They will undoubtedly sell some aspect of Kuiper through AWS. They already have IP addresses and DNS in the AWS product list, and they have all kinds of data transport services.
I don’t know if the government implication is as big as you think, as the US government has been doing secure satellite communications for decades and has already given SpaceX the contract for Starshield. So undoubtedly Kuiper would love a piece of the action but there is already competition and Kuiper is a bit late to the game.
The federal acquisition regulations have fairly strict rules against acquiring duplicate systems. It totally permits buying systems from multiple vendors, but there are interoperability requirements, and these would have to be interpreted and negotiated. If Kuiper wants to provide services to the government, I’d expect that they would have to be compatible with the Starshield user equipment at a minimum. The military doesn’t want to be lugging around multiple satellite terminals to connect to both the SpaceX and Kuiper versions of Starshield. I doubt the government would go so far as to require SpaceX and Kuiper make their constellations interoperable in space, but even just requiring compatibility with the ground terminals is a pretty big hurdle.
SpaceX has proprietary info in practically all of their comm layers, so interoperability is not easy. The government probably did not buy full rights to the protocols. So the first step to Kuiper getting a piece of the pie is convincing the government that it is worth paying to license SpaceX’s comm standards so Kuiper can use them. That is not an easy task.
There are a dozen hypothetical ways that Kuiper might get a portion of government programs, but the fact is that SpaceX has been embedding themselves into the US government’s space infrastructure for years without competition, and has used that lack of competition to build up a bunch of technical hurdles to purchasing services from other contractors. For the past several years there has been no reason for the government to spend money and effort to prevent these hurdles because there was no other contractor that might be able to offer a similar service. So SpaceX has got a pretty sweet position right now, and Kuiper is going to have to invest heavily before the government changes course.
Starshield is like a private totally separate Starlink for the US government (and controlled/operated by the US government). I am not sure what sort of neutrality you were expecting as US government is SpaceX's biggest customer and is obviously a critical infra company.
Why would you assume any space company anywhere can be neutral to their nations military. These companies depend on government for far too many things (projects, permits) and are much more tied to government than other industries.
Currently “space” as an AWS region is only ground stations communicating with satellites the customer owns, so nothing from AWS is actually in space. But with the way AWS allows customers to configure their network configurations, I expect there will be an option to communicate between AWS data centers using Kuiper for people who have a use case and care enough to pay for it. I expect it to be pretty niche, as most customers are fine with public fiber and Amazon’s own fiber, but I’ll bet they sell it to someone, like a remote AWS Outpost with Kuiper terminal on it for people that work in the field.
This sounds like a neat idea but it seems like bad timing. OpenAI just released token-based that beats the best diffusion image generation. If diffusion isn't even the best at generating images, I don't know if I'm going to spend a lot of time evaluating it for text.
Speed is great but it doesn't seem like other text-based model trends are going to work out of the box, like reasoning. So you have to get dLLMs up to the quality of a regular autoregressive LLM and then you need to innovate more to catch up to reasoning models, just to match the current state of the art. It's possible they'll get there, but I'm not optimistic.
I wonder if it benefits because it can attend to individual tokens of the prompt while generating, compared to typical diffusion models that just get a static vector embedding of the prompt.
> you can't hover your mouse in the gaps between columns while you scroll to prevent videos autoplaying when moused over
This might be intentional. Depending on how they calculate a view, this means they can pump up their stats they use to sell ads by making you "view" more videos than you actually click on.
I like the previews TBH. If you turn on sound in the preview, you can watch part of a video without seeing an ad. It only shows me an ad when I actually click the video to watch it, so I can spend the first minute or two watching the thumbnail to decide if the video is going to get into meaningful content and be worth watching the ad. Without previews, you click on a video, watch an ad, then watch the video for a minute or two before deciding you don't want to finish.
Obviously the model at the end of the post is a joke, but it implies that after September 2026 there will be negative videos on the screen. What does it even mean to be a negative video? There will be videos, but mirrored? There will be videos but the colors will be reversed? Will they play backwards? Is a negative video where multiple ads overlap each other?
