706: No One Is Getting a Good Deal Tonight

10:42John Siracusa:I'm so triggered by this because the group chat that I've brought up many times on the show, and I think I talked about this when I was in my car search, all they do is send 911s back and forth to each other. And one of the two people in the group, or one of the two other people in the group chat, hates the fried egg headlights. I understand what you're saying. I don't know what generation it was, and I don't care. I'm not a 911 person.

11:04Casey Liss:I think I got it right. 996. There you go. Anyway, if you want window widgets that used to be circular, but become not circular when you poke at them with something, maybe that'll come in a future beta. But yeah, real late for changes in macOS look and feel, but they're tweaking it down to the last minute.

11:20Marco Arment:And honestly, I think this is, I mean, so far it seems like they're not really breaking it by doing that.

11:26Casey Liss:No, I mean, they're just cosmetic changes, really.

11:28Marco Arment:Yeah, and I think this to me is, this bodes very well, I think, for the future of Apple's design, their software design, because you know i think obviously you know steve lemay's uh ascent to power here and whether it's related to john turnus's own ascent to power maybe who knows um this is all like relatively recent and so some of these changes i think are kind of coming in hot but they're good like so far all the changes they've made in golden gate uh to liquid glass into the system design they've all been for the better i can't think of a single change they've made that's been that's been worse

12:01Casey Liss:um and all the little siri blob that's next to everything well that's all the time although yes i just saw someone post had to get rid of that so maybe that'll be next week's episode

12:09Marco Arment:if i tried and it works yeah those little like siri like flaps that are next to all the text areas that has to go like those are awful and to have no option to turn them off no there is

12:20Casey Liss:that's what i was saying i think there's an option to turn them off i just system-wide like yeah yeah someone just posted about it i haven't had a chance to test yet because it was literally like minutes before we sat down but apparently it's screen time content privacy settings and turn off siri writing assistance and that removes them supposedly i haven't tried it yet i'll confirm

12:37Marco Arment:for next week yeah anyway so uh i i think it's i think it bodes well because these changes are all like not only are they good but i think they show more more of a return to what we love about apple and the design of what used to be called os 10 like it is getting a little bit more aqua-y but like a modern version of aqua you know last year when when 26 series oss came out and liquid glass was debuted and it was very controversial because it was some ways good in some ways bad there was a lot of it that we liked there and a lot of it was like oh wow this is nice to have a modern look we just didn't like a lot of the implementation details of it and granted some of the foundations of it i think needed to be rethought or thrown away but there was good stuff there and i think what we're seeing in the 27 generation now is the design team not getting rid of all the problems of 26's design but of actually giving us like here's the best of liquid glass here's here's like the promise of it being better realized it still is modern it still is glassy it still has some of some of the good things that liquid glass brought but they're sending off a lot of the rough of it and evolving it in a direction that not only is more Mac-like in a way, but is also just better overall and has a lot fewer of those major downsides.

14:03John Siracusa:Moving right along with more UI stuff, macOS 27 also has the new preview icon in beta 6.

14:08Casey Liss:Yeah, and this preview icon, I'm not a big fan of it, but it is better than the one that came before it. So yeah, moving in the right direction. All right.

14:16John Siracusa:Then last episode, I think it was, we were talking about how LLMs are not deterministic. And you had made a kind of offhanded comment, if I recall correctly, that even at temperature zero, they're not entirely deterministic.

14:28Casey Liss:Well, I gave temperature zero an example of them being deterministic and how that's not pleasing. We were talking about the AI watermarking thing, like Anthropics AI watermarking thing. That was the topic. And, yeah, we have discussed the role of temperature in LLMs in past episodes, including temperature zero. but we have some more feedback on that now, which is just basically reiterating something we've discussed before, but we didn't mention it last week. Right.

14:51John Siracusa:So when it comes to setting the temperature to zero, things are still non-deterministic. You need to consider that floating point is all over these things, and floating point is weird. And there's also a really good blog post that John Syracuse pointed us to by Horace He, who at thinkingmachines.ai, who writes, One common hypothesis is that some combination of floating point non-associativity and concurrent execution leads to non-determinism based on which concurrent core finishes first. While this hypothesis is not exactly wrong, it doesn't reveal the full picture. And there's a very long blog post that follows that.

15:27John Siracusa:That's very good. Additionally, you can check out the IEEE standard for floating point arithmetic, which is IEEE 754, which we'll link in the show notes. That talks about why floating point is weird.

15:37Casey Liss:yeah we got this feedback after maybe the first time we talked about llm since someone said hey even at temperature zero they're not deterministic because of weird floating point errors in the hardware and you're like oh yeah that totally makes sense because i mean a not floating point arithmetic being non-associative when the mathematical operation is theoretically associative is a problem floating point in general is really weird the more you know about how floating point works on computers the more you think how does anything ever work it's just you know there are finite resources we have to come up with a way to represent floating point numbers that can span a huge range.

16:08Casey Liss:And there's just, it's kind of like hashing, like that you can't get blood from a stone. Like you want to hash all the world's data into this number of buckets? Well, you're going to have hash collisions, right? And with floating point, it's like, you want to represent how many numbers in what range and how many bits do you have to do it? Well, guess what? There's going to be holes. And not only are there going to be holes, there's going to be weird, because there are holes, there's going to be weird rounding and there's going to be mathematical operations that aren't associative. I mean, you occasionally see this when someone naively programs a calculator and they use floating point and you can do certain additions and subtractions of simple decimal numbers that an elementary school student could do but your calculator will get it wrong because of the magic of floating point well uh all lms are doing is just tons of math all the time and it makes sense that these floating port weaknesses would accumulate uh so yeah temperature zero theoretically means completely deterministic according to the algorithm that lms are running but practically speaking because of the realities of the hardware and floating point representations and possibly also due to concurrency, it's not actually the case.

17:07Casey Liss:It's almost deterministic, but not quite.

17:10John Siracusa:All right, so with regard to the new photos shared album where you can get like full resolution photos in that in all the OS's 27, Tim Hardwick at MacRumors writes, for years now, Apple has allowed you to store up to 5 ,000 photos and videos in shared albums without impacting your iCloud storage limit. That's no longer the case in iOS 27, perhaps more significantly, if you invite someone to a shared album and they add their own photos and videos to it, As the owner, this too will count against your iCloud storage limit. On the plus side, the OS is 27 and introduce a free alternative. You can create temporary shared albums that allow users to collaborate on short-term projects without eating into your iCloud storage.

17:44John Siracusa:But after 30 days, any photos and videos not saved in your library are deleted. Apple has made it easier to invite people to shared albums. The company says friends and family can now join and collaborate on the web even without an Apple account or Apple device. Existing shared albums are unaffected by the changes, but the Photos app does include an upgrade shared album option for those who want to improve the quality of its contents at the cost of iCloud storage. Friend of the show, Greg Pierce, also writes, it's worth noting that if you create a shared album on iOS 27 or any of the OSS 27, it's not shareable with devices on older OS versions.

18:14Casey Liss:Yeah, I mean, I think all these trade-offs are reasonable and acceptable, although it has always annoyed me that the way they count storage, and I think, well, it doesn't say it here, but I'm assuming they're continuing the practice of counting it towards everybody's storage. You know what I mean? Like, oh, it counts towards your storage, but I think it also counts towards their storage, and even though we know Apple's only storing one copy of this. Anyway, yeah, I had totally forgotten that those 5 ,000 photo limits don't count against your iCloud limit. So they were lower resolution and they didn't count against your limit.

18:42Casey Liss:So Apple didn't have to, you know, give you too much free storage. But now that you get full res, it's back on you. And I guess it's encouraging more people to buy more iCloud storage.

18:51John Siracusa:You know, one of these days we'll get more than five gigs, right? Yeah.

18:56Marco Arment:Not this year, I don't think.

18:58John Siracusa:Probably not this decade.

18:59Marco Arment:I mean, well, honestly, like when you have a new CEO, this is on his list. Like, you know, Tim Cook had like had a list of things that he changed immediately, like, you know, like charitable giving, you know, stuff like that. Like maybe that's on Ternus's list. I doubt it. I can't imagine. I really can't imagine. But you never know.

19:17Casey Liss:Say they change it. Say they make it. They multiply it by five. So you get 25. Does that really make it? I mean, it's embarrassing. They haven't changed it in a long time. But basically, the way it exists is you get a nonzero amount for free. but that non-zero amount is basically useless if you really start using it for any of its intended purposes in particular using it for backups so it's like it's like they they want you to have storage but not enough to do anything particularly useful like if you no matter how stingy you are with storage if you use apple's anything like if you take photos on your phone like you'll fill it and then they'll want to get money from you so i don't think they're ever going to give you so much storage that it changes how many people need to buy the storage it will just change the timer on how long it takes before their phones fill up and they get frustrated yeah

19:59John Siracusa:so what's going to happen first uh apple is carbon neutral by 2030 or we get more than five gigs of iCloud storage carbon neutral easy yeah honestly I think yeah I mean one is a lot easier

20:10Marco Arment:than the other um but they're not going to do the easy one because it'll be too expensive

20:14Casey Liss:like because they haven't done it in so long obviously nobody cares I mean again maybe if it's on turnus's day one list then that's the only way it'll happen but if it's not on his day one list I don't see that bubbling up to his attention in the first five years or so that

20:26Marco Arment:he's then he's ceo hey you know what this is our last episode with tim cook as ceo it is oh that's true i didn't even think about that so we don't even know the day one list could literally be in time for our next episode like some stuff could start happening i mean it's unlikely that we would hear about it or that it would be done that fast but yeah you wouldn't think there would

20:44Casey Liss:be anything happening until like after the iphone event which we'll get to but uh yeah some people

20:49Marco Arment:did ask if we wanted to do like an exit interview for tim cook we already did yeah i kind of think like we've covered that well um i think mine would kind of be more like a you know don't let the door hit you kind of kind of thing but uh you know we should find maybe for next week we'll try

21:04Casey Liss:to find the episodes but we did we basically did an exit interview of tim kook over multiple episodes we really covered this topic thoroughly once we knew he was out the door and since then he hasn't done anything additional worthy of discussion i feel like all right continuing

21:16John Siracusa:with follow-up uh with regard to last week's overtime where we spoke about steven roblez's video showing how his son uses his first iPhone, a lot of people wrote in to say, oh, well, clearly that was sped up because the implication being nobody could use an iPhone that quickly. Steven wrote to us and said, I didn't speed up anything. I blurred some things for privacy, but everything was 1x, which was incredible to me. And then additionally, a lot of people thought it was very interesting, his son's preferences with regard to lowercase i versus capital I, especially in combination with the fact that autocorrect was turned off.

21:49John Siracusa:and Stephen said also on lowercase i versus uppercase i, that's not a text replacement. He actually presses the shift key to capitalize it manually every time. But I told him about text replacement and he seemed intrigued. We are sponsored this week by Tailscale. I know you've heard me talk about Tailscale a lot and I do that even times not like right now because I really and truly love it that much. Tailscale has compensated us for me to talk to you about them, but truly there is no amount of money that I will accept for me to say, This is possibly my favorite software of 2026. And I write software.

22:23John Siracusa:Tailscale is incredible. And what Tailscale brings to the table is imagine that all of your devices, your servers, your phones, your tablets, your PCs, God help you, your Macs, imagine they could all talk to each other always, irrespective of what continent they're on, what network they're on, what firewalls stand in between the two of them. Tailscale makes all of that go away. So all of the devices on your tail net, all of the devices on your network can always talk to each other pretty much no matter what. It is like freaking magic. I really mean it. Again, they've compensated us to talk to you, but truly, it is like magic.

23:01John Siracusa:They can't pay me to say that. It is absolutely incredible what Tailscale can do. Now, you might think, oh, I don't need another VPN. I already run WireGuard, or I already run OpenVPN, or I already have Ubiquiti Teleport or whatever. Well, Tailscale is WireGuard, yes. And so strictly speaking, it's kind of sort of a VPN, but it's a mesh VPN, which means instead of having a hub and spoke model, all your devices just talk to each other directly, which means throughput is very quick. Tailscale is truly incredible. It's absolutely worth trying. And it is free for personal use. Tailscale's personal plan will be free forever for up to six users and unlimited devices.

23:35John Siracusa:No credit card required. They will not ask you for one. Go to tailscale.com slash accidental tech to get started. And it's free. Why wouldn't you try? If you'd like to bring Tailscale to work, you can use the code ACCIDENTALTECH for three free months of any paid plan. Thousands of companies like Microsoft, NVIDIA, and the New York Times have already made the switch. So please, again, Tailscale.com slash AccidentalTech, T-A-I-L-S-E-A-L-E dot com slash AccidentalTech to get started. Thank you so much to Tailscale for sponsoring the show.

24:06John Siracusa:All right, we got some news earlier today as we record this. The Apple event Surprise and Shine has been announced for Wednesday. day september 9th at uh 1 p.m one true time zone uh your three intrepid hosts will be watching at home as per usual uh and i don't know i don't have too much more to say about that i don't know what

24:24Casey Liss:you guys have ruber pointed out that the reason it's on wednesday instead of tuesday is because of labor day so when labor day falls on the monday they shift it uh so that's why it's one day off it's a surprise and shine thing like i mean this event i've as usual with september events these days like everyone knows every single thing that's going to be announced except for like minor stuff that nobody cares about um you know right down to the the specific size measurement appearance potentially the color is everything about the foldable phone if it's not name ultra that'll be a surprise but anyway here's the invitation here what's the surprise and shine like i get it's a rise and shine pun but like is there anything particularly shiny or involving light

25:00Marco Arment:that is supposed to be announced is there anything even surprise i mean this is the least surprising apple event of the entire year usually like we almost we can almost always predict like it's almost to the day if not exactly to the day when it will be we know pretty much everything that's going to be announced in it like it's like because the iphone's leak more than anything else leaks

25:18Casey Liss:and even the leaks of like the airpods with the camera is going to be there german says no they're delayed until next year so anyway uh but yeah i don't i don't know what shine i don't know what the shine is about i don't i don't quite get are the screens really bright or something i don't

25:30Marco Arment:know we'll see i mean in the sun they can they can be reflective it's just actually not a great

25:34Casey Liss:thing usually but yeah i don't know i mean i mean i guess the main uh there's german had a whole big story about uh people who had supposedly used the foldable phone but like you know but we're never going to cover it because by the time we do another episode they'll have been announced uh but one of the open questions it's not that much of a question but anyway like uh the foldable phone the crease what does the screen feel like i mean the expectation is that it will be like all the other phones that use all the same components that apple's using and that they don't have anything special that anybody else doesn't have by this point samsung has already come out with their own exactly like the apple foldable phone phone you know like the same size and shape you know that same form factor presumably using the same screen technology they're selling to apple unless apple has some kind of exclusive deal so that'll be interesting like does apple even make an announcement that like our foldable phone is better than the other ones because of x y and z because at this point like not only is it leaked so much their competitors have had time to already release their iphone fold clones and apple hasn't even released theirs yet so when theirs comes out it's like, oh, that's exactly like the Samsung, whatever.

26:37Casey Liss:It's like, yeah, yeah, it is.

26:39John Siracusa:Then we have some big and semi-breaking news. This all happened yesterday. Thank you, Apple, for considering ATP's recording schedule. New Mac minis, new Mac studios. So we'll put links to the newsroom posts in the show notes. But John, I figured you were going to want to drive this. I will just try to quarterback as best I can. And you wanted to start with the chips. So there are a bunch of new chips announced, including the M6. m6 interrupt me when you're ready uh this is two nanometer it has dual 16 core neural engine up to two times the peak compute over previous generations this is one of the most interesting

27:16Casey Liss:things about the m6 again this is a plain m6 and remember there's not going to be according to the rumors there's not going to be an m6 pro and m6 max or m6 ultra this is it this is the only m6 and there are a bunch of interesting things interesting things about this but i think apple basically led with this is like they doubled the neural engines and i think this is the only one with double other than obviously the ultras which are two maxes stuck together so then you get double neural engines but uh yeah the littlest no suffix plain m6 has double the amount of neural engines in it i mean it makes sense like they use them a lot ai is popular what can we do to beef up the plain m6 on ai tasks without putting on a honking gpu that would make it too big too hot too expensive the neural engine is not that big put two of them in and so they did and again this is the the very first two nanometer chip from apple so maybe that gives them a little bit of

28:02John Siracusa:room and leeway to put it in a second neural engine right oh they have 12 core cpus which is two more cores than the plain m5 there are two super cores if you recall we talked about this a few months ago uh that's two less than the m5's four super cores but they the m6 has four performance cores, which is the same as the M5, or excuse me, I'm sorry, with the M5 has none, and then six efficiency cores, which is the same as the M5.

