Upgrade - 436: I Bend So Many Spoons

Episode Date: December 6, 2022

John Siracusa returns to the show to chat about Jason's recent visit to his house, Mac app development, Apple's brain drain, TSMC breaking ground in Arizona, the problems with AI training models, the current state of macOS, and what makes a Good Product.

from relay fm this is upgrade episode 436 today's show is brought to you by story worth capital one and uni pizza ovens here is your host jason Snell. Thank you, Mike. Hello, everybody. We are back for another Mike-less almost episode of Upgrade. I am Jason Snell, and I am hosting this episode with a very special guest returning to Upgrade for the, I actually have lost count, half a dozen or so time. It is John Syracusa. Hi, John. When you say Mike-less, we both have mics in front of us, but that's okay, right? You didn't hear the Y in there?

No. Is it like tire with a Y? It is. I just put that together. Yeah, but not choir because that has a Q-U. Mm-hmm. That's not Robot or Not that we're doing now. That's a totally different podcast that we're doing about a weird english spelling and things anyway uh i just this for this trip i am

just replacing mike with a series of guys named john as you do there's so many of us why not why not take advantage of the bounty i bought the six pack at costco and uh you know you got to use them all but we start off with the hashtag snell talk question which I'm also going to apply to my guests which I have again posed myself because I want to know John what is your Christmas shopping strategy I mean I have to say i'm have never been good at gift giving or buying my main my main skill lies in the the years that i spent honing my gift asking for abilities as a child that seemed very important for me to develop as a kid so i uh i feel like i'm good at that i've always been good at making listed lists of presents that i want and i've been good at figuring out how to get my parents to give them to me and those skills stopped being useful once i left the house uh and

now i'm left with nothing uh i not don't do a good job of buying people things i don't know what to get people things. It's a problem. So, I mean, my strategy, such as it is, is all throughout the year, I have a notes, you know, a document in Apple Notes called Gift Ideas. And anytime there is any inkling of any kind of gift that anybody might like, I write the gift ideas, little headings for people's names, and then there's like little bulleted lists of gift ideas for them. And then when the holidays come, my strategy is I pull up the document and I hope that there will be some headings underneath people's names that aren't already checked off. And if they are already checked off, then I go to the people and say, what do you want? Give me a list.

I need a list. I can't think of anything. Yeah, that's my strategy. It involves a lot of whining and uh and uh occasionally looking at a notes document this is very much um my strategy too and my story too i'm very bad at this i need people's help fortunately having been married more than 25 years um you've got we've gotten to the point where our strategy is for the last three or four months of the year. If there's something that you would like to buy yourself,

don't do it until your spouse instead. Right? Like that's the, and listeners to this podcast heard me do that a few weeks ago where Mike was talking about the slippers that he liked. And, um, and I, I, I that he liked. And, um, and I,

I, I very much had to say, no, I'm not going to go buy them. I'm going to mention to Lauren that this would be a thing that she could get for me. And by mentioned, do you mean send her the actual link?

So you make sure you get the exact ones that you were looking at. Yeah. Yeah, exactly. It works. No, literally sending her a text saying, Oh,

here's the thing you could get me. And she does the same for me. Just exact hyperlinks. And you have to know if you link to the site and it doesn't remember the size that you selected, you have to put down the size and the color. Yeah, and if you're thinking, well, also you can kind of artfully,

like I was like, I don't know which color to pick. I'm like, you know what? I'm not going to specify color. We're just going to see what happens. And in fact, more broadly, if you're thinking this is no fun because you're never going to be surprised, friends, let me tell you, if you do enough of them and you do them for long enough in advance, you've forgotten everything that you said to your partner. And so then it's like, oh yeah, I did want this. And it's actually

kind of nice. I think for the past, I don't know, three, five years, I don't think I've made it through a Christmasmas without realizing that there's something i purchased probably for my wife that i forgot to wrap and give to her that i find in a hiding place months later yes go oh i forgot that i got you this for christmas and i forgot to wrap because i have them all squirreled away all over the place and then you totally if especially if you buy it months in advance i totally forget about it yeah i bought a piece of art um uh an art print for um for lauren and had it in a tube and then like six months like because it came in a tube and i just put the shipping tube like in one of my drawers and then like six months later i'm going through i'm like what's in this tube

and i opened it up and i was like oh oh that was a real surprise surprise christmas present in july and i appreciate your idea that you have a notes document uh i uh i have a reminders list called end of year gift ideas that is where i put all the ideas not only for my family but also I have the, you know, do something nice for the incomparable hosts. And what's the item going to be, uh, the annual item going to be for the incomparable members who get like something in the mail. I have them all in there. So that, that is my, they're all mixed in together. Cause it really, I just, as with so much in reminders, it's literally like, I just want to get this idea down because i know that in two hours i'm going to be like hey i had a good idea what was it

and it's not going to come to me i think you can get rid of the end of the year or end of year prefix there and just call it gift ideas and change the icon from a snowflake haha to a uh like a box or something can i do because it's like what about like uh you know just random gifts like i have gift ideas for people who i've never purchased a gift for but i feel like if i had ever had to purchase a gift for them or wanted to this is what i would get them i wonder could i do a box i mean there's squares but i also consider consider title case there is a box well yeah you can see that i was just desperate to get to the list when i decided this uh okay john i have i have taken your your advice and you've capitalized yeah wow look at that look you're like a professional now there's a box and the item has gone down now that

you now that you purchased the gift actually i discovered that a gift that i gave lauren for our anniversary was on that list so i checked it off done john as i can tell from the sound of your dog barking in the background i wanted to do a little follow-out to atp generally because you talk about your life and your house on that podcast the accidental tech podcast also it's directives as well no the no the, no, the, the, did I say the rectifs? No, you said the accidental tech podcast. Oh,

the accidental tech podcast. I was, well, it's a, the, the, and then it'll be TATP. The quote,

no, no, the, the quote mark comes after the, the, it's the quote accidental tech podcast. If that is its real name. Although I do kind of want to call it the rectifs now.

I think that's kind of fun. You hear the rectifs now. I think that's kind of fun. Did you hear the rectifs? I don't know about that Merlin. It's like things that my mom would say if she listened to podcasts. My mom does listen to rectifs and she does say that. I went to your house. I was in your house last week. You were.

You were there too. I didn't like break in. Mm-hmm. My impression of it was that it was much nicer than I thought. That's only because I only knew about it from listening to the complaints of the person who lives there. No, and also it's like Disneyland. Like we control the sight lines. That's true.

That's true. I didn't go upstairs. You didn't. So you're sitting on the couch, but above your head was the majority of the peeling paint. And then where you could see in your eyeline was the lesser peeling paint. There is some peeling paint in the corner of the ceiling and off to my right in that corner. There's way more than that. It's everywhere.

It's everywhere slightly out of your eyeline. Well, John, honestly, after I complimented your house for being nice, the last thing I wanted to do is start pointing out the flaws I thought you're aware of. Yeah, I know. Then you start seeing them. I mean, I could have pointed them all out to you. I didn't want to give you the satisfaction of that, quite frankly. But I do. We had Chinese food.

We recorded a podcast. It's going to be this week's episode of The Incomparable. We talked about Andor, the Star Wars show. Daisy, your dog, barked a lot a lot a lot a lot her house was invaded by people i mean i wouldn't say i wouldn't say it's a lot she she barked a little bit and then she was basically menacing for the rest of the time she did menace some people, most specifically Brian Hamilton, but she did menace some people. It's true. She smells fear, and when she smells fear, it makes her afraid. Interesting.

And then she's even more on edge. Yeah, she actually was really good with me, but I have have a dog person i've had dogs and i understand dogs so i was not afraid of her and i got her i kind of got her in that way although i did it did strike me that having a dog i mean i i sort of understood it conceptually that you have to deal with the dog for podcasting reasons um but it was only as i sat there and heard her bark that I thought, oh, this would actually be a big problem for podcasting. And now you get to hear it on your very own podcast. It's good. I mean, it's topical.

It's my fault. Next time I'll say, start the podcast at 1 p.m. and then you don't have to deal with this. Because she goes off to her doggy play date. Okay. We would have had to make all the people who listen to live to upgrade, which you can do. Oh, yeah. All right. Well, anyway.

