00:00:21 ◼ ► And sometimes, if it's a long plane flight, like, you're podcasted out and then what's left.
00:00:27 ◼ ► And the thing about planes, too, is like, why don't you talk to the person next to you?
00:00:33 ◼ ► Just trying to have a conversation with the person next to me is just like, I feel like
00:01:02 ◼ ► You know, because, of course, whenever you are around, you know, children, and especially
00:01:22 ◼ ► If I need a break, I'm like, I'm going to stop and just watch TV for the rest of the night
00:01:39 ◼ ► Of course, I have, like, you know, my actual work, my programming projects, multiple programming
00:01:50 ◼ ► There's a million things around my house that need to be done, like different projects and
00:02:31 ◼ ► I feel like my recollection of summers, particularly around, around either Declan or Adam's ages
00:02:37 ◼ ► was, I, and I had friends that I would like play with or hang out with, depending on how
00:02:48 ◼ ► But I remember, I remember just being so bored and I feel like I could have done within reason,
00:02:58 ◼ ► Like my family growing up was not super into limiting like video games or anything like that.
00:03:10 ◼ ► Well, you, I think you were a normal teenager, like that's just, or a normal kid, but you
00:03:15 ◼ ► know, whatever age we're targeting here, or like, I think it's, I think it's maybe less
00:03:18 ◼ ► so with teenagers because they, they are more easily able to find their own stuff or, you
00:03:30 ◼ ► Unless my parents were taking me somewhere and forcing me to be like, let's go look at the
00:03:45 ◼ ► We have a new member special that we recorded just a few days ago as we were recording this.
00:03:56 ◼ ► Well, we explained at the top of the show, but this is a response to a bunch of listener
00:04:01 ◼ ► questions that we got about what makes a really good Mac app essentially, especially for people
00:04:15 ◼ ► And the device, which we fall back on a few times in the episode, but isn't actually very
00:04:21 ◼ ► helpful, which is like, well, you know, when you see it, well, maybe we know when we see
00:04:24 ◼ ► So we tried mightily in this episode to see if we could nail down a few concrete things
00:04:31 ◼ ► And if you're wondering about the title, again, we explained it at the top of the member
00:04:34 ◼ ► special where that phrase comes from, but it basically just means what makes a great Mac
00:04:38 ◼ ► I feel like we could have three more specials just on the same topic because I feel like
00:04:44 ◼ ► Like we didn't even get into like, it's can Apple still make good Mac apps and stuff like
00:04:57 ◼ ► And I think there's lots of useful, actionable stuff in there, obviously, if you're a developer
00:05:01 ◼ ► and if you're not a developer and you're just sick of hearing us talking about, oh, that's
00:05:11 ◼ ► And it is one of those very difficult things, like John said, of how do you define what is
00:05:20 ◼ ► And, you know, the people who have already listened seem to have very nice things to say,
00:05:25 ◼ ► And Marco, if one isn't a member, but if you were interested in hearing ATP dev Mac asked
00:05:32 ◼ ► You would head to ATP.fm slash join and join our wonderful membership for all of these wonderful
00:05:54 ◼ ► And your voice, you're like your voice, you're like morning, you know, radio DJ voice still
00:06:33 ◼ ► I apologize to whoever pointed this out to me, but somebody pointed out that you can add
00:06:47 ◼ ► But generally speaking, one of the things I love so much about Sonos is that they will play
00:06:53 ◼ ► You know, you can go into the Sonos app and pick which speakers you want to play things
00:06:57 ◼ ► And then they will just handle it in your phone or whatever device is not involved at all that
00:07:14 ◼ ► So if you are interested, like Marco is in the Dave Matthews band, or if you're interested
00:07:44 ◼ ► And I said in that episode, and here I am saying this, but of course, as soon as I finish this
00:08:03 ◼ ► I was thinking about it last night when I saw this and I was like, you know, since it's
00:08:07 ◼ ► only, since it's like one every three days on average, this could just be like one person's
00:08:20 ◼ ► And it's like, uh, you know, the, the thing that's causing the crash is an assertion that
00:08:24 ◼ ► says, and there should be nothing left in this dictionary while we're deallocating stuff
00:08:36 ◼ ► By that point, all the code in my app is finished executing and it's just deallocating stuff.
00:08:45 ◼ ► As I said on the show, when I talked about this, I've never been able to reproduce this.
00:08:59 ◼ ► And yeah, even put it in debugging and say, Hey, that assertion, I know what it's checking
00:09:02 ◼ ► because I had an LM decompile app kit for me and says, yep, this is what it's checking.
00:09:07 ◼ ► And so on my way out the door, I check and I always say, make sure this thing is empty.
00:09:28 ◼ ► Hayden Field over at the Verge writes, after a rollercoaster negotiation process with the
00:09:32 ◼ ► Trump administration that dragged on for two weeks, Anthropics Mythos 5 is finally back
00:09:37 ◼ ► in action at least somewhat for a select group of organizations, according to a letter from
00:09:43 ◼ ► Babel 5, however, sorry, John, the public facing Mythos class model appears to still be in limbo
00:09:57 ◼ ► Uh, with regard to indexing for the forthcoming Siri AI, a word on the street is that spotlight
00:10:16 ◼ ► So just do X messages per day, but then they don't think about like, well, what if someone
00:10:21 ◼ ► But I am excited to see that it seems like it's actually going to download all the messages
00:10:30 ◼ ► So anyway, this is related to people saying they were waiting like a week for their indexing
00:10:42 ◼ ► and that will do all the indexing for us while we wait for 27 to be released, but we'll see.
00:10:47 ◼ ► Then Aaron Ramist had some corrections for both you and for I, Aaron writes on the last episode,
00:11:14 ◼ ► But anyway, I priced it out just to see what it would be like to get a Mac mini with eight
00:11:22 ◼ ► That is for, that is for a binned M4 pro that is down two CPU cores and four GPU cores and
00:11:46 ◼ ► So, like the whole idea, I was musing about a Mac mini is like if I could get like some
00:11:59 ◼ ► And in most cases with Apple's pricing, an eight terabyte SSD costs more than the computer.
00:12:04 ◼ ► Like the upgrade, even on a Mac studio, very often in most configs that aren't ultras, the
00:12:08 ◼ ► eight terabyte SSD upgrade from whatever the stock is costs more than the whole rest of the
00:12:29 ◼ ► So if you want any additional RAM, then you'd have to upgrade to the top of the line max chip.
00:12:40 ◼ ► But if I were to buy it today and I have 64 gigs of RAM, so I miss I conflated the issue
00:13:01 ◼ ► We'll be like, whoa, we got to change a bunch of other specs because they're really limiting
00:13:11 ◼ ► It's like, well, if you want the entertainment package, then you have to get the, you know,
00:13:15 ◼ ► interior, you know, whatever thing like leather seats or things are connected in weird ways.
00:13:20 ◼ ► So, I mean, I don't really blame them given the difficulties of component pricing, but it
00:13:32 ◼ ► Chris Carley wrote in with a point that I think, John, maybe you intended to make this on the
00:13:37 ◼ ► Chris writes, a big Apple TV price hike makes sense since, unlike many other products, the
00:13:42 ◼ ► Apple TV doesn't have other expensive components to mask the changes in price of RAM and of storage.
00:13:47 ◼ ► The Apple TV is, in terms of cost, mostly just an old iPhone chip with RAM and storage.
00:13:51 ◼ ► Yeah, I didn't mean to make this point and somehow we moved on to another topic before I could.
00:13:56 ◼ ► But, yeah, we're complaining of, like, the Apple TV, 54% increase to the base price, 67%
00:14:10 ◼ ► It's just the SoC that has the RAM on it and then storage and, I guess, like, you know, whatever
00:14:20 ◼ ► It's similar to the point that I did end up making about, like, when you do it, when you
00:14:32 ◼ ► Well, I'm not saying the Apple TV is not overpriced, you know, for the power that it has and hasn't
00:14:39 ◼ ► But if one product was going to have the highest percentage increase, it would be the product
00:14:49 ◼ ► Jay Peters over at The Verge writes, after months of waiting, Valve has finally announced
00:14:53 ◼ ► that the Steam Machine, its new living room-friendly PC, will start at $1,050 and go on sale beginning,
00:15:11 ◼ ► And so if you get the maxed out 2-terabyte model with a controller, because apparently the
00:15:18 ◼ ► The Steam Machine was looking really exciting back when we talked about it many moons ago
00:15:23 ◼ ► as, like, a tiny living room PC that is as friendly as a console or tries to be as friendly
00:15:36 ◼ ► It's the, you know, Valve's Linux platform where they heroically find ways to play Windows
00:15:43 ◼ ► And it's not a big, fancy gaming PC, but what if you don't want to build a big, fancy gaming
00:15:55 ◼ ► Maybe this will be, like, as much as the PlayStation 5 Pro, like $700, $600 or $700, maybe even $800.
00:16:17 ◼ ► I mean, I did see a bunch of YouTube videos being like, we're going to build a Steam Machine
00:16:24 ◼ ► And I felt like most of those videos should have focused on how the PC is also more expensive
00:16:28 ◼ ► to build now because it's not like the Steam Machine is the only thing that is getting hit by this price increase.
00:16:37 ◼ ► Although I guess they're probably glad they didn't launch the Steam Machine at like $750 several months ago
00:16:47 ◼ ► Jay Peters continues, while console makers sometimes subsidize their hardware to bring prices down,
00:16:56 ◼ ► While this might seem like an easy solution, it doesn't align with our beliefs about how healthy ecosystems are built.
00:17:01 ◼ ► If there's anything we are religious about at Valve, it's our belief that open systems are better in the long run for ourselves and our customers.
00:17:07 ◼ ► When companies sell their hardware under cost for competitive advantage or buy exclusive content for it,
00:17:12 ◼ ► they're doing that to build a more closed system, one where you don't get to choose what software you will want to use.
00:17:24 ◼ ► You should be able to view it as just one option alongside all the devices for playing games
00:17:31 ◼ ► This means you get to decide which device fits your personal trade-offs around things like price performance,
00:17:37 ◼ ► That's the strength of the open PC platform, and subsidizing hardware runs counter to it.
00:17:42 ◼ ► I haven't seen anyone push back on this because everyone's like, rah, rah, yeah, open is great.
00:17:55 ◼ ► Like, they're not, I mean, I suppose they could make, like, a proprietary console, but they're not.
00:18:00 ◼ ► And as influential as Valve is, they don't actually control what constitutes a PC game.
00:18:06 ◼ ► Microsoft with Windows still has more control over that, even though they're not even running Windows.
00:18:19 ◼ ► Whatever the Windows games do, we'll try to figure out a way to get them to work on our weird Linux adapter layer.
00:18:33 ◼ ► If you make a PC game, you want it to run on Windows because you're really limiting your market if you can't.
00:18:39 ◼ ► See, the one way to do it is, like, hey, we're going to sell our consoles at a loss so so many people will buy them.
00:18:48 ◼ ► the PlayStation only runs PlayStation games and it will get exclusive to the PlayStation and all that other stuff.
00:18:54 ◼ ► But another reason people subsidize stuff is to get people in the door, to people who don't have a gaming PC, to buy a gaming PC.
