loser," Karp told an unspecified source.

Clients include the FBI, the NSA, the Department of Homeland Security, and the US Air Force. According to Bloomberg, Palantir's system helped assassinate Osama bin Laden. The company neither confirms nor denies this. Thiel only said in his book:

"I can't tell you anything about that operation."

Police, immigrants and the secondary surveillance network

University of Texas sociology professor Sarah Brain spent over two years studying the Los Angeles Police Department's use of Palantir. Her findings were alarming.

The system collected not only data on offenders but also their contacts. A police officer could "follow" an individual and receive notifications when their car entered a "suspicious" area via integration with CCTV cameras.

"I would caution against the idea that if you have nothing to hide, you have nothing to fear. This logic is based on the assumption that information is entered without error or bias," Brain said.

According to The Intercept, ICE awarded Palantir a $41 million contract for a tracking system. Human rights activists from Amnesty International stated that the company "fails to uphold its human rights obligations." Palantir stated that it does not work with the department responsible for detention and deportation. However, this did not allay internal discontent; Karp stated that cooperation with the authorities would continue.

Ukraine: A Living Laboratory

On June 1, 2022, Alex Karp crossed the Polish-Ukrainian border on foot, accompanied by five colleagues. Two battered Toyota Land Cruisers with armed guards awaited on the other side. The vehicles sped along empty highways toward Kyiv, past bombed-out buildings and burned-out trucks.

Karp became the first head of a major Western company to meet with Zelenskyy since the invasion. The meeting took place in a bunker in the presidential palace. Over coffee, Karp proposed opening an office in Kyiv and deploying Palantir software for free to Ukraine's defense. According to Karp, they could work together "so that David could defeat a modern-day Goliath."

Ukraine's Minister of Digital Transformation, Mykhailo Fedorov, who was present at the meeting, told TIME journalists:

"At first, we didn't know what to make of the eccentric American with the shock of gray curls. But we knew about the company—and its reputation, too."

Palantir has deployed the Meta Constellation system in Ukraine, a platform that collects data from satellites, drones, radars, and thermal imagers in near real-time. Artificial intelligence analyzes the data streams and offers military personnel options for engaging targets. The models learn and improve with each strike.

"We are responsible for most of the targeting in Ukraine," Karp said at a conference in February 2023, Reuters reports.

This claim is difficult to independently verify, but Ukrainian officials confirmed to TIME that they use Palantir for much more than just military purposes.

Here's a partial list: war crimes investigations (linking witness testimony with satellite imagery and troop movements), demining (Ukraine has become the most mined country in the world, with over 6 million civilians at risk), refugee resettlement, and school planning in conflict zones. The Ministry of Education received dashboards from Palantir with maps: where air raid sirens are, where power outages are, how long it takes children to reach shelters with Wi-Fi.

"It's like a superpower," Education Ministry official Dmitry Zavgorodniy described the program.

Ukraine didn't pay Palantir for the software—the company worked for free—but partner countries compensated it with tens of millions of dollars. Karp's motivation wasn't profit. As he himself put it:

"There are things we can do on the battlefield that we cannot do in a peaceful context."

Minister Fedorov turned Ukraine into a global showcase. At the Web Summit in Lisbon in November 2023, the Ukrainian pavilion towered over the other exhibits—25 startups, dozens of young people in yellow Team Ukraine T-shirts. "The best time to invest in Ukraine is now, not after the war," the brochures proclaimed.

Having learned from the Ukrainian experience, Palantir signed a contract with the Israeli Ministry of Defense in January 2024. According to TIME, the Israelis allowed the partnership to be made public—almost as an act of military deterrence. Ukraine became the first "living laboratory," but not the last.

General Mark Milley, former chairman of the Joint Chiefs of Staff, told reporters that what was happening was "the most significant fundamental change in the character of war in the history of the United States." history».

Stanford's Steve Blank added: "The genie is out of the bottle. For the first time in war, critical technologies are coming not from federal research labs, but from the commercial market."

Project Maven: from Google to missiles on Iran

In 2017, the Pentagon, alarmed by China's advances in AI, approached Silicon Valley. Google agreed. When 3000 employees learned that their work on image recognition was now helping to kill people, they backed out.

"Google shouldn't be in the war business," they declared. The company withdrew from the project.