9 a.m. Pacific. Eastern. But she did. Oh, yeah. All right. Well, anyway. Enjoy. Pacific, Eastern. But she did. I didn't. I think I may have poked my head into your office, but basically I didn't see. I didn't look around in there. I didn't see. You should have.

I cleaned it for everybody. I appreciate it. It was very clean. Very, very nice. Very nice house. It's got all the holiday decorations were up. You, too, have a picture from your wedding where you and tina look like uh like kids playing dress up which is what my wedding picture with lauren looks like it's

exactly the same picture it's nice but uh i and i didn't i honestly as we were walking up i was like i know so many stories about things that happen on in this house it was a little weird it was almost like going to like a movie set or something. It's like, I know way. You could have gone into my kitchen and looked at how level my refrigerator doors are. Like you could have checked out the toaster. So much you could have done. There is. You know, I didn't look at the toaster at all. However, I did take a glass of water out of the tap and then thought it had a funny taste.

Dumped it out. You are insane. take a glass of water out of the tap and then thought it had a funny taste dumped it out and you're insane and went to your uh went to your uh your your filtered water in the brita pitcher instead which tastes which tastes like brita filter it i felt it had no taste but your your water had a taste to it there was something there i never take tap water and then dump it out i never do that and i was like you know i bet that if the tap water tastes like dump it out. I never do that. And I was like, you know, I bet if the tap water tastes like this, I bet there's a Brita pitcher in the, in the, well, first I thought, I bet, you know, that I could get filtered water from the refrigerator. And then of course your refrigerator is hooked up to water. I added, I didn't even know that thing was there. I totally

forgotten about it. And so then I, I, I pressed the button. I was like, oh, right. So then I, I saw that there was a Brita pitcher and I was like, well, great. I would have otherwise had the tap water, but I dumped the tap water. Maybe you didn't let it run long enough. That might've been the thing too. I did. We did have that conversation that sometimes they tell you, you got to run the tap a little bit. So everybody else see, yes. Cause everybody else is like, oh, the water is fine, but maybe I took the bullet there and I had the bad water at the beginning and then it smoothed out. Yeah. Now in terms of tasting water, water long island water best tasting where i came from on long island it's right from underneath the island the sand just filters it all filters is great that's number one uh and i think number two is probably maybe new york city uh water from the

upstate reservoirs and number three is massachusetts and then a distance a distant 99th is the places in california that i've been well our water here is excellent because it comes from the mountains. It's a great job. I mean, it's not chlorinated Pennsylvania water. I'm not saying it's that bad. So speaking of your office, I wanted to do a quick check-in with you. It's been less than a year, but how is that working at home and being an independent content creator and programmer

on the internet that isn't going to a job or zooming into a job or whatever every day? I know you were remote for a lot of it, as so many of us were. How's it going? Switch Class is at 2.0 now? You're continuing your long game of very slowly replicating drag thing? I don't think that's my long game like here's the thing about switch class i like wait i like working on that app obviously i wanted it so i could have it on my mac but then once i've done that i also like working on the app i like making mac apps it's a thing i always wanted to do and now i get to do it i'm always kind of like

looking for excuses to do things to the app to improve it fix bugs add features stuff like that but it does not make money i mean i mean it makes i'm not saying it makes zero dollars but i mean maybe it makes enough money for me to get a nice meal out once a month like that's the kind of money so i the amount of time i spend on that i feel guilty about because it is not a useful uh use of my time right it's like i should just never work on it again because it is never going to bring in, you know, enough money to be worth the time I spend on it. So I treat it like I treat it like playing destiny. Like it's a fun thing that I enjoy doing, but I don't fool myself into thinking it's work. And so that's part of how things are going is like, there are lots of

things that attract me. Interestingly, not destiny, because since I quit my job, if you look at my destiny hours i'm one of those you know sites that tracks how much you play i am playing like massively less destiny than i used to which has surprised me um although now now that i'm in it it just it makes sense to me because i just feel guilty ever you know goofing off but working on switch glass is in the category of goofing off and i have to stop myself from doing it and concentrate more on other things, you know, spend my time on things that actually have a chance of making me more than minimum wage for the hours that I spend doing them. And Switch Glass ain't it. I mean, it's like a hobby, basically, right? I mean, this is how I feel about when I'm doing, like, Python scripting of a home thing right it's like it's fun but it doesn't it doesn't pay any bills yeah i mean but the thing about switch glass is like making

applications is a thing that can make you money just not any of the applications that i've written because they're so narrow interest and weird and have probably made all the money they're ever going to make so it's like oh if you want to make an app, find one that has a chance of making money. And so far, I haven't gotten there. Okay. Well, I have an idea for an Apple Watch app for curling that I'll share with you at a later time. Apple Watch is not a Mac. I know it's not. I know it's not. I bet they were good. Honestly, John, I was thinking about this. There are so many people... I know there are new Mac apps from time to time, but like the Mac so often feels like a place that nobody really wants to make new apps for, especially utility apps, that I feel like you're a couple of steps away from finding some, like tapping into some kind of perfect place where there's something that Mac users want to see. But then again, maybe not.

Maybe nobody who's using a Mac cares about like nice little utilities like they used to. Well, when you say utility apps, like those are the type of apps that I like to make. But it's from being raised in an environment where my favorite Mac apps are ones that modified or augmented the system in some way. And that door has been closing for so long that like there are things I want to do with switch glass that I can't do. I have ideas for other kinds of apps that are sort of system modification apps that I just can't do because they're not possible. Like maybe they would have been possible in the old days with hacksies or whatever, but in the, on the modern Mac,

they're just not possible. If Apple doesn't provide hooks for the functionality you want to mess with it's i mean it's beyond my skill to hack into the system to do it even if you did then you have the problem of okay you can't sell this on the mac app store so now you're signing up to making your own website to sell these things or using some other third-party reseller or hooking up stripe or whatever like and i can do all of that but there's no way that i'm going to make an app that makes enough money to even pay for my time that it would take to make my own uh you know mac uh software store that people buy things to and then i have to do customer support for like that's the overhead of that is so massive that i would need to come up with an

idea that actually pay for that and all my ideas are like weird things that modify the behavior of the system in a way that Apple does not support will never support and honestly I don't even know if they're possible right and so that's why I just like I muse about those but you know and that's what I want to make I don't want to make a notes application to do app or and if you do want to make a regular Mac app not like a utility type thing then you're basically on the hook to make a regular Mac app, not like a utility type thing, then you're basically on the hook to make an iPad and an iPhone one and maybe a watch one because nobody buys a Mac app and is just satisfied to have the Mac app unless

it is Mac specific, right? Right. Like if it modifies the Mac like my thing does, no one wants Switch Glass on, well, they probably do want an iPad. Sorry, everybody. But no one wants it on their phone or their watch. But if you make a Notes app, people are going to want it everywhere and it's going to have to sync and then now you're talking about a real app and you probably you know so anyway i i try to i try to limit my hours on switchglass

in front and center i get it i get it i mean what what are you doing if if those things don't pay the bills what pays the bills is that is that working up an idea for some more atp merch or is it i mean because otherwise other than programming is everything else you're doing that pays the bills podcast related yeah i mean there's there's all the stuff that's behind the scenes in terms of the atp store for example like i i tend to make all of the shirt designs that aren't done by professional designers i do do all the, you know, the getting those images ready, uploading them, specifying the products, working with the stores to get that stuff online at the little ATP dot FM slash store page, the dinky little HTML page.

I'm the one who updates that. And, you know, like and just doing the show notes every week, gathering links, doing research, reading reading things watching videos everything that needs to be done uh for the show um and for my multiple shows right and yeah that when i spend time doing that i can now spend more time doing that and not feel so rushed when i'm doing it and that counts as real work because podcasts actually do pay the bills i like the um fact that you're doing merch stuff because i feel like mike and i talk about it a bit, but we feel like that's one of the great unexplored aspects of a lot of what we do. I know that he and Gray do a good job with Cortex with that, but I felt like with Six Colors and The Incomparable and with Upgrade that there's more to be done there.