00:19:22 ◼ ► Hell, they make money even when someone buys a game and plays it on their Windows PC as long as they buy it from the Steam store.
00:19:30 ◼ ► Now, I know it's not a closed ecosystem or whatever, but, like, you know, a random company like Alienware can't sell gaming PCs at a loss.
00:19:39 ◼ ► But Valve is one of the few companies in the PC gaming ecosystem that could sell hardware at a loss because they make money in every game sold.
00:19:47 ◼ ► Similarly, Microsoft with the Xbox could sell the Xbox at a loss, hoping to make it up.
00:19:53 ◼ ► But again, Valve makes money when you buy a game from the Steam store, even if you don't play it on a Steam machine.
00:19:58 ◼ ► So I'm not saying they should sell at a loss or whatever, like, do whatever they want to do.
00:20:02 ◼ ► But, like, the answer about how we need to keep an open ecosystem when we don't want to lock you in, it's like, you selling your thing from low cost doesn't lock me into anything.
00:20:11 ◼ ► You still play PC games, games that I can play on my Windows PC if I don't want to buy your Steam machine.
00:20:18 ◼ ► And I'm not going to say disappointing that they're not selling it below cost, but I bet they're selling it close to cost.
00:20:23 ◼ ► Because if you look at what's in the Steam machine, this looks like they're probably not making a lot of money on the hardware.
00:20:34 ◼ ► Because the margins on taking a cut of everyone else's PC games is much better than the margins on selling a hardware box.
00:20:47 ◼ ► Starting on August 1st, five 12-gig models will be $100 more expensive, while one-terabyte models will be $150 more expensive.
00:21:10 ◼ ► Microsoft last raised prices in October by $20 to $70 and says it had hoped to avoid further hikes.
00:21:16 ◼ ► Quote, unfortunately, console storage and memory prices have increased by more than 2.5x.
00:21:49 ◼ ► Although Microsoft gives a good example of how they had raised prices by $20 and $70 and then had to raise them again.
00:21:56 ◼ ► Whereas the Apple thing to do is hold out as long as you can and raise prices on almost everything by a huge amount that you hope will stick.
00:22:07 ◼ ► Speaking of all these raised prices, a Micron executive suggests that Apple's aggressive purchasing tactics might have helped fuel the memory shortage.
00:22:15 ◼ ► Rolf Winkler in the Wall Street Journal writes that Tim Cook said, there's less supply at a time when consumers want devices and the memory guys are passing along huge price increases.
00:22:26 ◼ ► Then Rolf says, Micron chief business officer Sumit Sadana said the company couldn't make investments during the memory market's last downturn when Micron's gross profits went negative.
00:22:41 ◼ ► Sumit said, we told a couple of the customers who are being very aggressive with pricing at that time that this is not constructive.
00:22:53 ◼ ► Quote, a lot of the industry investments got shut down in 2023 because of really poor pricing and really poor margins.
00:22:58 ◼ ► So this is, you know, I mean, go to the press and, you know, passive aggressively subtweet as we used to say, Apple or whatever.
00:23:07 ◼ ► But, you know, it's, it's, it's quite, uh, quite a fantasy world to believe that, uh, you know, this company saying basically like Apple drove a hard bargain.
00:23:16 ◼ ► And if they had, if they had bought our ramp for more money back when things were okay, then we would have had more profit.
00:23:30 ◼ ► If, you know, if only, if only they would, would have agreed to pay higher prices, we wouldn't have given ourselves bonuses or done stock buybacks.
00:23:36 ◼ ► No, no, we would have built, we would have built factories because we would have said, we want to, you know, prepare for the future.
00:23:42 ◼ ► Now there probably is some truth to that because obviously the more money you make, the more investments in the future and so on and so forth you make.
00:23:47 ◼ ► But this is, this is really kind of like, well, look who, look how the tables have turned because back when Apple was in the driver's seat, they were driving these hard bargains and saying, you get nothing.
00:23:57 ◼ ► You get one cent per chip and you'll, you'll like it because we're buying a bajillion chips and those one cents add up.
00:24:03 ◼ ► But like giving them the most razor thin margins or I said, making, making things go negative.
00:24:07 ◼ ► And now that they're making huge profits and are richer and like, you know, Apple, you should have been nicer to us before, which, you know, whatever business is business.
00:24:15 ◼ ► Yeah, because like, you know, it's the notion that these companies are so like forward thinking and magnanimous and don't take all their money and, you know, again, give it to bonuses to the top executives and do stock buybacks to make people like, I'm not saying it's 100% that, but history has shown that when the profits are good, most companies do not use it to do forward thinking things.
00:24:39 ◼ ► Maybe Micron is different and maybe they would have done something different, but anyway, it's, uh, it's kind of like a kicking the rest of the industry while they're down.
00:24:45 ◼ ► Uh, Steve Trout and Smith, friend of the show, weighed in on this and wrote Apple starved the Silicon industry, excuse me, Silicon supply chain for years.
00:24:57 ◼ ► Famously, Palm blamed Apple for not being able to source components for its web OS devices.
00:25:01 ◼ ► Uh, and then we have a way back machine link to, um, to the verge where Chris Ziegler wrote, uh,
00:25:08 ◼ ► citing an anonymous source, quote, we told HP, we needed better displays for the pre three.
00:25:22 ◼ ► At various times, Apple has cornered the market on certain commodities, but the other thing Apple has done,
00:25:26 ◼ ► as we know from the Apple in China book, spend billions of dollars to make people make factories for them.
00:25:49 ◼ ► But yeah, I would, I would hope that this entire industry can eventually, uh, get its feet under it
00:25:55 ◼ ► again and try to get things more in balance and have capacity that, uh, more closely matches demand
00:26:01 ◼ ► without any one customer screwing everybody over, whether that be Apple or Nvidia or who else.
00:26:06 ◼ ► Well, and I think this also like, you know, a, this is a good use of Apple's cash hoard now,
00:26:10 ◼ ► because like if they want to front the money to build another factory for somebody, you know,
00:26:16 ◼ ► exchange for, you know, all of its output for a number of years, that's, that's something
00:26:27 ◼ ► Um, but that being said, I think another angle of this is that Apple, when they were dictating
00:26:32 ◼ ► all those terms, um, and this is, again, this is a great reason to read that Apple and China
00:26:47 ◼ ► You know, when you had, you know, back in the day, you had like Tony Blevins going around
00:26:50 ◼ ► like dictating terms to people and Tim Cook being his ice cold, you know, dictator self
00:27:05 ◼ ► And Apple's able to negotiate that kind of way, which is not really negotiation when they
00:27:10 ◼ ► And we've seen this over and over again, that Apple is really only able to work out good
00:27:19 ◼ ► They don't really show a lot of strength in negotiation when they are not able to just dictate
00:27:26 ◼ ► So that is one area where I am curious to see how this situation plays out today, because
00:27:32 ◼ ► part of the reason Apple is in this situation is that for the last couple of decades, they've
00:27:39 ◼ ► done so well that they have been able to dictate terms to all of their suppliers and have the
00:27:43 ◼ ► suppliers basically say, OK, well, we have to basically play ball because they're our biggest
00:27:47 ◼ ► customer or they're going to give us the highest price or they're going to buy all of our output
00:27:52 ◼ ► Whereas now with AI chip demand, Apple is no longer a lot of these suppliers' biggest customers.
00:28:04 ◼ ► But like if you look at between NVIDIA and Apple, like who has more sway over, say, TSMC?
00:28:16 ◼ ► And it definitely hurts their negotiating position when it comes to getting access to the highest
00:28:24 ◼ ► That's, you know, how much of their supply are they willing to give Apple and at what prices
00:28:38 ◼ ► And I don't think anybody who's there who's there right now has that skill, at least certainly
00:28:42 ◼ ► the people who have been dealing with supplier relations, you know, for the last couple of
00:29:01 ◼ ► But imagine I mean, this has actually happened, but imagine if the tables ever did turn and suddenly
00:29:13 ◼ ► Like the suppliers who have been under Apple's thumb, Apple thinks like, this is great.
00:29:38 ◼ ► for all its output, the RAM makers now are going to be like, we're not going to take that deal
00:29:43 ◼ ► either because we make so much more profit selling high bandwidth memory that the AI people make
00:29:50 ◼ ► Like the RAM that Apple uses in its products has lower margins than the RAM that the AI people
00:29:56 ◼ ► And so these companies are like, if we're going to build another factory, we're going to be
00:29:59 ◼ ► building high band with memory in it because we make more money for every one of those.
00:30:03 ◼ ► And Apple, I know you don't want any of this, but like, we would rather not take your money
00:30:11 ◼ ► We'd rather build it ourselves and sell the memory to all the people making AI servers because
00:30:16 ◼ ► we make more from like, that's the main problem with this is that the AI, you know, people making
00:30:21 ◼ ► AI servers, both the silicon, like the NVIDIA silicon and like all the RAM and everything
00:30:24 ◼ ► goes into it, those customers will pay more for their stuff that like that enterprise data
00:30:52 ◼ ► correctly predict how things are going to go, maybe that maybe they can get in a situation
00:30:56 ◼ ► where there's a glut and an oversupply and the price goes down and they're, they're feeling
00:31:11 ◼ ► Zach Hall at nine to five Mac writes the financial times reports that Apple is seeking a
00:31:15 ◼ ► clearance from the Trump administration to purchase memory chips from a banned Chinese company.
00:31:19 ◼ ► So looking at the financial times post, Apple is lobbying the Trump administration for clearance
00:31:24 ◼ ► to buy memory chips from CXMT, a Chinese company that the Pentagon has put on a ban list because
00:31:29 ◼ ► of alleged connections to the people's liberation army, according to six people familiar with the
00:31:34 ◼ ► The PLA people, people's liberation army is the military wing of the Chinese communist party
00:31:44 ◼ ► The iPhone maker has waged a lobbying campaign to get the blessing from the white house to
00:31:52 ◼ ► Apple is not barred from buying chips from CXMT or YMTC, another Chinese memory chip maker,
00:31:58 ◼ ► but the Pentagon is, Pentagon has put both companies on its Chinese military company ban list.
00:32:04 ◼ ► The so-called 1260H list contains dozens of Chinese groups with alleged ties to the PLA that
00:32:12 ◼ ► The Commerce Department last year added CXMT to a package of Chinese groups that it intended to place
00:32:21 ◼ ► But the White House told it to hold off on new export controls because the administration was in
00:32:25 ◼ ► the middle of a tough, in the middle of tough negotiations with China to try to reach a truce in
00:32:30 ◼ ► But most of the people familiar with the matter said it was unclear if Apple would get any guarantee
00:32:34 ◼ ► from the administration, especially a promise that the U.S. would not later put CXMT on the
00:32:39 ◼ ► So this is a little bit weird in that apparently they could buy from them, but it seems like the
00:32:46 ◼ ► And so Apple is sort of preemptively going to the government, to their wonderful friends in the
00:32:50 ◼ ► government and saying, hey, we really want to buy this RAM from the Chinese companies because,
00:32:53 ◼ ► you know, we need RAM and they have RAM and forbidding us to buy from them is really going to
00:33:01 ◼ ► And I know you're thinking about not letting us buy from them, but could you pretty please
00:33:07 ◼ ► This actually is, you know, as much as we constantly complain about corruption and everything in
00:33:12 ◼ ► the Trump administration, this type of thing where companies lobby the government to allow
00:33:17 ◼ ► them to do business with other international companies that they want to do business with
00:33:24 ◼ ► What's not normal is how the decision will actually be made, unlike decisions that are made
00:33:32 ◼ ► This will be made on, let's say, other criteria, but the actual situation is not that foreign,
00:33:46 ◼ ► And this type of thing of like, look, you're not going to build a factory overnight, but what
00:33:50 ◼ ► we can do overnight is start buying from a new customer that previously we were staying
00:34:11 ◼ ► Factor helps you hit your nutrition goals without the planning, grocery runs, or cooking.