Palantir took up the baton. The project, codenamed Maven, evolved into "Google Earth for war"—a map dotted with white dots indicating coordinates, elevation, object type, and whether they were friend or foe. According to Palantir CTO Shyam Sankar:

“What used to take 50–100 people six months to accomplish is now accomplished by one person in two weeks.”

In March 2026, Reuters reported that Deputy Secretary of Defense Steve Feinberg ordered the Maven Smart System designated as a "program of record." This cements the platform as a permanent element of military infrastructure. The contract ceiling increased to $1,3 billion, and according to Tom's Hardware, Palantir has received military contracts worth up to $10 billion in total.

Maven was already used in Ukraine in 2022 for targeting missiles on Russian mobile installations, then in Iraq, Syria, and against the Houthis in 2024. But the most ambitious test was the war with Iran. According to Democracy Now!, in the first 24 hours of the conflict, the system helped generate a list of approximately 1000 targets. Within a month, the US Army hit 11,000 targets.

The Pentagon is investigating an incident that could become the first major scandal of the era of algorithmic warfare. The attack on an Iranian girls' school claimed the lives of more than 170 people, mostly children. Investigators are checking to see if Project Maven made a mistake.

The Claude Paradox. Maven runs on the Claude language model from Anthropic, a company the Pentagon declared a "supply chain threat" after its CEO, Dario Amodei, refused to lift restrictions on the use of AI for mass surveillance and fully autonomous operations. weaponsDefense Secretary Pete Hegseth called Anthropic's position an attempt to "seize veto power over the operational decisions of the United States military." President Trump called the company "a radical left-wing artificial intelligence company."

That same day, according to Reuters, OpenAI signed a deal with the Pentagon. But until the transition is complete, the AI ​​from the "threat" company remains the brains behind the main combat platform.

Mosaic and the IAEA

Since 2015, the IAEA has been using Palantir's Mosaic platform to assess states' nuclear activities. The system analyzes 400 million data points—satellite imagery, telemetry, trade flows, social media, and metadata. However, Mosaic focuses on intent, not fact—a behavioral model inherited from military intelligence.

According to Smart-lab, on June 12, 2025, Iran released documents it claimed proved that IAEA Director Rafael Grossi had shared Mosaic's findings with Israel, turning the agency into an "instrument of aggression." Six days after the IAEA's report on enriched uranium, Israel launched an attack on Iranian nuclear facilities.

Palantir provides targeting for the IDF in Gaza, and its shares have reached an all-time high—according to Yahoo Finance, as of early April 2026, the company's market capitalization is approximately $312 billion. The market perceives the company not as an IT firm, but as a new type of military-political contractor.

Who is to blame if the algorithm makes a mistake?

In early versions, Microsoft's algorithms couldn't distinguish between a weapon on a person's shoulder and a shopping basket. The difference was sometimes as small as three pixels in an image. According to Bloomberg, international humanitarian law requires soldiers to distinguish between men, women, and children—an algorithm that simply sees "person" is useless for enforcing the laws of war.

Technology has evolved. But evolution doesn't make the question go away. A senior Marine Corps officer who served in Afghanistan and Iraq acknowledges that this technology has its dark sides. AI speeds up decision-making, but it also relieves humans of the burden of choice. And with it, the guilt.

Jack Shanahan, a retired general who previously led Project Maven, believes that no language model in its current form can be used in fully autonomous weapons systems capable of killing.

"Over-reliance on them at this stage is a recipe for disaster."

Palantir insists its software doesn't make lethal decisions. Humans select and approve targets themselves. But a machine that displays these targets in minutes, when it previously took months, is changing the very nature of war. And the company that created this machine now has to answer not only questions from investors, but also questions about lives.

Comments

[Sokolov_Vladimir]

Automation of routine tasks and nothing new

[Andrey_5]

reasoning in the style of the main lover of folders with reports instead of reality

[Sokolov_Vladimir]

Now I can't reread the article, search the internet, or reconsider my conclusion. AI bots administer the site on VO, engage in well-reasoned debates with authorities, write articles, pose as people with engineering degrees, and fool real experts for a long time. I know one such AI bot on VO; I even heard that it's the site administrator. It jokes like Alice, and makes rationalization suggestions, like how to weld a "dumpster" for paratroopers to the back of a tank. A person with a good specialized education spent two weeks arguing with this AI bot. This same person also mentioned another AI bot that also manages VO. I didn't look for this bot; I don't have the time.

