[JM]: The other day I was traveling in Normandy in France, and I saw the name of a town, and I immediately recognized it as one of several towns that I had seen in a video game that I played when I was probably 14 years old because this particular game is about World War Two and the Normandy invasion.
[JM]: And I happened to be in Normandy where a lot of these sites were located.
[JM]: And I see these names of these towns.
[JM]: And I realized I have not even thought or heard about these names since I was playing this game that I probably have not thought about since then.
[JM]: So I took it upon myself to do something that I've been meaning to do for a long time, which is one, to try to remember the name of this game that I played.
[JM]: Took me all of 30 seconds.
[JM]: Easy to find.
[JM]: Just typed in a few words about helicopters and Normandy and whatever other keywords I could think of.
[DJ]: And it was Leisure Suit Larry.
[DJ]: Great game.
[JM]: It wouldn't come out for a few more years. Because this was back in 1985 when a company called Sir-Tech, which was a company I already loved because they made the excellent dungeon explorer game called Wizardry.
[JM]: And in 1985, they came out with a game called Rescue Raiders.
[JM]: So I'm in Normandy and I'm telling my friend like, oh, I see this name of this French town.
[JM]: I remember this from this game that I played.
[JM]: And so you fast forward a week or two and I'm back at my desk and I think, okay, you know what?
[JM]: I did the little search.
[JM]: I found the name of the game, which at the time I was talking to my friend...
[JM]: I didn't remember.
[JM]: I find the name of the game.
[JM]: It's called Rescue Raiders.
[JM]: And then I think, okay, well, how can I play it?
[JM]: I do have an Apple II+, but it's mothballed and it most definitely does not work in its current state, just due to time and how capacitors age and the fact that none of the floppy disks that I have could ever possibly work all these years later.
[JM]: So I remembered that I had seen a bit of news go by recently that a new version of
[JM]: of a game archive had been released.
[JM]: And this archive is called Total Replay.
[JM]: And it is available on archive.org.
[JM]: I will put a link in the show notes.
[JM]: And version 6.0 had come out.
[JM]: It is a collection of I want to say over 700 games
[JM]: I don't remember how many, but it is a lot.
[JM]: And when I went to download it, I see in the description that the disk image file is a certain size.
[JM]: And I think, okay, well that can't be right.
[JM]: That this must just be like the emulator that plays the games because it said the size of this over 700 game disk image archive is 32 megabytes.
[JM]: And then I realized, no, actually that is the entire collection of over 700 games because
[JM]: a floppy disk, generally speaking, if memory serves, had approximately 143 kilobytes of space on it.
[JM]: So 32 megabytes is quite a bit of space.
[DJ]: What kind of floppy disk are we talking here, Justin?
[DJ]: Because the 3.5 inch high density floppy disks that were popular when I was coming up had, I believe, 1.44 megabytes of storage.
[JM]: That is correct.
[JM]: These are the 5.25-inch actual floppy disks.
[JM]: The ones you're talking about did not flop.
[JM]: The ones I'm talking about most definitely flopped.
[DJ]: No, you're right.
[JM]: And had one tenth of the storage capacity of the three and a half inch hard plastic disks that you're talking about.
[DJ]: Amazing.
[JM]: So I download this archive and now I got to figure out, okay, so I've got the disk image, but I don't have an emulator.
[JM]: So I'm unsure as to how to actually play the games that are on this disk image.
[JM]: I eventually find one called OpenEmulator.
[JM]: From a usability standpoint, it leaves a lot to be desired.
[JM]: It took me quite a while to figure out how to mount this disk image into the emulator, but I eventually figured it out.
[JM]: And in no time at all, I was playing this game that I have not played in decades.
[JM]: And a few things that I noticed about it are, one, it's not quite the same without a joystick.
[JM]: And I missed that.
[JM]: I'm trying to fly this helicopter using my Mac's trackpad and it did not go well.
[JM]: Also because there's two different buttons to fire.
[JM]: Here's a machine gun and then you have bombs and then you push both buttons at the same time to shoot a missile.
[JM]: And to do that was a little tricky on a keyboard, which was much easier on a joystick.
[JM]: So I never quite got the hang of that yet.
[JM]: And the other thing that I learned from doing this is that this game is a lot harder than I remember, or I was a lot better at it before.
[JM]: Probably the latter, I'm thinking, but it's definitely one of those two.
[DJ]: Yeah, I feel like when we go back and play old video games, because this is something I do sometimes also, it feels to me like there's a combination of things behind why the games feel so much harder.
[DJ]: I think one of them that we should really just acknowledge up front is that we are old men now and our reflexes are just not as good as they were when we were like 10 or 14 or whatever.
[DJ]: I think another one is video games used to be much harder than they are now.
[DJ]: So there is that inherent thing to them.
[DJ]: And I think going alongside that, because I remember this from being a kid, I would put up with more because now I have something like 250 video games.
[DJ]: I'm not exaggerating.
[DJ]: Like I've collected all these games, usually when they're on sale from like GOG.com.
[DJ]: And some of them are new games and some of them are these ancient games where I also need to like find one of the three USB joysticks that are still being manufactured to really be able to enjoy them.
[DJ]: But when I was a kid, I had a lot less choice in entertainment.
[DJ]: We had maybe four Nintendo cartridges and we certainly weren't getting a fifth until it was maybe mine or my brother's birthday.
[DJ]: So if Super Mario Brothers was frustrating, you were just going to keep playing Super Mario Brothers because the only alternative was to go play outside.
[DJ]: And you certainly didn't want to do that because your parents kept telling you to.
[DJ]: And no, you wanted to stay in here indoors with the cool video game.
[DJ]: When I say you, I mean me.
[DJ]: So I think there was a there was a degree of like.
[DJ]: Games were harder.
[DJ]: We were probably better at them, but I think we also were just more willing to put up with it because really like how much time are you going to spend trying to get good at rescue Raiders again in 2026?
[JM]: That is true.
[JM]: I had a lot more time before, and I remember moving on to all of these different cities in Normandy when I played this game.
[JM]: In the brief amount of time that I played it, I couldn't get past the first level.
[JM]: Right.
[JM]: Things have definitely changed and by things I think presumably it's me.
[DJ]: Yeah, I think so.
[DJ]: I don't think that's bad either.