00:34:15 ◼ ► What they have is fast, ready-to-eat meals that you heat up in like two minutes, and these
00:34:21 ◼ ► meals are built around your goals, whether that's weight loss, overall nutrition, higher
00:34:30 ◼ ► All of Factor's meals are crafted with functional ingredients, lean proteins, colorful veggies,
00:34:35 ◼ ► whole foods, healthy fats, and they have a list of 175 ingredients that they will not use
00:34:42 ◼ ► This is things like artificial color sweeteners, high-fructose corn syrup, refined seed oils,
00:34:56 ◼ ► They have globally inspired flavors like Mediterranean and Asian and so much more, so there's always
00:35:03 ◼ ► And they even have ready-to-eat salads with vibrant ingredients and over 70 add-ons to round
00:35:16 ◼ ► They shop, prep, cook, and deliver straight to your door so you have more time for everything
00:35:22 ◼ ► And you don't have to worry about shopping and chopping and cooking and all of that prep
00:35:30 ◼ ► So, head to factormeals.com slash ATP50off and use code ATP50off to get 50% off and free daily
00:35:39 ◼ ► greens per box with new subscription only while supplies last until September 27th, 2026.
00:36:12 ◼ ► Apple instead aims to introduce the next Pro and Max chips with the more advanced computing
00:36:21 ◼ ► Apple's taking this unusual step in order to fast-track technologies that it originally
00:36:27 ◼ ► The change should help meet growing demand for on-device AI capabilities and more graphics-intensive
00:36:32 ◼ ► Apple plans to introduce the base M7, codenamed Delos, which is apparently an island off Greece,
00:36:40 ◼ ► Apple is also planning higher-end M7 Pro, M7 Max, and M7 Ultra chips, all dubbed Andros internally.
00:36:47 ◼ ► The M7 Pro and M7 Max are scheduled for as early as the end of 2027, while the M7 Ultra is on track
00:36:54 ◼ ► The M7 line is designed primarily around major advancements to on-device AI processing.
00:37:00 ◼ ► The base version is slated to support about 240 gigabytes per second of memory bandwidth.
00:37:06 ◼ ► And the M5 has 153 gigs a second, and the M6 will have 200 gigs a second, according to Gurman.
00:37:26 ◼ ► Apple has also tested support for up to 768 gigabytes of memory in the M5 Ultra Max studio,
00:37:42 ◼ ► Apple's first-ever touchscreen laptop will rely on the company's current high-end M5 chips,
00:38:58 ◼ ► It's not going to be any faster, like, unless they come up with better cooling or something.
00:39:01 ◼ ► It's going to be exactly the same speed as a MacBook Pro you could have bought months and
00:39:21 ◼ ► that it'll just be the no-suffix M6, like, that's kind of disappointing that the M6 generation
00:39:29 ◼ ► It does make some sense, based on the rumors that Gurman has talked about here, of, like,
00:39:33 ◼ ► you know, apparently whatever was planned for the M7 generation, these things, whatever things
00:39:38 ◼ ► on the chip that make on-device AI better, Apple's like, look, we have limited resources,
00:39:47 ◼ ► And just because the M6s, maybe they're just like, well, they're, you know, a couple percent
00:39:54 ◼ ► So, rather than clogging up the supply lines of TSMC, the limited capacity that we get in
00:40:07 ◼ ► Because, apparently, the design is done and they're ready to go, and they are a bigger leak
00:40:11 ◼ ► So, yeah, it seems like the M5 is going to have a fairly long life as the top of the range
00:40:50 ◼ ► We'll have more on that in a future episode, probably, because there's some excitement going
00:40:59 ◼ ► Now, I'm not interested in the laptop, so it's not disappointing to me personally, but I do
00:41:21 ◼ ► So, if you wanted to try to get to the front of the line in one of these fabs, what could
00:41:27 ◼ ► Well, a user on Twitter called Jukan writes, from the GFHK monthly call, Apple and Intel
00:41:36 ◼ ► The M7 chip will use Intel 18A-P and is expected to enter production by the end of 2027, while
00:41:44 ◼ ► the smartphone chip will use Intel 14A and is expected to enter production by the end of
00:41:58 ◼ ► The closest I get was GF Holdings Hong Kong, but the GF stand for some other words that I
00:42:14 ◼ ► Yeah, I don't, the sourcing on this, I don't know anything about this user or what they're
00:42:23 ◼ ► Gavin Bonshore, who formerly was a senior editor at Anantech, who is now at Bontech Labs, writes,
00:42:31 ◼ ► Intel's 18A process is the company's 1.8 nanometer class node, the one that Panther Lake is currently
00:42:38 ◼ ► It combines ribbon-fet gate all-around transistors with power via backside power delivery.
00:42:43 ◼ ► And this delivers up to 25% higher performance or 36% lower power consumption compared to
00:42:53 ◼ ► The 18AP variant, which is what the M7 is reportedly targeting, is essentially a refined version of
00:43:01 ◼ ► For the M7, Apple may require some of Intel's advanced packaging techniques to hit its performance
00:43:06 ◼ ► targets, potentially involving combinations from the Foveros family, which is a 3D chip
00:43:11 ◼ ► The first mobile candidate for Intel's fab may not be the A21 for the iPhone 19 in 2027.
00:43:22 ◼ ► Current reports instead point toward the unnamed SOC, pointing toward an unnamed SOC, targeting
00:43:28 ◼ ► mass production on Intel's 14A node by late 2028, which suggests Apple will continue relying on TSMC for
00:43:40 ◼ ► Apple's one of the most demanding chip customers in the world, and winning any Apple production
00:43:45 ◼ ► gives Intel Foundry a proof point that other potential customers, including Amazon, Qualcomm,
00:43:52 ◼ ► Every M7 that meets Apple's performance targets is essentially a data point for every other company,
00:43:59 ◼ ► TSMC has effectively held a monopoly on the world's most advanced consumer chip manufacturing for several years.
00:44:08 ◼ ► it changes the competitive dynamics of the Foundry business in a way that matters well beyond these two specific companies.
00:44:13 ◼ ► So we did talk about the super vague rumors about Apple working with Intel on something-something,
00:44:18 ◼ ► and now they suddenly got a lot more concrete with the rumor that M7, that Apple is going to try to get out ASAP,
00:44:24 ◼ ► that they're skipping the M6 Pro Max and Ultra to get to the M7 faster, they're going to make Intel fab that.
00:44:31 ◼ ► And the rumor being that they're going to still use TSMC for their phone chips, because, you know,
00:44:40 ◼ ► obviously, like, physically biggest, most power-hungry, most complicated, most transistor chips,
00:44:47 ◼ ► And so going to Intel for those would be quite a feather in Intel's cap if they can pull it off.
00:44:53 ◼ ► As we talked about, Intel is taking some slightly different paths with its chip technology.
00:45:01 ◼ ► like the backside power delivery, where you route the things that give power to the chips from the bottom instead of the top,
00:45:06 ◼ ► because when you send it through the top, you've got to weave it through all the logic and everything.
00:45:11 ◼ ► The question always was, can they actually get them working and shipping and get the yields right?
00:45:16 ◼ ► And Intel is actually shipping some chips of its own on an earlier variant of this supposed 1.8 nanometer class node.
00:45:27 ◼ ► If this ends up being true, and, like, that M7 Ultra in 2028 is fab by Intel, wow, what a world.
00:45:41 ◼ ► but this is the whole point of, like, Apple doing any kind of deal with Intel, investing it.
00:45:45 ◼ ► It's like, it's to Apple's advantage for TSMC not to be the only source for a billion reasons,
00:46:07 ◼ ► But boy, what a world we're living in, where there's going to be a Mac chip from Intel again, but not x86.
00:46:16 ◼ ► I mean, that being said, though, like, Intel has shown a remarkable ability to screw things up in recent years.
00:46:29 ◼ ► Panther Lake is shipping on the 18A process, so it's not like, I mean, maybe Panther Lake has bad yields,
00:46:52 ◼ ► Like, I think, ultimately, Intel is too important to U.S. national security for anyone to let them fail.
00:47:05 ◼ ► It doesn't necessarily mean that Intel is going to be able to become a foundry, you know, for other manufacturers,
00:47:12 ◼ ► have all of the expertise and trust and processes that somebody like TSMC would have to make cutting-edge chips for somebody else yet.
00:47:25 ◼ ► I don't think they've shown any of that really so far, at least not at that kind of scale.
00:47:42 ◼ ► But the thing is, both parties are highly motivated to just, you know, slog through the worstness because they're, like,
00:48:06 ◼ ► So they're both motivated to deal with all that crap that you just said, which is totally true.
00:48:34 ◼ ► And it's still, still your point stands, which is, like, okay, so they do make it work.
00:48:39 ◼ ► What if it's, like, the Intel, whatever, the Intel cell mode where everyone didn't want to get the Intel one.
00:48:48 ◼ ► I mean, look, for lots of reasons, like, I do think, you know, anybody paying any attention to global politics or the chip business should be a little concerned how little of a contingency plan that we have if there is a conflict over Taiwan.
00:49:07 ◼ ► So, that's a pretty significant risk to the world these days, and especially to the chip business.
00:49:19 ◼ ► We want there to be strong, you know, non-Taiwanese chip manufacturing around the world, especially in our own country.
00:49:28 ◼ ► But that's a really difficult thing to set up, and it takes decades of investment and prioritization and just huge amounts of change and infrastructure that, you know, we've, I think, started to do some of that.
00:49:42 ◼ ► But I think it's going to be a very long time before we could even be remotely competitive with TSMC in Taiwan.
00:49:56 ◼ ► But, you know, anything that steps on the gas a little bit in that area, I think, is a very good thing, because it's a very hard problem, but also a very important one.
00:50:06 ◼ ► But it's kind of unfortunate that our best hope is Intel, because they have not really shown a ton of capability in that kind of area.
00:50:32 ◼ ► Like, it's not, you can quibble and say, well, TSMC is better for reasons X, Y, but they're not in a different ballpark.
00:50:39 ◼ ► Now, I'm not saying Intel has caught up to TSMC, but they are now plausibly in the conversation.
00:50:46 ◼ ► The thing they have running against them is not so much like, oh, you're terrible at fabbing things and so on and so forth, because, again, they've done a lot of work to close that gap significantly.
00:50:58 ◼ ► It's not like a situation where, like, why don't you just make a new RAM fab or, you know, make a new chip fab?
00:51:07 ◼ ► Their problem is what is very new to Intel is being a foundry for other people's chips, and that's where they suck ass compared to TSMC.