28:31Casey Liss:Yeah, I remember the confusion about this. I got myself confused and had to re-Google for all this again. When we say performance cores in 2026, we mean the M cores, the medium ones. They're in between. The super cores are the big ones. The efficiency cores are the little ones. The performance cores are the medium ones. And so this has fewer big cores, a bunch of medium ones that the M5 didn't have, and then same number of the small ones.

28:57Marco Arment:I still hate this naming of these cores so much. We should just say big, medium, and small.

29:03Casey Liss:Yeah. I mean, so the M6 has really thrown a monkey wrench into what I hope would happen, which is like, oh, there's this transition period where they rename them and some of them have the old ones and new ones. But like eventually they're all going on the same page and we're back to two sizes, the big and the little. But nope. Nope. They're keeping three sizes. they're keeping three sizes with the same stupid names we're never going to escape it right it's just it's big medium and small if they had just called the medium one something with medium in it like they're called m cores under the cover so it's clear someone was thinking medium but alas that is not the case so we have this these stupid names are going to be with us for a while because they're pumping out chips with three different tiny cores and it seems like they're not going to

29:37Marco Arment:stop yeah and and it just it still bothers me that like the p cores used to be like like the They used to be the big ones, and they renamed them.

29:47Casey Liss:The M5 was announced with P cores, but those same cores changed names to super cores once Apple decided the big ones have to be called super cores. Oh, my God.

29:55Marco Arment:Yeah. But anyway, I think I don't mind there being three different kinds of cores. It's just, you know, there's the temporary, and by temporary, I might mean like years long, weird transition period of us getting used to what these names now mean because they now mean something different. but actually having three different cores i don't think we i don't think we really know yet how much this matters like it's obviously like you know as you say the the m6 has actually reduced the number of its largest cores uh from four to two but then it has four of the mediums

30:27Casey Liss:and it had no mediums before and the mediums are pretty capable so like how does this i mean this arrangement like apple presumably knows what kind of workloads to these chips experience and what's the best combination of cores to handle those workloads with the minimum power so i trust them this doesn't look like a ridiculous combination for a plain no suffix m chip i think this will be good especially with the two neural engines but um yeah well we'll see when people start getting

30:50Marco Arment:these yeah because as far as we know there's no benchmarks out anywhere yet right i haven't i

30:54Casey Liss:wasn't able to find any i haven't seen them maybe they leaked while we're recording this but i

30:57Marco Arment:haven't seen any yet yeah because so we really have no clue how this actually performs besides Apple's kind of lofty claims. Apple's claims for performance usually are correct, but maybe narrow. It's hard to get a really good idea of how it performs broadly until you actually see it in lots of different applications.

31:17John Siracusa:There is a quote from Apple. The world's fastest single-threaded performance and up to 1.2 times faster multi-threaded performance as compared to M5.

31:26Casey Liss:And that makes it sound like the Super Cores are exactly the same speed as the m5 super cores because the only thing they're touting in terms of like a an increase is 1.2 times faster multi-threaded so you know it's the world's fastest just like the m5 was which is fine single core was good in the m5 it's good in the m6 and a little bit faster and multi-thread and a little bit faster and multi-thread is like it does have two more cpu cores granted it has two fewer big ones but it has four mediums so it works out to be better again that's i think that's a reasonable claim oh yeah because well

31:55Marco Arment:Well, it says the world's fastest single-threaded performance, but that could technically be true even if it's tied with the M5, right? Yeah. I mean, it's probably a little bit faster, but I mean, it depends. Like, you know, they have a process shrink here. How did they, you know, how did they, quote, spend, you know, the additional capacity of a process shrink? They probably, you know, they added cores, it looks like. So they probably spent some of the gain just putting more transistors down and having more cores. but you know did they also ramp up speeds in any different ways or is there more cash somewhere on the chip you know that i think that all remains to be seen high yield on blue sky right what's

32:34John Siracusa:out in apple s core or an all-new p core while size isn't everything silicon area per cpu core can tell you a lot that's why i made this handy chart and we'll put a link in the show notes uh the e core is roughly square maybe a little bit less than square than and not that big by comparison the p core is another like i don't know third bigger and then the s core is i don't know another 40 percent bigger than the p core maybe more these are quite different in size especially the difference between the e core and the s core quite a big difference there is a 12 core gpu which is two more cores than the m5 which is a roughly 30 increase in peak gpu compute for ai as compared to m5 50 increased geometry rates for complex graphics it also includes updates to the shader core architecture, dynamic caching, and hardware accelerated ray tracing.

33:24John Siracusa:You can get up to 32 gigs of RAM with 170 gigabytes per second memory bandwidth, which is up 10 % from the M5. Additionally, SME replaces AMX in the M6. John, can you make heads or tails of this for me, please?

33:40Casey Liss:Yeah, this was from a bunch of other people. I hadn't really thought about this, but there was a post from Taras Zaccaro saying the M4 was the first Apple chip to support Arm's scalable matrix extensions. That's SME, scalable matrix extension. Although Apple has included Matrix Accelerator in its devices since 2019, it used a proprietary instruction set called AMX. The developers had to use Apple-provided numerical frameworks to take advantage of the hardware. The M4 chip with SME changed this by allowing programmers to write low-level code that directly targets the matrix hardware. So AMX has been around forever.

34:11Casey Liss:And that has been a frustration of everybody. It's like, oh, AMX is a secret instructions. You can't know about them. It's Apple matrix instructions, but you're not allowed to like write to them like yeah the cpu supports it but i think they might reject you from the app store if you directly use them obviously you're outside the app store you can do whatever you want but they're like don't use these instructions like they're private we reserve the right to change them what you're supposed to use is apple has frameworks that under the cover use the amx instructions they're like apple just use these frameworks but obviously that's somewhat limiting uh especially if you're doing really low level stuff sme is arm like the people who make the architecture that apple chips are based on the arm company they have their own scalable matrix extensions and those are not specific to apple they're on any arm chip they're available to anybody who wants to make them and apple has incorporated them and now they're basically replacing amx so longhorn on uh mastodon said goodbye apple amx it was certainly interesting but sme takes things further from here m4 and m5 both had both sme and amx present hardware but amx and sme run on the same compute units but support for the amx instruction set has gone in the m6 so they have this transition period where they supported both of them but the m6 is say goodbye to the amx instruction set it's all sme from here out and i think the reason apple did amx is like either sme wasn't available or wasn't satisfactory to them so they made their own thing and they said and part of the reason they were trying to keep it private is like we would prefer to maybe switch to the standard, the ARM standard matrix instructions when they're available.

35:39Casey Liss:And so they kept AMX on and they kept it private. They said, please, you go through this framework. And presumably now, if you use the framework on a chip that doesn't support AMX, it will just use SME under the covers. So a pretty good technical transition for Apple, but it's worth noting that M6 is the first one that no longer supports AMX at all. So if you did write directly to the AMX instruction set and broke Apple's rules, your thing is not going to run on M6. Thanks.

36:01John Siracusa:So a very quick side quest introduction here. We are asked, or people imply to us on a regular basis, you know, you know that all the people at Apple listen to this show. And we know that there are some people at Apple that listen to the show. And my understanding is most of them are, you know, engineers that are just trying to get their job done. I don't think we have any, like, executives that listen to us or anything like that. But then, well, I don't know. We have. I mean, we've had executives on the show, but.

36:27Marco Arment:Yeah, I don't know if we still do, but I know we have. All right.

36:30John Siracusa:Well, either way, occasionally something will happen in the world of Apple and occasionally it makes me think, huh, that's weird. So Daniel Luce points out with a very, very specific chart from Apple, M6 compared to previous Mac Mini models. Faster long-form audio transcription, Mac Mini with M6, 10.3X, Mac Mini with M4, 4.8X, Mac mini with M1, not very good. I guess that's 1x. This, I genuinely have no idea if this is pointed at Marco slash us or not, but man, that is conspicuous.

37:08Marco Arment:No, it's not. I mean, the reality is that long form audio transcription done at large scale is a problem that not just Overcast has. That's true. Although I will say, for earlier today, I was looking at some of my numbers and Overcast transcribes about 30 ,000 hours of audio per day.

37:29John Siracusa:Gracious.

37:30Marco Arment:So it's kind of a lot that I do. But ultimately, there are lots of problems out there that need long-form audio transcription as part of their problems. Think about how many business meetings there are and how many products there are out there to try to take your business meeting notes and summarize them. And every, you know, Zoom and all those different platforms that host video conferencing and everything, they all have this. Like, this is a very common need. What about, don't forget the entire, you know, the intelligence community. They might have uses for this. There's lots of uses that are less than above board.

38:12Marco Arment:But there's a lot of need for long-form audio transcription. And it's a pretty exciting, you know, recent thing that computers recently have gotten a lot better at than they used to be. plus chances are it probably shows off the capabilities of the new you know the new neural engine architecture or the new gpus like you know whatever it is like they that probably happens to be a very very strong benchmark for these chips and that's probably why they chose it not because you know an and lal shimpy listens to this show and decided to do this for this part of the thing

38:45Casey Liss:like probably not yeah and this is the biggest benchmark for this is like m6 uh well i mean they're comparing it to m4 i guess but like it's double from m4 they don't compare it to m5 but it shows that they are getting dividends from that second neural engine uh presumably because i don't think this this could come from the cpu and gpu changes alone but who knows maybe it does like that's the thing about these benchmarks is like does audio transcription use the neural engine at all does it depend on what engine you're using we just see the bar charts we don't have the

39:15Marco Arment:real benchmarks from people yet so we'll wait and see yeah by the way it doesn't take 48 mac minis to transcribe 30 ,000 hours of audio a day. It turns out I have dramatically over-provisioned how many Mac Minis I need. You? Well, because remember I mentioned during the story that at some point I fixed a bug that was causing repeat work to be enqueued? Well, when I did that, the usage dropped by like two thirds.

39:37Casey Liss:Well, luckily those Mac Minis are just going up in value. So basically it's like getting a bunch of solid gold bars.

39:42Marco Arment:Well, the reason I was doing these numbers earlier, I was trying to figure out like, how can I best use this excess capacity that I have? So should I do things like summarization and auto-chapterization? What does that look like? How can I do that on a bunch of 16-gig M4 machines? What are my options for that? Should I outsource it? And that's why I was doing that math.

40:05John Siracusa:All right, let's talk M5 Ultra. There is now an Ultra for the M5 generation, which uses UltraFusion to connect two dual-die M5 Max chips to form the quad-die architecture. Now, is that the same as the M3 Ultra, or is that different?

40:24Casey Liss:Well, the M5 is the first one that uses the chiplets, the M5 Max and M5 Pro. They separate the CPU and GPU die. So this is really just a sort of cruel twist of fate of like, oh, you finally got the quad. No, not the way we wanted. It's still just two Maxes stuck together. it just so happens the max itself split into two pieces into a cpu and gpu die so they do tout it as a quad die architecture which is accurate but it's still two maxes they just can't stop sticking

40:49Marco Arment:those maxes together they love it i was briefly very excited and then i'm like oh wait no that's

40:54John Siracusa:that's not what i want it to be all right this has a 36 core cpu the m3 ultra has 32 12 super cores 24 performance cores uh no efficiency cores which is just like the m5 maxes that it's comprised of the m3 ultra has 24 of the old performance cores and eight efficiency cores

41:18Casey Liss:yeah i don't know why there were so many efficiency cores in the ultra back in the day but it's it's such a weird comparison i mean setting aside what apple compares to they love to compare things to the m1 to show big numbers or whatever but they also compared to the m3 ultra because that's the previous ultra there never was an m4 one so you get some weird comparisons like like the core names and the comparison don't even match anymore so uh yeah and and you know i it's pretty easy for the m5 ultra to beat up on the m3 ultra because you know it's a two generation later

41:44John Siracusa:those maxes that are stuck together but it sure does compared to the m3 ultra up to 1.25 times higher single threaded performance up to 1.3 times higher multi-threaded performance which is not bad uh in terms of gpu it is the same as the m3 ultra 80 cores 80 cores which is bananas up to four and a half times the peak gpu compute for ai versus the m3 ultra so that's such a like you look at

42:14Casey Liss:the other numbers 1.3 times 1.25 times for the cpu it's like all right the cpu is better but like you know not that much better but then you get to the gpu thing same number of cores up to four times that's because all these the cores have like the neural accelerators in them and it's a totally in GPU architecture. This is the big leap in the M5 Ultra is that the GPU, even though it looks, even though it has, quote unquote, the same number of cores, those are not the same cores at all. And there is a big difference in that, you know, in the Apple's press release, it's like, oh, everything's great for AI and local AI and blah, blah, blah.

42:43Casey Liss:And yeah, the M5 Ultra is going to be a lot better at that stuff with its GPU than the M3 was. M3 Ultra, rather.

42:49John Siracusa:There is a, quote unquote, more capable media engine. It has dedicated hardware enabled H.264 and HEVC in code to code for ProRes encode to code engines, hardware accelerated AV1 decode only. But John, what's up with that?

43:03Casey Liss:I read that in the press release and I was trying to figure it out. And then I went to Apple's compare page and it was like, this is exactly the same as the M3 Ultra. Like they do say like, oh, they're better or updated or improved. But like on Apple's own compare page, they use literally the same text for the M3 Ultra's media engine and the M5 Ultra's media engine. Maybe the media engines are the same. Maybe they just have the same number of components, but those components are clocked higher or are more capable. Again, we'll find out with the benchmarks.

43:29John Siracusa:Then there is up to a half terabyte of RAM with 1.2 terabytes per second memory bandwidth. So that means you can dump all of your RAM in a second. Am I reading that right?

43:41Casey Liss:I mean, sure, yeah. Dump it where? Into the CPU and out of it again. Yeah, right.

43:47John Siracusa:I mean, it's just the point is that you can move it so much. You can move faster than reading it. Yeah, one second is another.

43:54Casey Liss:Because if it only goes up to 512 gigs and it does double that per second of memory bandwidth, yeah.

43:58John Siracusa:That's bananas. Anyway, Apple says that it's 50 % higher than before. The M3 Ultra, John, you noted in our show notes, the M3 Ultra was 819 gigs per second. So this is roughly 32 % faster by John's math.

44:11Casey Liss:Yeah, I don't know where they're getting the 50 % higher than before. Maybe I'm doing the math wrong. Maybe they are. But again, we'll see when the benchmarks come out.

44:17Marco Arment:I think we are, though, before we leave this little segment, I think we are kind of seeing, though, maybe why we're not getting the quad chip that we wanted which i guess now would be the octo chip with this new construction method because when you look at how the cpu performance is scaling between the max and the ultra the m3 max to ultra instead you know it didn't get 100 faster by doubling the number of cores with that whole second chip it got about 58 faster on geekbench so if that's whatever that's representative of you know so you know there is some losses as as you scale up it doesn't scale linearly with cpus gpus scale a lot better cpus really don't um by that math though with apple's claim here that it's up to 1.3 x higher multi-threader performance when you compare the performance of the m5 uh max which we do have benchmarks for that's actually

45:11Casey Liss:scaling worse than the m3 well how can you yeah i think you have to do more math to figure that up because this is this is compared to the m3 ultra it's 1.2 times faster than the m3 ultra and 1.3 times faster and multi-thread than from the m3 ultra and then you have to go hey all right so the m3 ultra had 32 cores and this is 36 i haven't done all this math to figure out well

45:28Marco Arment:that's what i just did when you were talking about memory bandwidth so instead of listening to that i did this um so it 1.3 times the the m3 uh ultra multi-thread performance would be about 50 000 on Geekbench, but the same ratio scaling up from the M5 Max would be about 55 ,000. So it's actually scaling I think kind of significantly worse in that way than the M3 Max did if this math is alright and if these claims line up to what they should. But we'll know more with the benchmarks, but it does seem like the Ultra similar to how the Ultra has always been the Ultra really scales CPU performance fairly poorly.

