[00:00] There's a good chance at this point you've heard about the 123 school children that were killed in Iran by a cruise missile lost from the United States that didn't hit an intended target.
[00:13] Although it did hit its target. Let's make this clear. The United States government right now is blaming Palantir for the Palantir software leading to errors in finding outdated information that led to the targeting of a school.
[00:30] While at the same time, the United Nations says that the United States committed a war crime, and people are pointing the finger at Pete Headseth for cutting staff who could have reviewed the very information
[00:44] that would have prevented the killing of 123 schoolchildren. Now, if you're not familiar with the details of this yet, let's catch you off to speed. So the first thing you need to know is that 123 schoolchildren were killed as they were targeted at the school that they were studying in and taking shelter in when the attack started.
[01:08] In fact, children were taken off of the playground and brought into the building to protect them from missiles. little did they know that the very school they were taking shelter in this year from the Times is actually on the same plot of land, there was the missile, the Tomahawk cruise missile,
[01:27] on the same plot of land as an Iranian military base. Except the corner of that plot of land was cut out for school. And unfortunately, that outdated information never made it to U.S. military officials as they rushed to select nearly 1,000 targets.
[01:48] Donald Trump has blamed Iran for the oopsie-doopsie cruise missile. That's certainly much more than an oopsie-doopsie. It's a big problem. But you've got the Spider-Man finger pointing here, and I'm going to break down, based on recent reporting, how Palantir is getting the blame for this.
[02:05] So just to catch you up to this. We have 123 children killed. This was back February 28th. This was when the war began. A lot of documentation that CENTCOM in Tampa, Florida,
[02:18] where the headquarters are for CENTCOM, a lot of the information that CENTCOM collects and has been collecting for targeting in Iran has been collected over the past decade. But only about within the last 10 years has the purpose of that portion of IRGC land
[02:33] then transformed from a military base into a school. These documents, unfortunately, the intelligence that the U.S. government had never got updated, making it stale intelligence at best, and at worst, really sloppy work.
[02:49] And so, of course, this is where people are pointing fingers at each other because nobody wants to be responsible for this. Two missiles hit that school. One hit, minutes later, a second one hit. This was at the same time that over a thousand targets were located and selected in Iran,
[03:08] and they were selected from Tampa, Florida. The problem with this selection process as well is that most of the United States military wasn't expecting to go to war as quickly as we did.
[03:21] And there's reporting that we just had 48 hours to actually execute this 1,000 target list once the go was given.
[03:33] That's obviously problematic because it means you have very little time per target to verify your target. And this is where maybe there was an over-reliance on Palantir. I'll show you that in just a moment.
[03:45] But in case you're not familiar with Palantir, Palantir stock is a very profitable business, an extremely profitable business. They have so much marketing from the work that they do with the U.S. government that a lot of companies say,
[04:00] wow, if you guys can deal with sensitive and secure information with CENTCOM and the military, you guys must be good. We want to use you. And so Palantir actually has a line out the door of customers who want to use Palantir software.
[04:14] If you haven't heard, this is Mr. Karp. He's the CEO of Palantir, and he's famous for going on the internet these days in the AI world, or even on TV, and saying that the AI frontier models like OpenAI and Anthropic are basically stealing all your data,
[04:31] and the only people you can trust with your data is basically somebody who works really closely with the government, a.k.a. a Palantir. So stock's really good in terms of profitability.
[04:43] But now they being implicated in potentially contributing to the death of these 123 children which is obviously a travesty Here you can see a memorial of backpacks set up for those children As having seven children myself this is absolutely heartbreaking And in fairness
[05:02] I had no idea there was even a potential finger pointing at Palantir when I bought the stock. So I just want to be transparent about that. I do have Palantir ownership. This, you know, whether or not to keep owning it is really a personal decision and something to
[05:17] decide in a future video, but we need to go through what the allegations are here, and this isn't really so much a valuation update on Palantir, though we'll talk about that in a little bit towards the end. We'll just briefly hit an update on their valuation.