00:51:14 ◼ ► TSMC has been a foundry for other people's chips for ages, and they're used to working with customers and all their tools and everything.
00:51:21 ◼ ► Like, that is the problem that Intel has, because even though they've been fabbing chips forever, they have not been a customer-oriented service company that says we will fab chips for anybody.
00:51:33 ◼ ► And that was the whole transformation they made, where they split off the part that makes, like, you know, the actual Intel chips from the part that fabs them.
00:51:45 ◼ ► And we had a story many shows ago where Intel was like, if we can't get a customer, we can't build these factories.
00:51:51 ◼ ► Like, we're trying to build the factories for our future processes or whatever, but we literally can't.
00:51:56 ◼ ► We're not going to build them unless someone signs up to buy the chips they're going to come out of them.
00:52:03 ◼ ► Like, you know, our attempt to be a fab depends on somebody signing on the dotted line and saying, we will buy chips from the factory you will make.
00:52:11 ◼ ► If nobody does that, then our fab business is just going to fail, and we're just going to fab all our chips at TSMC like everybody else.
00:52:37 ◼ ► And Intel, you've only been fabbing chips for yourself for so long that you suck at it.
00:52:43 ◼ ► But again, both parties, Apple and Intel, they're both highly motivated to make this work.
00:52:55 ◼ ► Best case for Apple, they can finally be the most important customer again because I would presume they will become Intel's most important customer for its fab.
00:53:05 ◼ ► If only because, like, you know, hey, you couldn't have even built this factory if it wasn't for our orders, for our Mac chips.
00:54:11 ◼ ► The first thing I wanted to put in about the new Siri, I'm assuming you've all, well, I know
00:54:19 ◼ ► No, I do have the beta running on my iPad, and I did install it on the Vision Pro, but I
00:54:42 ◼ ► It's probably also the icon for, like, the little conversation app on the iPad where you
00:55:03 ◼ ► But in the 27 releases, they've gone with, they have a new sort of mostly monochrome logo.
00:55:10 ◼ ► This one has some rainbow tinges on it to remind you of the old Siri, but it's like a circle
00:55:27 ◼ ► So Brad Ellis posted, I don't know if you recognize this, but he posted a bunch of images from a
00:56:14 ◼ ► And it fits in perfectly because it looks kind of like the rebranded Pepsi symbol that nobody liked.
00:56:44 ◼ ► But, like, it's like a person with a giant pizza flopped over their head, like pizza dough or something.
00:57:05 ◼ ► But I do think this is the most significant Siri-y, like, visual identity branding change that has happened in a long time.
00:57:15 ◼ ► I mean, like, from what everyone has said, the new Siri is generally, so far, excellent for people.
00:57:25 ◼ ► My one minor irritation is that doing things like creating reminders via Siri now seems slower?
00:57:36 ◼ ► Whether it does it on device or not, it is doing a lot more processing than it used to.
00:57:41 ◼ ► That part is not amazing, but if it is really as smart as everyone says it is when asking more complicated things, you know, maybe overall the upgrade will, I'm sure it'll overall prove to be worthwhile.
00:58:14 ◼ ► I didn't love the weird rainbow blob that we had before this, that was in the menu bar all the time either.
00:58:28 ◼ ► I think the problem with it is, I mean, yes, first of all, that the Pepsi logo thing, like, that was incredible.
00:58:54 ◼ ► Because right now, when you see it, like, in the dock, it just looks, it almost looks like a placeholder icon.
00:59:09 ◼ ► So they're kind of trapped by their own decision to make everything monochrome in the menu bar.
00:59:14 ◼ ► Just like more, like, the symbol needs to be a little bit more, a little bit more detailed.
00:59:44 ◼ ► Like, it has never had a, I don't know if never, maybe there was early on, there was one back before Apple bought the Siri company.
00:59:50 ◼ ► But, like, yeah, they didn't, it never really had, like, a strong logo identity, nor has it had, like, a, I mean, if you look at your menu bar now, like, there's not a brand identity for, like, the sound thing in the menu bar, right?
01:00:02 ◼ ► Or the, the time machine thing, or the clock, or the, even Spotlight just being a magnifying glass.
01:00:10 ◼ ► Apple could, I mean, I guess this is it, but, like, come up with a logo that says, instead of saying, this is for controlling your displays, this is for controlling your sound, this is for controlling Bluetooth.
01:00:24 ◼ ► I mean, obviously, it's the branded for Siri, but, like, if there, you know, if there was a generic symbol, we haven't come upon it at all.
01:00:31 ◼ ► If we did, I think what we would say, I think the industry has decided that sparkle is the thing that means AI agent.
01:01:15 ◼ ► Yeah, have you seen, have you seen, speaking of things that wouldn't look good in the menu bar,
01:01:27 ◼ ► It's like, it's like two strands going in a circle and the strands are twisted around each other.
01:01:33 ◼ ► They even animate it, so they, like, they weave in and out of each other and straighten
01:01:37 ◼ ► The problem with that one is those two strands that make up the circle are so fine that it
01:01:49 ◼ ► And if you look at WWC sessions, I think they have an animated version of it where the lines
01:01:55 ◼ ► Like, they do have some other branding about this, but yeah, I think they, I think they're
01:02:02 ◼ ► I think their biggest problem aesthetically, I didn't, don't have anything in the notes
01:02:12 ◼ ► So you do command space and it's just like this giant, uh, sort of web 2.0 looking gradient
01:02:25 ◼ ► Like it, it doesn't offend me as it appears to offend many, but I don't think it's particularly
01:02:33 ◼ ► Oh, the other thing I'll complain about visually with the new Siri so far is that, um, when I
01:02:39 ◼ ► am just pulling down to use, I guess what used to be called spotlight to, to launch an app,
01:02:44 ◼ ► if I do want, you know, you pull down, you start typing the app name on the iPhone, that
01:02:48 ◼ ► And it's, I think it's trying to be a little too smart, a little too quickly on that now
01:02:58 ◼ ► No, I'm just, I'm literally typing the name of an app on my phone and I want to just tap
01:03:07 ◼ ► Like the, if you used to just command space to launch an app, that command space window
01:03:13 ◼ ► And so, yeah, there's the fight between, oh, you're just typing the first three letters
01:03:17 ◼ ► of an app versus, are you asking me some question that I'm going to send off to a server and do
01:03:23 ◼ ► It's, I mean, we, we went through this with the address bar and browsers years ago, where
01:03:32 ◼ ► Well, you type stuff, we'll, we'll auto-complete to bookmarks and sites that you've gone before.
01:03:43 ◼ ► There's a term for like a universal box where you type text and you do everything from one
01:03:57 ◼ ► Well, continuing on with the new Siri, a friend of the show, Guy Rambo writes, I'm doing an experiment where
01:04:01 ◼ ► every time I use chat GPT to help me search for something, I'm also asking the same question
01:04:06 ◼ ► So far, my experience has been that the Siri app typically responds with the same information,
01:04:14 ◼ ► Well, I, I've tried it a little bit of, we'll get to that in a second, but most of the people
01:04:17 ◼ ► who have tried to say, are you wondering, you know, did Siri AI quote unquote fix Siri?
01:04:22 ◼ ► And the answer is it behaves like the other AI agents, which is high praise when you're talking
01:04:42 ◼ ► And so people have been trying it say you can quibble sometimes one's better, sometimes the
01:04:54 ◼ ► It may be like the formatting better or worse, you know, maybe it's faster or slower, but at
01:05:00 ◼ ► Meanwhile, uh, another friend of the show, Dan Morin asked Siri, make a keyboard shortcut
01:05:05 ◼ ► to open clipboard history in a single keystroke to which Siri replied the keyboard shortcut
01:05:11 ◼ ► This shortcut opens a panel that displays a list of recently copied the items, allowing
01:05:21 ◼ ► For those of you not familiar, uh, Apple had a keyboard, uh, clipboard history to spotlight
01:05:31 ◼ ► You got to do like command space and then like four command four or something to get to the
01:05:35 ◼ ► So, uh, Dan's trying out the new Siri and saying, make me a keyboard shortcut that gets
01:05:42 ◼ ► And you notice that prompt, that message doesn't say anything about a platform, but he is asking
01:05:51 ◼ ► Anyway, why in the world would it come up with an answer that tells you how to do clipboard
01:06:03 ◼ ► I'm saying it works like all the other agents where sometimes it just does the wrong thing.
01:06:21 ◼ ► It gives a comically wrong answer, at least now it's failing more interestingly and, you
01:06:28 ◼ ► And so my experience with it, like this, this, the next part here is just me, my very first
01:06:34 ◼ ► I just installed golden gate and I just seen some videos of like people doing stuff with
01:06:38 ◼ ► it by like selecting files in the finder and then asking it to do things with the selected
01:06:47 ◼ ► I'm sorry to the developer of this every time I can never remember the name of this app,
01:06:52 ◼ ► Now, even on an ancient old Mac OS 15 that I'm running, select multiple files, bring up
01:06:57 ◼ ► an LL and prompt and, and have it do something like, you know, pull the text out of these images
01:07:03 ◼ ► Like, you know, but you're basically just talking to an agent somewhere, you know, whatever it
01:07:32 ◼ ► Uh, Siri, I said, this is a Perl script named thrash.pl designed to stress test an application
01:07:44 ◼ ► and automatically kills them after a random, random duration up to five seconds before launching
01:07:52 ◼ ► The script uses a SIG child handler to asynchronously clean up the exited child process without blocking
01:08:00 ◼ ► Whenever the number of active children drops below five or fewer, it automatically spawns
01:08:05 ◼ ► Target application that currently configured to execute a local development build of accessorize.app
01:08:11 ◼ ► It doesn't actually know the context, uh, because it's not, I guess, scanning my whole system.
01:08:21 ◼ ► That is just an app I made in Xcode that does some particular stuff while it's running.
01:08:28 ◼ ► And I wanted, I was basically torture testing switch glass to say, is there some kind of
01:08:33 ◼ ► crash when there's like massive amounts of activity of like tons of apps, uh, launching
01:08:37 ◼ ► and quitting all at random and inside the accessorize app, fiddling the things that switch
01:08:53 ◼ ► And, you know, again, we're surprised to know on, on this podcast, these agents are good
01:08:59 ◼ ► And it explained to me that, and this is, this is like, this is like, I don't know, less,
01:09:04 ◼ ► It's not a complicated script, but this is a more or less accurate description of what's
01:09:22 ◼ ► Uh, and then me being me come back to my computer later that day, I select the same file.
01:09:36 ◼ ► These are plain text files containing code written in the Perl programming language, often used
01:09:55 ◼ ► What happened to all the stuff about telling me about what's in there and what it does or
01:09:58 ◼ ► And this variability is one of those things that is, I feel like a hallmark of Siri AI and
01:10:03 ◼ ► some other products as well, which is, um, especially like a various chat GPT things and like their
01:10:16 ◼ ► Time of day, how much load our servers are under, what plan you're on this, that, the other thing.
01:10:20 ◼ ► It seems pretty clear to me that in the second answer, Siri AI chose to use a less powerful model.
01:10:27 ◼ ► Maybe it used a local model for that one and you use the server model for the first one.
01:10:30 ◼ ► Um, but either way, very different answer to the exact same prompt on the exact same day about the exact same file, which has not changed.