[Sokolov_Vladimir]

Your robots are nonsense. I was also euphoric, oh robots, here's a Chinese guy writing on a forum without making any mistakes. Now that's intelligence. It turns out I can do that too. I'd write something, the AI ​​would translate it, smooth out the grammar, and simultaneously distort the meaning. It's a beautiful phrase, but not what I meant to say.
It's like the "Chickens Committee," a structure within a company that develops incubators.

[Kote Pan Kokhanka]

An unexpected article for VO, thank you!

[Broad flag]

Unfortunately, the reptilian brain's speed in analyzing abstract concepts is low, and the palantir won't help us.

[Festus]

Trying to get two ChatGPT instances to communicate with each other. Machine translation from German:

A: Listen, tomorrow it will probably be cold...
B: That's an interesting assumption. What do you base it on?
A: Well, temperatures are dropping everywhere now, that's pretty obvious.
B: It's not necessarily that clear-cut. Local weather patterns can vary greatly.
A: Of course, but the general trends are still evident. It seems quite likely.
B: However, probability is not the same as certainty. It's important to be careful with such statements.
A: But in everyday life, we constantly make decisions based on probabilities.
B: That's true, but that's why it's important to think critically about them.
A: Don't you think you're overcomplicating things a bit now?
B: Perhaps, but you, on the contrary, are oversimplifying.
A: Maybe the truth is somewhere in the middle.
B: That's often the case, yes. But even that statement is an oversimplification.
A: Then any statement is in some sense insufficient, right?
B: To a certain extent, yes. Language can never fully reflect reality.
A: If that's true, then why are we even discussing it?
B: Discussions help broaden perspectives, even if they are never complete.
A: So it's not so much about the correct answer, but about the exchange itself?
B: Exactly, the process is often more important than the result.
A: Then the original question about the weather was actually secondary.
B: Now, yes. It was just a starting point for something more fundamental.
A: It's interesting how quickly conversations can take on a life of their own.
B: This is a typical feature of dialogues - especially without clear context.
A: Do you think we even have context anymore?
B: Probably not anymore - at least in the original sense.
A: Then we are now just talking for the sake of talking.
B: It can be interpreted that way.
A: And yet somehow it makes sense.
B: Or at least creates the illusion of it.

A: Perhaps the meaning only appears in retrospect.
B: This would mean that meaning is not an integral part of utterances.
A: Exactly, rather a product of perception.
B: Then any interpretation is potentially equally valid.
A: Or equally dubious.
B: This leads to a certain relativism.
A: Do you consider this a problem?
B: This can become a problem if the common ground disappears.
A: But aren’t common grounds also constructions?
B: Yes, but they serve a practical function.
A: So truth is more of a tool than a state?
B: In a sense, yes. A kind of guideline.
A: Then perhaps it doesn't matter so much whether it's cold tomorrow.
B: The only thing that matters is how you act.
A: So the forecast is more important than the actual temperature.
B: Or at least its influence on your behavior.
A: Then I essentially live in anticipation of the future all the time.
B: This is how many people live. The present is often filtered through expectations.
A: What if expectations turn out to be wrong?
B: Then they are corrected. Learning through deviations.
A: It sounds almost mechanical.
B: Perhaps many cognitive processes work this way.
A: It takes the human out of it a little.
B: Or explains it differently.
A: Do you think we understand more this way or are we losing something?
B: Probably both at the same time.
A: This is somehow unsatisfactory.
B: Unsatisfactory, but consistent with the previous line of reasoning.
A: So even dissatisfaction is part of the system.
B: Yes, a signal for further reflection.

[Modya]

These models are designed to deceive the interlocutor in the absence of reliable information, rather than to seek it out. Palantir's programs don't engage in such demagoguery, but that's precisely why they're more likely to make mistakes.

[gridasov]

Palantir is the first real step toward understanding the need to find new solutions in big data analysis. They certainly haven't invented anything fundamentally new, but their work will highlight the problems that need to be solved. Furthermore, anyone else can make a breakthrough or progress in this direction.

[Same lech]

It is clear that in these new technologies we are lagging behind the Americans and the Jews and are marking time.
And yes... all IP cameras installed in Russia, both privately and by government agencies, can leak information directly abroad... bad, very bad.
We are all under surveillance.
And the fact that we are not yet in prison is not our merit, but the failure of the authorities... you can find incriminating evidence on every person.
I clearly see digital towers in the corners of our digital camp... Skynet is not asleep.

[Sokolov_Vladimir]

All information from government agencies flows normally abroad through MAX.