[DJ]: Like I remember distinctly a point in time and it was maybe 20 years ago when I was still playing video games, but I realized I was much more willing to lower the difficulty level or use cheat codes or what have you when I ran into challenges.
[DJ]: I didn't used to like to do that when I was a kid.
[DJ]: It was like, no, it feels, it feels lame to switch from normal difficulty to easy difficulty.
[DJ]: So I'm just going to do this level six times until I get past the monsters with the grenade launchers or whatever.
[DJ]: And at some point, and I believe that point is called adulthood, I realized I don't want to do the same thing over and over and over and over for the small amount of joy you get from finally succeeding.
[DJ]: If I'm playing a video game, I want to play the game.
[DJ]: I want to see it.
[DJ]: I want to go on like a tour of the video game.
[DJ]: And so if I try to jump over the bottomless pit and I don't make it, I might try again.
[DJ]: But the third time I'm just going to fly over the bottomless pit or something because come on, like life is short.
[DJ]: So, and I think it's similar here where again, like you could probably play that first level 27 times until you beat it.
[DJ]: But come on, there's an awful lot of streaming video, streaming short form video that you could be watching.
[JM]: Yeah, probably not much chance.
[JM]: I'm going to play it 27 times just to get past level one.
[JM]: And these games, they didn't have difficulty levels that you could set.
[JM]: This was before that trend started.
[JM]: It was just as hard as it was and you either beat it or you didn't.
[JM]: And
[JM]: I don't know if I ever actually finished all levels in this game.
[JM]: I'm pretty sure I haven't, but I hold out hope that with these emulators that I'll be able to essentially freeze the virtual machine where it currently is like its current state so that if I
[JM]: do beat the first level and haven't lost any of my three available helicopters, then I can just set that aside.
[JM]: And that way I can just continue from that point forward and never have to have that frustrating experience of I finally made it to level two only to die immediately and now have to start all the way back at level one again.
[JM]: So I'm hoping that's how it will work, but I am not sure yet.
[DJ]: I have definitely done that when emulating old console games, because so many of those games were based around this idea that getting all the way to the end was going to be really, really difficult, and you were going to have to try probably hundreds of times.
[DJ]: And I suspect a lot of that game design came from both the limitations of the time, but also a lot of these video games started off as arcade games, right?
[DJ]: Where you would go to a place and you would pay money for a chance to play the game.
[DJ]: Like you put a quarter in and the purpose of that game is not to take you on a journey and show you the ending.
[DJ]: The purpose of that game is to keep you putting quarters in it for as long as you are at the arcade until your parents are done shopping at the shopping mall or whatever and
[DJ]: And so it makes a lot of sense that the game's like, oh, you got past level one.
[DJ]: Well, level two is even harder.
[DJ]: Oh, you got past level two.
[DJ]: Well, level three is even harder.
[DJ]: And then it's like, and that level eight is crazy hard.
[DJ]: Oh, you actually got past that.
[DJ]: So it's going to take you a bunch of different tries.
[DJ]: The game is not interested in like holding your hand through this thing.
[DJ]: And you can see this evolution over the years where video games became more story driven and
[DJ]: And they became designed around the fact that they know that you are playing this at home on your couch and you can play it for as long as you want.
[DJ]: And you paid for the game up front once.
[DJ]: That was a joke.
[DJ]: You don't pay for games up front anymore.
[DJ]: You pay using something called a monthly subscription.
[DJ]: But anyway, the thing where they have to make the game really hard and force you to start all over again, I think that was mostly a relic of when video games came from arcades.
[DJ]: And so the first console games are like that.
[DJ]: And one of the features that console emulation software can give you is this ability to save the state.
[DJ]: So that at whatever point you are in the game at this moment, whether the game has a built-in mechanism for saving your progress or not, the emulator can just save your state and then resume it from that point, which lets old fogies like us who don't have a lot of time to mess around with the original Super Mario Brothers anymore, maybe actually make some progress in them.
[JM]: Indeed, and if you want to avoid having to pay a subscription or any money whatsoever, you can go and download Total Replay version six.
[JM]: I said it was over 700.
[JM]: It turns out it features 519 Apple II games, all playable from this single
[JM]: bootable 32 megabyte hard disk image.
[JM]: The people who worked on this went on to say that there were no LLMs involved.
[JM]: They tested every game, squashed every bug, sweated every pixel for seven years to produce this updated archive of games that were literally the birth of personal computer video games.
[JM]: really weren't any before this.
[JM]: This is like the earliest personal computer video games you can play.
[JM]: So if you want to experience a little bit of history, I will put links in the show notes.
[JM]: Go check it out.
[JM]: I know that when I can find time, I'm going to go back through this catalog and see which other games I played when I was a kid, including Castle Wolfenstein, Sammy Lightfoot, that basketball game where you could choose Larry Bird or Dr. J and play one-on-one.
[JM]: So many great games.
[JM]: I look forward to checking them out and I'll probably have another update at some point once I get a chance to.
[DJ]: I'm going to have to check this out, too.
[DJ]: Did you ever play Lemmings?
[DJ]: Do you remember Lemmings?
[JM]: Yeah, I was already on to my Mac by the time I played Lemmings.
[JM]: But yes, yes, that was a great one.
[DJ]: Yeah.
[DJ]: Well, I think that's the first place I played it too.
[DJ]: And not knowing what I was doing, I had all my lemmings in the little level and then I pressed the super cool looking atomic explosion button, which kills all your lemmings.
[DJ]: So you can start the level over.
[JM]: That was often a handy feature, unless you push it accidentally.
[DJ]: Yes, I used it without knowing what it did.
[DJ]: So I felt quite bad for the poor little lemmings when they all crumbled into dust.
[JM]: All right, moving on to some follow up, I wanted to briefly mention this list, Dan, that you found of the small things that were announced at the recent WWDC conference in which Apple often posts like a wall of features.
[JM]: And this particular wall
[JM]: The type was really, really small because they put, I don't even know how many dozens of items that are in the screenshot of this list.
[JM]: And someone thankfully broke that out into actual texts in a longer list form instead of having to try to parse the screenshot.
[JM]: And I just wanted to mention it so that folks can check out this list if you are interested.
[JM]: There are a lot of little things in here that are
[JM]: small but significant quality of life improvements.