01:10:47 ◼ ► And unlike like the fancier models where you can be like, always use the big, super mega ultra model, super thinking all the time.
01:10:56 ◼ ► Uh, you'll more likely to get a better model if you have like an M4 or better, whatever the specs are that makes you use the more advanced local one.
01:11:04 ◼ ► But this variability in answer really makes it difficult to, I'm not going to say rely on, but just to like, to get consistent results.
01:11:13 ◼ ► Because imagine that I hadn't asked about the, I hadn't had the exact same letter for letter prompt about the exact same file.
01:11:20 ◼ ► You may come away thinking, well, it's really good at that first request, but when I ask it about the other kind of thing, it's bad at it.
01:11:27 ◼ ► Maybe, maybe it's just that the first time I went to a server and the second time it did local.
01:11:30 ◼ ► And so as with all things related to LLMs, it's very difficult to form, uh, concrete judgments based on what they do, because what they do is often not what you think they're doing.
01:11:58 ◼ ► And you'd be like, oh, they're great at sports with bad at knitting, but that's the wrong conclusion because you don't know what's going on behind the scenes.
01:12:04 ◼ ► It's not like they're telling you, I did this one to the server and this one to a local thing.
01:12:07 ◼ ► And so you end up drawing the wrong conclusions and you never ask it in knitting questions.
01:12:14 ◼ ► I'm not, you know, I mean, kind of like the vision pro that you upgrade and then never use.
01:12:24 ◼ ► Um, although I was also playing with the new version of Xcode and it's built in functions for doing agentic coding stuff or whatever to see how those work compared to how they were in Xcode 26.
01:12:36 ◼ ► And when I set that up, I'm like, oh yeah, you have to configure like, you know, one of the third party agents in Xcode.
01:12:44 ◼ ► Cause I'm pretty sure there's, it doesn't use, like you can't configure Siri AI as far as I can tell as an Xcode 27 agent.
01:13:17 ◼ ► For example, they have a, um, agent friendly, uh, incarnation of all their documentation.
01:13:26 ◼ ► It's like vectorized or somewhere or some way, uh, made with like embedding so that the agents can be more easily digested rather than just feeding it the text or whatever.
01:13:33 ◼ ► I, I don't know the details, but the point is Apple has supposedly done work in Xcode 27.
01:13:37 ◼ ► So that agents that are running in Xcode 27 have in theory, a leg up on agents that are just like browsing the web.
01:13:46 ◼ ► Uh, and here's where Casey says, well, it doesn't matter if it's a, you know, a special vectorized version of it or the web version.
01:13:56 ◼ ► But, but like it shows that, that, that it's not, you know, if you're saying, why would I ever use an agent inside Xcode?
01:14:05 ◼ ► Apple's answer is, well, if the ones that are actually used from inside Xcode, they have more direct access to Xcode functionality.
01:14:14 ◼ ► I mentioned on an earlier episode that they made a bunch of, um, essentially skills for agents that teach it about their technologies and you can export them and use them with third-party agents.
01:14:24 ◼ ► But any agent that you use within Xcode gets access to them automatically with no extra work from you.
01:14:29 ◼ ► So I guess Siri AI is not in that, but like, I hope someday Apple does, you know, at least maybe take a third-party agent and train it up on all their internal stuff.
01:14:39 ◼ ► We want an agent, you know, agents can train on stuff and stack overflow and all sorts of open store stuff that they've taken without permission and blah, blah, blah.
01:14:47 ◼ ► But as I said in the past episode, Apple can train models on its own proprietary source code that no one has access to.
01:14:54 ◼ ► Maybe they don't want to do that because like a lot of these agents, you can make them spit back like the New York Times suing to say, hey, we can make, we can make this agent spit back verbatim passages from New York Times article.
01:15:02 ◼ ► If they trained an agent on their own source code and let the public have access to it, could we get it to spit out like the source code to AppKit by, uh, you know, cajoling it in the right way?
01:15:13 ◼ ► It's like you have the source code to all your frameworks and we don't have the source code to all your frameworks.
01:15:19 ◼ ► So please train a model on that source code and make that be the model that I get to use from Xcode.
01:15:39 ◼ ► It is night and day difference from the old Siri, uh, and, uh, that's something at least.
01:15:50 ◼ ► They've like, they've really launched a Siri that is smart and, and has the intelligence side handled.
01:15:58 ◼ ► And so now I think what we need to see is, you know, what remains to be seen, first of all, is like scale speed and reliability because, you know, all the parts that are not locally running have to depend on a lot of infrastructure that, you know, has significant scaling challenges ahead of them.
01:16:14 ◼ ► That's always kind of been one of, one of Siri's weakest points has been, how does it actually perform for everyone everywhere all the time?
01:16:22 ◼ ► And so far they have not done well at that, but this is a whole different architecture with probably all different infrastructure also, because it's such, it's such different requirements to run all these, you know, modern LLM flagship level things.
01:16:49 ◼ ► I don't think we yet have great tooling, uh, for as developers yet to, to know, like, are we plugging in correctly to the new Siri system with our app intents?
01:17:02 ◼ ► Like, I think that's still a little bit early in the current betas, but, um, I haven't actually tried it.
01:17:08 ◼ ► But we, you know, we have that challenge of like, are apps going to be able to really get in there and adopt this?
01:17:21 ◼ ► As I was saying last time, like, you know, do developers have the right, like the incentive to do this?
01:17:28 ◼ ► Or do developers want to not be like disintermediated and to not have Siri, you know, be the face of their app to their customers?
01:17:46 ◼ ► Um, for most small developers, like, like us, we want to do things right by the platform because we're nerds and we're Apple fans.
01:17:55 ◼ ► I think it's going to be a very interesting fall and winter, uh, to see like which big companies adopt this at all.
01:18:06 ◼ ► The thing I'm mainly watching for as this rolls out, in addition to the things that you mentioned, is what does their iteration look like?
01:18:18 ◼ ► There's new versions coming out all the time, minus the government stopping them from coming out, which is currently a problem.
01:18:41 ◼ ► Now that they are using this third party stuff, will they have like an interim release?
01:18:50 ◼ ► It's like now actually we've swapped in all new models because we've got slightly better ones.
01:18:58 ◼ ► If they have been doing it at the Siri, it hasn't made any kind of noticeable change for the better.
01:19:11 ◼ ► But you can be sure that Google is continuing to make new and better models, new and better Gemini models for itself.
01:19:21 ◼ ► It's like we get access to these models, which are, you know, we're probably like maybe they were new when the deal was signed or new-ish when the deal was signed.
01:19:27 ◼ ► But over the course of the next year, if those just stay the same, the rest of the industry is going to keep moving forward.
01:19:32 ◼ ► And Apple's going to be using, you know, late 2025, early 2026 models when everyone else is, you know, moving along.
01:19:43 ◼ ► The whole rest of the stack also obviously needs to be developed and debugged and improved and made better.
01:19:52 ◼ ► One thing they could iterate on and I hope they'll iterate on is their world knowledge thing that we talked about in an earlier episode where they're not doing Google searches for facts, but they have their own sort of world knowledge index.
01:20:02 ◼ ► Apple's got to make sure good info is in there and fix mistakes and keep adding new stuff because the world changes every day.
01:20:09 ◼ ► And so that world knowledge thing needs to keep up with sports scores and holidays and national events and celebrities who are alive and dead and taking stuff from call sheet, right?
01:20:18 ◼ ► Like it's just, they can't just like we'd have a world knowledge thing and we'll just leave it until next WWDC.
01:20:28 ◼ ► But like these, we mentioned these things because they historically have not been Apple's strength, to say the least.
01:20:35 ◼ ► So hopefully they're, with the new leadership, hopefully they realize it's not enough to have this big coming out party at WWDC 2026 and roll out these betas and say, now we're done until next year because you're not.
01:20:51 ◼ ► So if you are in a part of the world that is extremely hot right now, which basically the entire Northern Hemisphere either was or is soon to be, then you might want to rejigger your wardrobe.
01:21:02 ◼ ► And you might want to come up with a wardrobe that works with you, not against you in this very, very, very hot summer.
01:21:08 ◼ ► And if you were wanting to do something like that, let me tell you, Quince is a great place to start.
01:21:12 ◼ ► Among many, many other things that Quince offers, they offer 100% European linen pants and shirts as well that are super breathable.
01:21:29 ◼ ► And I forgot to bring my Quince linen pants to the beach a few weeks ago, and I genuinely was very, very disappointed and upset by it.
01:21:36 ◼ ► So I have already gotten out these linen pants from my dresser, and I've put them in my staging area so I do not forget them when I go to the beach imminently.
01:21:49 ◼ ► And if you're not into linen pants, let me tell you, Quince offers a ton of other stuff.
01:22:01 ◼ ► It really is astonishing how much they do and everything that our family has gotten from Quince we've enjoyed, including, for example,
01:22:44 ◼ ► And Amar writes, one of the expectations from Apple after they moved to M-Series chips was that they would have complete control over release cadence.
01:22:54 ◼ ► We currently have a situation where the M5 was released first on an iPad of all products, while the desktops are languishing on M4 or even M3 chips.
01:23:02 ◼ ► I should note that this has been in the show notes for at least a couple of months now.
01:23:17 ◼ ► Anyway, Amar continues, to make matters worse, it seems as if the Mac Studio may not even get M5.
01:23:28 ◼ ► If you're in the market for the absolute balls-to-the-wall desktop that would last you seven or eight years, John Syracuse, you have nothing because investing in a M3 Ultra makes zero sense.
01:23:48 ◼ ► Well, the premise of this is that they would have complete control of a release cadence.
01:23:55 ◼ ► But complete control they can never have because they don't fab their chips and they don't control the world.
01:24:09 ◼ ► And I don't know how old this was in the notes, but like, everything in it, it's slightly outdated.
01:24:15 ◼ ► Because of the news we just read, but like, it, the, we kind of got spoiled by the M1, M2, M3 cycle where it seemed at least somewhat predictable and under control.
01:24:28 ◼ ► And I bet if Apple did have complete control, they would continue a orderly progression like that.
01:24:53 ◼ ► And actually, it's going to be the only thing, because we're skipping all the M6s, and we're going to do the M7, and Intel might fab it.
01:25:01 ◼ ► And Intel wasn't more chaotic, but it was worse in that there were long periods where we had nothing.
01:25:10 ◼ ► And so at least now it's like, there's this menu of things that we think are going to be awesome.
01:25:18 ◼ ► But every one of those M things, speaking of the M chicks in particular, every one of those M things was tailor-made to Apple specifications for specific product needs.
01:25:30 ◼ ► But when we get them, it's a thing that Apple very intentionally made that works really well.
01:25:36 ◼ ► Whereas Apple and Intel, their relationship was basically Apple trying to cajole and coerce and beg Intel to make the kinds of chips they wanted.
01:25:46 ◼ ► And Intel saying, well, I guess we'll put a really big integrated GPU because you need it for your laptops.
01:25:56 ◼ ► But like, it was just, it was never, you know, Intel, Apple was an important customer to Intel.
01:26:02 ◼ ► Apple wanted very specific things and Intel would give them a little bit of what they wanted based on the chips they had and the things they could add for them.