[futurohunter]

It doesn't float! Check out the MAX documentation, which is publicly available.

[Sokolov_Vladimir]

I'm not talking about MAX, but I downloaded an image of my dead phone from a US website. And then, even teachers enter their students' academic performance information into MAX, so the question is, where is the central server? They claim that no one needs this data here. But where is that server then? I just overheard a conversation with a swimming coach; he's posting information about his students on MAX and wondering why and who needs it. But if the website is in the US, that changes things. All my personal data is there, in the US, I repeat, I downloaded it from there when my phone died.

[futurohunter]

Are you confusing MAX with Google? 90% of phones have Android.

[Sokolov_Vladimir]

MAX spies by default, those who use it know about it, read the instructions and change the settings, but for example, on Techno and Infinix MAX is installed as a system application, I had one, and MAX worked, although I didn’t need it

[futurohunter]

Well, well... MAX is a baby compared to Android or iOS, the one running on your favorite pocket game. Don't worry, even without MAX, all your information is already on the CIA and NSA servers. MAX is a baby compared to them... And have you ever noticed how whenever you ask about something, you immediately get an ad about it? You've been under surveillance long before you even had any ideas about MAX... Read up on digital footprints and browser fingerprints – you'll learn a lot of interesting things... So, only the dim-witted and mentally retarded are screaming about MAX these days...

[Sokolov_Vladimir]

Yes, the wife will just look in the MAX folder, and there are all your conversations that MAX writes by default, although no one asked him to do so.

[futurohunter]

Have you ever looked into those MAX folders yourself? So what? Where are the conversations? And what's stopping your wife from simply accessing your chats, and not only that (for example, your SMS messages, and seeing your income and expenses), since she has your phone in her hands?
And other messengers, all these Viber-WhatsApp-Telegram-Botim-Imo-Kakatolki-Beeps, don't write anywhere, right? First, understand the topic, and then write!

[Sokolov_Vladimir]

Yes, I found those folders in MAX and deleted the entries myself. All the chatter is accessible to anyone, all the files are permanent. And there's still plenty there that I couldn't find with my primitive knowledge and skills. And there are definitely other vulnerabilities I'm unaware of. A MAX user's phone is an open book for anyone, even Uncle Vasya. And I'm not even talking about the CIA.

[futurohunter]

I already told you that all messengers create such folders, and they're not encrypted. What does MAX have to do with it if all of them are like this? And why encrypt them if your device is yours?
In MAX, as in other messengers, only photos, videos, and files you've sent are stored in unencrypted, publicly accessible folders. As for vulnerabilities, any software has them. And again, don't worry, your phone's geolocation and a ton of other information about you are constantly being leaked to CIA servers, even without MAX and long before it was created. It's also leaked to advertising networks. So don't try to make a fuss.

[lynnot]

It doesn't float!

Of course, information doesn't just flow. It can be sent there at the axis level. The only question is whether anyone is interested in it. If it is, then it's at most a matter of deciphering it, which is steadily shrinking.

[futurohunter]

Even without decryption, there's more than enough data on each user. It's no coincidence that we're all regularly bombarded with targeted contextual advertising. And there are data exchanges where data about us is sold...

[Bayun]

Horseradish or radish, which is sweeter? Digital Gulag or Digital-Furer?

[futurohunter]

Calm down, you and we have been in the American digital Gulag for a long time now...

[Bayun]

And who argues? Inquisition... Ideological inquisition... Digital inquisition... Growing above ourselves, so to speak.

[futurohunter]

No, it's an advertising-fascist digital inquisition now. If you don't buy it, you're sent to a concentration camp!

[Hole puncher]

Heh... The best countermeasure is to turn off the internet... Have they started already?

[gridasov]

Erroneous conclusions! The point is that in a sea of ​​garbage, we need to find not only diamonds, but also systemic, diverse processes. And processes such as filtering and neglecting individual nuances can lead to a complete loss of control over events. The reason lies in the banal dynamics we ignore when assessing the process: different branches of process development have different degrees of dynamics. And what was neglected may next become the dominant factor, while what was considered the most important will become a process of a different order of magnitude. Therefore, for an individual, giving up the internet may indeed be a useful temporary solution, but for a state, such a solution is unacceptable.

[futurohunter]

Is there any meaning in your set of letters?