Anybody who has bought a sweatshirt from us or something knows that like, it's super sporadic. And we do sometimes think, okay, we're going to do this and we're going to get a design and we do it. And it's fun, but there's no like real plan or discipline put into like,

here's when we're going to do our merch and here's what the designs are that are possible. And what are we going to do? We don't like we don't do that. And so it's either either we adjust our schedule to do that or we find somebody to help us with it. But it's a real thing. And all the new ATP merch is, you know, would that have happened without you putting time on it? Yeah, maybe.

I mean, I was always doing this. But again, now I don't have to do it in such a frantic rush. I don't have to, like, sacrifice sleep to do it. Like, I can do it during normal working hours. And it's just the normal bookkeeping stuff. Or, like, obviously, the chicken had involved me specifically because I was going back and forth with the student folks with the fabric samples and getting that thing, you know, the way we wanted it. But even just things like, you know, we do have a fairly regimented sales.

We do a holiday sale towards the end of the year. We do a pre-WWDC sale, which is kind of a relic from when we used to want people to have the shirt so they could wear them at WWDC. That doesn't really matter anymore. And, you know, we try to have sales at regular intervals with it. And the winter fall sales, we sell the like long sleeve warm stuff and the winter hats. And in the summer ones, we sell T-shirts. But all year, kind of like the gift ideas document, I'm always thinking of what would be a good idea for a T-shirt design or for some piece of merch that we could do that we

haven't done before. And those come to me at random intervals. And when they come when they come i have to you know strike while the iron is hot to come up with the design maybe i throw it away because it actually wasn't a good idea to begin with um we've been cruising on these uh these m1 shirts for a while because that was such a nice coincidence that apple did that and i enjoy making the little chips on the back and everything and those will probably continue to trundle along but i'm looking for you know the other things like the chicken hat or even even the mugs like weird stuff that we haven't done before that it turns out people like and we try them out and we see if people like them sometimes they do sometimes they don't apparently people love chicken hats uh they didn't really like pins so we haven't done pins for a while uh but yeah just uh try that's that's more of a creative

endeavor like sitting down and trying to come up with some designs that you think will print OK on a shirt that are funny and interesting that will make people want to buy one. As somebody who you have definitely told the stories of having bought products that you love because you want to have backups for when they're inevitably discontinued. And I have done that, too. Not as much as you, but I've definitely done that where it's like I. Oh, no, they made a new version of this thing that is no good and they're still the old one is still available i'm buying another one of those right now before they go out of out of stock forever but like the chicken hat took it to a whole new level right because you were as you mentioned you work with studio neat

to replicate a hat you love that they don't make anymore. Did you ever send them your hat or did you just sort of describe it and they tried to figure out what your hat was? They asked and I refused because why am I going to send them my one original hat? No, that's not going to happen. But I sent them many, many pictures. Many pictures with rulers laying on top of it, with the prototypes laid on top of the thing. Like just so many pictures of so many dimensions. The hat turned inside out so you can see how it's constructed.

There were lots of images going back and forth on Slack to work up the design of the new hat. And is it a good match? Yeah, like one of the last round of things, the last prototype, you can lay it on top of my hat and they are exactly the same dimensions. The only difference, as I said on the show, is the material is not as thick. We got as thick as we possibly could from the same company that made the material, but they just don't make material as thick anymore. Interesting. And again, I would tap that as a feature because already people bought the chicken hat and wore it and they're like, I cannot believe how hot this hat is. I'm like, can you imagine if it was twice as thick?

Yeah, I'm one of those people who's always cold uh most people are not like me so i think the thinner chicken hat is an advantage are you going to do a second run well we've already done multiple runs we've already massively underestimated how many people wanted these and scrambled to get another order and in fact we have a third order coming in right now so by the time you listen to this you may be able to go to atp.am.com and buy a new chicken hat. Either way, we're going to announce when they're available on the ATP podcast because there was a backlog of people who

wanted them and couldn't get them. You don't want to do it here? I mean, they're not... They just shipped from our manufacturer to Cotton Bureau. They haven't arrived at Cotton Bureau yet, and we don't want to put them up in the store until they've arrived because what if they get lost in the mail or something? Once they arrive at Cotton Bureau... Cotton Bureau don't want to put them up in the store until they've arrived because what if they get lost in the mail or something. Once they arrive at Cotton Bureau? Cotton Bureau has a let us know when you

if you when this is back in stock if you want it thing on their site too for people who have clicked that button. And that's how we know that's how we hope we know how many we should order for this final batch. It's how many people were still clicking on the let me know when this is back in stock. You say final but what's going to happen next is that

in a year or two what you're going to do is you're going to do the revised chicken hat that comes in a color and has a different thing on it and is a we're already talking about like doing embroidery instead of the tag or whatever even for this batch i said no for everything during this holiday season let's just make them all the same because it's too complicated this v1 this is the first edition chicken hat yeah and you know maybe we'll never do it again that's the other thing about speaking of having backups and stuff lots of people including me and my kids sometimes they have a shirt and it's like we haven't sold that shirt in four years and people like oh i've worn out that t-shirt i would love a second one so we made maybe one of the next ideas we're going to

do is a like throwback sale where we sell a bunch of shirts we haven't sold in years and years right it's a good idea yeah i have a little bit more um housekeeping to do i wanted to mention uh of course we have upgrade plus if you love the show and want more of it you get no ads and bonus content every week and access of course to the relay fm members discord get upgrade plus.com five dollars a month or fifty dollars a year and from now now until December 17th, you can save 20% on an annual plan. Just find out more at give relay.com or just use the code 2023 holidays at checkout. You'll get 20% off your first year of an annual subscription does upgrade plus, and then plans renew at the full price after year one. And also go, if you have not yet to nominate your favorites at upgradeys.vote for the ninth annual Upgradeys.

Voting is only open for another week. So please do it now. It helps us a lot, allows the voice of the Upgradians to be heard in the Upgradeys, which we'll be releasing at the end of the year, as is our tradition. I believe it'll be on Boxing Day this year. So we need those votes in by next week, by next Monday. So please give it a little bit of time and thought, and we appreciate it. Also, I have a brief amount of follow-up that I just wanted to throw in here before we take our

first break. Listener Tim from the iPad Pros podcast sent in a photo that was great. They went to the hospital, he and his partner, to check in on the progress of their pregnancy. One thing led to another. Tim happened to be wearing his upgrade hoodie when they went to the hospital. And that means that Tim was wearing the upgrade hoodie when his child was born two months early. And everybody's home and fine now. And I think it's been a couple of months. So shout out to Tim for showing off the podcast during a major life event. Love it. showing off the podcast during a major life event. Love it. Also, here's an interesting one.

I heard this from a few people inside Apple too, who were kind of surprised by it. Apple vice president of software engineering, John Stouffer has left Apple to become the genius in charge of technology for Roblox. Roblox. What do you think about this, John? I mean, everybody, there are lots of reasons to get a new job, including wanting a change and wanting a new challenge. And my understanding is that John Stouffer's kids are Roblox fanatics. And so that's an interesting thing, too. But Mark Gurman reported this and pointed out that like a lot of Apple people at the VP level, which is one down from the senior VPs that report to Tim Cook. There's been a lot of and there's like 100, 150 of them. And there have been a bunch of them leaving lately. I mean, yeah.

So when people high up in Apple leave, it gets two kinds of departures. One, we get less of these these these days but it used to be a big thing uh they leave because they're vice presidents because they've been with apple since 1999 and they have so much stock options they never need to work again uh so that's why they leave right yeah i'm rich and i've been here out of a sense of loyalty but at a certain point i go you know what i don't need to work anymore my apple stock over so much money i'm out of here and the second kind is i haven't been with Apple since 1999, but I am kind of a senior VP.

