> Whether it was overconfidence in their tools or a complete disregard for the lives of civilians that caused this lapse snip

it could be both, but we know no. 2, the complete disregard for the lives of civilians, is in play because, whatever else was going on, america was initiating war for the purpose of destabilizing a country, afaict at least, the reasoning has been unclear. destablize means to try to make things fuck up, and that tends to kill people. what people? how? who knows? things fucking up means out of control. at that point it's up to physics, not people.

it's like, if i set a house on fire, then later defended that action by claiming to have not known where i started the fire was a nursery.

back in the war on terror days america had a habit of blowing up weddings, and then claiming it was an accident. and i would think, accident how? did the missile fire itself?

I couldn't find a web site for the school when I searched for one and I also noticed that while schools are generally marked on Google Maps in Iran this school was not. Both are IMO not really relevant or reliable sources of targeting data anyways. I found very little evidence searching online for the school but I did find something that looked like a blog about a school trip. Again though the Internet is not a reliable source of data for targeting - should be obvious.

The main way targets should/would be selected is by direct intelligence. E.g. the targets should be identified through satellite or other observations. It's hard to imagine that a building that has operated for some length of time as a school would not have patterns that are visible from satellite vs. military facilities with different patterns. You also don't just randomly attack structures in this sort of surprise attack, you're presumably aiming for some specific people or equipment with some priority/military goal in mind, so you really want to have observed the targets and patterns and have up to date information on their usage.

I think what likely happened here is that the entire base was the "unit" of targeting and the mistake was in identifying which buildings were part of the base. In the satellite view the military buildings and the school look very similar (since the building as I understand it used to be part of the base but was repurposed as a school).

It's not true that whoever made the error had all the time in the world. Presumably once the order was given there was time pressure given that the strike was to be timed with the other intelligence.

In theory the US military should/is supposed to have good processes around this stuff. So we are told. Obviously failed in this case. It is a tragedy.

We learned that Israel was going to strike, so the US decided to jump on board. Do we know how long of a notice Israel gave the US? What you're attributing to as a thought out plan of attack seems to imply plenty of time. I don't think it's unreasonable for Israel to have learned of the meeting with little notice, deliberated internally for however long, and then told the US about it with not much time. I could totally see where under current Pentagon leadership, three clicks would have been the reaction. Yes, the US had been saber rattling and building pressure. That's probably part of why the Iranian leadership decided to meet. Whatever plans the US might have had went out the window when Israel called up and said we're striking now.

I recommend looking closely at the New York Times analysis. There were factors that might have mitigated this as a strike target, but it also really did look like a part of the compound (and it originally was!). Yes, with hindsight, we can definitively know, and with sufficient time each target could probably have been positively ID'd, but there was precisely one mis-strike in 1000s of sorties, so this already is a low error rate. TFA discusses 50 specific strikes all of which missed via automated analysis. That doesn't seem the same.

I don't disagree there. But this is not a case of hallucination, and an existing website is a signal, not a determinant, of the real situation on the ground. However, you have made a very, very strong assumption that these targets were not carefully evaluated. One that does not seem to be present in TFA or any analysis that I've read. In fact, the article itself quotes those in the know who believe this should have been eliminated as a target.

You have clearly articulated what I’ve personally explained to people. Thank you for that. The nature of the strikes as a part of a thoroughly pre-planned surprise attack lays the entire blame at the planners, approvers and those who executed the strike.

The lack of comprehension some people have baffles me, as I’ve had the displeasure of reading several dozens of online posts asking why kids were at school during the strikes. Even giving these people the benefit of the doubt that they do not know that not all countries observe the same weekday/weekend split as in the case of Iran, how in the world is a teacher or a child supposed to know when to hide from a surprise attack?

The easier it gets to give people the tools and power of lethal force, the more preventable injuries and death happen to innocent people. The cover of military conflict should not protect from consequences in cases like this.

Knowing the demographics of this website, it will not make anyone here safer that there is credible proof of Israel using Whatsapp metadata to source location data of adult men, and executing strikes based on that information. Western media already shared stories of how ordinary cell phone metadata was used to conduct strikes that killed innocent civilians. 15-20 years later the exact same deadly inaccurate methods are being used to quench the leaders’ and planners’ thirst for any results. One day a bomb might fall on any of our homes purely based on some circumstantial proof that wouldn’t even be enough for a traffic violation…

I agree with everything you said - but it's also the case that a set of parameters were created that, instead of requiring multi-person validation of target validity and provenance, prioritized speed to provide decision makers with options.