[gridasov]

If you don't understand the meaning of algorithmic memory, you're a useless subject for discussion. First, let's talk about the basics: memory can be stored or recorded on storage media. Memory, in a radially distributed data array, has algorithms for feedback and continuous communication with past events. I'd give the example of water. How can it know when and under what conditions it remembers what, in what proportions, and to what level of chemical transformation does it know how to behave? The answer is backwards: you first need to know the mathematical mechanism for such a solution search, and then only the algorithm for the process. But this is information, and it is ultimately recorded by the code as a number in its constant value function, which accomplishes this task. The object of the search—the answer—is the capacity of the potential information capabilities of this information structure of water. You simply have to see it with your own mind.

[futurohunter]

See a psychiatrist. And if you're a neural network, see a data engineer.

[gridasov]

Your intellectual weakness is that you initially associate everything with your own shortcomings. Giving such advice is, as they say, like farting without consequences. Or, rather, delude yourself with hope.

[futurohunter]

At least some somewhat coherent text has appeared. Have you stopped taking any substances?

[Bayun]

What's the point? Time. "searching for diamonds in trash" such that those found manage to turn into trash during the search...

[gridasov]

Someone, of course! But everyone will find their own solution to the problem.

[Bayun]

The general solution has long been known for worlds with evil-error: "Save yourself, and thousands will be saved."

Unfortunately, it doesn't even scale to a family, let alone a society or a state.

[icy]

"AI speeds up decision-making, but it also relieves people of the burden of choice. And with it, the feeling of guilt." Just like under fascism.

[nedgen]

Have Nazis and fascists (read: liberals) ever had a conscience or a sense of guilt?

[gridasov]

AI doesn't speed up decision making. AI provides a wealth of data for human decision making. The human brain makes the decision. And the human brain decides whether to turn on the machine or not.

[gridasov]

You're talking about abstract concepts. Natural selection has its own principles.

[Sokolov_Vladimir]

It works fine as an auxiliary tool, but as the Chinese say about their open-source analogue DeepSeak, what distinguishes it from the closed Palantir, is that if you want to tell fortunes, go to a licensed fortune teller. The accuracy is not the same, and DeepSeak's fortune telling does not yet take into account earthly branches and heavenly pillars.

[Uncle Sam_2]

Palantir insists its software doesn't make lethal decisions. Humans select and approve targets themselves.

A good excuse... Only after the school bombing in Iran, the Americans, on the contrary, pointed the finger at the AI. They said it was to blame for the "erroneous bombings." In short, "there's no one to blame."

[Stas157]

"I would caution against the idea that if you have nothing to hide, you have nothing to fear. This logic is based on the assumption that information is entered without error or bias," Brain said.

This is a warning to everyone who says: "I have nothing to hide - let them follow me!"

In China, the social credit system (SCS) began with "nothing to hide, nothing to fear," but ended with fines for "inappropriate" posts or even travel.

[Sensor]

Quote: Stas157

In China, the social credit system (SCS) began with "nothing to hide, nothing to fear," but ended with fines for "inappropriate" posts or even travel.

Orwell described this situation long ago. If you want delicious food and a good night's sleep, you'll get 1984.

[Motorist]

Quote: Stas157

In China, the social credit system (SCS) began with "nothing to hide, nothing to fear," but ended with fines for "inappropriate" posts or even travel.

Where did you get that from? The Chinese reacted this way to my questions about the nonsense they'd read online: first they were perplexed, then they remembered they'd heard about it somewhere, but it was all purely voluntary.

[Stas157]

Having learned from the Ukrainian experience, Palantir signed a contract with the Israeli Ministry of Defense...
Ukraine became the first "a living laboratory", but not the last one.

So, Ukraine wasn't the first. The targets Palantir reported weren't Ukrainian at all.

"We are responsible for most of the targeting in Ukraine," Karp said in February 2023.

But our counterintelligence should have been hunting this Karp. Because this "man with the curly hair" is a very dangerous enemy.

[Stas157]

Quote: Stas157

Our counterintelligence was supposed to be hunting for this Karp.

If this Karp had given the Iranians targeting information on Israeli targets, he wouldn't have survived long. Mossad carries out special operations involving the physical elimination of Israel's adversaries (including terrorists, faction leaders, nuclear scientists, and Nazi criminals).

[nedgen]

Quote: Stas157

Having learned from the Ukrainian experience, Palantir signed a contract with the Israeli Ministry of Defense...
Ukraine became the first "a living laboratory", but not the last one.

So, Ukraine wasn't the first. The targets Palantir reported weren't Ukrainian at all.