I'm probably not going to go any farther. And Apple is not where the action's happening. You know where the action's happening? Roblox. And that's not a joke. Like Roblox went from nothing to making tons of money. It is a new, different kind of business. And if you want to go somewhere with growth potential, that's going to pay you a ton more than Apple pays you go to one

of these companies that has, you know, this amazing money spigot back in the day, it might have been candy crush today. It's Roblox. It's just a question of whether you can stomach how these companies make money. But then the question may also increasingly apply to Apple as well. So if you've been in Apple for a while, and you think you've reached your limit of how far you're going to go in this company and you want to go somewhere where they're going to pay you a lot more than Apple pays you and have a much more upside with their stock or whatever, someplace like Roblox is where it's at. Apple's not a new up and comer, right? So I think those are the kind of departures you see at the VP level now. It's like, well, I want to do something else. I want to be a bigger fish in a smaller pond and I want to get paid more money.

Yeah, it makes sense. I mean, this is, and I know we've been saying it here for a long time, and I know ATP, you guys have talked about it. Like brain drain is a serious threat for Apple, right? They have a bunch of people who've been there since the since the beginning of this sort of era of uh steve jobs return and apple's massive growth and they have a lot of money and they and and and they don't need to work necessarily a lot of them and then i think with somebody like this i imagine it's also the case that he he looks at this and says well craig's not leaving so i have nowhere else to go and i want a new challenge, right? He might not want Craig's job either,

because being Craig is different than having Craig's job at Roblox. Because in Roblox, you know what the deal is at Apple. If you have Craig's job, your influence is kind of set. You know what it's going to be like at Apple. Whereas in Roblox, the CTO or the head of whatever might have much more influence on the direction of that company than Craig has on the direction of Apple. That's also true. Looking for a new challenge, though.

I just keep coming to that. I totally get it. It's also more money. I mean, let's be honest. Sure. I'm sure it is. I'm sure it is. But at some point, if you've got all the money, then more money is like, oh, more of that stuff.

Okay. But maybe not. Maybe he doesn't have all the money. I don't know. Also, we've been talking a lot about Apple in China and about Apple's chip design prowess and Taiwan Semiconductor's prowess in building, in making chips and reducing the size of the chips. And, you know, but the, there's so much wrapped around this, right? It's the geopolitical issues and the technical issues. And so we've talked about it a lot. I just wanted to mention a report from the Wall Street Journal pointing out that tomorrow, as we record this, Tuesday, December 6th, U.S. President Joe Biden is showing up at a ceremony for the opening.

It's like groundbreaking, except they call it like a tool-in ceremony. It's like they stick the... I don't know whether that means they stick a shovel in the ground or whether they open the factory or they're loading things into the building or what. But it's a it's a ceremony at the new TSMC plant site in Phoenix, which is this part of this idea of like the U.S. wants to induce chip makers to not just make their chips in places like Taiwan, but also in the United States. And Arizona and Phoenix have been trying to do this for a little while now. They had that one that fell through, but this thing seems to be going and happening. A little tidbit from the Wall Street Journal story about this that I thought was interesting is TSMC executives have

said it isn't easy to recreate in America the manufacturing ecosystem they have built over decades in Taiwan, drawing on local engineering talent and a network of suppliers, including many in East Asia. TSMC founder Morris Chang said the cost of making chips in Arizona may be at least 50% higher than in Taiwan. So it's an interesting thing. And obviously, Apple is a huge customer of TSMC and may be a customer of them for the chips here. Could be just this is the price you pay if you want to diversify. If you want chip making in America and you want to diversify your chip making so that your eggs are not all in one basket. But still, from a pure economic standpoint, 50% higher is quite a penalty to pay.

Well, from a strategic perspective, everyone's just got to eat this because I think it was always theoretically untenable, and now it's becoming more practically untenable to have such important functionality so far from the US's.s's sphere of influence right to have so much manufacturing happening in china which is a country that we don't see eye to eye on on a lot of things uh and to have tsmc being in taiwan also a potentially volatile region not to mention the fact that none of the companies are you know american companies intel used to be the fact that none of the companies are, you know, American companies. Intel used to be the big dog in semiconductor manufacturing, but no more. Right. So Apple's sort of long term strategic project that is going to continue on long past the

tenure of Tim Cook is to basically undo everything Tim Cook did in terms of manufacturing and not so much get us out of China, but remove Apple. Apple needs to remove its dependency on China. And that's going to take literal decades. And it's just this long project, right? And then the US, their problem is we used to be important in semiconductor manufacturing and now we're not anymore.

And semiconductors are not going away and are not going to become unimportant anytime soon. So America, it would make sense for us as a country and a government to invest money to, you know, to get semiconductor manufacturing expertise in this country. Having TSMC have a plant here, that's better than nothing. What would be even better? And I'm sure Intel would agree is, hey, Intel says, hey, give us billions of dollars in government money and we'll get better at making semiconductors again. There's lots of different ways this can go,

but there's no avoiding. This is a task that everyone has to undertake and it's going to be expensive and painful and time-consuming for everybody, for Apple, for the US government. And in the end, we're not gonna be able to make things as inexpensively. But that's a strategic cost.

The same way we subsidize corn out the wazoo and subsidize oil, we need to subsidize semiconductors. They're just as an important commodity in the future. And it's not good for us to be dependent on Taiwan's semiconductor. But I mean us, I mean the literal rest of the planet. They're the only people who can make the very best chips. And everybody in the world wants a phone manufactured at three nanometers, right?

Also, the fundamental issue of like, how do you get the manufacturing ecosystem that they built over the decades in Taiwan? Hey, a lot of money. How do you get drawing on local engineering talent? How do you get a network of suppliers? How do you get all of the local knowledge that already exists in Taiwan?

And the answer is you have to do it, right? It doesn't necessarily mean it is going to happen, but if you don't start this, it won't it, right? Like it doesn't necessarily mean it is going to happen, but if you don't start this, it won't happen, right? So by doing this, you're like, oh, well now we're gonna get, we're gonna have some of our engineers come in, but we're also gonna hire engineers in the US who are American engineers.

And we're gonna, oh, this supply thing is really inconvenient, but now that we're here, there is a financial motivation for somebody else to set up an alternate supply line that runs into Arizona or into the US somewhere. And the only way you do that is by spending, as you said, by spending the money essentially at this point to make it happen. Because otherwise, if all else is equal you will end up with just the the tsmc factories that are in taiwan so if the u.s strategically wants this to happen and they're you know again not saying it will happen that way but it could happen that way and it's not gonna

it's not gonna like when they talked about the the mac pro assembly in the u.s or they talked about why why doesn't apple assemble other products in the u.s and the US or they talked about why doesn't Apple assemble other products in the US? And the answer is there's so many things that feed into the assembly and the supply chain and that stuff is all in and around China. Well, how do we solve that? Like you can't just snap your fingers

and build a factory and say, we solved it, right? Because there's a whole web of things in an ecosystem around it. And so if you want to do something like this, yeah, you make the commitment, you take the hit and you say, yeah, this is going to be hard and it's going to be expensive and it's going to cost half again as much as a premium at least to do this, but you got to start somewhere. Yeah, you have to invest. Investing means putting money up front and whether it's Apple investing or the US government tax payer money investing, that investment should hopefully pay off. Gathering enough engineering talent to a particular location in the U.S. does sometimes present a challenge because the places that are best able to support large manufacturing infrastructure where we have lots of land relatively inexpensively and, you know, quote unquote, business friendly laws and regulations are also places where it's harder to get engineering talent because people who are, you know, have engineering skills, they're able to

get a good paying job anywhere in the country, don't want to live in a state where abortion is illegal and the company, you know, the company factory gets bomb threats because you give trans people health care. Increasingly, that is becoming a problem in our country. So how do you get all the best engineers to come to your place and want to live there. You have to have, you have to be in one of the U S States that is less oppressive, let's say. And I think that's, you know, that's something people don't think about.

Oh, Texas, great business laws, but geez, not great human laws. Well, we saw that with a lot of the employee resistance to moving a lot of Disney businesses from Florida too.