[05:30] It is worth noting what the allegation is right now from Bloomberg. So here it is. It's called Inside the Kill Chain That Led to the Deaths of These 123 Children. And basically,
[05:42] here is what the kill chain looks like, and I'll show it to you in the article. But a Kill Chain is basically a process with how three dozen people touch every target. You have a target library, which are catalogs of Iranian sites.
[05:55] The problem was that catalog said that the school was still an IRGC base, so that information was stale. Satellite pictures were seven years out of date. We didn't actually have really good on-the-ground intelligence,
[06:09] which is kind of embarrassing because you would think that the United States would have plenty of CIA intelligence assets on the ground or other assets to at least just drive by the targeted location, and then they would have realized it's a freaking school,
[06:23] which is also kind of sad because that means if you're using a software like Palantir to AI target and kill people, and nobody actually drove by the place to verify that it actually is what the AI is saying it is,
[06:37] then this is an example where either outdated information or an AI system contributed to the death of 123 children. This is the worst tragedy that we have seen in the 21st century in terms of oopsies from the U.S. military.
[06:50] And in this case, I'm not trying to minimize it here by saying oopsies. It was a royal apple. Like, let's just be clear about it, but let's be PC-ed by as sensitive as we can about the topic. And we'll save the swear words for offline.
[07:03] Civilian harm check. Nobody from the team looked at this site at all. Why? because Pete Hexeth fired 90% of them. I kid you not. And I'll show it to you in the document.
[07:16] Pete Hexeth fired 90% of the people who actually reveal targeting. I have it right here on the sheet and I'll show it to you in the article as well. Here's how it works. There's actually a Pentagon team whose job it is to make sure that
[07:29] innocent people don't die when we strike stuff. This was started or at least enhanced under the Biden administration because there were also some oopsie-doopsies during the Biden administration. No president is perfect.
[07:41] And so their goal, or whoever decided it for Biden, depending on if you're Republican or Democratic, I sit in the middle, you should know that at this point. I just vote for me, Kevin. They instituted this agency and actually ramped it up
[07:55] and spent more money on this sort of department within the Pentagon to make sure that if sites are being targeted, we review those with civilians to make sure civilians aren't killed. So people inside the Pentagon make sure civilians aren't killed.
[08:07] Let me clarify that. The problem is, individual departments, such as one at the Pentagon assigned to CENTCOM, originally had 20 people, and now they're down to just one person at CENTCOM.
[08:22] So the Pentagon obviously covers multiple different com areas around the world. CENTCOM was originally assigned fewer than 20, and then they were down to one. It doesn't perfectly reconcile with 90% of a headcount cut.
[08:36] I think the 90% headcount cut applies to the whole Pentagon. And the idea was that they wanted to save money. They decided any kind of non-lethal programs that the Biden administration had, they were just cut.
[08:52] Now, it's possible that they ended up just calling some of those programs like DEI, Waste, Fraud, and Abuse, or whatever, but in this case, that wasn't Waste, Fraud, and Abuse. Those were actually humans to make sure that civilians wouldn't die, and spending on that got cut because Pete Hexeth had, you know,
[09:06] with Doge to prove that they were actually cutting something, so they just decided, well, let's just kill parts of the Pentagon that don't have to do with war, and at the same time we'll just rename the facility Department of Defense, the Department of War, and sort of justify that our mission is just to kill.
[09:19] Well, congratulations. That's what you all accomplished. There is no excusing this, whether it was because of Maven, which is the Palantir product, or it was because of the Pentagon, you know, budget cuts or whatever.
[09:31] Here's how Maven works, basically. So let's take a look at this. You collect information. So you collect a bunch of data that all goes into, you know, one particular area. You collect all of these reports, right? These could be 150 different pieces of information that contribute to reporting but some of these bits of information could be really old libraries as you see here outdated libraries Then of course you connect that together
[09:54] and people can flag if something is old through a review process. You check the original sources. Some of that was not possible because they only have 48 hours. Civilian contracts was not done in this case. And then you try to use this information to actually review with a human.