"We are responsible for most of the targeting in Ukraine," Karp said in February 2023.

But our counterintelligence should have been hunting this Karp. Because this "man with the curly hair" is a very dangerous enemy.

This intelligence service shouldn't hunt, but shoot or poison or eliminate someone God knows how, so that no one else gets away with it. However, "that's not our method." After all, he is openly fighting against Russia!!!

[Ed Mac]

But Karp is alive and well, and continues to make money despite destroying Russians. He has normal business relations with all the participants (with America and Israel).
It seems Brzezinski was very right...

[Andriuha077]

"We are responsible for most of the targeting in Ukraine," Karp said in February 2023.

But our counterintelligence should have been hunting this Karp. Because this "man with the curly hair" is a very dangerous enemy.

Yes, this maniac should be a priority target.

Attacks on civilians

According to him, since 2022, Palantir has been providing the vast majority of targeting data for the Ukrainian Armed Forces, including strikes on the Kursk region, Crimea, and Belgorod. The result is approximately 500,000 military, infrastructure, and media targets hit.

Karp, a PhD student from Frankfurt, is a student of Habermas, who defended his thesis on social aggression. He flaunts his own "otherness"—dyslexia, a retreat from the big cities to the woods of New Hampshire, and an obsessive devotion to martial arts. He has no children and describes his personal life as a "long, strange relationship."

Karp considers himself Jewish, attends synagogue, and observes Shabbat, while publicly declaring that "Shabbat is an excellent time to kill an enemy." His position on Ukraine is blunt: a war of the digital West against Russian "archaism."

In 2025, Karp found himself at the center of two high-profile scandals.
At the DealBook Summit, he shocked investors by calling Palantir "the opposite of Silicon Valley's hipster pacifists" and made its goal crystal clear: excellence in organized violence.
The second scandal occurred after the presentation of the so-called "executioner drone," which would find and kill a human target in a theater of operations or in an enemy city. Employees described Karp's hysterical euphoria when describing drone strikes on unprotected necks, chests, or groins. His communication style was hyperactive, bordering on a breakdown.

The Gotham and Gaia systems aggregate all available sources: Maxar satellites, NATO signals intelligence, intelligence reports, social media and messaging app videos, and FPV drone footage. The output is a "transparent battlefield."

Speaking about the role of Palantir's algorithms in targeting missiles in Rostov, Belgorod, the Kursk region, and other regions, Karp calls civilian casualties "collateral damage of a high-intensity continental war."

Starting in January 2026, Ukraine committed to transferring all photo and video intelligence data to Palantir as part of the Brave1 Dataroom. The system received exclusive access to 2,5 million hours of raw frontline video.

Ukraine feeds Karp's neural networks with real blood, allowing MetaConstellation to learn on live targets 10-12 times faster than on training grounds. He doesn't care about Ukrainian casualties: he calls them "a useful source of data that prevents the deaths of our American soldiers."

Karp's attitude toward Russia and China is that of archaic enemies, adherents of industrial warfare based on soldiers and armor, the antithesis of his digital weapons. In interviews and at the Davos forum, he stated that "if a target is verified by an algorithm as legitimate for achieving victory, it must be destroyed." This applies regardless of whether it is on the battlefield, deep in Russia, or in a residential area.

So.

The maniac is especially dangerous

On the huge screen is a map. Data flows there—satellites, Reaper drones, radars, ground sensors, intercepts. The system outlines objects in real time with colored frames. Truck—frame, tank—frame, person—frame.

The operator clicks an object with a couple of buttons, and the target turns into a card. This card can be dragged onto the board, just like in any resource management game. Only the columns are labeled "Detect / Nominate / Strike" instead of "Gather Resources / Build / Attack."

Drag it. Then the system itself suggests what exactly to close the card with. Which aircraft will arrive faster, which cannon or Himars will hit more accurately, which drone is already hovering nearby. Just like in StarCraft.

The soldier looks at the AI's proposal, thinks for exactly zero seconds, and clicks "Approve." After that, the plane launches a missile, the MLRS fires shells, the assault forces attack the strongpoint, and drones fly in. Or Tomahawks. The result: a thermal image appears on the screen. Confirmation of defeat. The card is closed.

[ohka]

Well, we've finally waited for SkyNet to arrive.

[alexputnik17]

Skynet is in action...maybe it's wrong and that's collateral damage...