46:06Marco Arment:And so if what you need is maximum CPU that's it's not a very efficient way to get that it's much more efficient to cover larger resources

46:14Casey Liss:and larger gpu tasks but like what apple does with the ultra is they always say this and it's true where else are you going to go it's our fastest chip like yeah it's not as efficient as it as you could imagine it doesn't scale linearly but it still is the fastest so if you want the best performance this is what you have to buy and it's a shame and i don't agree with it that's why i was hoping so much that this would be the year that the ultra would not be two maxes stuck together but it totally is and with all the trade-offs and compromises that brings and apple gets away with it because it still is their fastest chip it is faster than their previous fastest chip and if you want the most performance this is the thing that gives it to you so tough luck

46:50John Siracusa:high yield has another diagram uh this time with a visual explanation of how the m5 ultra is designed uh we will put a link to that in the show notes for you to check i don't know if this is

47:00Casey Liss:based on knowledge but it shows basically the gpu sides touching each other so you get a cpu die and a GPU die on your Macs and you take the two Macs and you stick them together and it seems like the GPU side is the side that's touching with the silicon interposer. Presumably this is known because you could look at the interposer on the Macs or something. But anyway, that's the only diagram I saw of this, so I included it.

47:18John Siracusa:All right, the Macs. Both will ship with Mac OS 27 Golden Gate, as you would expect.

47:22Casey Liss:As you would expect, maybe not because that OS is not done yet. So now that puts a timeline on all this stuff. It's like they're going to ship with a beta OS? No, they're going to ship with the release version of the OS, which means Apple has to finalize the release version of the os in time to stick it on these computers in time to stick them in boxes in time for them to be sale on their for sale date which we'll get to in a little bit so this puts a line in the sand and says hey mac os golden gate it's shipping soon all right so they will have

47:48John Siracusa:the mac mini will have the m6 or the m5 pro in the m6 you can get uh with regards to ram 16 24 32 gigs or with the m5 pro you can get 24 48 or 64 gigs storage for the m6 256 512 one terabyte two terabyte for the m5 pro 512 124 in eight terabytes i mean as you go through these you can

48:10Casey Liss:see the strangeness that apple has brought on themselves which is the one with the higher number after the m is the low-end mac mini it's the m6 and the m5 pro but the m5 pro is the more capable more powerful more expensive has more ram has more storage the m5 pro is the quote-unquote good mac mini but it doesn't have the the double neural engines and the other you know the differences we just talked about so there may be some benchmarks i can imagine that might actually run faster on an m6 than on m5 pro if they take advantage of the neural engines that'll be interesting to see the benchmarks as well but yeah the m5 pro is the fancy good expensive more capable mac mini even though it's got a five after the m instead of a six with regard to connectivity

48:53John Siracusa:Both have the Apple N1, which is Wi-Fi 7, Bluetooth 6. They both have 2.5 gig Ethernet, but 10 gig is available. Two USB 3 front ports. Then the M6 has three Thunderbolt 4 back ports. The M5 Pro has three Thunderbolt 5 back ports.

49:09Casey Liss:This is usually the case with the no-suffix M chips. They don't get as much PCIe lanes or whatever, so they have lesser Thunderbolt capabilities. And so you get an M6 with Thunderbolt 4, but an M5 Pro with Thunderbolt 5. that's the way it shakes out.

49:25Marco Arment:Which means that if I get more base model Mac minis with the M6 I still can't do that thing where they cluster together to run big LLMs.

49:33Casey Liss:They want you to get a studio for that.

49:34Marco Arment:Or an M5 Pro version which can also work.

49:38John Siracusa:With regard to video, Tim Hardwick had a really good summary over at Mac Rumors. The M6 Mini can run two displays at up to 6K 60 or 4K 165 hertz up from 144 hertz in the M4 plus the third display at up to 5k 60 hertz over thunderbolt or 4k 60 hertz over hdmi the m5 pro can run all three displays at up to 6k 60 or 4k 165 which is up from 144 again in a two monitor setup the m5 pro model's second screen now supports 5k 120 or 4k 200 whereas the m4 pro supports 6k 60 or 4k 144 the m5 pro mini supports all three displays through one thunderbolt port while the m6 supports two per port uh single display support is unchanged both m6 and m5 pro support 8k 60 5k 120 or 4k 240 i like that apple continues to push this like oh we did 144 hertz

50:31Casey Liss:last time now we're up to 165 yeah keep going but apple sells nothing to help like they they are apple is not selling 240 hertz displays 200 hertz displays 165 hertz displays one they don't sell any of that at any price other people do and you can connect them to your max but it's fascinating that like the people who work on the video on macs and the video drivers and the video hardware are just like pushing as much as they can increasing the specs you know and these are and these are these are minis this is not a video powerhouse but like they keep increasing the numbers meanwhile the team that makes the displays is like 60 hertz is probably fine for everybody all right if you have three thousand dollars you can get 120 3 500 i forget how much that one costs it's so much uh luckily the rest of the monitor market exists with regard to pricing the m6 starts

51:16John Siracusa:at 900 bucks, which is 200 more than the M4 Mac mini. The M5 Pro starts at$1 ,700, which is$200 more than the M4 Mac mini. And some people have done some science and mathematics with regard to the quote unquote best config available for pre-order. The M5 Pro 64 gigs RAM, 8 terabyte SSD,$6 ,800. There's also a chart with pricing. Let's just pause there for a second.

51:41Casey Liss:A$6 ,800 mac mini what what are we doing here i mean i could figure out some things to do with it ramageddon is really just like distorting products beyond all reason because like in the old days when you know when ram and storage were not inflated by multiple times their normal price the mac mini kind of stayed within a mac mini ish range no matter how you expect it because like they were just they would confine the options right and you couldn't make a mac mini that expensive but now it's like i'm sorry if just you know as you add ram and uh storage it very quickly just dwarfs everything else that has to do with this product and the prices become ridiculous

52:21Marco Arment:yeah and and as usual with the mac mini the base model has always been by far the best value and as you try to scale things up like you very quickly explode the price of that computer into something way larger than the than the initial price suggests that it's like in the class of but i mean this is why i own a very large number of base model mac minis and almost none i think oh i have one my one that i use as like my archive uh server in my house has like the 512 ssd and that's it like i have only base models because the base model is a great value compared to everything else and you spec it up at all and that starts quickly becoming less of a good value

53:06John Siracusa:uh 95 mac also has a really great uh chart of pricing history which is worth checking out and then john you took the liberty of doing inflation adjusted versions of that same chart uh again both

53:17Casey Liss:of these will be in the show notes yeah so the like the pricing um the mac mini it started at 499 which is the lowest that it's ever been but that was back in 2014 right and the the chart just showing the number it shows it goes up and then down and then it goes back up again and you're like wow it's the highest that it has ever been or the base model being 899 and it makes sense you know the ram crisis storage blah blah the highest it's ever been and that's why i was interested in doing an inflation adjusted one inflation adjusted shows that no the 2018 model was the highest that it's ever been inflation adjusted the 2018 base model is 1076 dollars and uh yeah that's that's higher than the current 899 but it has gone up and down the the the lowest price ever base model is 712 in today's dollar so the actual lowest one is the 2024 model and inflation adjusted at 650 um so i mean that's that's kind of the changing face of what the mac mini is and how much it costs to field any kind of computer with any kind of chip especially intel chips in it but um yeah the base model is still trying to be a good deal but just everything is every every uh reasonable uh concept of pricing is thrown out the window because ram and storage are the only

54:23John Siracusa:thing that matters anymore in terms of pricing then uh this mac mini is not available for lease through Apple upgrade. Is that right?

54:31Casey Liss:Yeah, I checked. I saw people saying that I went to the Apple upgrade page and there's no Mac minis there. So I don't know why. Again, it can go up to like$7 ,000, but nope, can't lease it. That could be a yet thing. Maybe. I mean, I checked a couple times today. Maybe by the time you hear this episode, it'll be on there, but it wasn't when I checked.

54:48John Siracusa:You can pre-order right now as we record this. You could pre-order that same day when it was announced. They arrive starting on September 22.

54:57Casey Liss:So there you go. mac os 27 golden gate i'm assuming it will be released to the general public by september 22nd if not people who get apple hardware with a golden gate on it are going to have their own private copy of golden gate that nobody else has and nobody else can get uh so i would imagine yeah that's that's their deadline you know assuming people get their hardware on day one which is september 22nd uh golden gate should be released to the public by then yeah i'm guessing it's

55:25Marco Arment:release of the public probably on september 18th or somewhere around there because you know we have as we mentioned like the iphone event is september 9th and usually they give us yeah maybe about a week between the iphone event and the os releases so i'm guessing yeah sometime around september

55:39John Siracusa:18th probably for golden gate and all the 27 releases all right with regard to pricing uh the cpu for the new mac minis this is over if you want to get the m5 pro instead of the m6 that's adding $800 for the binned one or$1 ,000 for the unbinned one. That's a difference of$20 over the M4 Pro. RAM, 24 gigs is plus$200. Excuse me, 32 gigs is plus$400. 48 gigs is plus$600. That one is$60 more than it was for the M4 Pro.

56:11Casey Liss:And so there's other ones you're reading, like those offsets. Apple did, I think, a good job of cranking the prices on the previous model so that now we get the new ones. And it's like, oh, the upgrades are ridiculous, but they're they're almost the same as they were in the previous one like that you'll see this in a lot of the pricing comparisons it's 20 more expensive than the upgrade was to the pro chip before 20 what and this is 60 you can see apple like apple likes round numbers but they're they're also trying to like hold the line on these increases so again it's not increasing by you know 47 or something but 60 and 20 are weird increases and it's only the 48 gig upgrade that that is$60 more expensive than it used to be um but yeah it's it's like i see that and i see like apple's pricing wizards struggling to try to like because i think their plan was hey we'll increase the prices on everything and then when the new models come out we don't have to increase them again because we already did it that's not true of the phones by the way but for the max they did they did increase all the prices but it turns out they couldn't actually hold the line and for some of like the larger ram configuration they're like no actually we undershot a little bit RAM prices still go up.

57:19Casey Liss:We have to be$60 more expensive than that upgrade used to be.

57:23John Siracusa:With regard to storage for the M6, over the base 256 gigs, it's$200 for a half terabyte, $500 for a terabyte, which is$30 more than it was for the M4, and$1 ,000 for two terabytes, which is$80 more over the M4. For the M5 Pro, this is over the base 512 gigs. A terabyte is$300, 2 terabytes is 800. 4 terabytes is 1 ,808 terabytes. If you're a fool and did that last time when you bought a computer, that is now$3 ,800, which is$380 more than the M4.

57:59Casey Liss:Yeah, look at these deltas. 30, 80, 180, and 380. These are some weird numbers. Some weird increments. But that's what the market will bear at this point. They would have liked to have kept these increases the same as they were before, but they couldn't. And again, 8 terabytes, it's going to be$380 more than it used to be for a nice round$3 ,800 added to the price of your Mac Mini.

58:20Marco Arment:And to be fair, 8TB SSDs from Apple have always been absurdly expensive. This is more expensive than I think they have been, but not by... I think it used to be$2 ,400, wasn't it? 8TB? I forget what it was, but here's the thing. Or maybe$3 ,000? I mean, I think when they first were introduced, they were probably actually near this price, but they've always been like multiple thousands of dollars so this is more than they were but it's not we're not talking 10 times more we're not you know we're not talking that level of a difference here well

58:52Casey Liss:when this was first happening everybody's ram prices and storage prices were going up but apples weren't yet because they hadn't done their their price increase it was like hey apple's prices are starting to seem reasonable but apple fixed that all right with regard to

59:04John Siracusa:ethernet you can get 10 gigabit for 100 bucks which is uh 10 bucks more than it used to be with regard to Mac Mini pricing, Gruber writes, if you configure an M6 Mac Mini with 2 terabytes of storage, the SSD upgrade alone, which is$1 ,000, costs more than the entire base model computer, which is$900. So too with the 4 terabyte SSD upgrade for the M5 Pro Mini, which is an$1 ,800 upgrade for a$1 ,700 computer. Cool.

59:30Casey Liss:Just because you want more storage, you're like, oh, the price will go up a little bit if I add storage. And now it's like if you touch that storage tier thing in the configurator, you double the price of your computer as soon as you touch it. Don't do it.

59:41John Siracusa:There's also a new Mac mini ad, which is like the stop motion thing, which is very cute and very cool. There's also behind the scenes Instagram reel, which you should check out. Good stuff. I really enjoyed those. All right, Mac Studio. Maybe this will be John's future computer. The SoC is an M5 Max or M5 Ultra. With regard to RAM, the M5 Max comes in 364864 or 128 gigs with different CPUs for the 4864128. That requires the big boy CPU. M5 Ultra has 96 gigs, 256, 512 gigs if you have the M5 Ultra with the 80-core GPU. Additionally, the 512-gig memory option for the M5 Ultra is coming in late October for reasons.

1:00:24Casey Liss:Well, yeah. It's not available in the configurator now. You can't even see how much it's going to cost. The answer is more than you want to pay.

1:00:30John Siracusa:Exactly right. If you have to ask, it's too much.

1:00:32Casey Liss:yeah and they and so this when i was doing this price comparisons like obviously the like the mac studio the configurations that you could get the mac studio in have just been disappearing over the past month or two right and when i went to the comparison it was like there was nothing left of the m4 based mac studio it's like you get like one ram configuration per cpu and that's it and the m5 model has increased your options again but you're still constrained so if you want the binned m5 max which is like the one with 32 core gpu you get one ram choice and that's it the second you want any more ram you just got to go with the unbinned one which increases the price a lot so a lot of people have been tripped up by this gruber had a poster he's like oh my math on the pricing was wrong because it says oh add 300 for this cpu choice oh i should do that that's a good deal but as soon as you touch that option it's like well a bunch of other stuff is changing too uh you can't have the bin to one uh thing anymore anyway um but there are more i'm saying is there are more options that there used to be but there's still not as many options as there was back in the day when ram and storage wasn't messed up but it is a relief from the the end of the m4 days which is

1:01:33John Siracusa:like you got choices of like two computers and that was it with storage the m5 max gets 5 12 gigs 1 2 4 or 8 terabytes the m5 ultra gets 1 2 4 8 or even 16 terabytes i it i shudder to think about

1:01:46Casey Liss:how much 16 terabytes will cost we'll find out in a little bit but yes the answer is too much

1:01:51John Siracusa:all right connectivity thunderbolt 5 in rdma which is remote direct memory access that's the thing where you can chain them together right to make put put a big stack of mac studios and

1:02:00Casey Liss:connect them together so they can run a model bigger than any single one of them could run it works on thunderbolt 5 that's what you can do with your mac studios if you have a lot of money

1:02:08John Siracusa:the next generation ssd architecture is built on pcie gen 6 according to apple apple's n1 as we said wi-fi 7 bluetooth 6 these are the first time this and the mini are the first time these have come in a mac uh up to eight displays or up to four studio display xdrs at who who even cares about those anymore anyway at full 5k resolution and 120 hertz i actually by

1:02:30Marco Arment:the way on that topic oh no on one of my xdrs i actually don't even know which one it is whether it's the the land the mainland john one or the beach one but um i'm still paying for the annual apple care renewal because i have it like on the annual you should i wish i had not let mine lapse so terrified it's like 180 bucks a year i'm like and i'm wondering like what happens if it breaks like do they give me a new one somehow i bet they have parts maybe like i i'm just i'm gonna keep paying it as long as i can like how long will they let it go i i have no idea yeah i really wish i

1:03:04Casey Liss:was still paying for that but i'm i let it lapse totally forgot about it now i'm just living in

1:03:09John Siracusa:fear great uh let's see this also supports genlock over usbc for precise synchronization between a display in a professional camera capture like the iphone 17 pro i think the mini does too i think

1:03:21Casey Liss:genlock might don't be on both models but yeah this is not a feature that i'm interested in but

1:03:24John Siracusa:they're touting it on their new macs quote from apple simon simultaneously play up to 33 streams of 8k pro res 422 at 30 frames per second on m5 ultra which is cool Pricing. The M5 Max starts at$2 ,500, which is$200 more than the M4 Max. The M5 Ultra starts at$5 ,500, which is$600 more than the M3 Ultra Max Studio. Lease pricing. M5 Max starts at$49 a month. M5 Ultra starts at$110 a month. The quote-unquote best config available for pre-order. Everyone better sit down. m5 ultra 256 gigs of ram 16 terabyte ssd the just eminently affordable 18 300 holy that is an expensive computer i do wonder i mean we're not aware of these prices because we don't shop for these things and uh two of us don't but no no not these prices the price is

1:04:22Casey Liss:about to say like the the prices that people pay for like the big giant cost as much as a car Nvidia things that run LLM training and inference, you know, those things, those have always been tens of thousands of dollars, but I'm assuming those have also increased. I mean, I don't, I don't know how it works. It was like the whole reason the RAM crisis is there is because they bought up all the RAM for those. So maybe those prices are steady. But anyway, when I see like an 18,$20 ,000 M5 all trick and fig, and it's only got 256 gigs of RAM, I'm like, well, if I have $20 ,000 and I want to do LLM stuff, why don't I buy like one of those rack mount Nvidia things that has way more RAM and is way more powerful and that's like$30 ,000 maybe you're getting more bang if like if that's what you care about if running obviously like they don't run mac os like you can't use them as your desktop computer so i get it but at these prices once you start saying$18 ,000 for 256 gigs of ram and and 16 terabyte ssd which is probably you know the main thing is inflating this price it's like that's not that great a deal if you just care about ai stuff you're and you know like to marco's point earlier you're probably better off getting some lesser M5 something Mac studios and connect them together with that RDMA thing than buying the one big honking one.