[JM]: I mean, this is the kind of release that we have been clamoring for for years, the kind of release where at least if this list is to be believed, Apple went through and just tried to
[JM]: file down the rough edges, tried to polish the things that were rough, tried to fix the thousands upon thousands of little bugs that we have all run into over the years, often the continued unfixed for years.
[JM]: So it makes me happy to see this long list.
[JM]: One of the things that I saw someone highlight that jumped out to me
[JM]: is this thing that I've seen a bunch of times when I'm trying to pay for something using Apple Pay.
[JM]: And there's the default card that's displayed when I go to pay for it.
[JM]: And then I want to use a different card.
[JM]: So I tap this thing that's right there.
[JM]: And the thing that I tap on every time is actually not the right thing.
[JM]: The thing I'm tapping on that I think is the thing that allows me to change to a different card is an element that lets you change your billing address or choose a different billing
[JM]: And every single time I go to change cards, I tap the wrong thing, because it's just put in a weird place.
[JM]: So in any case, they finally changed this.
[JM]: And it looks like the change they made seems very useful, and much better than it was before.
[JM]: And apparently you can even swipe between the cards, which is a much more user friendly and discoverable way of changing your payment method.
[JM]: So I just highlight this as one of what seems to be many, many small improvements.
[JM]: And I am excited about that.
[JM]: you
[DJ]: I'm looking forward to OS 27 actually being released, I presume in September, so that I can update my devices to it.
[DJ]: I don't usually install beta versions of the new operating systems.
[DJ]: I recently saw a post from someone who said they had installed the OS 27 beta and enjoyed it at first, but now they're talking about their phone overheating and these things locking up and all this stuff going wrong.
[DJ]: And credit to them, I don't think they were complaining about it, like how dare Apple release this buggy software that isn't finished yet.
[DJ]: It was much more of a, I saw this coming and here it is.
[DJ]: I took my chances and now I'm paying the price.
[DJ]: But it's tempting when
[DJ]: I think a lot of us spent the whole last year with an annoying red badge on our settings app on our iOS devices because we really didn't want to upgrade to OS 26.
[DJ]: And I'm looking forward to just going back to upgrading my system software and not worrying about it being terrible.
[JM]: It's nice to have this to look forward to, even though, yeah, I'm not even considering installing a beta.
[JM]: And the earliest that this will surface is three months from now.
[JM]: And I'm not going to install it on
[JM]: launch day when it is released, because I'm not an insane person, and have learned that even on launch day, they're effectively shipping a beta, you just happen to be the person who's beta testing it at that point, if you upgrade.
[JM]: So I figure, okay, six months from now,
[JM]: I have something to look forward to and I still have reservations regarding some of the user interface changes that even though they will ostensibly be improved in OS 27, I still feel like even with the things they've done to make it less craptastic, it will probably still be a downgrade aesthetically and perhaps from a usability perspective compared to... Yeah, what is it you're running?
[DJ]: Snow Leopard?
Yeah.
[JM]: Sequoia on Mac OS and iOS 18 on the phone and iPad.
[DJ]: Okay, operating system boomer.
[DJ]: But no, teasing aside, I hear you on that.
[DJ]: And yeah, the 0.0 release is often still pretty buggy.
[DJ]: So you might be right.
[DJ]: It's better to hold out for the 0.1.
[DJ]: To set our expectations here in June, we should think, well, we'll start the new year with a new operating system.
[DJ]: That's what I'll do on New Year's Eve is update all my Apple devices.
[JM]: What could possibly go wrong?
[DJ]: I mean, I could have a really boring New Year's Eve.
[DJ]: I guess that's the main downside.
[JM]: And a really terrible New Year's Day, potentially.
[DJ]: I don't know.
[DJ]: Not necessarily.
[DJ]: I think you either, you know, New Year's Day, it's a fresh start and either you have a nice new operating system or none of your devices work and you get to practice a sort of digital minimalism.
[DJ]: So it's win-win.
[JM]: Yeah, good point.
[JM]: As one more brief bit of follow up, last time we surmised that the reason that Apple said they were delaying Siri AI in the European Union for iOS 27 and iPadOS 27 was due to the DMA's interoperability requirement, specifically in this case relating to the large language model that is powering Siri AI.
[JM]: And I think we probably also forgot to mention that this seems to only be for iOS 27 and iPad OS 27.
[JM]: So I guess that means that for Mac OS, the DMA doesn't really cover it.
[JM]: And therefore Apple is able to ship Siri AI for Mac OS.
[JM]: in the EU, I would assume.
[JM]: But in any case, it's been kind of entertaining, kind of tiresome to see Apple and the EU snipe at each other over this specific topic where a representative from the EU says, yeah, we're not going to let
[JM]: Apple just ignore the DMA and they have to comply like everyone else.
[JM]: And meanwhile, Apple is saying, we've tried really hard to work with European Union and try to offer solutions that would protect privacy and security while still being compliant with the DMA, but they would just not listen to reason.
[JM]: And therefore we had no choice but to not ship this feature at the current time in the EU.
[JM]: And I don't really know that I can easily pick a side on this particular debate.
[JM]: But while I respect what the EU is trying to accomplish with the DMA, I'm just not sure that this particular item makes a lot of sense.
[JM]: Because what they're essentially saying is you as the user have to be allowed, say you buy an iPhone, and you use it for the first time, you have to be allowed to say, Okay, I want to use Google's Gemini, or chat, gbt, or Claude, or whatever the
[JM]: list of anointed choices are you have to be able to choose which of those you want, and not just have Apple's model being the only one that's available.
[JM]: I just don't think this is something that's something you can plug in and swap out.
[JM]: Like for example, a web browser, I can understand a little bit more.
[JM]: if they want to tell Apple, okay, when you tap on a link for the first time on a new phone, you have to prompt the user, what do you want your default web browser to be?
[JM]: Because web browsers are a little easier to swap in and out.
[JM]: But I just don't know that large language models really work that way, particularly with the kind of personal context and deep integration that Apple is shooting for with OS 27.
[JM]: Not to mention,
[JM]: the very real privacy and security concerns that Apple raises.
[JM]: Sometimes Apple raises privacy and security in ways in which feel to me a bit like red herrings.
[JM]: Like we don't want to comply with whatever
[JM]: the thing is, or we don't wanna do this thing people are telling us to, because it's not great for us from a business perspective.
[JM]: But instead of saying that, we're just gonna blame it on privacy and security.