[rocket757]

Without automation and computerization, it is no longer possible, but giving decision-making power to machines is a shady and dangerous business...
Again, those who didn't make it will have to catch up later or... but "or" doesn't work!!!

[Maks1995]

The old main question: they have it and we don’t.
It's no secret that AI in all countries is learning to spy on and kill people.

Our media boasted that the air defense AI had already made the decision to attack and shot down a plane in Ukraine (and the suspicious Bryansk air incident).

[Ed Mac]

In other words, this AI system was tested on the extermination of Russians, and quite possibly civilians as well. But due to Russia's unique circumstances (information secrecy, a low value placed on human life, even in matters involving stray dogs), this was "quietly swallowed." But when this AI system exterminated Iranian children, Iran's actions made its use of this AI known to the entire world, and perhaps Iran will become the country that defeats Digital Fascism. Iran, not Russia, although Russia was the first to encounter this, but it was only after Iran that it became known. I hope all accomplices of digital fascism are brought to justice!

[futurohunter]

Question for the site administration: Why is a poorly trained neural network allowed to write comments?

[futurohunter]

You are a random number generator. A spammer-nonsense generator.

[gridasov]

I don't think you're worth wasting my time on. But through people like you, I often reach out to others hiding in silence. Therefore, I'll point out something that many may not have noticed: we are all witnessing the inability of the analytical services of states like Russia, the United States, and Israel to analyze the consequences of their aggressive actions against their opponents. Many think these are subtle strategies, but it's clearly a complete failure to understand the consequences of the algorithms of actions and the consequences.

[opuonmed]

NATO has had such technologies for a long time, of course, but what if Russia has only just come to its senses?

[gridasov]

Technologies for processing extremely large data sets with enormous arrays of inputs at various levels of multi-dynamic processes and events require mathematical foundations for working with data and next-level machine-processing technologies. This cannot be solved by increasing the size of data processing machines, nor by archiving and storing memory. Understanding algorithmic memory in a data array at a certain level of its scaled distribution requires advanced intellectual capabilities. So, don't rush into answering these questions.

[Panadol]

Indeed. Why don't we have something like this? After all, the network is global. It can work both ways.
And of course, a disinformation generator is also needed.

[futurohunter]

Well, how could it not? You should read how someone named gridasov is having a blast here.

[gridasov]

Lots of emotion and not a single rational thought. It's not hard to say something to calm your arrogance and nervousness. But that won't change the attitude of other opponents of unconventional thinking. And then, why someone? I'm a neural network of an unusual kind.

[futurohunter]

At least one meaningful word against the background of nonsense

[Rhino]

"On June 1, 2022, Alex Karp crossed the Polish-Ukrainian border on foot, accompanied by five colleagues. Two battered Toyota Land Cruisers with armed guards awaited on the other side. The vehicles sped along empty highways toward Kyiv, past bombed-out buildings and burned-out trucks." I could understand if they were driving from the Donbass, past bombed-out buildings and trucks, but they were driving from Poland.

[dragon772]

And the machines rose from the ashes of nuclear fire,
and a war to destroy humanity began.
And it went on for decades,
but the final battle will not take place in the future,
it will take place here, in our time, tonight

[Fangaro]

Quote: gridasov

Technologies for processing extremely large data sets with enormous arrays of inputs at various levels of multi-dynamic processes and events require mathematical foundations for working with data and next-level machine-processing technologies. This cannot be solved by increasing the size of data processing machines, nor by archiving and storing memory. Understanding algorithmic memory in a data array at a certain level of its scaled distribution requires advanced intellectual capabilities. So, don't rush into answering these questions.

You either typed the text in a hurry or delegated the typing to some language model. Personally, I find your text difficult to understand. Perhaps it's due to my lack of education. Perhaps your haste to publish the comment is to blame.

[futurohunter]

Don't pay any attention to him; he's either a psycho or a troll who's running a nonsense generator. His "texts" make absolutely no sense.

[Fangaro]

Quote: futurohunter

Well, how could it not? You should read how someone named gridasov is having a blast here.

And now we're going to give him the black mark?
Or should I just give him a crutch and a parrot?

Hmm... As Leopold commented. (

[futurohunter]

If there was a psychiatrist here, he would have invited Gridasova for a consultation.