This certainly doesn't absolve the person implementing those parameters, but it is equally the responsibility of the very top of the decision-making structure.

And fundamentally, this is aUS doctrine issue. The US is willing to strike targets in foreign soul with no boots-on-the-ground confirmation of target nature.

It's how the Obama administration drone-struck a wedding before this and how a missile got dropped on a Chinese embassy before that. The doctrine itself is flawed.

I'm not sure how true that is. Enemy factories and command centers don't grow out of the ground overnight.

Nor do planes get maintained, armed, fueled and flown to the target zone in the matter of minutes.

In preparing such an operation, I'm sure the critical path even with traditional planning methods, is in other places.

While I agree, that there are certain scenarios where an important enemy commander or an expensive mobile launcher gets detected, and you only have a window of minutes to hours before its gone, this is not one of those cases.

I feel like the military bought some fancy new hammers, and wanted to show the purchase was justified.

Was the Iran War Caused by AI Psychosis?

https://houseofsaud.com/iran-war-ai-psychosis-sycophancy-rlh...

I agree with your overall sentiment, but how realistic is it? Israel/US says they've been hitting thousands of targets (so reality might mean ~hundreds, still a lot), how are they supposed to verify this at all?

> Humans should have been double and triple checking every target by other means.

How practically would this happen? The US/Israel don't want people on the ground, and people on the ground is exactly the only way you can actually verify stuff like this, not every place in the world is on Google Maps or have a web presence at all, so the only realistic way to verify this would be to visually inspect it in person, something neither parties who started this war want to do.

Even better, don't make attacks against other soverign nations that don't pose an immediately and critical threat to you, and this whole conflict could have been avoided in the first place.

But no, the president has to be involved in some sort of child-trafficking scheme, so pulling the country into a war seemed preferable to being held responsible, and now we're here, arguing about fucking details that don't matter.

I can't find any sources for this that don't link back to this tweet.

More importantly, I think you forget how bad human intelligence is. These kind of targeting failures have been commonplace in every war since the airstrike was invented.

I think what happened here is they searched on Google Maps for the word 'police' and didn't spend too much time verifying the results, because bombs are cheap.

Give me a source that says Police Park was hit. AFAICT while the place exists, this story was made up.

It's definitely been reported before that Claude was used for Iran attacks, at the beginning of March or earlier:

https://www.theguardian.com/technology/2026/mar/01/claude-an...

Edit: Also, https://www.washingtonpost.com/technology/2026/03/04/anthrop...

I have heard the claim everywhere.

there is a lot of confusion about all this stuff

you, today, can use Claude in Amazon Bedrock, and the way that works is, if you want it to be this way: the piece of code and model weights and whatever other artifacts are involved, they are run on Bedrock. Bedrock is not a facade against Claude's token-based-billing RESTful API, where Anthropic runs its own stuff. In the strictest sense, Bedrock can be used as a facade over lower level Amazon services that obey non-engineering, real world concerns like geographic boundaries / physical boundaries, like which physical data center hardware is connected by what where / jurisdictional boundaries, whatever. It's multi-tenancy in the sense that Amazon has multiple customers, but it's not multi-tenancy in the sense that, because you want to pay for these requirements, Amazon has sorted out how to run the Claude model weights, as though it were an open-weights model you downloaded off Hugging Face, without giving you the weights, but letting you satisfy all these other IP and jurisdictional and non-technical requirements that you are willing to pay for, in a way that Anthropic has also agreed.

This is what the dispute with the Pentagon is about, and what people mean when they say Claude is used in government (it is used in Elsa for the FDA for example too). Anthropic doesn't have telemetry, like the prompts, in this agreement, so they have the contract that says what you can and cannot use the model for, but they cannot prove how you use the model, which of course they can if you used their RESTful API service. They can't "just" paraphrase your user data and train on it, like they do on the RESTful API service. There are reasons people want this arrangement ($$$).