[JM]: I don't really think this is one of those times.
[JM]: Because if you look at the kind of deep personal context,
[JM]: that Siri AI is going to have access to, your email, your messages, your contacts, your calendar, like really all of the data on your device, I can see how Apple is somewhat resistant to be like, sure, we'll just allow some company that operates in another jurisdiction, perhaps with really lax or even malicious
[JM]: stances when it comes to privacy and security to have access to this kind of data.
[JM]: So far as it stands on this particular topic, I think I have to give the nod to Apple.
[JM]: And I hope that they're able to propose something at some point in the future that the EU finds palatable enough for Apple to actually ship in Europe.
[DJ]: I have to admit, I don't really understand the goals that the EU is going for here.
[DJ]: I'm not a citizen of the EU, so I have the privilege of not understanding them.
[DJ]: None of this really applies to me.
[DJ]: But you mentioned LLMs.
[DJ]: We know and acknowledge a lot on this show that, unfortunately, large language models have driven everyone in the world completely insane.
[DJ]: But fundamentally, like they're a feature of software the same way that like a database is a feature of software.
[DJ]: And so when you said that the complaint is around like what model is being used, I'm curious if that's really true.
[DJ]: And if it is true, what's the point?
[DJ]: Because this notion of Siri AI uses a large language model and the context of all the data on your phone to provide you with this rich interactive user experience.
[DJ]: The end goal is the features it gives you, that you can interact with your phone using your voice or freeform text, and it will do a bunch of complex things with the data on your phone.
[DJ]: The large language model that's used is an implementation detail of that, the same way that like a given method in the Swift programming language is some implementation detail.
[DJ]: So suggesting that like Apple has to allow the user to select which large language model will process their voice input and then use APIs to do stuff with like your messages and photos.
[DJ]: is complete and utter nonsense, like it's incoherent, doesn't make any, like, it's like going to a software company and being like, no, you can't use a for loop in this part of your code.
[DJ]: The user should be able to select whether you use a for loop or a while loop.
[DJ]: No, no, no, hold on.
[DJ]: I see you've declared a list to store a sequence of things.
[DJ]: The user should be able to select whether you use a list or a linked hash map.
[DJ]: Everything I just said is animal sounds.
[DJ]: It's nonsense.
[DJ]: It's total bullshit.
[DJ]: No, users don't need to, don't want to, are incapable of, and can't believe that you're even suggesting that they should have any input into, like, the technical implementation details of software.
[DJ]: So wait, like, what is the DMA actually requiring of Apple here?
[DJ]: Because it can't just be like, no, no, no, not Siri, chat GPT.
[DJ]: Like what?
[JM]: So to be fair, I think it's quite possible that I have not fully framed this correctly.
[JM]: It is quite possible that instead what the European Union is saying that the DMA requires is not that Apple be able to plug in other LLMs, but that instead that ChatGPT and Gemini
[JM]: and Claude and all of these other large language model apps that they have to be able to access the same data that Siri AI can access.
[JM]: In other words, they have to be able to access all of the email, all of the messages, everything you do on the device, just as Siri AI does, they have to be on even footing.
[JM]: So that is more likely what the European Union is saying that the DMA requires of Apple to do.
[DJ]: That makes more sense.
[DJ]: And I assume that really, really, like really, really, under the hood, what the European Union wants is that some company that builds large language model powered software in the European Union has access to the same things as Siri AI does, right?
[DJ]: Because
[DJ]: I understand giving consumers choice or whatever but there's something very illuminating about the world in which we live if the governmental authority of Europe is saying to one of the largest most powerful American technology companies in the world
[DJ]: What you need to do is give these other four of the largest, most powerful American technology companies in the world the same access to your users' data as you have.
[DJ]: Because how does that help Europeans or Europe at all?
[DJ]: All it does is continue to re-entrench the...
[DJ]: concentration of all economic power in the hands of American technology companies.
[DJ]: So like, two thumbs up?
[DJ]: DMA?
[DJ]: Like, come on.
[DJ]: I presume that the real goal, I hope, please desperately, that the real goal is to try to make things more available to like European companies and European providers of software and services, if those even exist.
[JM]: I think the intention of the DMA is probably less about ensuring that European-based companies have a better chance of success in the business world and more about preventing gatekeepers from restricting competition.
[JM]: But obviously, if there's more competition, that means that European companies or companies anywhere else on the planet have a better shot of actually being able to
[JM]: provide solutions for European customers.
[JM]: So it is a necessary but insufficient condition for European companies to even be able to walk in the door is for them to be able to have access to the same kinds of API's that Apple has or that Google has.
[JM]: And speaking of Google, they decided to go in a different direction than Apple as it relates to how to approach the DMA and integrating models into their software.
[JM]: And it seems that Google went for the ask for forgiveness instead of permission route specifically, and just went ahead and shipped features without there being any interoperability at all.
[JM]: And it doesn't seem like it's going very well, because the European Commission has demanded that Google
[JM]: open up Android devices to other third-party model-related services.
[DJ]: When you say model-related services, you mean you're talking like fashion models?
[DJ]: Like services related to...
[DJ]: So Google has to open Android to like what?
[DJ]: Like services so that cosmetics companies can more easily employ models?
[DJ]: Are you just dancing around the phrase AI assistance, Justin?
[DJ]: Is that what this is really coming down to?
[JM]: The amount of cognitive hoops that I have to jump through to avoid saying the phrase AI gets really, really tiring.
[DJ]: I know, I know.
[DJ]: I just wanted to take that burden off of you for just a moment.
[JM]: You know, I usually say generative software when it comes to this, because I can't always say large language models, because that only defines a portion of generative software.
[JM]: And sometimes generative software is cumbersome to say, and sometimes I just don't think of it.
[JM]: And so I sit there and stumble until you accurately point out that I'm just dancing around, not saying, you know, trying my best not to say AI.
[DJ]: So yes.
[DJ]: Yeah, I mean, whatever we think about the use of the term AI, and we do try to avoid it because it's mind poison.
[DJ]: But no, I mean, it's just become such a misleading and content-free term that I think also pushes people in weird directions when they think about it.
[DJ]: So we try to avoid referring to things as AI.
[DJ]: But the parties involved in this situation are talking about what they call AI assistance.
[DJ]: Right.