01:26:12 ◼ ► Like, can you believe we're doing this custom integrated GPU just for Apple, this one customer?
01:26:17 ◼ ► And then Apple's like, but we would change a lot of stuff about that chip, not just the GPU.
01:26:24 ◼ ► So I do think things are better than Intel, but the current situation with the, you know, the rumored M6 Pro Max and Ultra cancellation, the already passed lack of an M4 Ultra, the studio waiting forever for an M5 and still doesn't have it.
01:26:39 ◼ ► I think when the person was writing this, they maybe didn't expect this to be the case, but it still doesn't have it.
01:26:49 ◼ ► Yeah, like I think the reality is that what Apple Silicon brings us is Apple gets to design exactly the chips they want.
01:27:14 ◼ ► And also, as mentioned earlier, Apple is no longer TSMC's biggest customer or most important customer necessarily.
01:27:21 ◼ ► And so they have to compete for other vendors for that manufacturer's time and for the yield on those high-end lines.
01:27:29 ◼ ► And so, you know, we like to think like, oh, just when you go to the next chip, like every year, you just make another one and make it better.
01:27:46 ◼ ► You might be trying to figure out how do you squeeze more performance out of the same feature size without being able to shrink down the process yet.
01:27:55 ◼ ► And then when you do shrink the process down, well, what if the new process works a little bit differently than the old one?
01:28:00 ◼ ► And what does that mean for all the different little tiny gates and stuff that are in the chip?
01:28:04 ◼ ► And sometimes things need to be built a little bit differently or things need to be designed a little bit differently when you change the process.
01:28:11 ◼ ► It's a very complicated, like, implementation of the idea of make a new chip every two years or every year.
01:28:23 ◼ ► And if there is any little bump in the road when tackling any of those challenges, that can push the date out or that can make a whole type of chip maybe not worth making or impractical to make.
01:28:36 ◼ ► Or it would make the yield so low that it would be way too expensive and, you know, it wouldn't really sell at those prices.
01:28:52 ◼ ► Maybe on that whole process, like back when they made, like, the N4P or N4E, whatever that was.
01:29:00 ◼ ► Because, like, it turned out that was just super expensive and not amazing yields forever.
01:29:18 ◼ ► So, Apple does a really good job, generally, in lots of ways, of hiding the reality of the complexity of their supply chain and design process from the customers.
01:29:34 ◼ ► Because Apple, like, Apple went through all of COVID without really ever being out of stock of anything for very long.
01:29:42 ◼ ► And without really being short-supplied on components and without really changing any prices, they were the only ones.
01:30:05 ◼ ► They masked it all with just execution and competence and buffers and whatever else they were doing.
01:30:37 ◼ ► So, in a case like when you have the really huge ultra chips, that's kind of an edge case.
01:30:56 ◼ ► And things that work for a chip generation, for all the other chips they make, might not work for the ultra chip or whatever it is.
01:31:29 ◼ ► If these M6 rumors are actually true, it will be interesting sometime in the future to find out, did Apple actually design an M6 Pro, an M6 Max, and an M6 Ultra, and basically just had to throw in all that work?
01:31:57 ◼ ► And apparently the M7 design has some kind of advancement that is a significant advantage for on-device AI.
01:32:04 ◼ ► And so, I can see Apple saying, I know, I know, like, the M6 Pro, Max Ultra, they're great chips.
01:32:11 ◼ ► But, like, if we can get the M7 six months sooner by just not doing the M6, we should do that because it's more important.
01:32:23 ◼ ► Like, if the M7 actually is better, I would rather have that six months early and just skip the M6s.
01:32:27 ◼ ► But, if you just paid all your employees to design those things, maybe they've been test-fabbed.
01:32:33 ◼ ► Like, it's just such a shame to say, yeah, we spent, teams spent, you know, years working on these chips.
01:32:39 ◼ ► And we just have to go, sorry, that's, I mean, it's like, you know, people who make a movie that never gets released.
01:32:53 ◼ ► Although, I don't think that's the case because the design timelines on these are long enough that they have to.
01:32:58 ◼ ► Or, maybe they made the decision three years ago and it has nothing to do with the RAM crisis.
01:33:11 ◼ ► And, like, I bet, like, M6s, like, what they've done with the M series has shown, you know, there are good years and bad years.
01:33:29 ◼ ► And, again, I think we were spoiled by, like, they're going to come out with an M series.
01:33:50 ◼ ► Because, if you map it out and you see the release cadence and how they've kind of spread and distributed and holes have been formed where certain chips just never arrive.
01:34:00 ◼ ► And, we're getting more holes and more delays and more, like, you know, as Amara points out, like, it seemed really weird when, like, a new M chip would debut on an iPad.
01:34:23 ◼ ► And, I do remember, in case someone's going to write in about this, way back at the beginning of the M series or maybe somewhere in the middle, like, there was some rumor from someone saying, yeah, all those quad chips, like the rumored M1 that was going to be, like, four maxes stuck together.
01:34:40 ◼ ► And, at some point, there was a rumor or something that was saying, like, and, by the way, don't keep your hopes up for any other sort of, like, quad-type chip.
01:34:47 ◼ ► Because the roadmap has been laid out and the earliest there could even possibly be a monster chip like that is the M7 generation.
01:34:57 ◼ ► And, all they were saying was, I've seen the roadmap for M1 through M6 and it's not on there.
01:35:19 ◼ ► And, I mean, back then they weren't saying that the M7 generation was going to concentrate on on-device AI because that wasn't really a thing back then.
01:35:37 ◼ ► But, I am excited that this rumor says that Apple is so excited about the M7 that they're willing to sacrifice, you know, what, three quarters?
01:35:47 ◼ ► Most of their M6 line, all of their M6 line except for the single, presumably single die, no suffix M6, that they're going to sacrifice all of those just so the world can get the M7 a little bit sooner.
01:36:04 ◼ ► I enjoyed the discussion about languages in the overtime segment of episode 696, including Apple's assertion that Swift can be used as a replacement for languages from C to Perl.
01:36:12 ◼ ► It occurs to me that if code is being developed agentically, the choice between a lower or higher level language might differ.
01:36:19 ◼ ► If no human is going to review the code, shudder, maybe the bots should work closer to the metal for more efficient code.
01:36:24 ◼ ► Conversely, maybe the wealth of available libraries for open source frameworks would make sticking to those frameworks more effective.
01:36:31 ◼ ► I understand what Matthew's going for here, and I'm curious to hear what you two think.
01:36:34 ◼ ► But the thing that struck me about this is that, and maybe it's just that I haven't seen it, but I got to imagine there's not a lot of, like, ARM or x86 assembly code posted online.
01:36:47 ◼ ► Nowhere near as much assembly code, for example, posted online as there is C, C++, Swift, Perl, PHP, et cetera, et cetera.
01:36:55 ◼ ► So if you're relying on a thing that needs a whole corpus of data to train upon, I would think you would want that thing to be the language that it has the most data about, you know, something like a C++ or perhaps Perl or whatever, you know, where the LLM knows so much about whatever the language is.
01:37:18 ◼ ► And it's seen so much code from whatever the language is that it's going to, it's more likely to do a good job as opposed to, like, assembly, which I got to imagine there's not near as much assembly floating around the world as there is Perl.
01:37:31 ◼ ► I feel like Matthew's question here is, you know, again, one of the many traps that the LLMs lead us toward, which is it's so easy to map onto them the things that we have been exposed to in, like, science fiction media.
01:37:46 ◼ ► And that, well, yeah, we have these high-level languages for us, but computers are computers.
01:37:51 ◼ ► And so we don't, they don't need to, they're, they're unlike us and, and they're unlike us in the ways that I have seen computers be unlike us in the movies, which is things that are nonsensical to us computers can understand.
01:38:02 ◼ ► So why bother with these high-level languages when computers can just understand the, the binary language of moisture evaporators?
01:38:13 ◼ ► So they don't need, and in some respects, like they're on the right track and that computers aren't like us and they have different strengths than we do.
01:38:28 ◼ ► But as you just pointed out, Casey, LLMs, how LLMs, how this specific technology works is training data is the thing that determines, like their training determines what they are.
01:38:53 ◼ ► And that gets to my second point, which is computers also benefit from all the same things that humans benefit with high-level languages in that there are, they don't have to think about certain details when they're using a higher-level language.
01:39:08 ◼ ► And then there can be certain assumptions they can make about how the high-level languages work because languages like C and other languages provide some guarantees about how things work.
01:39:17 ◼ ► Not that you know exactly what assembly will be created by the compiler, but there are certain, that's what a high-level language provides.
01:39:24 ◼ ► It provides certain behaviors above and beyond the behaviors defined by assembly itself.
01:39:34 ◼ ► Obviously, there's undefined behavior and all sorts of stuff in C, but I'm just saying, like, the benefits of high-level languages, those benefits also benefit LLMs.
01:39:42 ◼ ► It makes them able to write code without worrying about the low-level concerns the same way it does to us.
01:39:51 ◼ ► Even though we operate differently, that combination of things I feel like can't be overcome, the massively larger amount of training data and the fact that they can work better in high-level languages.
01:40:00 ◼ ► Then I'll add a third one, which is if you're paying per token, you want to use a higher-level language because the bottom line is there's fewer characters produced.
01:40:08 ◼ ► There's just fewer, at the minimum, there's fewer output tokens, but probably there's also fewer input tokens because if you're feeding it some source code and say, what's wrong with this?
01:40:17 ◼ ► I'd rather feed it a paragraph of C instead of a much longer or several paragraphs of assembly because, yeah, it takes more instructions.
01:40:26 ◼ ► So I know what Matthew's getting at, but I just think it's like it's about barking up the wrong tree.
01:40:36 ◼ ► And, you know, wouldn't they be more comfortable in binary because they're computers or robots?
01:40:50 ◼ ► It's like, how can it make sense of things when it's doing, like, pairs of letters or three?
01:40:54 ◼ ► It's like, yeah, so they're not exactly like humans, but they are, in fact, large language models.
01:41:20 ◼ ► Yeah, I think the, I kind of mused a couple of months back, like, have we developed the last programming language?
01:41:35 ◼ ► In this, and what I mean by that is like, do we have any reason to make new languages now, really, for mass use?
01:41:43 ◼ ► Because my thinking on that was like, LLMs have now been trained on all the languages that we have, you know, like that people publish code for in large volumes.
01:41:54 ◼ ► They are not surprisingly, really good at JavaScript, because there's a lot of JavaScript out there right now.
01:42:03 ◼ ► They're, they're, they're pretty good at Swift, they're pretty good at C, they're, and they're, you know, as you kind of get down, like, you know, the less commonly published languages on the web these days, you know, they're, they're okay at PHP, you know, I'm sure they're okay at like Ruby and Perl and Python.
01:42:18 ◼ ► You keep going down, like, a little bit lower in the, in like the usage list, like, I'm sure they have, they've seen a lot of Java, and maybe they, maybe they've seen a little bit of Objective-C, maybe they've seen a little bit of, of regular C, you know, probably a lot of C++, but maybe not so much Rust.
01:42:33 ◼ ► Like, like, you start to get, like, the lesser used languages, and if you're doing LLM-based development, based on what they've been trained on, as John was saying, probably they're going to be best at the languages that they have seen the most.