The vendor (Palantir) can use, whatever model it wants right? It chose Claude via "Bedrock." I don't know if they use Claude via Bedrock. Ask them. But that's what they are essentially saying, that's what this is about. Palantir could use Qwen3 and run it on datacenter hardware. Do you understand? It matters, but it also doesn't matter.

It's a bunch of red herrings in my opinion, and this sort of stuff being a red herring is what the article is mostly about.

Time for a massive boycott to all US-American products? Or what else will stop the thugs?

What AI does best is remove accountability and ownership

I 100% believe this was done by somebody inside our administration. It’s totally in character for them. Look at what they do to immigrants and our own citizens. They intentionally target children.

But if you wanna look externally, you can’t rule out Israel. They have intentionally bombed a school to kill children in the past, well before Gaza.

Before you take out your pitch fork, remember what the US did in Vietnam. Ugly stuff happens in ideological wars. It is not controversial to say Israel has done similar things.

Also, someone in our very pro-Israel administration claimed they got us into this war. Israel manipulating an ally is completely unsurprising.

But it doesn’t stop at Israel. I think every single ally we have in the Middle East would do the same thing. Everyone they’re fighting already does.

Buddy

You're not wrong but what we can tell from open sources is:

- The building does seem to have actually been a school and "detached" from the rest of the military complex.

- The school the Iranians claim it was does seem to exist even if it's not 100% clear that's the identical location.

- At the time of the attack school would have been in session.

- The signature of the attack seems similar between all the buildings attacked and we have footage showing a Tomahawk hitting the area.

Another thing we can tell is that the US has to know the truth here and isn't coming out with an official statement.

And I'm saying this as someone who thinks the Iranian regime is evil, needs to be struck down, was trying to acquire nuclear weapons etc.

As to the numbers I agree they are to be treated with suspicion. The Iranians are obviously motivated to lie, inflate them, and treat all casualties as civilians. But we can still try and estimate given the size of the building what would be the number of students. We can also estimate the outcome of the missile hitting the building and correlate with the photos and satellite imagery, and until we have better data use those estimates.

Everyone acknowledges that the US killed a whole bunch of kids, including the US

> To this day we don't know what actually happened.

I feel like we know enough already. A school was bombed, the ones who did it sucks big time and should be held responsible. Currently, the US and Israel is waging a war against Iran, and one of them dropped the bomb(s), unless suddenly Iran got their hands on American weapons, then that needs to be investigated too, because someone surely dropped the ball at that point.

The basics remain the same, investigations have to be launched to figure out where exactly in the chain of command, someone made a mistake, and then hold that person(s) responsible for their fuck up.

Have those investigations been launched?

>> Again, the basic facts on the ground are not known

Think for a second WHY that is! They can find and kill the Iranian leaders who will be doing the utmost to conceal their location and yet that can't tell us whose bomb blew up a specific building? Of course they can. They're waiting until people forget and they can final release the result of their 'investigation'.

I think its fair to treat things that the Trump administration and the Iranian military agree on as facts. If they were distortions that favored one side, we would see pushback from the other. Maybe there are distortions that somehow benefit both of these parties, but it seems unlikely. At minimum, then, this was a school, the Americans bombed it, and children died as a result.

Yes. They funded Hamas' invasion of Israel, a close ally of the US, which intentionally killed over 1000 civilians.

They have also repeatedly threatened to use nuclear weapons on Israel and were in the process of developing such weapons.

The Royal United Services Institute (RUSI) has an updated tally on defensive and offensive munition expenditures. It's likely not 100% accurate due to the sensitive nature of those figures.

> 11,294 munitions in the first 16 days of the conflict, at a cost of approximately $26 billion.

Several detailed tables are in the link below.

https://www.rusi.org/explore-our-research/publications/comme...

It’s a tale as old as time: start a war to support the military industrial complex. Imagine a $4 billion investment into public transportation or parks. Every 10 years we can invest into a new city instead of bombing some kids overseas (whose siblings, fueled by hatred, then commit terror attacks on the west).

We'll run out long before 2029. The 850 fired so far is about a quarter of the entire supply.

https://www.reuters.com/business/aerospace-defense/us-uses-h...