1:05:28Casey Liss:But if you want to, Apple will sell it to you for almost$20 ,000.

1:05:32John Siracusa:You can pre-order now, arrives September 22nd. Gruber has some comments here as well. It kind of stinks that there are no RAM options for the studio between the 96 gig base and the $4 ,000 256 gig upgrade. Jesus.

1:05:46Casey Liss:I mean, 96 gigs is basically enough for anybody who's using it as a desktop computer but if you want to use it for ai stuff that extra ram really pays and yeah you jump from 96 next choice and the little configurator is add four thousand dollars to go up to 256 which again 256 gigs of ram used to not be that obscene like the computer i'm sitting next to in theory supports 1.5 terabytes of ram obviously a lot slower than the ram that they're talking about here but still if just you know i hate going backwards i hate having things that are the The numbers are lower than they used to be, even if, again, it's not the same kind of RAM.

1:06:21Casey Liss:And the prices are way higher. So that's not great.

1:06:24Marco Arment:Yeah, I recently was looking, I was looking earlier, like, I wonder, could I, with my next purchase, could I fit comfortably into 96 without noticing a degradation of performance? Because I currently have 128, because when I bought this, you know, M3 Max MacBook Pro, whatever it was three years ago, like 128 at that time was a lot, but it wasn't like obscene. It was like just, oh, that's it. That's like a high end desktop. um and i currently use about 92 gigs most of the time so it's like well i can i can fit into 96 like i it would it would still feel like annoyingly like a downgrade though so i i just what what a shame to have to buy a computer in this day and age john yep it must really suck

1:07:06John Siracusa:uh then before we move on there's a new max studio ad which i actually kind of liked um and they do a lot of like play with the the words on the screen it's a brief ad and i think you should check it out not as fun as the mac mini one but still not bad yeah i think that mac mini

1:07:21Casey Liss:mad is really fun and interesting i mean i don't know how well it sells the computer or even relates to the computer but it's a cool looking short piece of animation uh do we want to save your

1:07:34John Siracusa:trip down the configurator for overtime no i'm kidding i'm kidding i'm kidding yeah most of our listeners don't care about this right right i'm so tempted to legitimately do that but i'm not that evil uh so john all right you can handle this however you choose i know i speak for me and probably marco to say i am dying to know if you've pre-ordered anything um i i did ask you to walk us through your thought process with regard to configurations if you want to jump to the end and tell us what you've done or if you want to walk through wait wait before he does casey do

1:08:03Marco Arment:you want to take bets on what he did if anything and and you know the specifics of that oh i i

1:08:09John Siracusa:do i do but i'm not prepared to answer this question so i'm gonna have to walk through

1:08:12Marco Arment:the configurator real real fast all right let's do it together okay so okay first of all i i'm gonna bet that he did place an order today or at least he probably okay so indecisive he's going to buy this mac studio but he probably wants to wait until somebody gives him some kind of like insider discount and that might not be available yet this second that might that might take like a week or two or something i'm guessing though he has decided he will buy this and he has decided what configuration he will buy so let's step through the configurator now together um so

1:08:47John Siracusa:i think m5 max not m5 ultra oh i disagree he'll absolutely go for the m5 ultra really well let's

1:08:54Marco Arment:see before we get there let's see because like maybe some of the storage options might require okay so i'm gonna say memory i'm gonna say he probably wait hold on john how much memory is

1:09:04John Siracusa:in your Mac Pro?

1:09:04Casey Liss:Are you clicking? You should both be clicking on the options because as Marco said, the stupid configurator changes stuff all the time. It throws up, oh, you want to change that? Well, you got to change these. It's exactly like buying a car. You need the premium package now. My Mac Pro has 96 gigs of RAM.

1:09:18Marco Arment:Yeah, so I'm going to say he went either 96 or 128. Let me see. Well, there is no 128. There is 128 on the Max chip, not on the Ultra.

1:09:27John Siracusa:Okay, so I'm going to say

1:09:29Marco Arment:he went 128 Max.

1:09:31John Siracusa:I disagree. I think he's going 256 Ultra.

1:09:35Casey Liss:What are you clicking on? You clicked Max, but then you have to pick, I hate this, you have to pick options before the other one's like ungray. Yeah, yeah. Did you click on the bin or unbinned Max?

1:09:44Marco Arment:I think I just clicked unbinned because I think unbinned is required to hit 96 or to hit 128. See, if he goes, the problem is if he goes, let's see, if he wants 96, he has to get the Ultra. Oh my God.

1:09:57John Siracusa:I'm telling you, he's going Ultra and it's unbinned.

1:09:59Casey Liss:This is just like cars, man. You need the leather package and the all-weather package if you want the heated mirrors. But if you want the heated steering wheel, you can't get the sunroof. No, you're exactly right.

1:10:07Marco Arment:All right. So, Casey, you're betting ultra...

1:10:11John Siracusa:Unbend with 256 gigs.

1:10:13Marco Arment:Okay.

1:10:14John Siracusa:And 8 terabytes. That's my vote.

1:10:16Marco Arment:Yeah, I think the 8 terabytes is probably a given either way. So, you're going ultra, unbend... I already lost track. 256. 256. 8 terabytes. What's your total on that?

1:10:29Casey Liss:Wait, wait, wait. What about trade-in? Are you clicking the trade-ins thing?

1:10:33Marco Arment:We're going to skip that for now. You can't skip that.

1:10:35Casey Liss:That's an essential part of the thing.

1:10:36John Siracusa:Well, I don't have your like serial number and all that jazz.

1:10:38Marco Arment:I'm going to say no. You don't trade anything and you keep it in the attic.

1:10:41John Siracusa:Oh, that's true. Yeah. Either way, you're not going to trade in. So mine says buy for$14 ,300. And it's so funny to me because John is the cheapest. Sorry. One of the more frugal people I know. but when it comes to this and only this john knows no limits and so i think we are looking at at least fourteen thousand three hundred dollars for his computer it would not surprise me if he actually

1:11:11Marco Arment:spends more than that all right i'm gonna go with m5 max unbinned so the 18 core and 40 core it's like the best m5 max because i don't think john can accept a bin chip i think he wants the best version of whatever chip he has so m5 max full cores 128 gigs eight terabytes for 8059 dollars

1:11:36John Siracusa:that is a much better deal bang for well it's both of these are hilariously overpriced but no one's getting a good deal here tonight no no but given that you are a little bit less or a little bit more than half of what i've got that's a much better choice but i know i feel like i know john And I feel like as frugal as John is, there is no turning back when it comes to his computer. And this is how he's going to get over the fact that there is no Mac Pro anymore, is by buying a truly asinine Mac Studio. So, all right, I will leave the configurator up because knowing me, I will forget what I've selected in the next five minutes.

1:12:12John Siracusa:John, walk us through what your thought process was. If you are ready or if you'd like to take this a different direction, that's fine too.

1:12:17Casey Liss:All right, well, so the first thing I did was I looked at the configurator as you were doing and got frustrated with the limitations and the options. But anyway, I immediately went to the storage section to see what are the storage options. And the storage options on both of the models of Mac Studio are unchanged in price. Like if you want a 4TB, it's at$1 ,800. If you want 8TB, it's at$3 ,800 if you get the Macs. And the Ultra is at$1 ,500 and at$3 ,500 because of the vagaries. But anyway, these are exactly the same as they were for the M3 Ultra and the M4 Macs. so storage prices for the mac studio are the same as they have ever been they're awful but when apple increased the storage prices on the m4 mac m4 slash m3 ultra mac studio before they were able to hold the line on those and so they're not worse than they were but they're not good obviously um for the um for the the cpu and stuff mostly unchanged like there are things tied together so if you get like an unbinned 64 gig with four terabytes um it's like 5300 bucks or eight terabytes is 7300 bucks if you bin the chip you get minus 300 for the bin cpu which is not that big of a savings but then you have to go down to 36 gigs of ram like you can't you know the ram is tied to the it's just it's super annoying um for the m5 ultra uh if you go with the binned m5 ultra it's seven grand which is more than the m3 ultra was if you go bin 96 gig it's seven grand these are configs that i was trying um it's 200 more than the m3 ultra config was and if you go eight terabytes it's nine grand which is again 200 more than the m5 ultra so these combos of like bin 96 gig if you want an ultra but you're not going to get the fancy one but you want 96 gigs of ram like that's 200 more and then unbin 96 gig the prices are the same as they were on the m3 ultra which is nice because like m3 ultra to m5 ultra is a big upgrade in the cpu um you know four terabytes is 8300 and eight terabytes is uh 10 ,300 so these are these are like the configs i first looked at m5 max binned uh unbin 64 and bin 36 because i have to like i'm never going to get a bin 36 but just fyi they make you do that um and then m5 ultra bin 96 and unbinned 96 and then i looked at all those numbers and then i looked at everything over um and like i i don't know i mean i thought about this a lot before these computers came out but it really depended on what the prices did and having the prices not move too much with the exception of a couple of$200 more expensive options um and again those$200 more expensive options are for a big upgrade from an m3 ultra to an m5 ultra like i'm you know i didn't buy the the m4 mac studio because i waited too long by the time i realized there was going to be nothing else by the time the mac pro was killed so on and so forth uh by the time i realized nothing new was coming anytime soon it was like now it's too late to get an m4 mac studio although i tell you i did think about it a lot because for reasons i'll get into later in the show like i'm doing a bunch of development stuff mac os 27 i have to test my things on i can't do it on my mac because mac os 27 doesn't run on my mac so i had 27 on the macbook air and then as we discussed in the earlier episode now i've got 27 my wife's computer so i'm using her computer and she doesn't like that i love like you could solve this problem by buying a macbook neo and you won't do that well i thought i mean that's what i'm saying i thought about that i was looking at mac minis what's the cheapest arm mac mini i was looking at ebay like m2 mac minis you know like i was looking for like literally anything they're not that cheap you'd be surprised i know that's what i'm getting at so and even and even the neo like the neo yeah it's like is that the cheapest arm mac even that one increased in price and i don't want that with a laptop and it comes with not a lot of ram and if i'm going to be doing stuff i mean i just the neo was kind of out and the minis like again i'm looking for like used mac minis like i don't care i just needed something arm on my desk that i could just plug into my monitor and have a thing but they were just not that they're super expensive but it's like well if you're thinking about getting a new computer sometime soon don't spend 800 bucks on a computer you're going to use for two months you know what i mean yeah but i'm kind of desperate for something arm i need something arm in my life um and the price is not changing is like oh this actually kind of makes it more difficult because like i don't know like i mean a couple things make this more difficult for my decision making one that the m5 ultra is just two maxes stuck together so it's not interesting or novel the scaling is not better than it used to be it's you know it's not exciting it's the kind of chip that's probably not going to get a chip design on the back of a future t-shirt just saying if they if they had done something exciting i might have drawn another chip on the back but now it's not happening there'll still be a shirt though don't worry thinking i won this bet i think you did win this bet i'm shocked i'm not so uh and then i'm looking at these configs and i'm like you know this is this this goes into my well two things one as discussed on past episodes i screwed up i waited too long i made poor decisions i think

1:17:10John Siracusa:marco and i have told you that for the last three years running right sure but like it just like i

1:17:16Casey Liss:mean and the thing is like oh my god casey he bought a used m4 version the excuse is no you know that by the time we saw the ai ram crisis coming it was already too late you know what i mean like by that point it wasn't too late to buy like i could have bought you know a mac at that time but it was too late for me in the cycle it was like i said it was too late to buy an m4 mac studio because the new ones were coming out but it was also like hey if you wanted to save money you should have bought them before they increased all the prices right so i was i was in a fix and it was it's like in a way that i remember everyone on the internet and youtube are all making fun of me like oh you bought a mac pro and then they changed chip architectures haha this thing has lasted like a champ like longer than most people's computers last ironically apple transitioning chip architectures was less of a disruption to my normal buying patterns than the ai induced ram and storage crisis because that has screwed things up screwed me over way more than the

1:18:06Marco Arment:chip architecture change ever did do you buy enough macs to qualify as a buying pattern

1:18:11Casey Liss:yeah i do i mean if you go over from 1984 till now you know it definitely counts as a pattern word and the pattern is i keep them for a long time that's the pattern um yeah i mean this is this mac pro is gonna end up being like six seven years seven years older so by the time i get something new um so all this is to say kind of like with my constant monitoring of the tv world and the uh the automotive world this is the not not the right time to buy a new computer it's the from in my from my perspective you know from with my rule set the way i buy computers i'm always looking is this the right time to buy is this the right you know with my tv is this the right one to buy should i wait till the next one or should i do it like when is the right time to buy for me this is not the right time to buy a new computer on multiple fronts even even before you get into the ram and storage you know crisis with pricing it's still not the right time to buy you apple kind of agrees that it's not the right time to buy a new computer because apple's like hey we think the m7 and like the ai stuff we're doing according to the rumors that the ai stuff we're doing in the m7 are so good that we are just as casey would say poop canning the the m6 pro max and ultra just so we can accelerate the timeline on the m7s because we think we did such a good job on those that we want to get those out asap and just forget about the m6 pro max and ultra that that's like unprecedented and part of that is probably the ram crisis and chip crisis and blah blah but also So I feel like Apple, given the timelines on their chips, like the whole, you know, the ascendancy of the importance of doing local inference and stuff on Macs, if you extrapolate that timeline backwards for typical silicon design timelines, it makes sense that the M7 is the first one that's like, oh, no, local inference.

1:19:54Casey Liss:Like that's like that's where all we need to spend our transistors. And the rumor is the M7 is really good at that. this is exactly the wrong time to buy a new powerful computer because apple itself according to the rumors is like yeah our our good our good super powerful computers this isn't it it's great that we upgraded it's great that you know we never had the m4 ultra but now we have the m5 ultra and it's fine and it's good and it's better than the m3 one and it is the best most powerful chip we have and if you need one right now now now this is the best we have so tough luck but the m7 would

1:20:22John Siracusa:be the one that i would want to get are you gonna wait is it not that you're gonna buy a used one

1:20:26Casey Liss:but you're just going to keep frigging waiting so i could clearly not choose the wine in front of me

1:20:30Marco Arment:casey he's going to switch to a desktop laptop and get the new ultra touch screen because you love laptops and tuck screens so much i love them so much um so all that said it it demands a specific

1:20:44Casey Liss:strategy from me what should i do in this situation again i screwed up i'm in a bind i i'm in a tight situation this is the wrong time for me personally to buy a new computer according to everything that i believe in in terms of new product buying again cars tvs computers everything uh and yet somehow i need some new thing should i go back on ebay and find one of those mac minis and buy i was scouting i almost like a lot of these on ebay are scams obviously so i got to be careful but there was like a and i think it was like an m2 ultra mac studio with a huge amount of ram and like a like a four terabyte ssd or something for like a thousand bucks or something i'm like this has to be fake but i was interested right like not that i'm you know because again you can't find a mini like you feel you feel bad buying a crappy like m2 based mini or something for like 900 bucks but then you're like m2 ultra max studio for like less than two grand you know like with a buy it now but it seemed like it was a scam so you know you gotta be careful out there people and if it looks too good to be true it probably is uh but i looked into a lot of that and so here i am should i go back to ebay and find something different and let me tell you things have not gotten better on ebay in the time since i've been making these decisions um so here's the strategy i came up with i need something that i normally never get which is a laptop no never which is a short-term mac

1:22:05Marco Arment:a tide me over computer oh my god oh no i need something to just get me over this hump we're in

1:22:12Casey Liss:a difficult period sometimes when the chips are down you gotta do whatever it takes to survive like in goodfellas when you're in the slammer you did whatever it takes to survive but when you come back on the outside you got to leave the junk alone so paulie says right you know understand you got to do and so i'm i'm in the slammer i gotta do what it takes to survive until until my m8 or m9 ultra mac studio whatever arrives because that's the computer i want to buy i don't want to buy an m5 anything mac studio i want to buy something that's after the m7 after that it's Like they really figure out what these things are going to be for.