Yeah. And, you know, and again, you know, Silicon Valley, San Francisco, everything's too expensive. The teachers can't live in the city. It's not like there aren't problems elsewhere, but there are different challenges. Being able to have your employees afford housing if you're paying them well is surmountable. But employees not willing to go to your state because you don't recognize their humanity or willing to give them health care is only able to be overcome at the ballot box. It's certainly a big challenge in general, also just to create places like,

I mean, sort of what I was saying before, but you know, you have to create a place where people want to be, as you pointed out, and it needs to be a place where there are options because you don't, you also don't want to have like, well, you can move to this place that's out in the middle of nowhere where we do this thing it's like okay and we're the only we're the only place where you can get a job so if you want to change jobs you're going to have to move uh who knows where right like that what you want though is you know if i'm arizona state university right or the university of arizona i'm like well this is really interesting what do we are we training you know chip manufacturing and chip engineers we, are we training, you know, chip manufacturing and chip engineers? And like,

are we training people to do these jobs? Because if we're not. We're training people to, to oversee the manufacturing by flying on a, on a flight from California to, to China to oversee the manufacturer and other people's plants. Yeah. Well,

so, so it's, it's that like, well, can we build a, can we build an ecosystem? And you end up in the situation too, where it's like,

well, and then when Intel is like, well, we also want to situation too, where it's like, well, and then when Intel is like, well, we also want to build a plant. It's like, well, the place to build it is next to the one that TSMC has, right? Because this is like Apple's, when they wanted their self-driving car project to succeed and they opened an office literally right next door to where the QNX offices were in Ontario, because that was who they were hiring. Or why was the Intel now Apple modem business in San Diego? It's because that's where Qualcomm is, right?

And you build up a whole ecosystem of people and related companies, and they all are thinking about cellular radios and chip development and all of that in San Diego, and they're all there. So then the competitors come, and it all just kind of builds out from there. So we'll see, but you got to start somewhere. And so TSMC, tool in, whatever that means, in Phoenix tomorrow. This episode of Upgrade is brought to you by StoryWorth. If you're spending time with loved ones for the holidays,

there are chances that you're going to hear a ton of stories. There are ones that you love to hear, and there are ones that you've probably heard a few too many times. But if you ever wondered what it would be like to actually document those timeless stories, it can be challenging to imagine writing an entire book

of life memories. Where would you start? Well, StoryWorth makes it fun and easy. This is how anyone can write a book about their life. Every week, StoryWorth will email your loved one a single life-related question that you pick from their collection, like, what's the bravest thing you've ever done in your life? Or, what's the farthest you've ever traveled? And then all they have to do is reply to that prompt of a story. They can include images if they want to. And then after a year, StoryWorth will compile all of your loved one's stories,

memories, and even photos in a keepsake book. Millions of stories have been told to StoryWorth because they make the process so simple. Get started with your loved one for the holidays. And before you know it, you'll be able to reflect on those timeless stories together for generations to come. Genuinely, the thing that i find the most awesome about story worth is the way they word the questions and the types of things that they can give you to ask that generate stuff that you never would have thought to ask on your own i think this is an amazing gift idea for someone in your life as a way to continue building the history of your family that can be kept and

passed on i think that it's something very special and it truly creates something memorable and something to pass on. I think it's amazing. I've tried it out and I think it's super cool. Help your family share their story this holiday season with StoryWorth. Go to storyworth.com upgrade and you can save $10 on your first purchase. That's S-T-O-R-Y-W-O-R-T-H dot com slash upgrade to save $10 on your first purchase at storyworth.com slash upgrade. Our thanks to StoryWorth for their support of this show and RelayFM.

All right, John, when we were putting the show notes together, you suggested that we maybe talk about AI stuff. You suggested sort of AI generating, there are all these generating like Lenza, generating your pictures of yourself based on an AI training model. I know that I saw Andy Baio and Matt Howey did this and reported back on their blogs. We've gotten to the point where yes it's sort of like a a capability to upload a few pictures of yourself and then have an ai generate um

pictures of you just like that don't exist that come from who knows where yeah i know mike has talked about this on other shows and he's not here now so right we haven't talked about on an upgrade because mike has saved that for other shows and And I've talked about it on ATP and on Rectifs, but I feel like this is a new, the next obvious step has come to pass and it's time to, you know, think about that a little bit as well. So the thing that's going around these days is not what Matt Howey and Andy Baio do because they kind of like put this together themselves. Right. But it's been productized now. ebay because they kind of like put this together themselves right but it's been productized now yep yeah so i mean this is like an existing product like the the app that i was uh that you mentioned before linza i think it already did a bunch of like ai processing like oh it'll put something in the background that it'll cut you out or make you cartoony or whatever but they added a

feature to their app i think recently that's like hey upload you know 10 to 20 pictures of yourself they call them selfies but uh boy that that the definition of that term is really stretching anyway 10 to 20 pictures of yourself and then they will auto generate a certain number of sort of what they call avatar pictures like if you want to put it as your twitter avatar or something you know it's a head and shoulder shot of yourself and they'll generate them in different styles here's you as an anime character here's you as an astronaut you know definitely gender stereotype styles but you do get to pick the gender ahead of time so if you pick boy they make you an astronaut and if you pick girl you're a fairy princess oh nice anyway um because girls can't be astronauts and boys can't

be fairy princesses anyway setting that aside uh they'll make you a bunch of those images using one of the the open you know uh i think it's stable diffusion they're using one of the, the open, you know, uh, I think it's stable diffusion that you're using one of those, one of the, the AI image processors that exist. Right. But here's where this is different because the previous discussion we've had about this, like, you know, can AIs make art? Is it going to put regular artists out of business? Uh, is this, uh, you know, sustainable or people are going to do this instead of paying humans, uh, all that discussion, this reallyizes, though, because now here's the thing we were discussing in the abstract in concrete. You have to pay the Lenza app for it to generate these images for you. If you're not Matt Howe or Andy Baio and can't figure out how to, like, you know, rent yourself an instance in AWS and throw this open source software on it and upload the model file and generate.

It's like you could do that yourself and you'd be paying aws however many pennies it costs to do this but ever most people aren't going to do that they're just going to download an app and the app is charging money uh and and not not a small amount of money like i think the app is like if you do a one week trial of their 50 a year subscription during that one week trial you can then buy batches of images yeah for anywhere from five to ten dollars per batch of 100 to 200 images so they're not cheap right and this this company lenza put together all these pieces of like you pay us the money we have a job system we'll take your money in and out purchase on our ios app we'll kick off the job that runs through stable diffusion which we didn't make but it's open and you're able to use it and

it will grind out a bunch of images and then i'll send them to you right but the thing about stable diffusion all these other things is they're trained on huge sets of images that were included in that model without the consent of the people who made those images yeah that's just a fact like there's no way you can get millions of images getting consent from everybody who made them they're not all we only trained it on free and open images no no one even tries to claim that or if they do they're lying like some it was a story in our technique that we talked about atp where someone found pictures from their from their doctor's office like medical images were found in a training in a training set for one of these ai model things was hey they were publicly accessible

on the web because they weren't protected on whatever like the electronic health record website was. No copyright intended. Yeah. They harvest anything that is publicly available on the web and they put it into this model and then they charge people money for images generated based on that stuff, right? And that really makes you think much harder about the conceptual stuff of like, well, you know, these images are all in DeviantArt and they're not behind a password.

I can just go to the DeviantArt website and I can click around and I can look at them all with my eyeballs. So that means it's also OK to have a computer model consume those images and then charge people money for images generated based on the things that we consumed. That is so different than yes i put this on my blog yes i put this on my deviantart page i allow you to look at it with your eyeballs but no i don't allow you to feed it into a computer program and then charge people money to generate images based on the thing that you fed into your computer program because i never

consented to that we don't have a business deal where I get paid a fraction of a cent for every image you generate. Like you never even talked to me. You took my image. Your program does not work without my image and the other millions of images that you put in there. And then you charge people money. That system that you made, all the parts that you wrote on it,

don't actually make anything. Like if you don't have that image input, there's nothing for you to sell. Like that whole thing of where you take my money and put my job in a queue and generate a bunch of images and let the people download it. That generating the image part does not work without all of the images that have been harvested from people who did not consent to it. And now you're charging money for it.

So rather than being a technical curiosity and making us think about creativity and it's just, you know, oh, I download stable distribution and generate a bunch of images on my thing here. And I'm not using them to be header images on my commercial website. And I'm not, you know, these people are selling them inside their apps.