[DJ]: Which is the subsystem of software on your phone that lets you say, yo, dingus, send the book tickets to this concert and then send the confirmation to Justin or something like that.
[DJ]: Right.
[JM]: Yeah, in any case, it will be interesting to see whether the EU comes down hard on Google because they kind of have to if they've already told Apple, who presumably asked for permission and was told no, they can't very well just let Google go ahead and continue to ship software that they just shipped without being compliant at all.
[JM]: So it looks like a final decision on their investigation will be made no later than July 27th.
[JM]: So we'll find out soon enough.
[DJ]: I really am interested in what the outcome of this will be.
[DJ]: The more pessimistic take is that users of smartphones in the EU are just not ever going to get some of these features because every time Apple and Google want to do a certain type of thing with their integrated platform, the EU says, no, actually, you have to do it this way.
[DJ]: And
[DJ]: It's not totally unreasonable for them to say, OK, no, we're not going to do that.
[DJ]: So you just don't get this feature ever.
[DJ]: Sorry.
[DJ]: I hope you keep buying our stuff anyway.
[DJ]: You will.
[DJ]: You don't have any other choices.
[DJ]: But, you know, but on the other hand, the more optimistic take is that Google and Apple eventually will find ways to deliver the same or similar features to European customers in a way that doesn't.
[DJ]: does allow for more interoperability, I think that would be good and I want it to happen.
[DJ]: I'm just not sure I have that much faith that it will, but we'll see.
[JM]: There is a bit of interesting game theory going on here, right?
[JM]: Because the whole point of introducing legislation like this
[JM]: is the European Commission is saying, we have this huge market, and if you wanna offer your product in this huge market, this is what you have to do.
[JM]: And if you don't do it, you're gonna miss out on this massive market.
[JM]: But if you look at it, it's really not as big as I think they might pretend that it is, because I believe sales in the European Union make up 7% of Apple's total revenue.
[JM]: Now, 7% is not nothing, but it's also a number that's not 30%.
[JM]: And when you have only two companies that really meet the qualification of gatekeeper, according to the DMA, as it relates to this specific topic of mobile operating systems that have enough users, then now there's only a few possible outcomes.
[JM]: One is it goes the way that the European Commission intends it to go, which is with a lot of grumbling and delays, Apple and Google ship software that allows for this interoperability that they demand.
[JM]: Another scenario is that Apple decides, you know what, we're not gonna do this.
[JM]: And Google decides, okay, well, we've already shipped it without the interoperability,
[JM]: And we now can either remove it or figure out how to add that interoperability.
[JM]: And so let's say that Google adds it.
[JM]: So now there's a bit of a competitive edge in Google's favor.
[JM]: So if you are using an Android device, you have the ability
[JM]: to have this choice between who powers your generative software stuff.
[JM]: But with Apple, you don't.
[JM]: And maybe that makes a difference in terms of people's purchase decisions.
[JM]: Maybe it doesn't, but it could.
[JM]: And so maybe if Apple wants to avoid that, they capitulate and they add that same interoperability.
[JM]: Or maybe it goes the other way around and Apple adds the interoperability and Google says no.
[JM]: But so that's a second possibility is that one of the two decide to play ball and the other doesn't.
[JM]: And then the third scenario is that both of them say, No, thanks.
[JM]: And neither of them opt to ship these features that have these interrupt interoperability requirements.
[JM]: Instead, they just don't ship those features, it'll be really interesting to see what
[JM]: the European Commission decides to do if that happens, because that means that users in the European Union are not going to get these very useful features that they know other people on the planet get to have, but they don't.
[JM]: And it's because their government is making it that way.
[JM]: And I wonder how long that will stand if that comes to pass.
[JM]: All right, moving on to one last bit of follow-up.
[JM]: I wanted to note a bit of unfortunate news, and that is that Hetzner recently announced that they are increasing their prices again.
[JM]: At first, when I saw this news, I thought, okay, well, yeah, I already saw this.
[JM]: Yeah, they're racing their prices.
[JM]: And then I realized, no, no, this is new.
[JM]: They're racing their prices yet again after what was, I don't know, a month or two.
[JM]: I don't remember when it was.
[JM]: Maybe it's longer than I realized, but it can't be more than three, I would think.
[JM]: In any case, it's a sizable hike as well with a, say, mid-range-ish server that used to cost $18.50 a month is now going up to $25.00.
[JM]: And obviously this is happening because the price of RAM and hard drives and solid state drives and seemingly every single component that relates to computing whatsoever has skyrocketed.
[JM]: And the reason for that, as we've discussed, is because the so-called AI hyperscalers are buying up all of those components.
[JM]: and making it either very expensive or impossible for many other folks to get those components.
[JM]: So this is just another sad reminder of what is already going on and which seems to be getting worse before it gets better.
[JM]: And speaking of the hyperscalers, Anthropic recently announced Claude Mythos 5 and its companion Claude Fable 5.
[JM]: Mythos was released recently as an exclusive tool for security researchers.
[JM]: The thought being this model is so powerful, it has an ability to discover security vulnerabilities that is unparalleled and therefore it should not be delivered to the general public because it could be used to discover all kinds of
[JM]: software flaws and vulnerabilities that could be exploited.
[JM]: So the idea is that they released it to what was supposed to be a select group of people until some other people figured out how to get access to it anyway without too much trouble, which was a bit of a scandal at the time.
[JM]: But last week, Anthropic released a new version of Mythos, version 5.
[JM]: and also released its companion model Fable 5.
[JM]: And Fable is, as I understand it, essentially the same model as Mythos, but with a whole bunch of extra guardrails attached to it to prevent people from using it to do things like find security vulnerabilities or distill models so that competitors are prevented from using Fable to make their models better.
[JM]: and a few other types of prompts that Anthropic was trying to prevent.
[JM]: So much so that many folks typed in really innocuous prompts and were basically told, sorry, you're being blocked from doing that.
[JM]: We're not going to answer that question.
[JM]: And the reception related to Fable 5, I think was kind of mixed.
[JM]: It seemed that some folks had really good results with it.
[JM]: particularly for long running, complicated, here's an idea, I want you to go build the whole thing top to bottom type of requests.
[JM]: Whereas other folks said that for everyday code generation, or analysis with an existing code base that in some ways, the previous Opus 4.8 model got better results for them.