1:22:48Casey Liss:Because like this is basically the turn here. You know, unless something dramatic changes, Apple is still not particularly interested in gaming. But Apple seems to really be pointing their high performance computing at AI related stuff, which makes sense in the world that we're living in. And it just so happens that Apple's hardware is actually pretty good at AI related stuff. And I can see a future for myself where instead of paying whatever dollars per month to some company for their coding agent, I can run something that is as good as what I'm currently paying them to run on their servers on my M9 based Mac Studio at some point in the future.

1:23:24Casey Liss:Right.

1:23:24John Siracusa:Obviously, you're not going to buy a computer for like five years. Is that what we're getting there?

1:23:30Casey Liss:Oh, my gosh. But this is this is this is what I'm thinking. This is my strategy here, which is like, obviously, by then, the ones that run the servers will be better still. So I'll still probably want to use them. but there is a certain baseline utility of like let's say like a you know current informer sponsor anthropic fable five level model that you could run on your mac locally in some years in the future again the server ones are going to be better by then right but something fable caliber locally basically for free a fable caliber thing can do a lot of mundane tasks coding wise like not maybe your super hard ones not maybe find the obscure bug or whatever but like just run of the mill like write me this automation script do this thing whatever i think that is going to be attainable locally on max at some point in the future but only for people who have the big honking max so it's like why do you ever buy a big honking max my answer used to be so i can run pc games at high frame rates when i reboot into windows that answer is not the case for so many reasons now but my answer in the future i think will be and this is this is a use case that was not a thing before coding agents existed and were good right now it's like i would like to be able to run powerful coding agents locally at some point in my future that's the big hunking computer i want to buy again m8 m9 m10 like who knows what the future but i feel like in that timeline that's the one i want to buy but i do have to buy a computer now because i can't keep running intel i have to i have to buy something so what do i buy i need to tide me over computer you bought one of marco's mac minis no he should probably if i had known it i had all these spares i should have just seen if

1:25:09Marco Arment:i get the friend price that would have really helped right i mean we could talk about it but they're all base models you wouldn't want any of them yeah that's the problem that's the problem

1:25:15Casey Liss:so and oh and as we've discussed on past episodes like i got the eight terabyte problem like i have a four terabyte ssd and it is absolutely full which is why i made a program that saves disk space and i use my own program all the time to save me from death constantly i'm at the ragged edge of constantly every time we do an episode it's another few hundred megabytes it's not helping like i'm so close to filling my drive and i have a four terabyte drive and the only choice above four if you want an internal drive from apple is you got to go to eight and then you're mortgaging your house so how do i deal with this okay so i'm guessing you already had your guess you don't need to guess now after i've explained everything all right all right all right so here let me just tell you what i've done i have ordered a computer and i ordered it like very early in the day because i I had sort of gone through this process in my head multiple times.

1:26:00Casey Liss:Then once I saw the reality, I'm like, it's slotted into one of my decision trees. I ordered myself a Mac studio because I'm not going to get a laptop. We're not going to get a Mac mini. I ordered myself a Mac studio, the M5 max, because it tied me over computer. 64 gigs of Ram, which is less. Wait, which M5 max,

1:26:25Marco Arment:the bend or unbend?

1:26:26Casey Liss:Well, so i got the unbind because that's the one nice thing i did for myself because you're correct marco i cannot you cannot abide a partial processor i cannot stomach a bin cpu if it comes with parts they should all work yes exactly so and and in the grand scheme of things going from bin to unbind with my other specs wasn't that bad so yeah the the unbind one 64 gigs of ram which is less than my my mac pro has but i mean the thing is my wife's computer has 64 gigs of ram okay and so i'm using golden gate with 64 gigs of ram it constantly has four people logged into it all families logged into it because we're all home right now um and with everybody logged into it and everybody doing what they normally do on the computer and then me using xcode 64 gigs i'm able to get buy-in with four people logged in so i think i mean again 96 what i got on the mac pro that was part of my like indulgence for myself and i basically said it means i never have to think about ram and it's true what i do on a mac pro ram becomes not a thing that is in my mind at all i don't even know if i even touch swap so i don't need 96 unless i was doing lm stuff which again i'm not um 64 and you know basically it's your only real option because of the configurator and a four terabyte ssd what okay why because that's what he has now yes but you hate it i know so here's here's the strategy okay so first of all let me just tell you what that that config costs.

1:27:45Casey Liss:MacStudio M5 Max Unbin 64 gigs of RAM, 4 terabyte SSD is$5 ,300. Obviously, I'm going to get discounts. Okay. Leasing, it's$106 per month for three years. And if you did that lease, at the end, you'd have about$1 ,500 left to pay on it. I didn't really seriously consider the lease mostly because I don't know how it works with like Apple discounts. You know what I mean? And I didn't want to navigate that, like getting other people being nice and helping me get discounts i'm like oh but can i also do the lease i didn't like i don't know what the deal with that is so i just didn't bother figuring it out also because i don't know what the delivery timelines are going to be like which is why i ordered this computer within like an hour and a half of it being announced and i did get september 22nd delivery okay now here is the strategy how am i going to live with this have i just screwed myself yeah you got to tie me over computer but you haven't improved anything you're getting computer with less ram and the exact same amount of storage now you're screwed well here's what i did because i looked at that eight terabyte option a lot but it's just it just so much it becomes not a tied me over computer it becomes like a you know a big beefy computer but it's not a big beefy computer that i want so what i did was i bought a samsung ssd samsung four terabyte ssd for like 580 bucks god that's so much cheaper than it is from apple now that ssd used to be like 300 bucks before all this happened so it's doubled in price from what it used to be but compared to apple's prices compared to getting a four terabyte internal and a four terabyte external from samsung it's like a rounding error compared to getting an eight terabyte internal and my plan is photo library on the external disc so right away because my photo library is more than half of my four terabyte ssd it's more than 50 of my drive is just the photo library okay so it's a lot of sense immediately i have a less than half filled main drive and i have a less a roughly half filled photos drive so i got the uh it's a uh samsung 990 evo we'll put a link in the show notes i think there's only like one or two left in stock but it may be lower priced than it used to be i just put in the chat room if anyone wants to buy the lone remaining one um it's not a particularly fast fancy ssd but you can't beat the price it was it was like 40 something percent off on amazon that's why i got that okay and i got uh i had to find a case for it which was way harder than finding the ssd um i wanted a thunderbolt 5 case i didn't care about usb the case like you can get a case for these like a thunderbolt 5 enclosure for for an ssd that cost like 200 bucks so you're paying 600 bucks for ssd and 200 bucks for the case it's rough so i found a cheaper one i got this like 150 bucks from owc it's it is thunderbolt 5 but it doesn't like the i don't either doesn't have usb at all or if it does it's super slow i think this is the one that doesn't have usb at all so if you don't have a thunderbolt computer you can't use this but i don't care it's going to be permanently attached to the back of my Mac Studio.

1:30:36So I don't care about it.

1:30:38Casey Liss:So yeah, that's my tied me over computer. And I can tell you with the discounts that I ordered it, I got out the door for the, just for the Mac Studio, not counting like the external SSD. I got out the door for less than five grand. This is my tied me over M5 Max, Mac Studio with the only indulgence that I gave myself was an unbinned SoC. Everything else is cut down to the bare minimum. I think 64 gigs of RAM will be fine. Again, based on my experience, I think the split SSD thing will be annoying to me. Again, my strategy with that is like, the reason I always like to have it all in one is because it's just so much simpler from a backup perspective, but I'm just going to have to deal with the complexity because the savings are significant.

1:31:17Casey Liss:If I had, if the only change that I had made from on this computer has gone from four to eight, it becomes a$7 ,300 retail computer from a$5 ,300, just$2 ,000 difference. And I took, you know, less than like a quarter of that$2 ,000 difference and put it towards the external drive and enclosure and everything like that. so yeah not a great time for me to be buying a computer but i did have to buy a computer and i did buy a computer and it's not the computer of my dreams but it is a computer and oh here's my other strategy with this this is my wife's next computer she's got a max studio now it's fine for her now by the time the m8 or m9 is out that probably won't be fine anymore maybe macOS will drop it she's got she's got an m1 max max studio with 64 gigs of ram performs does everything she needs perfectly fine it's a great computer it's faster at xcode than my computer obviously right like everything about it is great she has no problems with it but someday support for it will drop this is her future computer an m5 an unbidden m5 max max studio with 64 gigs of ram and a four

1:32:14Marco Arment:terabyte ssd all right that that i was gonna try to talk you out of doing this but that last bit with replacing your wife's on that cycle i think that actually makes a lot of sense and that's

1:32:22Casey Liss:also the reason that i'm not too uh too worried about not having the lease because the lease would make sense oh it's a time you ever give you you should totally lease it and then give it back that's perfectly what you do but it's like no this computer has a future purpose in life and And it will be fine for that purpose for her computer, especially if three years from now, one or both of my children are out of the house. We'll see how that goes.

1:32:42Marco Arment:Yeah, I think the the only thing I would caution you against is what if your assumptions about three years from now or whenever that is like, what if one of two things happens? Either it doesn't go in that direction and, you know, and they they're just incrementally better with three regular years of progress or and or what if three years from now computer resources are even more expensive and the the expected fall from from, you know, an increase in supply either hasn't happened or demand has just continued to increase. And so we still haven't caught up with supply.

1:33:20Casey Liss:I mean, I mean, you kind of know the answer to that because I do feel like an M an unbind M5 Max with 64 gigs of RAM and a total of eight terabytes of storage attached to it can last me a pretty long time. Again, I'm on seven years of using an Intel Mac Pro that was like similarly spec'd with like, you know, half the storage and, you know, ancient CPU and everything. I could use this computer. This tied me over computer. I could use it for seven years easy.

1:33:46Marco Arment:I'm not saying you can't use it. I'm just saying, like, if your plan right now is, all right, I'm going to sacrifice, I'm going to compromise. John, you are not a man of compromise.

1:33:58Casey Liss:I am. I totally am. Not with your computer. You're not. Not with the computer. Here's the thing. Waiting. Waiting is a compromise. When I'm not buying stuff, I'm not getting fancier, faster hardware. That is a compromise. Like, I always want to buy at the right time to buy. but in the times in between especially the end of it i'm using older slower crappier stuff that is a compromise like there's no there there's no free lunch here you're always taking trade-offs like i will be i mean if it's not three years if it's four years if it's five years like i'm fine with that um because that's what i'm forced to do in this situation like i i think put it this way i would be less happy with like a mini or something because i do want something first of all i want it to be you know faster than any of the max that we have faster than like because i do feel like the m5 max is going to be faster than you know this is the other thing people like why didn't you just get a laptop at that point my my own brother even said to me you what you end up buying this computer you could have just bought a laptop like three months ago yeah but here's the thing if you if i was to get the 14 inch version of this same specs that's 800 more and if i was get to 16 inch it's uh 1100 more obviously because they have screens and keyboard like it makes sense that they cost more but there's no way in hell i'm going to pay even 800 extra for a screen keyboard and trackpad that i despise not the screen the screen is good but like whatever i was never going to get a laptop so with me getting this m5 max it is the fastest especially if you can't sustain performance mac that you can buy that's not the ridiculous ultra and so i think i'll be happy with it in that regard i mean and part of the reason i know i'll like this is using my wife's m1 max and xcode i'm like wow this computer is fast like because i'm using my intel a computer like certain things in xcode takes so much longer um but you know again my mac pro is slower than an iphone right so i i'm i'm sure when i get this i will be overjoyed and everything will be fast and the store that's the other reason i didn't want the m4 model because i knew the m5 models have faster storage the storage is faster i don't care about the ultimate speed of the photo library that's more about bulk storage so i think this will work out yeah i guess i guess my my my

1:36:01Marco Arment:worry here is that your tied me over computer is just one of your computers like this is just going to be your computer for the next x years and if x is you know more than a few uh i feel like you you might be disappointing your future self by compromising now in ways that if you were buying it to be your your next six to seven year computer you would not necessarily want those same

1:36:26Casey Liss:compromises well so here's the thing uh i also have expenses even though i only have one child in college i still do have two and a half more years of paying for that college and her college is more expensive than my son's college uh there are other there are other expenses in my life that make it not the right time for me to buy a eight to ten thousand dollar computer or anything like that as much as he thinks i would love to get that computer like it's just it's just not financially responsible for me to be doing that i'm paying for a whole bunch of things especially like in these, you know, this part of the year.

1:36:57Casey Liss:And it's just maybe, you know, it's just, it's not, it's not the right time for multiple reasons. So this compromise is something I have to do. And if I, if I use this computer for five years and then replace it, I will still consider it a tied me over computer. If I use it for seven, I'm like, well, tied me over lasted longer, but I still think I'll be, I won't be as in dire straits as I am with this Mac pro now in seven years, because seven years from now, first of all, the M5 max is not in any danger of being obsoleted or no longer supported by mac os because it is basically their second newest chip that they have or third newest i guess because you've got the ultra and the m6 um and it'll certainly have enough power and i don't i expect there not to be another architecture transition in the next five to seven years so yeah this basically gives me the leeway to wait and like if the m7 comes out and it's not exactly what i want i'll be like oh i'll wait to see what the m8 does and if the m8 isn't what i want to go wait till the m9 once the n10 comes out i'll be like well maybe i have to buy this one but it's not ideal, but sometimes you got to do what you got to do.

1:37:56Casey Liss:And this is what I have to do right now.

1:37:59John Siracusa:I am quite shocked to be honest with you. I'm not saying that you're making the wrong choice. I don't love the way you're handling storage, but I do understand what you're saying. And it is certainly the more frugal, well, not even frugal. It's the, it's the more intelligent approach I would argue is to do what you're doing. Uh, I just, I wouldn't want to be bothered.

1:38:20Casey Liss:And if it external storage for my drive ends up being a disaster. It'll be good content for the show.

1:38:27John Siracusa:That is true. No, I think this does make a lot of sense. I am very surprised this is the route you're going, but I think it makes a lot of sense. So you think sitting here today, and obviously you don't know what the future will bring, but you think this is a two to three year computer would be your guess?

1:38:43Casey Liss:No, it's at least three years. I would say three to five is my guess, somewhere between three and five. And if it goes to seven, then I will have miscalculated. But you know, like i said even if it goes to seven i think i'll be fine because i won't be in a crisis like i am now you know what i mean like in seven years like it's not gonna be like oh my god i need a new computer i can't get anything done in this i'm using my wife's computer it's too slow it's not like i won't be like that in seven years it'll just be like oh i guess this wasn't a tide me over after all but like that's that's kind of my line between what's a tide me over and honestly after the you know the discontinuation of the mac pro like that dream ended with the mac pro and like any computer i was going to get was going to be like well i'm forced to get the lesser offerings like the same thing with the m2 ultra mac pro like that would have i would have considered that a weird compromise and a lesser offering that i never would have bought anyway because it's the same as a studio for my versus right so i'm kind of out of that phase for a while and like we said before with your believe shirt hey in a decade if they reintroduce the mac pro maybe i'll be back in business but i'm it's i'm in a down period now i'm like it's like the rebuilding decade for like sports teams it's like yeah you're not gonna win any championships anytime soon so just hunker down do what you got to do and then wake up in a decade and see if apple has changed

1:39:50John Siracusa:its mind about high-end computing wow well i'm shocked i am genuinely shocked clearly marco gets the win on this one even though both of us were wrong i feel like marco was much closer to the

1:40:01Casey Liss:right he had the right instinct but he just didn't lean as heavily into how much this is not the right time for me to buy a computer like how much i feel that how much this is not it's wrong it's

1:40:10John Siracusa:the wrong time yeah i don't know if i can do the rest of the episode i'm truly shocked but i will soldier on for you, the listener. We do have a few more things that were announced. There are new keyboards, Mark Gurman writes. Apple also launched new Mac external keyboards, but they're virtually identical to the prior ones. Minor labeling changes on the keys to match the recent MacBook Neo or Air updates. John, you've done some science and some research. The old one is the, which I just bought like a month ago, maybe two months ago, is MXK83LL slash A. and you can tell that that's the one because the home button this is for the um extended keyboard the home button has the word home on it uh page up and page down have the words page up and page down and so on the new one is mjm 74 ll slash a where the home end page up page down among other things are all symbols now um this is the magic keyboard with touch id and numeric keypad for mac models with apple slogan usb c u.s english black keys we'll put a link in the show notes i miss i

1:41:15Casey Liss:don't uh i'm assuming this is what they did on the neo like again they added fn on the fn key whereas like the older model doesn't have it it's just got the globe uh both of these don't have control in the lower left corner they have the stupid globe slash fn key there i hate really hate that they don't say home end and page up and page down anymore because first of all the home and end symbols it's like they don't read as home and end to me but second page up and page down are up and down arrows and i know they don't use arrows on the arrow keys they use triangles but it's like that doesn't say page up or page down if someone looked at that keyboard they'd be forgiven for thinking that was the up and down arrow but it's not just below it is the actual up and down arrow i think this is a big regression i don't like the symbols i i mean maybe this hap helps with like internationalization or whatever like clear doesn't say clear anymore it has like a a box with an x in it looks like someone typed a like the wrong unicode character or it's like the character encoding like there's no character in this font like it's just shift doesn't say shift it's just got the up arrow like it just seems like a savings for money where like they don't have to put words on it and there's i don't like it at all but i will reiterate as we discussed on past episode i in anticipation of me getting a new computer saw this when it came up recently i was like oh yeah they screwed up the keyboards didn't they they got rid of like in particular i can't stand not having the control key in the lower left i mean it's there but it's half size and it's not in the corner anymore so i bought on ebay uh the model that i like which is mmr 3ll dash a or slash a uh it's the a2520 model but that's the particular uh apple part number for the particular color and blah blah but the point is is an extended keyboard it's got touch id the control key is full size the full double width thing in the lower left corner and it says home and page up and page down on it that's what i want and that's what i got it sitting here next to be sealed in the box when my Mac Studio comes, that is what I'm connecting to it.