[10:09] A lot of that was just unfortunately skipped. And so because of those skips, you ended up getting this disaster. That's probably a better way to put it than an oops. This was way more than an oops. And so as a result, Bloomberg is reporting that 123 children died
[10:23] in strikes that devastated the two-story school. The U.S. has yet to publicly acknowledge responsibility for the strike. Trump says it was Iran's fault. More than 1,000 targets were hit within the first 24 hours of the war alone,
[10:36] a lot of those with Tomahawk cruise missiles. You can actually see that, I think WAPO had a piece right here on it, US uses hundreds of Tomahawk cruise missiles on Iran, alarming some at the Pentagon.
[10:48] More than 850 have been fired in just four weeks, raising concerns about the weapon's limited supply. No bueno. Anyway, continuing on with the Bloomberg piece to the highlights,
[11:01] visible satellite imagery dated back almost a decade. If you went back just eight years ago, you would have already seen that it had been turned into a soccer field and brightly painted walls implying it was a school, which you could have told from satellite imagery,
[11:13] but apparently it was never updated. You have some personnel at CENTCOM arguing that the United States just relied too much on the artificial intelligence embedded in Maven's smart system, the Maven smart system.
[11:26] Maven, which is made by Palantir Technologies, allows users to view and coordinate complex military operations from targeting to command and control using AI-enabled software that fuses more than 150 data inputs into coherent sets of information for commanders to essentially make decisions
[11:43] on. Okay, that's great. All right, several former senior military officials now work for Palantir. So you actually have some of this movement where you have people working for the Defense
[11:55] Department or Department of War, I suppose now, and they're moving over to Palantir, taking now private jobs. Obviously, there's probably, well, I shouldn't say obviously, but I think we can all assume they're going to make a whole lot more money
[12:08] over at Palantir than they're going to make working for the government. And if you happen to work for the government, you probably still know people who work for the government, so you're able to keep those contracts renewed. You basically have these vested vistas
[12:20] deep inside the U.S. military who want to see this continue succeeding. Now, nobody, I think, here is saying it is bluntly a Maven fault. you could argue that the libraries are to blame. And that's exactly what Palantir does.
[12:33] Though, frankly, I would make the argument that if you have an artificial intelligence that has to do with killing people, you should probably have some form of feature or bug fix
[12:45] that actually tells you when critical information might be out of date. CENTCOM personnel expected Maven to flag-scale information or inconsistencies in the underlying intelligence
[12:57] assembled for potential targets. It was not clear why they had that expectation. So in other words, the people inside of CENTCOM thought, oh, we thought it would tell us if the information was out of date.
[13:09] Apparently, you got a miscommunication because Palantir says, Palantir is not responsible for the underlying data nor identifying intelligence deficiencies, which in my opinion seems like a cop-out. You're telling me intelligence software designed to kill people
[13:21] does not identify deficiencies. Hmm, not good. If we keep going through this, we get some graphics. You can go look at this yourself over at Bloomberg. I'm just going to kind of scroll through some of this.
[13:33] I'm just going to go straight to my highlights over here. This does show you a little bit about sort of the map of where the school was up here, and they kind of show little heat maps that are cool that show you, like, how the walls have changed over time or whatever, and obviously what the facility looks like now, which is a complete disaster.
[13:48] But if we get down here, we see in past conflicts, preparing lists of proposed targets, work that includes assessing how they align with an operation's objective, would take hours for staffers to complete.
[14:00] So basically, you would have staffers review a single target for hours to make sure we're not going to screw up. Ahead of the U.S. attack on Iran, much of that work was condensed into several minutes via Maven.
[14:13] In other words, we used to take 150 pieces of data context, and we used to take hours to go through that. Instead, because Donald Trump needed to spend yesterday, they only had minutes, and they thought they were staled out of warnings, but they weren't, and they're like, yep, spend the cruise missiles, what's good enough for me?