So far, they are not funded to do this for that long. They have floated a $200B bill to congress, which made national news coverage. It would start a huge, prolonged fight over the war and actually force them to ask permission from congress to fight it (barring totally disregarding the constitution which is still a possibility).

Unfortunately I can very well imagine several more months and years of this. We are still fighting a forever war that started in 2001. This is all a generation of Americans will know, and that is sad.

Don’t worry. The Saudis and UAE will happily pay all costs of the war.

I think more than one. One and then another 1 hour after

The immediate concern isn't really fully autonomous systems, it's that the nature and design of recommender/suggestion systems prompt humans to sleepwalk through their responsibilities.

https://en.wikipedia.org/wiki/Computer_says_no

That theory seems unhinged to me. Somebody is divorced from reality. Hopefully not me.

There are rumors that the IRGC has started to recruit adolescents (again). I'd have an easier time believing that, but where the Middle East is concerned, there is an endless amount of propaganda and hyperbole, so I don't automatically believe what I read. Pardon my condescension, but it would make the comments sections better if everyone did likewise.

kakacik's razor - Never attribute to incompetence of IDF/Mosad that which is adequately explained by laser focus intent to murder and exterminate enemy at all costs, via all avenues, all is allowed.

That is not what this article is about at all.

Yeah no big deal right?

Check out this example: https://www.bbc.com/news/articles/cjw4egp7lwno

"As we pass through Khan Shaykhun, we come across a street painted in the colours of the Iranian flag. It leads to a school building that was being used as an Iranian headquarters." "On the wall at the entrance of the toilets, slogans read: "Down with Israel" and "Down with the USA".

It was evident that these headquarters were also evacuated at short notice. We found documents classified as "highly sensitive"."

This is a BBC reporter reporting from Syria after the fall of Assad.

It is strategy for the IRGC and Hamas to operate from civilian infrastructure like schools to gain immunity. That's what's "not a human error".

> This is intentional genocide

Do you think it's genocide when the IRGC kill 30-40K Persian civilians? Or only when Americans missiles aimed at a military base miss their target?

Sufficiently advanced negligence is indistinguishable from malice.

This is not to say that this administration is definitely not targeting civilians or infrastructure on purpose; just that the end result, and the moral culpability, are the same in either case.

Replacing one java tool with another doesn't solve anyone's problems. If they'd only used Rust then lives would have been saved.

Yes, the US has 160 schools on military bases:

https://www.militaryonesource.mil/education-employment/for-c...

Yeah man. Who's ever heard of a military base with a school on, or near it besides every military officer with a family ever?

Both the Washington Post and the Guardian articles agree that the system used here was Maven.

The key sentence in that Washington Post article appears to be:

> The Pentagon began to integrate Anthropic’s Claude chatbot into Maven in late 2024, according to public announcements.

As far as I can tell this is the public announcement - a press release from November 2024: https://www.businesswire.com/news/home/20241107699415/en/Ant...

> Anthropic and Palantir Technologies Inc. (NYSE: PLTR) today announced a partnership with Amazon Web Services (AWS) to provide U.S. intelligence and defense agencies access to the Claude 3 and 3.5 family of models on AWS. This partnership allows for an integrated suite of technology to operationalize the use of Claude within Palantir’s AI Platform (AIP) while leveraging the security, agility, flexibility, and sustainability benefits provided by AWS.

I assume you actually read the article and didn't just post this after a quick skim, yes? Because saying this:

> The distinction between Maven and Claude is futile

Doesn't make any sense at all when you read the article and understand what Claude actually does in this equation. From the article:

> Neither Claude nor any other LLMs detects targets, processes radar, fuses sensor data or pairs weapons to targets. LLMs are late additions to Palantir’s ecosystem. In late 2024, years after the core system was operational, Palantir added an LLM layer – this is where Claude sits – that lets analysts search and summarise intelligence reports in plain English. But the language model was never what mattered about this system.

The whole point here is that whether an LLM is involved or not is immaterial to the system as a whole, and it's a disservice to the public to focus on LLMs here.

The article you're responding to is making specific operational claims about Claude's (basically non-) relevance. I'd be interested to hear if you're directionally correct, but forgive me if I need more details than "but it integrates Claude".

This is not a correct take at all given the contents of the article.

Better than carrying water for people who blame inanimate tools for their own personal and professional failures.