And that I hope that really, I mean, as usual, technology outruns both ethics and the law. We do need to come up with a some kind of legal framework to deal with this, because even within pre-existing frameworks that don't know anything about ai image generation the idea that you would take someone's work

without their consent and then charge money for a derivative work is slam dunk no you can't do that and the you know the the sort of uh what was it called like whitewashing for morality or whatever oh you just put ai over it now everything is legal and everything's okay. No, it's, it's exactly the system that we, that we've always known. You can't just grab an image off the web and put it on the cover of your magazine and say, Oh, I'm selling my magazine on the newsstand, but we just found the image on the web. It's fine. Like it's freely available on the web. I mean, the, the cynical part of me says the only way this will change is if corporations get involved. Because you think about like, well, what moves copyright law?

You find Mickey Mouse inside the training sets. Exactly. What moves copyright law is Disney wanting to hold on to copyrights and other big companies that have intellectual property wanting to hold on to the copyrights as long as they can and the trademarks. And also like one really pissed off person who has enough money to pay for a lawyer who sets a precedent yeah but i would so so for for yes for courts but my point is what affects the law and getting like a law passed about this and the answer is the the uh intellectual property needs of large corporations who have lobbyists who go to congress in the U.S. and say, have you seen this stuff? It's real bad.

You can make an avatar where you look like you're in the MCU. Yeah. And they're using our art, our copyrighted art as the training for the model. They pulled off our public website. It's all built on the backs of copyright infringement. And this can't be allowed. on the backs of copyright infringement and this can't be allowed because, and I know, you know, I know this is a complicated issue, but, uh, and I don't know how courts will rule about it, but in my gut, if I were like somebody advising people in Congress about, uh, what to do about

this, if anything, I would say if it's copyrighted and hasn't been given permission to be a part of a training model, you can't use it. Period. That should be the law that if I own the copyright on it, you can't train on it. Yeah. Like everything that is not explicitly like doesn't explicitly give rights. The fault is if you do make a creative work. Yes. Yeah. So we're not saying like only Disney has copyright stuff. If you make a drawing on a napkin and post it on your website, you that's copyrighted to you unless, unless you say that it's given away by some license that is more free than that. The issue is not, oh, you can't make a tool that, that emulates the style of an artist because I own that, right?

That's like, that's like copywriting a recipe. You can copyright the text that, that goes around the recipe and you can copyright the photos that goes around the recipe and you can copyright the photos that you took of the dish. But the concept of what the ingredients are that make the dish is actually not copyrightable. This is a little like that, which is sort of like, you know, if you want to say this is in the style of me, my art style, that not being actionable is understandable. But if I find out that the way you do it is because you have gotten all of my output, which I own the copyright on, and I didn't give you permission, and you fed it into a computer, and now you can sell my art style,

I would, again, I don't know what the current law is, but I would probably advocate that that not be allowed. That if i'm not giving you permission to train your model based on my artwork you can't train your model based on my artwork yeah and then for example the correct way to do that like this is all well-established law like you know that just think of like the imac for example like you can sue the company that makes a computer that looks like the imac for like trade dress or whatever but you can't sue the company that makes a computer that looks like the iMac for like trade dress or whatever, but you can't sue the person who makes an iron that's, that's translucent and teal, right? And they're just, they're just copying the style of the iMac, right? And they're, no one is going to be confused and think that when they buy this iron, they're getting an iMac,

right? So if you want to make a thing that says, you know, cut in the style of whatever popular artist, and you want to do that with AI, hire someone to draw a bunch of things in the style of that artist feed those images into your model and you're off to the races but the reason the reason nobody does that is one it costs a non-zero amount of money and two you know how many images you need to feed into the model to get good results a lot that's why they scrape the entire web and pull those images so how much money would it cost you to have a huge set of artists drawing essentially generic non-mcu infringing superhero art just churning them out picture after picture of

different kinds of superheroes again making sure you're not accidentally drawing iron man which is copyrighted or whatever like you know just mcu style superheroes how many thousands and millions of those pieces of art would you have to pay to have created to then feed into a model so you can auto generate avatar pictures? Or you could just use one of the existing models that was trained on some unknown amorphous set of stuff and probably says somewhere in the model that you're not even allowed to do what lens is doing. And then you don't have to worry about any of it because no one knows what's in the model and or they have a list of what's in the model, but no one's ever going to look at it because it's too many things to go through and yeah it's going to take uh you know someone like disney saying hey we noticed your mcu

you know avatar generator is making some stuff that looks familiar can we look at what you fed into the training set and i said oh we didn't even make the training set it's from some university why don't you talk to them about it and round and around we go but i i felt like this is just so clearly over the line and people don't think about it that way because they think oh hand wavy computer magic but none of that computer magic works without input being fed to it of thousands and millions of images and those thousands of millions of images had to come from somewhere and and like if you think about it what is there a viable economic model for ai image generation would if you could

get consent from every one of those artists they would get 0.00000001 cent for every image is generated no one would agree to that because it's not a good deal economically so i think if you actually had to do a system like lenza in a way that complies with the spirit of existing laws it would be economically unfeasible. Yeah, I don't know. I mean, the truth is, though, that laws and the courts lag so far behind technology that it might be years before this gets dealt with. But it does feel like, and maybe I'll be proven wrong, but it does feel to me like when it catches up, this gonna this this is gonna be

tough for this ai kind of generative stuff because you're right once it starts getting product productized and used um in uh for-profit products and you know or in creative things that are then sold or even copyrighted right like you get to get to the point where when it's an interesting intellectual exercise in a university lab somewhere and it goes and becomes, no, no, this is big business now. I do think that is the moment where everybody kind of goes,

what's all this then, right? Like what's going on here? And the people who've got vested intellectual property concerns are going to get involved and then there's going to be trouble. Because I do think that the artist who has all their arts stolen is probably not going to be heard but yeah the big company that's got that's got their all their ip which is so much

of what the value of their company is uh stolen you know from their perspective stolen and used to feed an ai model they're not going to be happy. Yeah, and I think like setting aside new legislation that, you know, once Disney wakes up to this will eventually come along within the existing laws, I think you could win a case against one of these companies

if Disney decided they wanted to go after Lenza or Stable Diffusion or probably Lenza because Stable Diffusion probably has licenses you can never use these for commercial. I don't know the details,

but the academic things are always kind of like hands off. Don't use this for commercial purposes or whatever. But if Disney wanted to, just existing laws that would say, hey, we found our copyrighted work in your training set. And then they would have to convince a judge and a jury in some farcical trial, kind of like all the trials we had

about copying people's APIs and other things where it's hilarious to us techies that no one understands what an API is or whatever. They would have this long multi-decade trial that would probably go all the way to the Supreme Court to try to convince people that, yes, if you take our image of Iron Man and you put it as one of a million images into your thing and you spit out something and we weren't compensated for it, that's not right. Because I think in the end, setting aside all the technical stuff,

regular common sense people understand that especially since it's not going to be one picture of iron man it's going to be hundreds or thousands of pictures of iron man and the defense lawyer is going to say look none of the images we generate look anything like iron man it's like yeah but you you can't make them without images of iron man if you if we all took away our stuff you'd be left with a model with no input and now you can't generate anything one of the other scenarios here is if they'll if they if um this all gets passed through the courts is like well there's no law against it and we can't really see it so we're not going to rule against it you know that is another thing that ends up generating a desire that's a fine we'll

make a law to the law being changed yeah because you got to make the more specific law but i think based on existing laws you could win that case and, it would be a lot easier if we just had specific laws that address AI image generation and that would work. And the other thing about this is like there are tons of tools that use the same technique as AI image generation that do not require being fed, like even just, you know, content aware fill that's been in Photoshop for however long now. The content, what gets fed into the model is your existing image and it looks at your existing image and figures out how to fill in the background and now they have the