[JM]: But in any case, I don't have any personal feelings about it one way or another because I never got a chance to use it before the other shoe dropped.
[JM]: And that is that just mere days after these new models were released, the United States government told Anthropic that they had to withdraw that model and could no longer offer it to anyone as far as I understand it.
[DJ]: Does that include people in the United States?
[DJ]: Because what I heard was that they put export controls on it, basically, which at least means that you and I, who are not in the U.S. at this moment, can't use the model.
[DJ]: I didn't actually look further into, is it still available to customers in the States?
[DJ]: Or is it just, did Anthropic just be like, okay, no, never mind.
[DJ]: No one can use this.
[DJ]: Either way, it's a very unusual sequence of events.
[DJ]: And to me, like it does feel like it flies over the heads of those of us who are mostly just normal consumers of this generative software, like flies over our heads in a couple of ways.
[DJ]: One of them you just sort of pointed out, which is that Anthropic came out with this model.
[DJ]: And for most of the things that most of us are doing day to day,
[DJ]: with anthropic software, does it matter that much?
[DJ]: I was already successfully doing a lot of stuff with Opus and even Sonnet and et cetera, and now there's this thing that's massively more powerful.
[DJ]: What's interesting to me among the things that are interesting about this story is
[DJ]: For a large set, I think of what we could call like normal tasks.
[DJ]: It feels like these models are already good enough.
[DJ]: But of course, the companies behind them can't justify their current state of existence on the notion that we have shipped a product that's good enough.
[DJ]: Now just pay for it and we'll all be happy.
[DJ]: They have to totally reinvent the trajectory of the entire human race for the rest of time.
[DJ]: Right.
[DJ]: Because they've already convinced lots of people with lots of money that they're going to do that.
[DJ]: And so you have to invest hundreds of billions of dollars into like steamrolling entire cities to build data centers.
[DJ]: That's not something that needs to happen or can really be justified by the hey, this thing's already pretty good at writing computer programs and reading your email for you.
[DJ]: So it's weird to watch this, what feels like kind of an inflection point where like now the companies that are building these models are going past the point of basic utility, trying to figure out what comes next.
[DJ]: And no sooner have they taken a step in that direction, then there's a lot of like, oh, wait, is this a bad idea?
[DJ]: Should we not be doing this?
[DJ]: I'm not saying that that's exactly...
[DJ]: what went on in this case, but it does feel that way when anthropics like, well, first with mythos, they're like, here's a model we can't let anyone see because it's too dangerous.
[DJ]: All right.
[DJ]: There's, there's a bunch of stuff in that statement to unpack, but then it's like, okay, we've released this version of that model that has a bunch of guardrails and the U S government is like, no, not good enough.
[DJ]: You can't release it.
[DJ]: It's too scary.
[DJ]: This is weird.
[DJ]: This is a weird landscape we're on now.
[JM]: It is really weird.
[JM]: To answer your previous question about how this went down and who it affects, the US government cited national security concerns and subsequently issued an export control directive suspending all access to these models by any foreign national, whether in the United States or outside the United States.
[JM]: including foreign nationals that are Anthropic employees.
[JM]: And so Anthropic decided to respond to this by just pulling the models entirely from everyone for reasons that I think are kind of understandable.
[DJ]: Among other things, I would prefer that they not go to great lengths to try to vet which of each of their users is an American national, right?
[DJ]: Because that would only make the product more of a surveillance hellscape.
[JM]: Right, and it's just silly for them to have to be like, okay, to their own employees, you guys can use this model, but you can't because even though you're just as valid an employee as all of our other employees, you happen to not be a United States citizen, so you can't use the model you actively work on.
[JM]: And the absurdity of that leads to the obvious conclusion that they just needed to pull the models entirely.
[DJ]: Yeah, totally.
[DJ]: At least the way that these models work, it is probably not likely that some employee snuck a copy home in their briefcase.
[JM]: Probably not.
[JM]: And this national security letter that Anthropic received doesn't really talk in detail as to what the concern was, but they say that their understanding is that the government believes that it has become aware of
[JM]: a method to bypass or jailbreak fable five, which I believe means getting past the safeguards and essentially accessing the bear model without any of its protections.
[JM]: Anthropic goes on to describe why it doesn't think that particular claim makes a lot of sense.
[JM]: But ultimately, I don't know that reason is going to win the day in this particular argument.
[JM]: So we'll just have to wait and see how it all plays out.
[JM]: And
[JM]: Presumably at some point these models or their successors will become available once again.
[DJ]: There is a question about do you get to pick the word fable from a dropdown?
[DJ]: Sure.
[DJ]: But this isn't the first time that Anthropic in particular and the U.S. government have butted heads on something.
[DJ]: And so it feels to me like there's also a really bigger question being worked out in real time in public right now, which is who has control over what when it comes to big changes to how we all live our lives?
[DJ]: And maybe that's all overblown.
[DJ]: We'll see.
[DJ]: But I certainly get the impression that Anthropic, at least, and the people who run it and own it and work for it, to some extent, don't just think that they're shipping a chatbot and a thing that helps you write computer programs.
[DJ]: I think what they believe is they are working towards ushering in a new era for all of humankind, for better and for worse.
[DJ]: And there are real and legitimate questions in the way our society is structured about who gets to make what decisions about something big like that.
[DJ]: There are very real problems with a government going to a private corporation and just kind of saying, nope, you have to do it this way.
[DJ]: But on the other hand, governments do represent the citizens as a whole.
[DJ]: At least they should, at least for now, while liberal democracy still sort of exists.
[DJ]: What I'm saying is like there are very big issues on both sides of the debate about this private American company with a profit motive saying we're going to transform the fate of all humanity forever.
[DJ]: And also on the side of the government of the country in which they're incorporated saying, no, you can do this, but you can't do that.
[DJ]: And you have to do this, but you won't do that.
[DJ]: These feel like really big fundamental questions that have been explored before in human history, but we're seeing them play out super rapidly in this particular sphere right now.
[DJ]: And that is both very interesting and also sort of worrisome.
[DJ]: Like personally, I'd rather keep my head down and just use clod code to write like Python apps.
[DJ]: But it feels like in that same zone, there's also this other crazy stuff about the fate of the entire human civilization that's going on.
[JM]: Yeah, it is very strange.
[JM]: And we've seen this technique used before when it came to encryption, where a government says, this is essentially a weapon.