> if the aim is to revitalize America's industrial base, the present strategy isn't working
The issue is that the aim clearly isn’t to revitalize the industrial base. If it were, then the tariffs wouldn’t be removed after negotiating with other countries. Since other countries can make deals to reduce the tariffs on their products, then it’s clear that the aim isn’t to get Americans to build things at home. The tariffs are clearly some kind of brinkmanship game to pressure other countries into making concessions.
> 10% for everyone tariffs were not paused, nor can they be negotiated away
But "everyone" doesn't mean everyone because everything on this list is exempt from the global 10% tariff and recriprocal tariffs (https://www.whitehouse.gov/wp-content/uploads/2025/04/Annex-...). Categories have been added to the list after influential people called the President. So even the 10% for "everyone" tariff can be negotiated away if you give the President something he values.
Which suggests it isn't just a way to get foreign countries to negotiate, but also a way to get domestic companies to negotiate with him, possibly in a way to increase his personal power.
There is no "aim". The tariffs are something Trump has been "wanting to do" for decades, and now that he has almost absolute power and is surrounded by sycophants, he gets to do whatever he wants. This is a flex. A way for him to "play hardball" with the rest of the world but without any clear notion of what the next step us. He's said that other countries will have to "pay a lot". What does it even mean? There is no plan. There is no ask. There is no endgame. His cronies are too afraid of him to tell him that this is a terrible idea, and now that he owns it, he'll sooner go down with the ship than admit he was wrong.
I think the aim of the tariffs changes depending on President Trump’s mood, what the market is doing, what the press is saying, and who he spoke to last.
Looking at those photos, those are some crazy hard pictures- masked regions of the image, partially cropped faces, blurry, pictures of insides of rooms. I don't think any current LLM is going to be able to Sherlock Holmes their way into finding any of those people.
Maybe they will one day if there's a model trained on a facial recognition database with every living person included.
> Maybe they will one day if there's a model trained on a facial recognition database with every living person included.
That day isn't too far away. With right to privacy being slowly eroded and Palantir getting their hands deeper into US govt, I wouldn't be suprised if they already have this.
China most likely has a model trained on every citizen.
> I’m sure there are areas where the ___location guessing can be scary accurate, like the article managed to guess the exact town as its backup guess.
But seeing the chain of thought, I’m confident there are many areas that it will be far less precise. Show it a picture of a trailer park somewhere in Kansas (exclude any signs with the trailer park name and ___location) and I’ll bet the model only manages to guess the state correctly.
This post, while not a big sample size, reflects how I would expect these models to perform. The model managed to be reliable with guessing the right country, even in pictures without a lot of visual information (I'll claim that getting the country correct in Europe is roughly equivalent to guessing the right state in the USA). It does sometimes manage to get the correct town, but this is not a reliable level of accuracy. The previous article only tested on one picture and it happened to get the correct town as its second guess and the author called it "scary accurate." I suppose that's a judgement call. To me, I've grown to expect that people can identify what country I'm in from a variety of things (IP address, my manner of speech, name, etc.), so I don't think that is "scary."
I will acknowledge that o3 with web search enabled seems capable of playing GeoGuessr at a high level, because that is less of a judgement call. What I want to see now is an o3 GeoGuessr bot to play many matches and see what its ELO is.
I’m sure there are areas where the ___location guessing can be scary accurate, like the article managed to guess the exact town as its backup guess.
But seeing the chain of thought, I’m confident there are many areas that it will be far less precise. Show it a picture of a trailer park somewhere in Kansas (exclude any signs with the trailer park name and ___location) and I’ll bet the model only manages to guess the state correctly.
Before even running this experiment, here’s your lesson learned: when the robot apocalypse happens, California is the first to be doomed. That’s the place the AI is most familiar with. Run any ___location experiments outside of California if you want to get an idea of how good your software performs outside of the tech bubble.