1:43:05John Siracusa:But that's lightning, isn't it? Or is that also USB-C?

1:43:08Casey Liss:It's lightning on one end, but it's USB-C on the other. And the Mac Studio doesn't know or doesn't care what's on the other end of the wire.

1:43:14John Siracusa:That's gross.

1:43:15Casey Liss:And I can use it through Bluetooth if I wanted, but I don't.

1:43:17John Siracusa:That's gross. I don't love it. But I mean, I hear the logic there. I don't love it.

1:43:21Marco Arment:Especially if you're using it wired, it doesn't really matter.

1:43:23Casey Liss:Yeah, it's a desktop computer.

1:43:24John Siracusa:It's like, the wire is not impairing my keyboarding. Believe me. And then we have possibly the biggest news of this entire episode. There is a new polishing cloth. Emma Roth at The Verge writes, Apple just dropped a new polishing cloth and unlike its refreshed Mac Mini models, it comes with price cut as spotted earlier by MacRumors. The polishing cloth is still made with soft non-abrasive material but costs$9 instead of$19. It doesn't look like anything else has changed as the description matches the archived version of Apple's old polishing cloth, which it no longer lists on its website. And we'll put a link to both The Verge and MacRumors in the show notes.

1:43:59John Siracusa:uh starry horizon says that the new polishing cloth is about the same size as the old one but a lot less stiff they say it's more like the nanotexture or apple vision pro polishing cloth

1:44:10Casey Liss:that's my complaint about apple's polishing clothes like i got along with my protos play xdr and they were just so stiff like who you how do you even use this it's just like it's you know normally you take a polishing cloth and like you wrap it around your fingers or hands so like you're you're touching the screen with a small area so you can get some force on it but this thing is so stiff It's like you have to put the entire pad against the screen or something. It's just, it's weird. I'm not a fan of their polishing cloths, but yes, they continue to update them and sell them, apparently.

1:44:33John Siracusa:I would be remiss if I didn't mention our friend Quinn Nelson's shop, Copeland.supply, not a sponsor, but a friend, where he has the Cloth Pro Max for$18, which is very, very good. I have one, and it's great.

1:44:46Casey Liss:He just got undercut by Apple in price, though. Yeah, sure did. Because he was 18 and they were 19. Look, it's cheaper, but now they're down to nine. You know, if there's one thing getting cheaper on Apple's website, it's the polishing cloth.

1:44:55Marco Arment:well it doesn't have any ram that's true all right thank you to our sponsor this episode tail scale and thank you to our members who support us directly you can join us at atp.fm slash join one of the many perks of membership is atp overtime our weekly bonus topics in addition to all those specials that we talked about earlier we also do a weekly bonus topic after every episode a little bit more that we just couldn't fit into the main show this time we're going to be covering a recent benchmark to suggest, is macOS slower than Linux when performing file system operations? Join on to listen, atp.fm slash join.

1:45:29Marco Arment:Thank you so much, everybody. We'll talk to you next week. Now the show is over. They didn't even mean to begin. Because it was accidental. Accidental. Oh, it was accidental. Accidental. John didn't do any research. Marco and Casey wouldn't let him Cause it was accidental It was accidental And you can find the show notes

1:45:58Casey Liss:At ATP.FM And if you're into Mastodon You can follow them

1:46:05Marco Arment:At C-A-S-E-Y-L-I-S-S So that's Casey Liss M-A-R-C-O-A-R-M And T Marco Arment S-I-R-A-C U-S-A-C-R-A-C-U-S-A It's accidental Accidental They didn't mean to Accidental Accidental Tech Podcast So long

1:46:35John Siracusa:John, you've been busy apparently and I genuinely know almost nothing beyond that. So what have you been up to?

1:46:40Casey Liss:I apologize in advance for having a Casey-like wind-up for this but I feel like it's important for context. All right. Calm down. it's just it's a that you said point of pride it's just uh anyway so uh we'll link in the show notes to my blog post about uh my three uh existing mac apps front and center switch glass and hyperspace and if you read through those blog posts like they're like the announcement posts of like here's this app and here's why i made it and stuff and with the exception of hyperspace the especially explanation for front and center switch glass it's basically like um i used to use a bunch of apps on my mac and then uh you know and one particular drag thing by james thompson it stopped being developed and i kind of needed the functionality was provided because i'm so used to using my mac with that functionality and it was hard to find in other apps and so i'm like well i mean that's what got me to do mac development in the first place that's what finally got me to do it after all these years it's like no one is making this weird little app that i want i'm going to make it myself i used to use drag thing for this drag thing is no more i used to use a bunch of other apps for it they're also gone i'm gonna have to make this myself so front and center was the first one like the simplest little app you can possibly imagine and switch glass was the second one and drag thing uh people who don't know back in the day both on classic mac os and on mac os 10 um let you put make like these little pallets on the screen with stuff in them and one of the kind of pallets you could make was an application switcher palette and also just as a side inside the drag thing settings there was like a checkbox that said and by the way make mac os 10 look work like classic mac os when clicking on windows so the features front and center used to be in drag thing and the features of switch glass used to be in drag thing and drag thing did tons more stuff too but i'm like these are the features that i miss the most from drag thing i miss the front and center thing which is when you click on a window for another app what happens is it bring just that window to the front or all the windows they change that behavior in mac os 10 i could never get used to it both behaviors are useful but I wanted to have the option to use one or the other and switch between them by shift clicking, which is what drag thing gave me.

1:48:40Casey Liss:So I made front and center switch glass. I wanted a little application switcher. Why do you want that? Don't you just use the dock? It's just, I just could not live without it. I just got used to it. The dock has stuff besides other apps in it. It's got the trash. It's got folders. I do use those features. I just wanted an app switcher. So I'm running switch glass with my app switcher in the upper right and the dock also visible. It's just the way I want to do it. These are the weird little apps I made, but you can see how the first two apps i made were just entirely motivated by drag thing goes away blame james thompson because i needed that functionality uh and they were just very focused like it was like why don't you combine switch glass and front and center you know drag thing just had a checkbox in there why not just make front and center a checkbox and switch glass like no i just want little turdy apps that just do one particular thing they're paid up front there's no subscription you buy them once people bought front and center five years ago for eight bucks or five bucks or whatever i was charging for it then they just have this app forever and switch glass it's just an application switcher uh and it's ridiculously configurable in appearance because this is what i wanted but whatever people would probably just use the defaults and probably nobody uses it but me but i like it and i run these apps 24 7 when i'm using my mac front and center and switch glass are running and hyperspace of course i'm using to try to save my disk space which is and hyperspace was like hey this is my first real app that's not super weird because it's like nobody wants front and center and switch glass except for me okay but hyperspace that's potentially useful to a lot of people a lot of people have want to save disk space without deleting files so i'm like finally i'm out of this stupid corner of making weird little apps that are just for me hyperspace even though it is still you know it's a mac app first of all so right away i'm eliminating most of the population of the planet because it's just the way it is um and also it's a pretty nerdy mac app but it's not as you know narrowly focus is front and center in switch glass so i thought i had gotten out of the the dark alleyway of making tiny little apps that only satisfy some need for me but i i was switch last out i guess wow six years ago 2020 geez anyway switch glass apparently has been out for six years and during that time lots of people would have feature requests switch glass you should add this you should add that and i was like no it just switches apps it's right in the name it just it's a switcher yeah i could add other stuff but i'm not going to but here's the thing there's one more feature from drag thing that switch glass doesn't do because it's narrow and focused that i miss so i basically just use the app switcher but i also had like another drag thing palette that had like some folders in it and in the mac os 10 age without drag thing what i do instead is i put folders on the lower left corner of my desktop i generally keep a mostly clean desktop but i always just have like permanently like the stupid on-screen climate controls and your touchscreen ev permanently in the bottom left corner of my screen i have a set of folders um but you know unlike a drag thing palette or something else the desktop gets covered sometimes and you can't get to the folders and i was like i should just extend switch glass with a feature that says oh and by the way you can also make a little palette with some folders on it and put it somewhere else and And I like basically implemented this like in a branch of switch class.

1:51:45Casey Liss:Like, I don't know. I'm so bad at guessing. If you'd asked me how old is switch class, I would never have guessed six years, but probably three or four years ago I implemented in switch class this feature. It's like, oh, and you can also have one other palette for the hell of it that you can put some folders in and click them and they open the folders and find her like that's it. That's the whole freaking feature. Okay. But I was like, oh, what is like switch class is supposed to be focused. I don't want to add this on. And like, how confusing is it going to be? is like so you have an app switch glasses and app switcher and also a little place where you can put folders like what the hell like it doesn't make any and like how do i fit it into the preferences do i make another dialogue box for it how do i explain the functionality like it's incoherent and so even though i implemented it i'm like this is not the right place for switch class also i don't remember but switch glass again 2020 it's swift ui like the view itself is swift ui but swift ui in 2020 on the mac was garbage it didn't do anything it's still not great but like all of like the drag reordering of like collection views and stuff that you take for granted and like ui kit and app kit and stuff did not exist in any way so i hand rolled every single part of it it's like an entirely custom ui hand rolled at like great pain and expense to look and behave exactly the way i want it to and this is this is like this is ancient swift ui code this predates at observable it is like remember at state object versus at state remember at published that's what switch classes it is the oldest creakiest uh swift ui code you've ever seen it supports os is like backed really far okay it was like obviously it's more than the os that was that was new in 2020 but it supports even back even farther like and i don't want to cut off support for that it's like because you should modernize switch class it would be so much easier to write with modern apis but like i don't want to just like abandon the old os's at this point and maybe i should eventually anyway but like that still doesn't solve the problem of where to add this functionality so i was just stuck it was like i'll just go back to the folders in the desktop like i've been using the little folders in the corner of my desktop forever i'll survive without it and here's the thing folders in the corner of your desktop nobody wants like that's even worse than front and center it's like you're gonna make that a separate app you're not gonna put it into switch class you're not going to modernize switch class switch guys like does what you wanted to do why would you go in there and add new bugs by modernizing and like so i was just stuck along comes the magic of vibe coding i'm like this is the solution to my problem i'm just going to vibe code a new modern app to make a little palette with like five folders in it and stick it in the corner of my screen that'll be my solution no one else never needs to use this app it'll be just for me i won't muck up switch glass it's not going to screw up like my the purity of my apps it's not going to be hard to explain no one else even needs to know it exists and it will solve my problem and you know and i know this is totally within the realm of vibe coding because how hard is this it's the easiest thing in the world so i did i vibe coded two seconds like okay there you go it's a little thing and you put make a little palette and you put folders in it and it's great um and then you know because vibe coding is fun and you're like oh you know i went off and did it and did some other things then you argue with it about stuff and blah blah blah and then eventually all right so it's fine you're like you know i've got some other ideas that i always was interested in trying is i was i was doing this in mac os 27 so i could use all the fancy new apis and i told the uh the my little uh coding agent i said you know the materials that you get in mac os that i give in switch glass and everything of like the various like different tints of backgrounds and stuff like in the and in tahoe they added clear glass and regular grass and tinted glass to add to that collection none of those materials is what the dock looks like like apple does something custom for the dock so i'm like oh coding agent i know you love to decompile figure out how the doc oh my god what it does and then make me a material that does what the doc does and it went off and did its thing and it found like there's like these recipe files and materials files in the doc and there's also a whole bunch of private apis as you could imagine that only the doc uses that are not available you know like and it found all this stuff and it was like hey uh i've implemented the doc material and it was close and i was like it's not quite the same and i you know i worked out this whole thing of like here you can tell whether you're exactly the dock like make it make it just like the dock and eventually it did obviously using private apis and all sorts of like you know unholy things that would never fly in the mac app store or whatever um but now i have this modern app with modern swift ui that gave me the little feature that i wanted that i could also make it look like the dock which was fun and then i got a little bit mad with power and i was like but that's the dock in mac os 27 the doc looks different in mac os 26 and it looks different again in mac os 15 so i need to make like a little toolkit to let it uh figure out with how the doc worked on all those other computers and simulate those and so i did i said hey do that i want to be able to pick doc on the current os doc on mac os 27 doc on mac os 26 doc on mac os 15 this is the fun of vibe coding because i'm like i'm not sitting in front of the computer while it's doing this i'm like doing other things and it did that i was like oh what am i doing here this is supposed to just be like my toy project for me and i already have what i want out of it and all this cool doc stuff like no one's ever going to see this like i can't put this in the mac app store obviously it's all private apis nobody cares like basically what like what the l what these lm coding agents did is they made me like one ten thousandth of a gi rambo you know what i mean like this this is not a skill that i have like i can't i can't reverse engineer these things i don't know how to go spelunking in their files and figured it out he could do it and obviously the lm can do it somewhat but like like this is just like a glimpse of what it's like to have like again one ten thousands of the skill of someone who actually knows how to do this stuff but it was fun it was a thing here's i'll put it this way it's a thing that i personally could not have done without a coding agent because i don't have those particular skills at all and yeah you could learn them but like i feel like at this point in life like i'm never going to become one ten thousands of a gear rambo i've just i've got the coding agent for that but it was fun and i had this thing i just didn't know what to do with it and so like i could go in a couple of different directions here i could say well that was fun i'll use it myself and it'll be fine but like at that point i felt like i had built up a significant amount of functionality in particular i did like i didn't tell you how i started this project but the way i started it was i said hey coding agent here's the source code to switch glass do that but over here and use modern swift ui so i'm basically i'm pointing it to my existing all handwritten you know ancient all handwritten code and i'm saying see that do exactly that like i want it to look like that i want it to behave like that i want it to everything but don't use you know at published and at state object and all this other crap and i said and if you can use the modern like reordering apis in mac os 27 you should do that um i had to support back to 15 because i wanted to use it on my mac my mac is stuck on 15 um so i said support back to 15 so you can't use 27 only apis entirely but use them if you can and anyway i told it to do that and that was that was an interesting just a brief aside on that that was interesting llm experience because you're like i'm pointing it at my own source code i'm saying basically copy that like i've already figured it out and like it would take a cut at it and then i would try it and like i don't know if it's a context window thing or something but it was like you missed all of like the performance enhancements i spent so much time tweaking my stupid switch last thing so it has decent performance on slow old intel max and the like and i'm i'm developing it on m1 max so the first time i brought it over to my intel mac and i like resized the icons it was like a half a frame per second like what the hell lm like i gave you the source code to a thing that performs amazingly on intel cpus and you missed all of my optimizations and did like the dumbest, most naive thing.