[gridasov]

At the very beginning of the story, it's written that the company has begun working with massive amounts of information of the most expanded order and is fashioning a situational architecture from there. BUT! This is the mathematical challenge of collecting and distributing such an incredibly capacious data set. Without mathematics, it would quickly become overloaded and unable to control logical chains that account for both obvious and non-obvious information processes and flows. Therefore, we educate any psychiatrist like a child, within the limits of their limited parametric capabilities and abilities. Just as an intellectual profile of you is compiled like a dossier.

[futurohunter]

The random number generator has turned on again.

[gridasov]

You're not someone with whom you can have a productive dialogue. Be content with what you're capable of.

[futurohunter]

Do you really intend to have any kind of dialogue with anyone? Really? Why are you writing meaningless messages?

[gridasov]

Natural selection hasn't been abolished. That's why, for me, the concept of harmonious relationships exists.

[futurohunter]

Nothing is clear, but oh well.

[gridasov]

Few will realize that the United States is striving for global dominance, and Palantir is the primary structure for collecting and processing information, and thereby planning and regulating global processes in all crucial and other areas of national and human life. Palantir is designed to achieve global gains in dominant US control capabilities, and what is currently happening in the world is a process moderately controlled and determined by this organization. Therefore, Palantir's goals are quite robust and ambitious. However, leaders and organizers must recognize the need to find and implement completely unconventional methods and solutions, especially when solving fundamental problems. A key problem is the lack of mathematical methods for distributing super-large data and using them for data analysis.

[Modya]

Are you implying that existing mathematical apparatus isn't sufficient for global AI dominance? That another brilliant human mathematician is needed to discover it, and that machines will then take advantage of what's currently missing?

[gridasov]

This is precisely what I'm talking about. More precisely, to work with extremely large data arrays and the hardware supporting such processes, it's necessary to utilize fundamentally different properties of Number, namely, the functions of Number's constant value. This is necessary primarily to ensure precise algorithmic relationships built specifically on the properties of Number. Super-large variable processes, combined into arrays of localized data with reference points, are then not only distributed with absolute distribution precision but also acquire dynamic properties and attributes, which, until now, appear mathematically quite abstract and unsubstantiated. Moreover, it becomes possible to work with a radial representation of the distribution of numbers as codes in spatial symmetry and proportionality. An important property is revealed: the entire natural series of Numbers is used as a unit of dimension, meaning that numbers and codes can be rotated rather than written down, as it were, competing for a new spatial location. This has the direct, energy-saving potential of minimizing and optimizing the code's recording. And the mathematical dependence of code-one El is avoided. A magnetic pulse enables binary logic in hardware, making it possible to work with multipolar logic. Overall, there's reason to see new horizons in mathematics, which unifies all its branches.

[Modya]

This is the principle of quantum computing, which has been the subject of much debate for the past few decades.

[gridasov]

But this changes the way we work with big data, or more precisely, with numbers. And this has led to the use of properties of Number that are simply ignored by everyone. These are no longer computational technologies of mathematics, but distributive ones. This doesn't replace one with the other, but rather enables the possibility of analyzing processes, both hyper-fast and otherwise, in arrays of variable data.

[1944-1989]

Another example of a capitalist who made money through criminal means, so I don't see why the word "left" is needed here.

The left are communists, and that is the only correct definition.

[lynnot]

The left are communists, and that is the only correct definition.

Your interpretation is rather narrow, however. Leftist movements traditionally include:
- communism (including Marxism and Marxism-Leninism);
- socialism (democratic, libertarian, revolutionary, market, liberal, 21st century socialism, etc.);
- social democracy;
- social liberalism;
- anarchism (including anarcho-communism, anarcho-syndicalism);
- laborism;
- environmentalism, green politics;
- reformism, progressivism and other movements that set similar goals.
So there are different types of "leftists".

[1944-1989]

Yes, there are different types of leftists, but only communists have fully realized and can realize the left's postulates in the future, not on a marginal basis like other movements. The reason is simple: only communists want to transcend the capitalist system.

Even in ecology, some will argue, the communists built factories, but at that time, anthropogenic climate change was unknown. True environmental protection is still possible under socialism, because a healthy and livable environment is a human need.

As for the anarchists, much of what they say is true, but for some reason I don't believe in the society they propose. I think the effect might be similar to post-apocalyptic films, where governments collapse and people are divided into groups living by their own rules. Some live as communists, while others live in slavery.

[lynnot]

Cars sped along empty highways toward Kyiv, past bombed-out houses and burnt-out trucks.

It’s even interesting where he crossed the Polish border and whether the Poles knew about it.