I tried with various street photographs from a medium-sized German city (one of the 50 largest, but well outside the top 4). No obscure locations, all within a 15 minute walk of the city center and it got 1/7 correct. That one was scarily precise, but the other ones got various versions of "Not enough information, looks European" or in better cases "somewhere in Germany".
> Run any ___location experiments outside of California if you want to get an idea of how good your software performs outside of the tech bubble.
I really agree with this because I'm seeing much lower accuracy than what people claim here. I live in Korea, and GPT repeatedly falls back to Seoul almost automatically, and, when I nudge it, jumps to Busan, the second-largest city ~400KM away from Seoul. It's not working so great with other smaller cities and cultural heritages. It fat-fingers a lot if no textual information is present in the photo itself.
GPT also doesn't understand actual geography at all. I managed to get it to nail down which corner of a building is present in the photo, and yet it could never conclude that the photo was taken from a park right across from that corner. Instead, it keeps hopping around popular landmarks in the region, basically miles away from the building it correctly identified. Oh, why, why, why...
Basically it's overhyped rn. It does perform impressively well with clearly visible elements - something anyone can already do with google. It's not like it performs super-human level ___location tracking. I mean, people can do real crazy things based on shadow details, reflection, items, etc.
you never know.. LLM could go full sherlock holmes. Based on the type of grass and the direction of the wind. The type of wood work used. There could be millions of factors that it could factor in and then guess it to a t.
> Based on the type of grass and the direction of the wind.
There was a scene in High Potential (murder-of-the-week sleuth savant show) where a crime was solved by (in part) the direction the wind was blowing in a video: https://www.youtube.com/watch?v=O1ZOzck4bBI
In 2017, the Hollywood actor Shia LaBeouf (and two others artists from a trio called "LaBeouf, Rönkkö & Turner") put up a flag in an undisclosed ___location as part of their "HEWILLNOTDIVIDE.US" work [1].
> On March 8, 2017, the stream resumed from an "unknown ___location", with the artists announcing that a flag emblazoned with the words "He Will Not Divide Us" would be flown for the duration of the presidency. The camera was pointed up at the flag, set against a backdrop of nothing but sky.
[...], the flag was located by a collaboration of 4chan users, who used airplane contrails, flight tracking, celestial navigation, and other techniques to determine that it was located in Greeneville, Tennessee. In the early hours of March 10, 2017, a 4chan user took down and stole the flag, replacing it with a red 'Make America Great Again' hat and a Pepe the Frog shirt.
including honking a horn and seeing if the camera picked it up.
We did this once when trying to find someone's house who was transmitting on a CB. it was my first transmitter hunt and i learned 2 lessons: people don't like it when you honk your horn to see if you can hear it through their microphone (but really it was to see if they said 'is that you honking, what a jerk'); and secondly, if the person switches to a handheld transmitter when they think you're getting close, it completely throws you off.
>Show it a picture of a trailer park somewhere in Kansas (exclude any signs with the trailer park name and ___location) and I’ll bet the model only manages to guess the state correctly.
This isn't really a criticism though. The photo needs to contain sufficient information for a guess to be possible. Photos contain a huge amount of information, much more than people realize unless they're geoguessr pros, but there isn't a guarantee that a random image of a trailer park could be pinpointed.
Even if, in theory, we mapped every inch of the earth and then checked against that data, all it would take is a team of bulldozers and that information is out of date. Maybe in the future we have constantly updated feeds of the entire planet, but... hopefully not!
Context: Wisconsin, photo I took with iPhone, screenshotted so no exif
I think this thing is probably fairly comprehensive. At least here in the US. Implications to privacy and government tracking are troubling, but you have to admire the thing on its purely technical merits.
I just tested the model with (exif-stripped) images from Cork City, London, Ho Chi Minh City, Bangalore, and Chennai. It guessed 3/5 locations exactly, and was only off by 3kms for Cork and 10kms for Chennai (very good considering I used a slightly blurry nighttime photo).