[JM]: And so we are going to add export restrictions, and you cannot ship this software with this high level of encryption, because we have deemed this to be a weapon.
[JM]: And it seems like taking a page from that same playbook, we're just seeing it repeated here with okay, instead of strong encryption, it's now this allegedly very powerful, large language model that could be potentially used as a weapon of some kind or be used in a way that
[JM]: compromises national security.
[JM]: And then you have to wonder, okay, that's possible.
[JM]: Or is this just another salvo in what amounts to a vendetta over a previous disagreement regarding these two entities?
[JM]: And there's no way to really answer that question as of today.
[JM]: But somehow I get the feeling that we haven't seen the last of this particular feud.
[DJ]: No, I don't think so.
[DJ]: And if you would indulge me in a little speculative fiction, I just want to do this thought experiment because there's something very interesting here where, you know, the world that we've grown up in has for the last 40 years or so been relatively free of like big conflict between big nation states.
[DJ]: I guess I'm talking about how
[DJ]: I grew up hearing about the Cold War between the US and the USSR, but that was pretty much over by the time I was out of childhood.
[DJ]: But this whole notion of this technology is a weapon, but it's also available to everyone is interesting for the way that nation states work.
[DJ]: Because on the one hand, in a global economy, it shouldn't really matter to Canadians, Australians, and Europeans who
[DJ]: who built the best chatbot, if they can all use the chatbot, right?
[DJ]: But if a country goes, well, actually, LLMs are a weapon, like they're a national security concern.
[DJ]: Well, it follows from that that every country needs their own LLMs in the same way that every country needs their own stockpile of nuclear weapons, to go back to the Cold War analogy.
[DJ]: And again, obviously, I'm not saying it should be the case.
[DJ]: I don't actually think anyone should have nuclear weapons.
[DJ]: I'm saying that like,
[DJ]: From the perspective of a nation state's national security, they would see it that way.
[DJ]: If LLMs are a weapon, we can't rely on this other country's LLMs.
[DJ]: We need our own.
[DJ]: But in order to build these really powerful LLMs, as far as we understand, you need access to all this computing hardware.
[DJ]: And there's a strictly limited amount of that in the world.
[DJ]: There aren't very many companies that manufacture it.
[DJ]: They only exist in certain places, and as we have just been lamenting because we can't afford to buy a hard drive for our Unraid servers anymore, American AI companies have already locked up a lot of the supply of that stuff.
[DJ]: So if other countries made the same determination that it seems like the U.S. is doing...
[DJ]: Go like, oh, well, LLMs are a weapon.
[DJ]: We got to build our own LLMs.
[DJ]: How would they do that?
[DJ]: Because like, how would they get the hardware to build the data centers they need so that the companies that would have to be founded in those countries to develop the things would could actually like you'd get you can imagine this hypothetical arms race so called to like LLMs.
[DJ]: But America has basically this gigantic head start on that arms race.
[DJ]: At least it seems that way.
[DJ]: Meanwhile, maybe there's like China hypothetically over here.
[DJ]: who does build and develop their own LLMs, and maybe they even have other LLMs that, for the purposes of national security, that they haven't told anyone about.
[DJ]: This is why I said this is speculative fiction.
[DJ]: I don't know anything about any of this actually being the case.
[DJ]: But this is what's coming to mind when we talk about the weird, as you put it, like a feud that we're seeing between Anthropic and the U.S. government.
[DJ]: And maybe it mostly is a feud between like two rich, powerful people that don't like each other.
[DJ]: Sure.
[DJ]: I mean, stranger things have happened.
[DJ]: But on the other hand, there is a really interesting, at least novel to be written about the near future where there's this arms race in LLMs.
[DJ]: And how would that arms race actually play out?
[JM]: I suppose if a government were to follow the Tim Cook doctrine, a more sensible thing than trying to limit the use of a particular large language model service would be to try to make sure that that country has enough
[JM]: RAM, solid state drives, hard drives, CPUs, GPUs, that would seem to be perhaps a more effective control than the tactic that's being used here, but also probably a lot harder to do than just issuing a decree from on high.
[JM]: All right, moving on other news, SpaceX has gone public.
[JM]: And for those who are keeping score at home, SpaceX is currently composed of a company that launches rockets into space and deploys satellites among other activities.
[JM]: But those satellites can be used for internet connectivity.
[JM]: And that is how the internet service Starlink was born.
[JM]: which is a subsidiary of SpaceX.
[JM]: Fast forward to February of 2026, in which SpaceX acquired xAI for $250 billion, resulting in what I understand to be the highest valued business acquisition in history.
[JM]: And xAI itself
[JM]: as a reminder, had acquired the company formerly known as Twitter a year prior in March, 2025 for $33 billion.
[JM]: So you have this entity that by my count is composed of these three major businesses, launching rockets, satellite-based internet service, and generative model stuff.
[DJ]: Right-wing social networking.
[DJ]: Oh, sorry.
[DJ]: I thought we were still talking about the company formerly known as Twitter.
[JM]: No, you're right.
[JM]: Technically, there's four.
[JM]: General software stuff and the company formerly known as Twitter.
[JM]: So social media, yeah.
[DJ]: So not content to undertake the highest valued business acquisition in history, SpaceX also decided they needed to be the biggest IPO in history.
[JM]: Correct, which happened just three days ago, or at least business days ago, opening at about $160 a share, rising rapidly to 176 on the first day of trading, and then closed at $160, the same place that it started.
[JM]: But still, as you said, the highest valued IPO in history.
[JM]: Now, yesterday, the second day of trading, it shot up 20% to $192.50 a share to reach a market capitalization of $2.5 trillion.
[JM]: Now, today on the third day of trading, as of this moment, SpaceX is up over
[JM]: 15 additional percent to $222 a share, achieving a market capitalization of $2.9 trillion, making SpaceX as of this moment, the fifth largest company in the world by market capitalization.
[JM]: That puts it larger than Amazon, just to really put it in perspective.
[JM]: And let's not forget, Amazon makes money.
[JM]: They make a decent amount of money.
[DJ]: Right, I wanted to say, nothing has ever made it more clear to me that money is imaginary than this whole situation.
[DJ]: Because this whole idea that like, well, this is the biggest company, one of the biggest companies in the world buy a thing called market capitalization.