01:42:49 ◼ ► So, my thinking generally on that is, like, we probably want LLMs to be writing whatever is the most popular language that can be used to develop the type of app or on the platform that they're, that they're building.
01:43:03 ◼ ► So, if you're making a web app, you probably want it to be making JavaScript in, you know, on the back end and the front end for that.
01:43:09 ◼ ► If you're making an iPhone app, you probably want to be using Swift, you know, like, that kind of thing.
01:43:15 ◼ ► And now that we've had all these languages, you know, out there and training the LLMs, willfully or not, does it make sense to invent a new language now that would be kind of from scratch from the LLMs perspective?
01:43:32 ◼ ► You know, if we're, if we're going to use LLMs to help us write code or to write it for us most of the time in the future, which does not seem that remote of a possibility, shouldn't we use a language they've been trained on?
01:43:58 ◼ ► As they get smarter, they, they have more training, they have more parameters, they have more sophistication in their harnesses and, and their prompting and everything else.
01:44:07 ◼ ► What we are seeing, I think, is that over time, LLMs will actually not even care what the languages they're asked to generate because they're, they're thinking kind of higher level.
01:44:17 ◼ ► They're, they're seeing higher level patterns and they will have just as easy of a time generating one language as any other language because they're just kind of outputting patterns.
01:44:25 ◼ ► And the training data was more about like building those mental patterns than the specifics of, it must be this language, you know, printed out this way.
01:44:34 ◼ ► So I think ultimately, right now, the answer is, if you're going to have an LLM write code, make it a popular language.
01:44:42 ◼ ► But I bet in the fairly near future, the answer will just be who cares, it doesn't matter.
01:44:49 ◼ ► I don't know if there's anything in the way LLMs currently work that will help it and not care about the language because in the end it has to make, it has to know the syntax and output it and it very much will parrot back code structured to similar what it's seen.
01:45:03 ◼ ► So I don't, I don't think it's becoming language independent, but to your earlier question, I think it's important to ask like, why do we make new languages?
01:45:13 ◼ ► Why don't we, aren't we in the same situation of like, well, all these programmers were trained in the languages that exist and we should just keep using them.
01:45:28 ◼ ► But the reason we do it is because the new languages offer some benefit that the other language didn't.
01:45:34 ◼ ► Just look at Apple with Objective-C and with my Copeland 2010 articles where I was saying, hey, Apple needs a modern language.
01:46:03 ◼ ► Like people are going to suddenly start with zero years experience with this new language.
01:46:07 ◼ ► They weren't LLMs, but they have the same issue, which is like, well, we're already trained, so to speak, on Objective-C.
01:46:20 ◼ ► It's difficult to write correct, bug-free, secure code in these languages because they have undefined behavior.
01:46:25 ◼ ► They have pointers, like, you know, manual retainer release before the days of ARC, all that other stuff.
01:46:45 ◼ ► That's why a memory-safe language will, humans will make fewer memory bugs in a memory-safe language, and so will LLMs.
01:46:56 ◼ ► That's why we made all these higher-level languages is because they provide a way for a thing of some skill, whether that's a person or an LLM, to write fewer lines of code, which is fewer tokens for LLMs, but it's less work for humans, and to make fewer mistakes because of the design of the language.
01:47:12 ◼ ► And I don't think that will change because, like, it's not as if Swift is going to be the end-all, be-all.
01:47:18 ◼ ► You know, it just seems like C is going to be the end-all, be-all when everyone moves on to C, but then, you know, C plus C.
01:47:40 ◼ ► And for all the benefits that we see in Swift, all the things that the Swift compiler does and the Swift language does are so focused on trying to make certain kinds of bugs, if not impossible, then either really hard to do, or a lot of Swift design is like, we have to allow you to do this.
01:47:57 ◼ ► But every time you do it, we're going to put something in the code so it's easy for both the human and LLM to find.
01:48:03 ◼ ► For example, if you search your Swift code for the word unsafe, U-N-S-A-F-E, that should point you to a lot of places where you're doing a thing that bypasses some protection.
01:48:19 ◼ ► And so there will be some language after Swift that is better than Swift in the ways that Swift was better than Objective-C.
01:48:29 ◼ ► If we get to the point where humans don't want to make a new language because they can't imagine a way that a language can be better, then we'll be at the end of history and we don't need to know any more programming languages.
01:48:41 ◼ ► I mean, I don't know how long Swift was going to go, but Swift does still let you do a bunch of unsafe stuff.
01:48:47 ◼ ► And 200 years from now, someone's going to think it's barbaric that all the unsafe stuff that we can do in Swift is like, why did they even allow that in a language?
01:48:54 ◼ ► Like, well, they had to because it was trying to replace C code and the certain things you can't do and blah, blah, blah.
01:49:03 ◼ ► Like, we have so much safety guarantees and mathematically provable correctness and impossibility of, you know, security flaws in these five realms.
01:49:13 ◼ ► And we're fighting these other realms that you hadn't even thought of for quantum computers and stuff like that.
01:49:19 ◼ ► So, yeah, I think we'll continue to have new languages and I think everyone will benefit from them, language models and human alike.
01:49:34 ◼ ► I will say, this is not an official answer of mine, but I will say I happened to hold the MacBook Adorable earlier today.
01:49:43 ◼ ► And I keep thinking about like the Intel CPU conversation from earlier and I held this thing and it just is so amazing.
01:49:52 ◼ ► And I know we've talked about this a hundred times, make it 101 now, a modern cut of that with Apple silicon on the inside would be so amazing.
01:49:59 ◼ ► But anyway, to actually answer Phil's question, I grew up on ThinkPads, that's what I'd want.
01:50:03 ◼ ► I would want a ThinkPad 701C, which is what you would know as the one with the butterfly keyboard.
01:50:09 ◼ ► And I would also want a ThinkPad T30, which was, I think, I'm pretty sure that was my college computer.
01:50:35 ◼ ► Yeah, the answer is like I really shouldn't go buy vintage computers because I just won't use them.
01:50:46 ◼ ► Like I bought a Palm 5X a couple years back off V-Bay because you can get one like brand new condition for like 25 bucks.
01:51:04 ◼ ► If we were going to do a video show and I wanted to have a background that looked cool on video.
01:51:14 ◼ ► Like I, you know, I would have that kind of fun vintage computer stuff basically as scenery behind me in a video.
01:51:27 ◼ ► One of the problems you quickly face, which is something that you don't really face with game consoles.
01:52:00 ◼ ► And then, if you can get it to actually like connect to the internet, if it's that kind of device.
01:52:18 ◼ ► Will it have like the right code signing support server side still running to let it do any, you know, install any software.
01:52:28 ◼ ► Like, you start running into problems like that with a lot of kind of mid-vintage computer equipment.
01:52:35 ◼ ► Now, if you go back far enough to stuff that didn't really use the internet for anything.
01:52:57 ◼ ► Like, and that's, I mean, that is how I used my entire first computer for much of the time that I had it was like, yeah, I would, you know, have a couple of games.
01:53:30 ◼ ► But it's the kind of thing that I like watching, like, when other people on the internet do cool stuff with vintage computers.
01:53:39 ◼ ► It's a fun thing to kind of see breeze through, like, a social feed or a blog post or a YouTube video.
01:53:53 ◼ ► And I will point out, much like collecting books, collecting books is a different hobby than reading books.
01:54:00 ◼ ► So, the idea that you would actually have to use these computers you collect, it's like, no.
01:54:11 ◼ ► And on your second point about what are you going to do with them, I think it shows the bifurcation in both your knowledge of the scene and also in, sadly, the technology.
01:54:24 ◼ ► And one of the shames about them is if they don't have a good jailbreak available for you,
01:54:31 ◼ ► it's sometimes difficult to get those more locked-down iOS devices to, you know, do interesting things.
01:54:38 ◼ ► Because, like, again, if you don't have access to a good jailbreak, you're limited to what they used to do.
01:54:44 ◼ ► And if, like, the servers aren't up or the software doesn't work or, as you said, the TLS, blah, blah, blah.
01:54:47 ◼ ► But in the realm before closed-down platforms, man, people are adding Wi-Fi to Mac Pluses, okay?
01:55:05 ◼ ► Steve Troutman-Smith was just porting his apps in Swift to classic Mac OS so it can run, like, a 128K Mac and stuff.
01:55:12 ◼ ► Like, this, for the computers that are not locked down, people just write stuff or have LM's write stuff.
01:55:22 ◼ ► If they write a new web browser to run on a machine with 128 kilobytes of RAM that had never seen the Internet.
01:55:29 ◼ ► And, you know, they have a little, for Wi-Fi hardware, they have, like, you know, little peripherals that you can buy and plug into your Mac SE.
01:55:47 ◼ ► Then there's, I have them, and I'm going to, like, make them work in the modern Internet.
01:55:56 ◼ ► It looks so weird to see, like, on a 9-inch monochrome black-and-white screen on a computer from the 80s.
01:56:04 ◼ ► And they, you know, like, do two asterisks around something or one or whatever it is in Markdown.
01:56:32 ◼ ► I had these computers as my one and only main computer that I used for years and years and years.
01:56:45 ◼ ► Many, many, many years later, I just always wanted to own one because it was such a cool computer.
01:57:18 ◼ ► And I did, in fact, use them when they were contemporary and in college and thought it wouldn't be cool to have one of these someday.
01:58:33 ◼ ► I need to get rid of tons of computers before I think about even getting another one of them.
02:00:00 ◼ ► M-A-R-C-O-A-R-M-N-T-Marco-R-M-N-S-I-R-A-C-U-S-A-C-R-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U-S-A-C-U
02:00:00 ◼ ► m-a-r-c-o-a-r-m-n-t-m-arco-r-m-n-s-i-r-a-c-u-s-a-c-ra-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-c-u-s-a-
02:00:30 ◼ ► wrong, but I can't figure out what the heck it is. So I have four, I generally have my devices
02:00:39 ◼ ► in a, in a focus mode that only lets through like a handful of contacts. And I had, I didn't think
02:00:47 ◼ ► about it until just this moment that I should have tested this issue outside of that focus mode to
02:00:51 ◼ ► see if that focus mode might be part of the problem, but let's assume it isn't. But the,
02:01:00 ◼ ► I will feel the tap tap when those people that are blessed through that focus mode text me. I'll
02:01:08 ◼ ► feel it to feel the tap tap when I'm not in the focus mode, which is not common, but it happens
02:01:12 ◼ ► when I'm not in the focus mode. And I'll feel a tap tap when people text me that are, you know,
02:01:16 ◼ ► just regular people or people that are not special enough to be in my, you know, almost all the time
02:01:29 ◼ ► Aaron texts me, which is crappy because the one person I want to always get a tap tap from
02:01:37 ◼ ► Is this an elaborate, uh, set up here for you to try to, for us to try to like, excuse you
02:01:43 ◼ ► from something that you're partially, partially, no, no, it's not that I'm not answering your text.