1:59:29Casey Liss:So I had to go back to the computer and be like, hey, look at this file in this area. See where I'm doing this? I was just like, you're not, I did this already. Just copy it. Just like, it's just, anyway, it's like a black box, but it didn't, like, it got the gist of what Switchglass does. But even given the source code, it did not literally copy all of my options. So I had to go back in for each thing and say, do this like this, do this like this, do this like this. Like, yes, copy all these optimization, a file at a time. Look in this file, that's where this thing is. look in this file that's for this thing is the animations weren't as smooth or as good look in this file see when i did this see this is the reason i did this over here like it's all in the comments anyway eventually i got it to work like switch glass and it just it just felt like that i had something that seemed like it might have some kind of value besides just this weird thing that works for myself but of course it's all private apis and crap so what am i supposed to do with this and so i said i could have gone two directions which is one just keep it for myself and whatever but the other direction i felt like i could go in is what if i turn this into an app that basically supersedes switch glass that it does everything that switch glass does but with modern apis but also makes it have little folders in the corner or whatever like basically a general purpose way to make a bunch of palettes around your screen and i know you're thinking are you just making drag thing is that what you're doing it was the whole thing of like You're just going to reinvent drag thing.

2:00:50Casey Liss:You're just keep making Mac apps till you reinvent drag thing. And I think I'm not doing that. James Thompson, correct me if I'm wrong. Yet. I think I'm not doing that.

2:01:00Marco Arment:What do you think, Casey? Next year? Yeah, seriously.

2:01:02Casey Liss:But I'm really approaching it. Because once you can make a bunch of arbitrary palettes all around your screen that have different things in them, that's basically drag thing. The only difference is the drag thing can let you put tons of stuff there and do all sorts of interesting things. And I'm doing a much smaller set of less interesting things, at least for now. so i just decided i have to make this into an app and also by the way this is related to the computer that i just bought my tradition seems to have been my tradition my one-time tradition which is buy a really expensive computer and then make a bunch of mac apps that eventually pay for that computer over the course of the next several years which is what i did with my mac pro that's my plan here is like make a new app and try to pay for my computer over the next three to five years in terms because that's the kind of income i think make your apps make enough to pay for a computer yeah over the course of several years like that's what we're talking about here okay and so that's my plan i'm like i'm gonna ship this as an app and it's i it's gonna be an app that lets you put a little bunch of little pallets all around your screen it could if you use switch glass this will do exactly the same thing as switch glass but like again better with more modern apis and stuff with better code and stuff um but also it'll let you have a bunch of folders in the corner and you can have another little pallet over there with other stuff in it and you could have like whatever like it's just you know it's it's not a great idea again it's not drag thing it's still fairly limited but it is no longer like switch class which is like this is an app switcher and that's it this is instead arbitrarily arbitrary little pallet-y turds

2:02:30Marco Arment:all around all your screen you really need to work on that messaging yeah i mean that's basically what

2:02:36Casey Liss:it is like i don't even know if anybody wants this but i i once i had the ability to do this i'm like well i could those folders over there that's not the only thing i might want on the screen especially if they're auto hidden and they just pop up when they come when i go to the edge of the screen like i can think of other potential uses for this um and then i just kept adding stuff to it because and by i i mean almost entirely my coding agent the only code that i wrote in this is basically all the ui because i don't know what it is with coding agents but they cannot lay out a dialogue box to save their life and like i'm like don't even bother just i will do the dialogue boxes there it's it's the dialogue boxes are not complicated it's like you know a label and a toggle control like that's it but they cannot handle it so i did other than that they're doing everything and i'm just directing them um but what do i do about the doc material like what like i was all excited about that obviously i can't ship that in the mac app store and i'm absolutely not going to send this direct to people and make my own payment like it's not that was exactly what i was going to

2:03:29John Siracusa:ask is are you going to do some sort of other payment processor if this was an app that i

2:03:34Casey Liss:thought people were actually going to buy in any appreciable numbers i would have considered it but this is not that app so no i'm not doing that it's mac app store and which means i can't do any of the private apis and stuff like that but i have some precedent in dealing with this which if you remember in switch glass um again it's just a little app switcher uh and it does have little right click menus on all the apps and stuff that have various actions you can perform and the obvious thing that you would want them to do like when you right click on a on an app icon in the dock you can quit the app right right click and click on an app in the dock that's a reasonable thing to want to do mac app store says no you can't do that um oh so you're gonna do a little daemon thing yeah there's an entitlement that you need to do that that apple will not give you and did not give me uh and so you can't do that um and so my solution to that was okay well i am just going to make an unsandboxed notarized direct download free app that switch glass will look for if it happens to be there launch it if it is there and talk to it until it to quit the apps and so that app is called quit class it's like quit but switch glass quit class it is a free download if you have switch glass and you want to quit apps you just download switch glass launch it once it is totally invisible it doesn't have a dock icon doesn't have menu bar icon you never have to launch it manually switch glass will do all the management for you the point is if you have switch glass installed to have launched it one time ever switch glass will have a quit in the menu that will work because switch glass communicates to switch glass and says hey you do the quitting for me and it does the magic of not being sandboxed so how can that strategy help me get a doc material uh on uh on my app not quite as easy because it's not like you know i'll send you a message and you'll do the work like it's the material that's showing behind the little palettes and that's i mean again if i was gi rambo or somebody clever there's probably some way to like have things rendered by another process and put back into yours and blah blah but i'd i'm never going to figure that out but here's my strategy and this is probably why i mean if apple again if apple does listen to this they're probably going to shut me down and this will never be a thing but until apple stops me which may be very soon here's my strategy ship my app in the mac app store with the things that mac app store apps can do so no quitting no dock materials just tons of other limitations that suck and even more new ones that i discovered as i submitted builds and they kept rejecting me and i had to remove entitlements remove entitlements remove entitlements they don't like me i've been rejected more on this app so far it's not even released just on test flight builds anyway get something that will go up on the mac app store then also make from the exact same xcode project another version of the same app with all the the forbidden features and forbidden entitlements and all that good stuff and make that a free download on my website but the the trick is that that free download will not run unless you have purchased my Mac App Store app.

2:06:30John Siracusa:Oh, that sounds unfun, but like a pretty decent solution to the problem.

2:06:34Casey Liss:And I mean, I'm sure Apple hates this, but like as far as they're concerned, the app on the App Store is perfectly legitimate. It does only the things that it's allowed to do. And that's it. I'm not going to link to the obviously from the app. I can't tell anybody from my Mac App Store app. I can't say, hey, go get that. I can't say mention it. I can't link to it. I can't acknowledge that it exists. I can't do anything. And the other thing is my Mac App Store app has no idea that other app exists doesn't communicate with it doesn't look for it doesn't do anything it just has to be there on the disc with its like mac app store receipt you know the cryptographic receipt that like it happens when you buy an app like a paid up front app meanwhile if you were to wander over to my website and find this other unsandboxed app that is free for anyone to download but is entirely useless unless you have installed this mac app from the mac app store just run that app and that app can quit apps can resolve aliases to servers can do dock materials can do all of the fancy features that you'd always want to do but couldn't do before and so this is my first foray into non-sandbox mac app development and once i knew i could do that i basically made i'm working on making this app not a dock replacement but i'm doing stuff that i didn't even think was possible like uh i've always said people always say hey switch glass should show the notification badges that are in the dock there's no api for that there's not even like there's nothing for that but i'm like well if i'm out of the sandbox i'll do it let's go nuts yeah hey coding agent find a way to do that and the answer is always as always is accessibility accessibility exposes the badges and so my app will look at the accessibility view of the dock and when it puts a badge i'll put a badge same thing for like the quick command is easy um yeah so you can make a very convincing doc lookalike with my app that a regular person could even forgive him for thinking was the doc there are still some features that things the doc does that i can't do because there's they're not exposed from accessibility like basically i'd have to inject into the process or something but anyway this is my plan i'm going to put this app out when it's not even close to done but like i'm going to put this mac uh out of app out on the mac app store and i'm going to have an unsandboxed version of the app that is free for anybody who has bought the actual app and then if apple doesn't stop me i'm just going to continue doing that until they do stop me if they do stop me i'm not sure what i'll do but like honestly this is not a very mass market app but i just i just had so much fun making this and adding functionality to it i'm still having fun um so yeah and my plan is as casey pioneered on atp i'm going to offer the access to the test flight to any atp member i will warn you atp members this app is rough it's not done the good thing about it is unlike hyperspace like it doesn't shouldn't really destroy your files or anything like the beta is not it's like it's a little palette a bunch of palettes on your computer where you put stuff in it like it's not it's not modifying your files like it's it's the the danger that's probably one of the reasons that i'm very comfortable vibe coding this is like it's not it's not a consequential app it is just a ui frivolous little ui thing that you may or may not like but if you want to try it and find bugs and report bugs or whatever um at some point maybe by the time you listen to this but certainly before next episode if you log into atv.fm and go to your member page it's just atv.fm slash member or click on membership in the navigation you will see a section appear on your member page as it has in the past that says hey do you want to join the test light for john's new app click here and then you will join it i'll also have a link to my slack for the app if you want to join and complain about stuff there um yeah i'm not sure how i'm going to describe this app or pitch it or anything like that but it is basically you can make a bunch of little dock things all over all your screens and have them contain whatever you want including the trash can like you can put a trash can in it and all that stuff like you can you could i'm not suggesting that you make a dock like i'm not going to use it for that but i have replaced switch glass in my daily use with this because it is like switch glass what better oh the other one i forgot to mention is custom icons man that their situation i've found so many feedback saying please make an api for this but like even without being sandbox there's no way to get a lot of these things i did get the calendar because if you use like a doctile plugin and if you're apple i can load apple's doctile plugins but i'm not going to load doctile plugins from third-party apps because i just feel like loading third-party libraries into my process is just a bridge too far i could do it but i'm not gonna but anyway calendar will show the right icon and i might do custom stuff for things like ivory that use like the catalyst set custom icon but the the icon apis the private icon apis on mac os just do not expose this stuff but still my switchgrass glass replacement is currently showing a badge for slack right now if i had calendar in there it would show the correct calendar date and it uses all modern apis and my little folders in the side they're looking beautiful they will raise when my cursor goes of them over them so they have auto raise on them so they'll go in front of other things if i slam my cursor against the bottom of the screen it's all looking good so yeah that's my uh that's my new app wow all right so i have a question

2:11:42John Siracusa:what's going to happen first you completely subsume all of drag thing functionality or apple is carbon neutral i'm never going to do everything that drag thing did like that's the

2:11:52Casey Liss:problem with this app people are going to see it and they're going to be like i want to do this i want to do that and now i no longer have an excuse other than i don't want to do that no i'm just

2:11:59Marco Arment:messing with you yeah i honestly i don't think i foresee this drawing attention from apple negatively because what you're doing first of all it's like yeah you're doing a whole bunch of like private api crap but like not in a way that would blow up in their face um and second of all with like the mac app store thing like where you're gonna have this app require the mac app store purchase that like apple cares if you try to bring purchases out of their in-app purchase system right you're trying to bring purchases into it you're trying to you're doing a hack for them to make more money so i don't think they're going to care about that either here's here's where i think

2:12:37Casey Liss:they'll get tripped up and for the first release of this i'm going to avoid this but for subsequent releases i will not avoid it and here's what it is when you submit to app review and you have like you have to have like urls in your app for like your support url and your privacy policy and crap like that and also like the help but like in the help menu of all my apps there's a help item and if you select that help item it launches you to a page on my website oh yeah then they go poke around to see if you're trying to suck purchases out yes and on my website you can click on the logo at the top and get the most recent blog post and the most recent blog post is like hey look at this unsandbox app you can download they're gonna be like sorry you can't link to another app to download like you basically that is an app store rule you cannot in your app say hey download this other app from my website you can't do that at least not in the u.s even if

2:13:21Marco Arment:you don't have your own purchase system outside of their system no i know but like i what i'm

2:13:26Casey Liss:going to do is i'm going to for the first release of this i am not i'm going to make sure there's no links on my like the question is how many links do you follow because remember the horror stories of like i linked to my website from my about screen on my app and 20 clicks from that you can get to this post that mentions this thing and apple's like sorry we found our way to this thing that mentioned another app rejected like that's my fear so one thing one thing that i've

2:13:46Marco Arment:done um even though i don't think i am at risk for any problems with with my with the website but like for unforgettable the privacy policy and terms of service pages which have to exist for all subscription apps um those when you open those in the app there is no header or footer or navigation to go to other pages it only is those pages um could you do something like that

2:14:09Casey Liss:yeah i probably i probably will when i put them up like but the problem is the help page i want there to be a header on the help page like it's part of my website like we'll see how it goes like if i have to and also it's like as you know with the app review it depends on what reviewer you get is this the reviewer that clicks three three follows three links deep from your help page or four or you know it's just there's no consistency so i'm gonna do what i can apple don't listen to this podcast as marco says see i'm bringing money into your system i don't know what they think of it but this is my current strategy if it doesn't work out i'll think of something else i mean the good thing is this is never going to be a mass market my app anyway so who cares it's just mostly a curiosity and a fun thing to do on the show, but I'm having fun making it.

2:14:46John Siracusa:No, this is very clever. I did not see this like a wholesale replacement of Switchglass being where this conversation was going because all we knew was John's new Vibe Coded app. I'm surprised this is where it ended up, but I mean, I think it makes sense. And I think if it's working for you, that that's all that matters.

2:15:04Casey Liss:It's a regression for me. Like I said, it's a regression for me trying to make useful apps going back to my, I'm back on my BS as they say i'm back to making apps that nobody wants that are weird they're just bigger than they used to be and also mostly not coded by me but that's just the place where we are the place where i am right now so maybe someday i will return to making apps that other people might want to buy but uh yeah in the meantime i'm doing this one you haven't asked me about the app's name yet okay what's the name john it's bad not switch glass i just assumed you were subsuming switch class no no no switch class will continue on um as as you know again it's it's a six-year-old paid up front one-time app that if you paid for six years ago it will keep i'm gonna keep it running right um but i will probably never modernize the code so at some point if like if this version of swifty y no longer compiles i maybe i'll have the lms modernize it or something but anyway i i named this app when i thought no one else was ever going to see it been there yeah yeah you know how it is um also on that front i made the icon myself and it's not good oh been there too right we've all been there and here's the thing with this i'm probably not going to change the name because like who cares like whatever it's not it's not it's not as bad what was your past one uh fluk up it's not fluk up bad but it's not good all right calm down the icon on the other hand it's not good but my plan is if i release this in the mac app store and apple does not kill it uh once it makes enough money for me to pay someone to make an icon that's like the first x amount of money from this app is going to pay for its own icon so it will not be in the black until it is paid for its own icon yeah that's that's where i'm and i honestly making the icon was fun but it just like it's not again it's not in my skill set and i and i used ai help like so the icon well i did i first problem is i have no good ideas for the icon so that's not a good place to start if you have no ideas like if i have an idea i can fumble through and like make something but i have no good ideas so i went with bad ideas because you got to do something right and then i was like can you ai generate an icon no you can't they just look terrible so i i did what i've done it a few times like i did this for the beta icon for hyperspace i bought stock photos from istockphoto.com you can did you know you can buy stock photos you can you pay for them and they're yours free and clear and you can use them in your app icon and so i bought stock photos uh and then i ai generated one element and then i photoshopped it with a bunch of crap that i did in photoshop and affinity designer plus my stock photos can you see how this is making for a not good icon so that's how i ended up with what i have so what's the name i will show you the icon first and see if this helps you get the name i will try to show you a small version of the icon so because it looks worse the bigger it gets uh here you go so this uh the the inspiration of this icon which i will put in the oh my god is this air power no no so the the menu bar icon because like switch glass has a menu bar icon and this one has a menu bar icon the menu bar icon is essentially like the sf it's the sf symbol for

2:18:25Marco Arment:like a computer screen with dock at the bottom of it like like a lower third yeah it's like a

2:18:30Casey Liss:computer screen i forget what the sf symbol is called but it's like computer screen with dock or i don't even know but that's that's what it's supposed to be supposed to be a like a round rack with a line at the bottom and i'm like that's my app it makes docks i'm not going to use docks i'm using pallets for a reason like i don't want a confusion looks like inset filled bottom third square yeah uh but i'm like but yeah but like obviously that's so fine that's the monochrome menu bar icon but what is the actual app icon and so here uh this is what i've got for the app icon and it makes sense with the name kind of sort of i tried a lot of iterations in this marco you have any guesses of the name of this thing keeping in mind that it makes a bunch of little uh uh rectangular uh long skinny things around the edges of your screen and and given that you're looking at the icon, which one of you can feel free to describe the icon.

2:19:15Casey Liss:I won't do it.