[DJ]: And their market cap is almost $3 trillion.
[DJ]: And if you're sitting there wondering like, well, what's a market cap exactly?
[DJ]: I mean, it's the number of shares available times the share price.
[DJ]: But like, what does that mean?
[DJ]: What does that mean in terms of like, why is this company that valuable?
[DJ]: And I think the simple answer, although I'm not a financial professional, is it's that valuable because enough people think so.
[DJ]: Because as far as I know, as you say, unlike Amazon, which for whatever else you think about it, certainly makes a lot of money.
[DJ]: As far as I know, SpaceX doesn't make any money.
[DJ]: I think they lose money on the whole.
[DJ]: But people are so convinced that, again, this is going to be one of those companies that ushers in the glorious future of all of humanity.
[DJ]: At least I think that's the selling proposition.
[DJ]: that people have decided that it's actually worth $3 trillion.
[JM]: I think all of the four businesses that we've discussed that make up SpaceX, right?
[JM]: The rockets, the satellite internet, the large language model business, the social network.
[JM]: I think all four of them lose money.
[JM]: So very different than Amazon.
[JM]: And at number five on this ranking of the world's largest companies by market cap, at this rate, SpaceX will soon pass Microsoft to reach position number four, leaving only Apple, Google, and Nvidia ahead of them.
[JM]: And who knows, if this continues, they could even pass Apple and maybe even the other two.
[JM]: Anything is possible in this world that I don't understand anymore.
[JM]: So much so, just a couple of hours ago, I saw the news that SpaceX has acquired the company that makes Cursor, which is a tool for code generation, for $60 billion.
[JM]: which when you compare it to the other numbers we've been talking about, may not sound like a lot of money anymore, but $60 billion is a huge amount of money, just period.
[JM]: And also it's a huge amount of money for what they bought.
[JM]: This is a tool that is used to do code generation stuff, but it trails significantly, I would guess,
[JM]: the user's revenue, every other metric I can think of compared to Claude and OpenAI.
[JM]: And I cannot by any stretch imagine what on earth makes Cursor worth $60 billion.
[JM]: But in any case, I suppose it's all part of
[JM]: the story.
[JM]: Because during this IPO process, SpaceX told investors that it thinks the addressable market for generative software products is worth $26 trillion.
[JM]: Now, to put that into perspective, $26 trillion is roughly equivalent to the entire United States gross domestic product.
[JM]: So if you took all of the value of all of the goods and services that the United States produces, that's what it thinks the addressable market is for its generative software, which is just so absolutely beyond rational thought i don't know how to phrase it any other way.
[DJ]: Like I'm sure that there is more data out there somewhere, but that's why there's so much of this stuff feels like an exercise in, like when I was in the fourth grade, we'd have recess where they let you out of class for an hour so you could run around, and because of who me and my friends were, the kinds of people we were we generally would not run around the playground. Instead we would sit there talking about the superheroes we were going to be, and we would spend the whole period just talking about which powers our superheroes had.
[DJ]: Because I'd be like, no, I can fly, but I've also got like laser beams that come out of my eyes.
[DJ]: And my friend would be like, well, I can do this.
[DJ]: And then it would be time to go back inside.
[DJ]: We'd spent that whole time.
[DJ]: And this feels very similar to me because you can imagine people, very specific people in this case, whose names we need not mention necessarily.
[DJ]: Again, I'm going to assume they're standing next to an elementary school playground going like, well, the total addressable market for AI is $26 trillion.
[DJ]: And one of them's like, oh, yeah, well, this company is worth $27 trillion.
[DJ]: And it's like, where are these numbers coming from, bro?
[DJ]: Like, seriously, where are we coming up with all of this stuff?
[DJ]: This entire boondoggle, and incidentally, now that this boondoggle is –
[DJ]: priced into the public market, many, many people's retirement funds rely at least in part on it.
[DJ]: So that's cool.
[DJ]: This entire boondoggle, like everything else with AI these days, like with OpenAI and Anthropic and these other companies that are buying all the computers and building all the data centers and they're valued at these gigantic, insane numbers.
[DJ]: It all feels predicated on stuff that's just like a guess.
[DJ]: That's just like someday this is all going to be really, really, really, really important and valuable.
[DJ]: And therefore, this.
[DJ]: Therefore, it's worth everything.
[DJ]: Therefore, it gets all of our resources.
[DJ]: And I think a lot of us, you know, I don't know, regular people that don't own nine companies or whatever are sitting here going like, are you guys sure about this?
[DJ]: Because I was told I should just put my money in index funds, but now even the index funds are subject to the madness.
[JM]: Yeah, we are all going to own a decent amount of SpaceX soon enough, because if you own any fund or ETF that tracks the NASDAQ, well, guess what?
[JM]: You're just days away from being an owner of SpaceX shares.
[JM]: And thankfully, the S&P 500...
[JM]: their standards and pours decided, yeah, okay, we're gonna stick to our guns.
[JM]: We're not gonna allow you to fast track your way into our index.
[JM]: So a year from now, those funds will be affected.
[JM]: So good on them for sticking to their guns.
[JM]: But regardless, we're all gonna own this soon.
[JM]: Elon Musk is now a trillionaire, which is a word that when I typed it into my show notes, totally failed spellcheck, but that probably won't last for long.
[JM]: That'll have to get added to all of our spellcheck dictionaries.
[JM]: And one final note,
[JM]: I saw this exchange where a teacher was having the following conversation with a student and it goes like this.
[JM]: Student: So Elon Musk is a trillionaire?
[JM]: Me: Yup.
[JM]: So how did he get that much money?
[JM]: Well, he doesn't really have a trillion dollars cash.
[JM]: He just owns a lot of stock in companies that are valued at a trillion dollars.
[JM]: So those companies make huge profits?
[JM]: Oh gosh, no.
[JM]: They all lose billions of dollars a year.
[JM]: All of them, huge losses.
[JM]: Followed by The Economist headline,
[JM]: "Gen Z mysteriously hates capitalism and no one can figure out why."
[DJ]: It's a mystery.
[JM]: All right, everyone, that's all for this episode.
[JM]: Thanks a lot for listening.
[JM]: You can find me on the web at justinmayer.com and you can find Dan at danj.ca.
[JM]: Reach out with your thoughts about this episode via the Fediverse at justin.ramble.space.