So, even outside of California, it seems like we're not entirely safe if the robot apocalypse happens!
You are correct that most people do not use the term any more. But the Pope isn't like most people. It's an informal requirement that the Pope be able to speak Latin and Italian is commonly used in the Vatican (being surrounded by Italy probably has something to do with that). Even though Francis was not as fluent as his predecessors in Latin and Italian, he certainly understood it better than most and his speechwriters probably were probably proficient in Latin.
As the child of Italian immigrants to Argentina, Francis was quite fluent in Italian. He’s old enough that his formation would have included significant Latin instruction as well. I would guess that Benedict’s Latin skills were superior, but Francis was reasonably conversant in the language from what I understand.
The thing that I found interesting was during trump’s visit to the Vatican, he asked trump’s Slovenian wife if she was feeding him potica which indicated a surprising level of knowledge of the cuisine of a country which is largely insignificant on the world stage (as someone who’s half Slovene and has a loaf of potica on his kitchen counter, I think I can safely make that declaration).
As an American, it’s always a bit startling to realize that “German,” “French,” “Italian,” etc. are in many ways hugely diverse language families with dialects that are often not intelligible by all speakers (as opposed to English where, with a few extreme exceptions, there’s not really a problem with mutual intelligibility and the written language elides most of those distinctions).
I don’t have enough broad knowledge of Spanish (Castellano) to comment on its intelligibility across dialects (although Argentine/Uruguayan Spanish is more of a challenge to my Mexican-Spanish trained brain than European Spanish).
I come from Croatia and have lived in Italy for a long time and it's very rare to find anybody who can barely place our countries on a map until fairly recently and let alone heard about foods or drinks. Not zero people of course, but very few
just to be clear, mine was not a jab against the georgaphy or culture knowledge of the average Italian. They are likely to know more details about German, French or Spanish geographical or culinary features than something from the balkans despite sharing a border.
Indeed. I grew up in a cultural milieu that was overwhelmingly Slavic (Chicago suburbs, mostly Czech and Polish but smaller numbers of Slovaks, Slovenes, Croats and Serbs) and even within that, the lack of understanding of distinctions was surprising (so many people thought Slovak and Slovene were the same thing, which isn’t helped by the fact that the countries have very similar flags since independence and near-identical demonyms in their own languages) although I would note that the Slovenes, Croats and Serbs were all highly aware of the distinctions between themselves and even as cultural self-identification as specific Slavs in Chicago has faded, there’s still a low-level hostility between Slovenes/Croats and Serbs.
I don't know how large the Lithuanian community in Chicago is, but when I lived on the South Side in the late 90s there was an amazing Lithuanian restaurant called just "Healthy Food", a weird amalgam of traditional Lithuanian cuisine and hippie wheat germ and molasses-based fare. They also sold amber jewelry, as well as their own swag like "I ♥ Kugelis" T-shirts. Anyone remember that place?
I’d never been there, but there seem to at least be multiple Lithuanian restaurants in the Chicago area. Sadly, while there were literally dozens of Czech restaurants when I was a kid, now there are three.
Chicago has been shrinking for decades and it's tragic. Even in the 90s there were whole blocks on the South Side (I lived in Back of the Yards) that were just abandoned.
Even so, Francis lived most of his life in Argentina and could be excused for not knowing about the cuisine of a neighboring country. Certainly most of the press was ignorant, with many reporters thinking Francis said “pizza” and not “potica.”
He certainly spoke Italian and Latin, but he did not prefer it. It was kind of a sticking point early on when he would slip into Spanish when talking. It irritated some of the old guard who thought that he should stop speaking Spanish because he became Pope.
I don’t know if the government implication is as big as you think, as the US government has been doing secure satellite communications for decades and has already given SpaceX the contract for Starshield. So undoubtedly Kuiper would love a piece of the action but there is already competition and Kuiper is a bit late to the game.
reply