2:19:16Marco Arment:So it's like a wave crashing onto a beach, like the foaminess of the end of a wave onto a beach.

2:19:22John Siracusa:With air power sitting on the beach.

2:19:23Marco Arment:Yeah, with a big black sunglasses case sitting or like a big, like a polished oval stone. So I'm guessing you probably didn't call it Docker, even though that would be a great name for it. That would be hilarious. I think that's taken. Yeah, maybe. Hmm. And the icon, it's like a polished stone from a beach.

2:19:45Casey Liss:Is there any long, skinny, rectangular things related to beaches that you can think of, Marco? Docs. That's right. But I'm not using that one, so what's wrong? Right.

2:19:55John Siracusa:The answer is right there in front of us, and I'm cheating because if you look...

2:20:00Casey Liss:I know, I know. Casey, don't tell him.

2:20:02Marco Arment:Oh, the file name.

2:20:04John Siracusa:It is obvious once you hear it, though.

2:20:05Marco Arment:Jetty.

2:20:05John Siracusa:Yep.

2:20:06Marco Arment:Yeah. Jetties.

2:20:07Casey Liss:I grew up going to a beach, Long Beach on Long Island on the North Shore that has a jetty. I played on that jetty all the time. For those who don't know, a jetty is like a long skinny pile of stones or other things that goes out into the water and directs where the erosion should go so that one people on one side of it are mad and people on the other side are happy. Yep. That's what a jetty does. And it's a long skinny thing. And, you know, what's a bunch of long skinny things? I can't use dock. I can't use anything having to do with dock. calling the app jetty and the uh the unsandboxed version is called jetty pro

2:20:42Marco Arment:this is honestly i don't see anything wrong with this name i mean i haven't done like you know a trademark stretch or anything but like at first glance i don't think this is a bad name at all

2:20:50Casey Liss:i agree i was kind of surprised it wasn't taken it's who even knows what a jetty is and i the reason this thing is black is because i tried a bunch of non-black rocks but the problem is small sizes they just become mud and this this at least is when you shrink this down to tiny sizes is it looks like the icon in the menu bar so you can be like what is the icon in the menu bar oh it's this app is i can see that's why it has the thick black outline around it like the and it's it's jet black like i guess like whatever like this is a bad idea for an icon it's a bad idea for an icon not well executed like i'm not happy with this icon but honestly i don't have a better idea but when i get enough money so if i get enough money from this app i'm going to give it to someone to come up with either a better idea or a better version of this icon if we can't come

2:21:33Marco Arment:up with a better idea is there a reason you didn't like orient it to be more like a jetty uh because this is what the menu bar icon is oh yeah that's all reason right right the menu bar like i picked

2:21:44Casey Liss:the menu bar icon first and i got used to it i'm like and i couldn't think of any other ideas for the things i know what you're saying like but the thing is the the ui in this app doesn't ever stick out from the screen they're always flat against the edge and i know that's not what a jetty does but like again it's not i can't i can't i can't uh defend it like it is what it is no this icon

2:22:04John Siracusa:this icon is straight trash but the name i like yeah and and it doesn't even matter like you know

2:22:10Marco Arment:like if you know people who don't live near these things don't know what jetties are who cares there's lots of names that like people don't know the origin of the name they just associated it with the app and it's fine yeah you know how many people know what a google is even if even if you

2:22:21Casey Liss:spell it the other way although i will say that jetty jetty with no colon suffix was in fact taken on the App Store because everything is that I ever try.

2:22:29Marco Arment:Well, I mean, yeah.

2:22:30Casey Liss:No, I think this is fun. What is your suggestion for my colon suffix?

2:22:35John Siracusa:What did you say? Little turdy palettes everywhere?

2:22:38Casey Liss:I'm not going with little turdy palettes.

2:22:40Marco Arment:See, what you actually want to do, like if you want people to actually find and download this, what you actually want is figure out what keywords you want that people would be most searching for for apps like this and figure out some subtitle that would that would include those keywords like that's that's the actual answer to this but since you are really good at at not making money with your apps because you're a

2:23:07Casey Liss:purist i have to i can only go where my heart leads i mean it leads me to non-profitable places

2:23:14Marco Arment:um i mean for for your subtitle i i don't know i i can imagine you being very pained at the idea

2:23:20Casey Liss:of even having to make one no i i accept that it's the way it is i had to do with hyperspace to like it is what it is like i'll do it you know as long as i as long as it's not on my icon in the app in the applications folder i'm fine yeah i don't have a clue what i would write oh here's what i've got and i'm uh as usual not happy with it but uh it will actually probably help with seo is i've got jetty colon custom docs because people are going to search for doc yeah again i'm not fair i'm not going to call things doc because in my ui i have things like material doc which means make it look like the dock right or preset dock which means make a thing that looks like the dock so i do not use dock for the name of my things i always use palette which is i think what drag thing might have called them i don't know i'm using palette is what i'm going with but you can't put palette on the headline so jetty colon custom docks is terrible and looks gross and kills my eyes but it will come up in people's searches maybe i'm will apple let you use the word dock we'll find out i mean they did let me test flight it and let me tell you they were rejecting me like crazy in test flights i'm like it's a test flight like reject me an app review you fine but like i'm just trying things out here and they're like nope we can't even let this test flight thing up you're not getting that entitlement forget it wow so they let they let it up with the

2:24:28Marco Arment:subtitle we'll see how it goes well good luck i i think i think that is a perfectly good subtitle for this app but uh i would be honestly i'd be surprised if apple let you use it are you gonna

2:24:38Casey Liss:capitalize doc yeah it's title case capital j capital c capital d jetty calling custom docs other apps in there say doc stuff okay there's doc door for example i don't know if that's on the mac app store we'll find out if i have to change the subtitle fine but i can't i would have just used plain jetty obviously if i could have gotten it but someone else has got it this

2:24:55John Siracusa:is this is exciting i think this is fun um i think you should i think you should stick with it and see where it goes and you know listeners this is not an official perk of being a member so uh if this never if the test flight never ever shows up in membership show up well all right just make it just trying to leave you an out trying to leave you now nope i've got the test flight build up

2:25:15Casey Liss:i've made the group it's all ready to go i just haven't pushed it to the website yet oh all right like i said the app is uh filled with bugs but you know the consequences of those bugs is the animations are glitchy in your custom docs it's not a big deal real-time follow-up there

2:25:27Marco Arment:are actually lots of apps in the mac app store that include the word doc including some like things that seem to be doing somewhat similar tasks like adding extra docs basically so yeah

2:25:37Casey Liss:there's lots of apps that do this and it's like why don't you just use one of those apps obviously i need the app that works exactly specifically like i want it yeah of course well yeah i mean look i just made a reminders app there's lots of reminders apps including reminders exactly but that's the whole thing about vibe coding like i never would have bothered doing this whole rigmarole for just this one little thing but vibe coding makes it makes it possible to do this with the amount of effort that with the amount of effort that makes sense for the amount of money this is not going to make you know what i mean yeah because like the amount of effort i put in a switch class was tremendous hand rolling all that learning learning on the fly like my very first swift ui thing huge amount of work into that but that was like a learning opportunity blah blah and i didn't i didn't make it back on the app like but it was but it was a good learning opportunity but now when it came time to do like oh revise switch class to like you know generalize it to do this thing i implemented and i realized this is not this is a waste of my time this is not a good use of my time which is good because like i didn't do that i did hyperspace instead so but anyway but this is one of the beauties of vibe coding that even for developers it opens up the potential to do things that would have been a waste of your time otherwise and this

2:26:37Marco Arment:definitely qualifies oh yeah i mean like like just literally in the last you know two weeks i i not only you know do i make my regular software but i've i've vibe coded two different apps just for myself like one of them was uh i have an app that decrypts audible books and there's been like you know command line utilities to do this i want to listen to my audio my audio books in overcast okay of course like i want to use upload them to my uploads and use my audio engine with my effects and my speed to listen to audiobooks so i buy my audible and i use my crappy vibe coded app that i'm never going to release because i don't want the hassle um you know but it just took two seconds you know for for this app to come together great and i don't have to i don't have to remember like all the command line incantations for how to use the different tools and where do they drop the little dirty configuration files and how do i keep all the logins straight and everything nope i just had a vibe coded mac gooey app that stores stuff in keychain and it's great and i didn't have to look at any of the code ever and that's great and then my second one is an app that um you know how like in the app store people want you you basically want to have screenshots that aren't really screenshots they like show the app kind of in device frame they have a bunch of text crap to brag about stuff about the app and i years ago like i tried this with overcast i even i think we didn't even didn't under the radar episode about it um where i did see like yeah they those do actually convert like a little bit better it wasn't a huge difference i think at the time it was something like seven to ten percent better conversion rate um but it's you know that matters and i just hated having to like make them and i hate having to make all the different size screenshots to optimize for different devices and some of that's easier now but not all of it so i in like one giant 15 line prompt i had claude make me my own stupid little admin app to make those screenshots it's basically a gooey editor for me to add text and symbols and crap onto screenshots and have them put into a device frame and then to be able to export them in this nicely, you know, nicely sized and named list of ping files that I can drag into App Store Connect.

2:28:41Marco Arment:And it took that whole thing took a total of, I think, three prompts and maybe a half hour. And now I have this app that, again, like I'm not going to release it. It's not it doesn't make any sense for me to release it. But like it's vibe coding is really helpful for like making little things to scratch your own personal itches. you know stuff that used to like either only like you could only do it if you had a lot of free time and a lot of you know skill in that area um or if you could find another market for it and now you can just make your own stuff like it in the same way like people can make their own spreadsheets you know to solve their own problems for for all these years now you can make your own little apps to solve your own problems like for for internal tools for admin tools it's very very valuable

2:29:24Casey Liss:that screenshot frame thing part one of the reasons it probably did such a good job is there are so many of those apps out there and some of them are probably open source so it has an idea of how to do it like that was my hope with the um it didn't work out as well as i hoped it would that was my hope with feeding it switch glass as the initial prompt which is like look at the source code to this app here's the repo you have all the source code just do that it didn't work as well in my experience as when these things are trained on open source repos of like a thing that makes a device frame screenshots with text for your phone like there's so many of those apps out there and i got some of them have to be open source it makes me think that like training if you know because obviously switch glass source code is not public it's not an open source app okay but but if these models had trained on the switch glass source i feel like they would have performed better than me saying here's the entire switch glass source code in your context window because my experience with doing that is it was like it wasn't even looking at the source code or it was like looking at just certain portions of it and i had to direct its attention look at this file look at this function look at this thing it was a little bit frustrating and counterproductive productive but it was a good learning experience in how to use these things but yeah for the device frame thing so many apps are out there to do that i think the reason everybody makes them predating vibe coding people were making these little apps for themselves people who did have enough free time to do it because everyone wants to do it in their own way and again some of those must be open source and some of them must be in the training data so when you ask for one it's like i totally know how to make that i've seen this app before well i i mean i

2:30:47Marco Arment:did actually specify a lot of the behavior like i specified the data model i specified the ui and and a lot of that stuff. But I don't know, I don't think most of the coding agents now are like replicating entire apps from their training data. I think it's more piecemeal and more involved than that. But, you know, I think it's a pretty well-known problem set. Like you can just, if you, you know, if they look in their training data or do web searches for just what these kinds of apps need to be, what are the App Store Connect requirements, stuff like that, like that's all easy to find.

2:31:21John Siracusa:that's good stuff sean i am excited to see the test flight and and kick the tires the app once and never look at it again probably but you know what i'm excited to see it nevertheless yeah

2:31:31Marco Arment:congratulations like i as i as i have you know recently gone through this process of launching an app like launching any app is a significant achievement and it's a significant amount of work it's way like just you know my stupid audible decryptor and screenshot annotator like that took no time if i wanted to release either of those that's a that's a totally different ball

2:31:54Casey Liss:game yeah even if you're not writing the code for the app right the the hassle i mean you could get the lms to do that too but i'm i'm not going to do that but like yeah there's so much like overhead that's why like you need some you really need some motivation to get something over the hump of like am i going to actually release it to people because it's just such a hassle and there's so much stuff to deal with especially if you're dealing with app store and app review and the website and support and bug tracking and it's just so much that like keeping stuff private like my a little trimmer app or these other you know i have a bunch of other little like command line things that i made for myself that like give me info about files that i use as like debugging tools while developing hyperspace i have all these vibe coded tools that like spelunk in the file system like that stuff is great but once you release anything to the public it's like well now you've taken on this giant mound of crap into your life so it better be worth it for what you

2:32:41Marco Arment:want to do yeah exactly and that's that's that's partly why too like you know i see i see a bunch of people around our circles. I'll see somebody who we followed in years for Mastodon, and they'll be launching apps for the first time. And these are people who, maybe they were power users, maybe they weren't necessarily programmers, or they weren't this kind of programmer before. I don't have any problem with people doing this. Obviously, people feel very differently about using AI to code stuff. Because I have mentioned that some of the code I write is AI code, I'm now getting regular one-star reviews on both of my apps because it's quote ai slop even though i mean it's like i spent three hours today writing writing code and unforgettable myself which i've written most of and it's like and people you know people will accuse whatever yeah i'm sure i'll get those reviews on this app

2:33:32Casey Liss:too but you know it's like yeah you want handwritten ones i've got three other apps

2:33:35Marco Arment:right but and the thing is like i i really don't have any problem with people 100 % vibe coding an app and releasing it as long as it works. And as long as, like, you know, if there's a bug, they'll be responsive to it. You don't have to be handwriting the code to have an app out there that works and that you're responsive to making sure it continues to work. So I think, like, to me, I know this is controversial, but I guess I've already lost all these people or angered them. This is a very enabling technology. lots of people can now make apps who couldn't make apps before and lots of people could make apps before like us can now make more apps better apps and apps that just wouldn't have been worth the time to make before so again not a perfect technology without its downsides but overall a massively enabling technology and i'm able to take joy in the fact that it can do stuff like this that like these like an app like this that you just made that like that couldn't really exist or that wouldn't have existed before now this exists that's a cool thing yeah i mean the problem is

2:34:42Casey Liss:that you know people hate the technology but like we're talking about this one very narrow corner of the technology where it excels and the technologies are not only used for programming and that's part of the big problem it's like okay that's great for you in your little programming world but like people use ai for other things and it's extremely problematic in many of those things that's even before you get down to like okay but what about any you know so that whole wad of crap is there but like we have such a narrow focus in the show because we're programmers and we're always talking about programming and i think sometimes people hear you guys love ai it's like i mean it's we're using one very specific thing which is a popular thing that people use and people are getting you know utility out of it but boy people are like the majority of people who are using chat gpt are not making apps no definitely not let's put it that way and and you're right like there are a lot of

2:35:29Marco Arment:i think more problematic uses for it that that i think uh you know we're not we're not advocating for everything all those things but in this case uh you know using it for software development it's

2:35:40Casey Liss:yeah which is why like i'm excited about being able to run like local like open weight models in the future on your mac without paying some giant company without you know running in data centers and like all that you know like i can see that future being attainable but uh we're not there

2:35:53Marco Arment:yet yeah and and like that's what i was literally looking earlier that's what i was saying about like with you know trying to figure out how to use my excess mac mini capacity i was looking at like okay well what if i want if i want to do like you know summarization and stuff like that with local models running on mac minis that i own it's like that's that's pretty low impact really like for the world like it's not that's not that bad but yeah apple was touting that by the way with

2:36:14Casey Liss:the new mac studios they're like look in this giant stack of mac studios that are connected it takes so little power like i think snell was saying they said look it's just plugged into one outlet like i mean okay great but like people usually plug computers into one outlet like but for ai it's like look this is this is so energy efficient you can you know run this incredibly big model plugged into a single household outlet it's like i guess this is the

2:36:36Marco Arment:bar now you can have this computer in your house right you don't even need three phase industrial

2:36:43Casey Liss:power yeah instead of whatever whatever you have in your like a 240 volt or whatever dual power

2:36:48Marco Arment:strip thing in your data centers right and and again and i do want to just while we're on this topic briefly i do want to reiterate everyone saw that big list of features that apple did on these os releases right you know apple's using claude right like everyone knows that out there i i think i think you need to know that like that huge massive list of things that they're adding in this os release that that's not a coincidence like they're they're definitely using we know that they're using claude widely they have been for a few years and that's why they're able to get this much stuff done so if you're like taking a super hard line against using anything ai generated i i have i have some bad news for you to use mac os ios probably windows and yes also linux like i i think you're going to have a hard time uh avoiding ai generated code forever