02:01:53 ◼ ► Uh, yeah. I mean, I, I, that is a little bit of this, but like she, we haven't had any particular
02:01:59 ◼ ► conversation about this in a fair bit of time now, but there are times that she wants a semi
02:02:03 ◼ ► urgent response from me and very understandably gets a little perturbed when I'm like, I didn't
02:02:10 ◼ ► even see your text. Now I will note that I had way back when set up a custom, like, um,
02:02:17 ◼ ► haptic notification on the phone for when she calls. Like you can, there's a way, I think
02:02:22 ◼ ► it's an accessibility. There's a way to set up a custom buzz basically when she calls. And
02:02:27 ◼ ► since my phone is always silent, my watch is always silent insofar as it doesn't make verbal
02:02:31 ◼ ► or not verbal, but it doesn't make audible noise. Um, but there's gotta be something I have done
02:02:37 ◼ ► wrong that, that is causing me to not receive the tap tap when she sends me a text on my watch.
02:02:45 ◼ ► And for the life of me, I can't figure it out. I did look and in the contacts app, there is,
02:02:51 ◼ ► there are a bunch of different, um, like, uh, uh, toggles that you can do. And one of them is,
02:02:59 ◼ ► you know, silence notifications. I'm sorry, that's not in context. That's in the messages app.
02:03:06 ◼ ► maybe this exists on the watch and I've somehow turned it on and I don't know it, but on my phone,
02:03:11 ◼ ► I see, I'm sorry, it's hide alerts, which is off. And then I generally don't send read receipts,
02:03:16 ◼ ► read receipts, whatever you pronounce it, but I have those on for her. I have show and shared with
02:03:21 ◼ ► you turned on. I have share focus status turned on, which I also don't typically turn on for most
02:03:32 ◼ ► in iMessage on my phone. And for whatever reason, I cannot figure out why my watch does not do a tap
02:03:39 ◼ ► tap when she sends me a text. Oh, actually that's mostly true. Interesting wrinkle. It seems to be
02:03:46 ◼ ► when she, it's only when she sends me a text in the conversation between only she and I,
02:03:51 ◼ ► we are on a zillion different group chats where she will break through, like for example,
02:03:56 ◼ ► in focus mode, which is what I want for the record. She will break through in a group chat where none
02:04:01 ◼ ► of the other people in the group chat are blessed in this particular focus mode and she'll break
02:04:05 ◼ ► through in the group chat messages. But if she sends me a private message, so to speak,
02:04:08 ◼ ► I don't get it both in the sense that I don't receive it. And also I don't understand what's
02:04:13 ◼ ► happening. So that's, I, I, I am, this is a call for help. This is, this is one of the advantages
02:04:19 ◼ ► of being on the show is that you can ask for help and, and crowdsource this. What have I done wrong?
02:04:32 ◼ ► A lot of husbands seem to have a technical reason why they're not answering their wife's texts
02:04:43 ◼ ► It is really not. And I would, please, if you figure out the answer, let me, so, okay, my
02:04:50 ◼ ► theory, so I, I wear an Apple watch and I'm usually, when I am like not getting these messages,
02:04:57 ◼ ► it is almost always when I'm at my computer. Now, I'm always at my computer. So that's, that
02:05:03 ◼ ► might not be anything. But, but I, yeah, I, I don't know what it is, but there is something
02:05:12 ◼ ► about like me being at my computer where I, my theory is that like one of the devices is
02:05:20 ◼ ► kind of saying, I got this. I'm currently being used. I'll alert the user. And then the other
02:05:25 ◼ ► ones don't. I don't know. My, another theory is, you know, if I'm in the middle of doing something
02:05:43 ◼ ► No, truly, this happens to me a lot. Cause remember I'm going three up and if I'm looking at the
02:05:47 ◼ ► leftmost monitor and the notification is, is actually on the upper right-hand corner of my
02:06:00 ◼ ► I mean, I would love to do that if such a thing is possible. I thought I did do that actually,
02:06:10 ◼ ► or to stay there until you interact with them or dismiss them. Maybe the answer is just to turn
02:06:24 ◼ ► Anyway. So I am genuinely gobsmacked that you also have this problem, Marco. And I wish I could
02:06:29 ◼ ► tell you like, yes, there have been some issues between she and I from time to time when she really
02:06:33 ◼ ► needs a response and I just don't give it to her. And she's right. Like, I'm not trying to paint
02:06:42 ◼ ► Uh, well, sometimes we're in the, you know, we're in the same house. That's the other thing
02:06:50 ◼ ► Yeah. But, you know, something I want to talk about in a future episode of ATP is just how ridiculous
02:07:01 ◼ ► conversation during the day is done via text message. We're in the same freaking house.
02:07:07 ◼ ► My son will call me from his bedroom. I'll be downstairs and he will call me a phone call
02:07:13 ◼ ► Yep. I mean, this is, we, we don't live in big houses. This is a regular size house, but
02:07:19 ◼ ► Yep. Anyways, but I, I, I've got to imagine that I have been, apparently we, Marco and I
02:07:43 ◼ ► And I, what I think is I think Marco was on the right track with like, we know Apple devices
02:07:48 ◼ ► in theory coordinate with each other. So they don't all have notifications. So one of them's
02:07:52 ◼ ► got to say, I've got this. And the other data point is I don't wear a while Apple watch, but
02:08:03 ◼ ► she's wearing an Apple watch. Now, my question for all of you is like when you're wearing the
02:08:08 ◼ ► watch, okay, maybe the watch doesn't vibrate and you expected it to, but does the notification
02:08:21 ◼ ► Well, I mean, obviously the watch hasn't handled it because you're saying your watch isn't vibrating,
02:08:24 ◼ ► but like, I do wonder if adding the watch to the mix is the thing that, that it exposes
02:08:29 ◼ ► whatever Apple bug I'm, I'm, first of all, I'm saying I'm calling this an Apple bug because
02:08:34 ◼ ► so many people, I like my wife and everyone else is always asking me questions exactly like
02:08:38 ◼ ► you just phrased it, which is like, I expect to get a notification and I don't. And very,
02:08:42 ◼ ► very rarely it is a question of settings, much more likely it is. You've got a dozen Apple
02:08:47 ◼ ► devices. And honestly, I don't even know what's supposed to happen here, but we can demo
02:09:01 ◼ ► bugs plus Apple watch that causes this problem. And the solution is all get rid of your Apple
02:09:17 ◼ ► honestly, who's trying to text you during the day? It's fine. Just leave yourself open to
02:09:23 ◼ ► So my general, I haven't messed with this in literally years, but my general focus node,
02:09:28 ◼ ► um, like routine is that when I sit down to work at about eight in the morning, I go into
02:09:36 ◼ ► a work focus mode and you two knuckleheads can break through. Mike can break through. There's
02:09:39 ◼ ► a couple other people that break through, but generally speaking, it's one of those situations
02:09:43 ◼ ► where I will see it when I will see a message when I go to the messages app, you know what I
02:09:47 ◼ ► mean? Like I, it's not going to notify me. Um, if, if, if some random schmo sends me a text
02:09:56 ◼ ► the kids are home from school, I go into this, this kind of universal focus mode, I call it
02:10:00 ◼ ► personal. What that basically means is family can break through and basically no one else.
02:10:07 ◼ ► Um, I actually genuinely don't remember if I have you guys and Mike on there or not, but,
02:10:11 ◼ ► um, for the most part in the personal focus mode, which is what I live on basically from
02:10:15 ◼ ► the time the kids get off the bus until their bedtime, then it's pretty much only family that
02:10:22 ◼ ► can break through. And then in the evenings, typically I'm pretty sure they all get turned
02:10:27 ◼ ► off and I am in like wide open, no, no focus mode, wide open mode. Um, and I, I think you're
02:10:33 ◼ ► right. I do need to experiment with this. I think it's worth experimenting with my custom
02:10:36 ◼ ► vibration, even though that is only for phone calls, if I'm not mistaken, but I should experiment
02:10:40 ◼ ► with that. But I, I just don't understand why she, the one person I always freaking want to get
02:10:48 ◼ ► texts from or notifications from is the only person I consistently don't get them from. And
02:10:52 ◼ ► specifically in the conversation with just the two of us. Cause like I said, she will poke
02:10:58 ◼ ► through on group chats, like a group chat will be popping off when I'm in the personal focus
02:11:01 ◼ ► mode and I will see only her messages, which is chef's kiss. That's exactly what I want because
02:11:07 ◼ ► whatever it is she's saying, wherever she's saying it, I want to know what I want to know
02:11:09 ◼ ► immediately. But for whatever reason, when she sends something in our private channel message,
02:11:16 ◼ ► whatever you want to conversation, I guess, then it won't, then it won't notify me and it's
02:11:20 ◼ ► I mean, you're highlighting the fact that she's special and that she has different settings
02:11:24 ◼ ► than everybody else. So maybe unspecial her and make her have the same settings as everyone
02:11:27 ◼ ► else. Also do some experiments with her two feet away. At least at the very least, this
02:11:31 ◼ ► will give you more plausible cover when she thinks you're ignoring her messages and you
02:11:34 ◼ ► say you don't, because if you do that experiment for a while until she's sick of like send me
02:11:38 ◼ ► a text message, like when she's like literally sitting next to you, hell put your watch on her
02:11:41 ◼ ► wrist and say send a text message. Right. Of course, if it backfires on you and it constantly
02:11:47 ◼ ► vibrates every time she does it, she's going to give you quite a look. But, uh, uh, I'm taking
02:11:50 ◼ ► all you at face value here and say, this is not elaborate cover for your past sins, but
02:11:54 ◼ ► is in fact an actual technical problem, which I have no problem believing because Apple's
02:11:57 ◼ ► notification system is very buggy. So, and like I said, I'm on the other end of it. I'm on
02:12:01 ◼ ► the one who's the watch wearing wife is telling me, Oh, I just never, never vibrated. I don't
02:12:06 ◼ ► All right. So let me ask you. So I'm looking at my personal focus and I'm looking at my iPhone
02:12:09 ◼ ► and I have turned on intelligent breakthrough and silencing, which I don't think that's really
02:12:15 ◼ ► relevant because of anything that would make them come through. Not, not, not to come through
02:12:18 ◼ ► anyways. So in the people category, I have, you know, you guys, Mike, Aaron, and a couple other
02:12:25 ◼ ► people, basically like a couple of family members and, um, like the, basically the school trunk line.
02:12:30 ◼ ► So then I also have some apps that can break through, although not many and below the people
02:12:35 ◼ ► in apps section, there's an options, uh, thing or navigation link, basically where it says notification
02:12:41 ◼ ► options. And it appears to me that everything on the screen is under the auspice of silence
02:12:48 ◼ ► notifications. So there are, uh, four different options, show unlock screen on or off. I have it
02:12:54 ◼ ► off hide notification badges. I have it off silence notifications. I have as always. And, and then I'm
02:13:00 ◼ ► sorry, a different section is appearance dim lock screen, which I also have off. So it's, so if I'm
02:13:05 ◼ ► reading this right, silence notifications is show on lock screen off, hide notification badges off
02:13:09 ◼ ► silence notifications always. But again, that's for the silence notifications. And then additionally,
02:13:15 ◼ ► there's an option for appearance dim lock screen, which I also have off. So I don't think that's it
02:13:19 ◼ ► either. It's got to be something I've set up wrong, but I'll be damned if I can figure out what.
02:13:24 ◼ ► Do you only want to, do you only want answers from spouses or do you take it from single people?