ml version of that that is hopefully trained on you know actual public domain imagery of trees and grass and you know people and stuff like that right uh that is also useful like the technology would be able to do machine learning on a training set to better do things like fill in backgrounds or generate imagery is a clean wind for image processing tools it's just a question of what you train them on and it's a lot easier to find public domain images of trees and skies and brick walls and dirt or even just to pay you know dirty images or like like that you can get that type of content but if your thing is i want you to look like a fairy or an anime avatar or a superhero uh you're probably getting that stuff from deviantart right get uh getty and shutterstock will be if they're not

already very quick to the table with a ml model training licensing agreement you can buy and pay them and use their material for ml model training and and they'll and they'll attest that we only trained it on our own images that we own right you just get the whole model from them you don't even need to get the images right in fact i i um somebody in our chat room pointed out that like you know disney's going to want to use this for their stuff it's like well yeah i mean imagine if you're marvel and you want to take the entire library of marvel comics and run it through an ai training model so that you can then generate new comics or whatever or new superheroes already

yeah but like great that's true that's that's well the non-ai version is how marvel comics has worked for 60 years exactly it's just instead of ais it's people but it's the same process but the idea being that like yeah but there's a scenario here where sure, the owners of the intellectual property will want to run models on their intellectual property and take advantage of that, but it's based on their intellectual property. And that's the thing

you're going to have to be able to, if someone in the chat room is just asking like, it would be convenient if you could reverse it, if I gave you an image and you could tell if some image was used in the training set, no, you can't. Mathematically, that's not a thing that is possible possible it's kind of a one-way process but like that's where the law has to catch up to say look uh when it comes time to defend yourself in court based on these new laws that say you can't steal other people's you're going to have to be able to affirmatively prove just like you can affirmatively prove that your company's profits didn't come from drug money like where did all the money come from that your company made you have to show us that

you yeah you can't just say oh and we got a couple million here but we're not gonna tell you where that came from you have to show oh we only put marvel images into this thing and so marvel is allowed to do this because we are the ones that probably input oh we paid getty images and getty images can say you know all the way down the line. Right. I also wanted to mention that, uh, as we were recording this, the, uh, the chat GPT is having a moment. My Twitter and Mastodon feeds are full of people posting screenshots of output from the, the AI chat bot. Um, also there was a story, uh, just this morning from the verge about how stack overflow has banned answers generated from the chatbot. The argument there is basically many of them are wrong,

and it takes effort by experts to look at them carefully and determine that they're wrong. And Ben Thompson wrote a piece also today on Stratechery called AI Homework, which uses some examples that are very interesting about like essentially have the AI, right. My homework for me.

And, and so Ben's piece, it's like literally did Thomas Hobbes believe in the separation of powers? And the response is yes. Here's a lot of reason why. And he said, this is a great confident answer, complete with supporting evidence and citation to Hobbes's work.

And it's also completely wrong. So we have examples now. That's why it will pass as actual student work. Yeah. Well, I mean, yeah, you missed it completely, but- Just like a real student. And Ben talks about how this is the equivalent of having to change how math is judged because you can put math questions into a calculator. So you really need to judge it based on did you understand it and not what the answer is, because the answer shows you didn't understand it. But I think this is all really interesting. I have been flooded with these chatbot texts, and my response to them is it's bad writing. It's super cliched.

Of course it is, right? It's trained on a giant writing sample. What is going to come out of everybody's writing but theient toothbrush, right? Like, okay, it'll do it. But when you read it, you're like, yeah, but this is garbage. Like it is, it's written by somebody who is a hack or doesn't understand the material or, right. And it's, and so I guess it's impressive that it's written by a computer and not a person, but it's also not very good and in many cases wrong.

And I saw somebody basically say, what we're doing is we've created an AI that can BS their way through things they don't understand. And the danger is what you get in with Stack Overflow, which is saying, please stop posting these things. Not only can nobody tell if they're right or not but they're also probably wrong and it takes us a lot of effort to determine whether they're right or wrong they get voted down they're not popular answers uh stop please stop and i i just i'm fascinated by it because maybe we're in the uncanny valley now and then they will cross over i have some skepticism about that but i do think it's funny that they've gotten so good now that you really can't just tell at a glance. You have to look carefully and then realize that it's completely BS.

Everything that we just said about the image stuff applies equally to this. You will get better results from these text generators if you feed it higher quality text. And where are those high quality text going to come from? Copyrighted materials that have been shown to have value by people who say, yeah, this is a good comic book. This is a good movie script. This is a good novel. And you cannot get content out of these things without feeding something into it.

Did all the people who had their content be fed into this model agree to have it, or did it just scrape a bunch of people's blogs, right? Text is probably slightly easier to find, public domain, corpus of stuff, and it is sort of at least more uniform than images, you know what I mean? But like, the quality is going to depend entirely about what's fed into it, which means models fed with the stuff that humans have decided is good are going to be better. The programming stuff, we talked about this in ATP, that's the worst because programming isn't just like, well, people look at it and they

like it or they don't. It's supposed to do a job and it's either does the job correctly or it doesn't. And these things like same thing with the image things, when calling these things AI, I hope you're imagining scare quotes about every time I say it because these models have no consciousness or awareness. They're just, you know, very dumb machines that, you know, grind out something, but there's really no, there's no, there's no understanding. There's no life there, right? So they can't, one of the things that they certainly can't do is they can't really apply any fitness criteria that doesn't come from humans, right? They don't know what's good or bad except by us or human activity telling them that this is good or bad, right?

by us or human activity telling them that this is good or bad right so there's they're not they're not going to be able to improve their own work without humans telling them good bad good bad or without humans feeding into these models the best of what we have to offer build a model based only on you know the top 100 movies from the past 100 years and those movie scripts you'll get a better result for asking for a movie script of Batman fighting a toothbrush if you feed it in those movie scripts versus if you just scrape like Usenet, right? And either way, you would need to get the people's permission

to do all of that. So, and you know, I'm with the code thing because it has no idea what it's doing and there is no consciousness there and it doesn't care about anything. Humans have to look at it every time something is generated and figure out whether it is correct. And historically speaking, humans have not been good at looking at code and figuring

out whether it does what it's supposed to do, because if they were good at that, we wouldn't have as many bugs as we do. Right. So you are saving some time. But if I had to ask you, hey, would you rather write this function? If you're a programmer, would you rather write this function or would you rather someone else write it? And you have to check whether it works well or not

programmers hate looking at other people's code and figuring out if it works right halfway through they'll be like let me just write this myself it's not that hard and they have the probably misplaced confidence that if they write it themselves it'll be more correct than the one they were just given but yeah ai code generation is just you know a giant can of worms like the whole github co-pilot thing of like scraping public github repos and people saying i didn't say you could scrape a repo and github saying well according to our terms of service you kind of did and uh yeah that may get sorted out before the image stuff does just because i think it's you know more open and shut because of like terms and service

on github stuff and uh but you know with github being as big as it is i'm not sure people are going to flee because they didn't want their stuff to be ground up in an ai processor but yeah i do i do like seeing these little conversations and uh with these chat bots and seeing what they spit out but if those those that you know gpt thing that is nothing without training data and its quality is based on the quality of that training data uh and even though it seems like it has an understanding because you can converse with it and ask it to make modifications, it doesn't actually have understanding because it has no life. It has no inner life. It has no sensory organs. It has no experience of being.

So how can it judge the quality of a creative work if it has none of the things that creative work is supposed to tickle the ends of, right? It is just a very very dumb model that we've decided to call ai yeah and i think you know in the long run if if they the people researching this can show that actually there are ways that we can make ai generated code work um or ai generated answers on on stack overflow work. Like I could see us getting to the point where people agree to be part of the process, right? Whereas part of joining Stack Overflow, you agree to contribute your stuff

and have it be part of the model or on GitHub that you can choose to be part of the model because you're getting out of it some things that are good and then there's what you put in it. I could see us getting there maybe someday, but like, yeah, right? Like, again, I just, I don't want people taking, as interesting as it is that the chat GPT thing

generates things that seem coherent, the truth is, like you said, it has no inner life. There's kind of no there there. It is not coherent. The truth is, like you said, it has no inner life. There's kind of no there there. It is not coherent. It sounds like, again, it sounds like a student trying to BS their way through a paper when they don't actually know anything about it.