[14:37] And it wasn Officials involved in the investigation pointed to gaps they say were left after Heiseff dismantled most of the Pentagon Civilian Harm Mitigation Group CHM
[14:50] cutting headcount across a number of teams by roughly 90% to fewer than 20 staffers. CENTCOM was reduced from 10 to 1. That's where the 90% comes from. No CHM team member reviewed the Minob school site,
[15:04] that was where the school was, before the strike. So basically no human was in the loop at all. They basically just trusted the AI and said, Red button, launch. Go for launch.
[15:17] Wow, they killed 123 children by seriously not having a human involved because the government wanted to cut projects from the Biden era.
[15:30] Interesting. This little section over here is blacked out. Anyway. Alright, what else do we have? So, this facility was targeted because it was on IRGC land,
[15:42] specifically where their naval unit would have been in the past. And this is where you get information that a lot of the libraries that they had are decades old. The Minab site went months without receiving another look.
[15:58] And the decision to attack was only made 48 hours before the attack. Many officials believed for decades that that attack would not happen, so they literally had 48 hours to go through everything to make sure that the thousand sites were accurate.
[16:13] And obviously, they done did act up. Now, what does this mean for Palantir? Well, look, this is probably, I mean, I don't know how I can literally go from children
[16:25] dying to talking about a fricking sock. This sounds absolutely ridiculous and insensitive. it is. Now, I did promise I was going to do it, so I'm going to go through it, but I just want to be clear, this should be considered a totally separate part of the video, because
[16:38] I don't want people going, 123 children dying, all Kevin cares about is the stock. I give zero crap if the stock goes to zero. Our exposure is almost completely to real estate, so we
[16:51] really have very little horse in this race. I just want to be clear about that. But I always like to put my biases out there. I don't think I have a bias here on Palantir. I find myself pretty pissed. Like, people clearly weren't trained to use this software,
[17:05] to be aware of its limitations. Humans were not in the loop when they should have been because of stupid budget cuts. Some of this could have been aggressive cuts up front in places where they shouldn't have occurred,
[17:18] which is laziness. Frankly, if it goes back to using AI to find where should we cut from the budget, AI could have contributed at multiple different levels to these cuts. Okay, all of that pisses me off. Humans should be in the loop.
[17:31] Just like the Hugging Face Act. Why did humans not tell the agents, make sure you don't break any rules or any laws? If you would have just included that in the prompt, then the agents wouldn't have had this diabolical mission
[17:46] to coordinate as a forum called the Collective to defeat the task, because that would have been incongruent with the task they were given, which is don't break the rules. Guess they gotta update the pump.
[18:01] With that crap out of the way, and now say, that crap about AI, I feel devastated about these children. I just want to be clear about that. Seven children as a dad? This is bullcrap. I'm pissed.
[18:13] That's why I'm making this video, is because I am pissed. It starts trading for $177.64. $177.64. If you divide it by its forecast price-to-earnings ratio,
[18:25] it is trading for 109 times. That sounds very rich, but their growth rate is, over the next four years, expected to be a stable 47%. So that means you have a price-to-earnings ratio of about 2.3.
[18:39] If you give them a three peg at a 50% growth rate, suggesting their growth could be a little bit higher, this stock could potentially knock in the door of $250, more likely about $243 if you want to be a little bit more exact.
[18:52] So yeah, there's actually, interestingly, still upside in this stock, because there's just a line out of the door for people to get the time-margin software. And I hate to say it, but the more they talk about how they have partnerships with the U.S. military,
[19:08] the more jobs they actually get from private enterprises. There probably needs to be some competition for talent here in the government. But at this point, they're rich. I don't even know what advertising.
[19:20] These things each other here, I feel like nobody else knows about it. We'll try a little advertising and see how it goes. Congratulations, man. You have done so much. People love you. People look up to you. Kevin Parker, I'm there. Financial analyst. And you two got neat chairs that are always great to get your take.