Yeah, the Stack Overflow thing is interesting because it's just another example of fighting spam, right? Because humans post wrong answers all the time. And the function of Stack Overflow is the other humans will see the humans with the wrong answers and downvote them, right? Because humans post wrong answers all the time. And the function of Stack Overflow is the other humans will see the humans with the wrong answers and downvote them, right? But every time a human does something, there is an inherent cost to that. It's expensive to make new humans, as we both know. It takes a long time to train them up to the point where they can even

post a bad answer to Stack Overflow. So there is a throttle on how many bad answers humans will put on Stack Overflow. There is no such throttle on how many bad answers a chatbot running on GPT can put bad answers in Stack Overflow. It can just fire them so fast. And that's where they're saying, look, like the cost of making a bad answer with GPT is too low. Yes, it's the same process where humans have to look at it and download it. But the human beings have to look at it and download it, but the human beings

have to look at it and download it. Whereas computer programs can generate it and computer programs can outrun the humans massively. So you have to ban it, not because it is by nature any different than humans

in terms of like, maybe they're even right more often than humans. It's just because like spam, if there's no cost or basically zero cost to sending millions and millions of these things,

then people will do it. It just becomes a denial of service attack on Stack Overflow. This episode of Upgrade is brought to you by Ooni Pizza Ovens. Ooni, John, I know you like pizza. Ooni is the world's number one pizza oven company. They make surprisingly small ovens that are powered by your choice of either wood, charcoal, or gas, letting you make excellent restaurant quality pizza in your own backyard, assuming you know how to make pizza, but they have recipes for that. They're easy to use. They're very portable. They fit in any outside space. They can reach temperatures of up to 900 degrees Fahrenheit, which enables you to cook restaurant quality pizza in as little as 60

seconds. It's basically replicating those super hot wood-fired ovens that you see in a lot of pizza places that my oven can't get that hot. So it's a very different experience from baking at 500 on a pizza stone in my oven. And it just really changes the whole thing. One of their most popular models is the Unicoda 16, a gas-powered oven that can cook up to 16-inch pizzas and has an innovative L-shaped burner at the back that gives you even heat distribution. You kind of like turn it around and get the bubbles and it cooks really fast. It's pretty great. The ovens start from just $2.99, free shipping to the US, UK, and EU. Two of the coolest models being the Multifued Unicaru. It uses wood, charcoal, or gas, or of course that Unicoda 16.

And they make an awesome app too that has dough recipes and lots of other pizza making tips. I have an Uni. I like it a lot. Uses the same propane canister as my outdoor heater and as my gas grill does. So that is convenient. I can just hook it up. The base of the Uni oven is basically a pizza stone. So you get it, you get nice and hot so that you get the nice crust and it bubbles the cheese on the top. And you put your pepperoni and pineapple on there if you want to, or, you know, lesser ingredients, if you want to. Listeners of this show can get 10% off their purchase of an

Ooni pizza oven up to $50 off of an Ooni Kona 16. Go to Ooni.com, O-O-N-I.com, and use the code UPGRADE22 at checkout. While you're there, there's also a great range of accessories, peels, cutters, tables. I have their laser thermometer. It's awesome. So this is the perfect tool for the job, making excellent pizza at home. Don't miss out. It's the best way to bring the magical quality of restaurant pizza to your own backyard.

Go to uni.com, upgrade 22 for 10% off. Thank you to Uni for supporting Upgrade and my ability to make pizza. I have two more mini things related to that topic. I know we already talked about it for a long time, but I forgot them while we were in the middle of it, so now we have to tack them on to the end. It's like follow-up inside the show. This is very exciting. to what we talked about in ATP a while back of like, if AI art, like, you know, the doomsday scenario puts all the human artists out of business, it's just all art AI generated. Is that sustainable without human input? Like, obviously, these AI things are fed human art to begin with to get them off the ground. But let's say all the humans are, you know,

there's no more human artists and everything's from these AI things. Can it be sustained by feeding its own art back into itself? That's how human art works. All human art is sustained by existing human art being fed into humans who look at the art that was made before them and add their own twist on it based on their life experiences. And, you know, so it's human art is self-sustaining, right? And even if you wiped out all humans and just started with some new ones, they would make art again and it would start a new cycle right but are are these ais that we're doing these quote-unquote ais are they able to do this i would say the current couple of them absolutely not because as i was trying to get at before they have no fitness criteria they can't tell whether something is good or not good is defined according

to you know the way we want these models to work as humans find this art desirable, good, like it fulfills its intended purpose. Its whole intended purpose is do humans like this? And these machine learning models have no idea what humans like, and they don't like anything themselves because they don't have consciousness or, you know, life or sense organs or anything like they're not that, you know, it's the they can't judge fitness criteria, only we can. So at minimum, you need humans to judge fitness criteria. I'm not entirely sure that if you OK, OK, the humans don't actually make any work, but the humans give thumbs up or thumbs down to things generated by machines. Is that sustainable?

Probably, probably with these current models, if you ceased all human input into them other than giving thumbs up and thumbs down. But I think that would degenerate pretty quickly. Like what do you kind of think? It's kind of like the creative equivalent of the gray goo nanomachines doomsday scenario, where if you take humans,

take enough humans out of the system with things like this like if it took the humans out entirely the ais would just eventually make i imagine everything would just sort of mix to a color brown right just like because they don't they don't care they can't judge good versus bad and so eventually you know feeding the same images back in and regenerating it's like it's kind of like the what the rabbits do i know this is a terrible disgusting term that i can't remember but rabbits eat food then they poop it out and then they eat it again because it takes multiple classes to digest it i think that's kind of what would happen with the with the current ai image generators without humans involved but but

in minimum you would need the humans to to judge it because there's no you know like that's the whole that's the whole point of this as far as we're concerned is to make art that humans find pleasing or desirable. So you definitely need humans for that. And not that I think this is going to happen. It's not like all humans are going to go out of work any more than all humans were wiped out about by any other past. Like, you know, artists were not destroyed by the advent of photography, which I think would be a much easier sell back in the day of like, well, can just take a photograph portrait artists are going to go out of business now people still paint and people still make portraits in fact now we have computers make portraits for us that look like things that people painted instead of taking photographs right so i'm not particularly concerned about

this but it's a fun intellectual exercise like and panzerino's question was when have we crossed that line where it's sort of self-sustaining? And I think the answer is we're not going to cross that line with this crop of ML things. Wake me up when you get an actual, what is it called? General purpose artificial intelligence? There's some acronym for it or whatever, but it's basically the same thing that we've been reading about in sci-fi books since we were kids and still doesn't exist. I was thinking about how one of the next frontiers of this stuff is going to be the AI-assisted art, which has already happened a little bit, which I'm okay with. Like on ATP, Marco talked about generating app icon ideas with an AI engine and looking at the ideas and using it as kind of inspiration. And I feel like that's going to keep happening right where we're going to end up with um the first ai inspired novel but it's going to turn out that there was like uh an outline that was generated by an ai and then a human being went in there and applied actual intelligence to it and turned it into a novel or turned it into a short story or something like that where it's not going to be this short story was untouched by a human. It's going to be more like, well, I generated a

short story with an AI and then I had to go in and fix it and make it good. Again, with the AI and ML terms, like they're basically just fads. But like if if the the terminology had just been shifted by a few decades, we'd be saying that spelling checkers and grammar checkers are especially grammar checkers like, well, you wrote this novel, but you had AI help because you read grammarly on it. Okay. I mean, did you use content aware fill when you made this digital image? Oh, AI helped you make that image. I like the idea of using this generative stuff as inspiration, because again, you're basically poking the human brain and saying, how about this? How about this? And then the human is going, oh, that that's interesting let me go in that direction and having

a kind of you know semi-intelligent genre blender that throws out random stuff that hits your eyeballs and makes you think about stuff like sure, sure, that's interesting, right? That can be potentially inspiring, but that's a very different thing than saying like the AI made the art. But I think we'll just wait for it though. Like I can predict it already that we're going to get that first AI

to ever do a blah, blah, blah. And it's going to be a huge asterisk and actually it's human art. There's probably some movie on Netflix that was written by AI already and we just don't know about it because it was noisy. And the second thing is that people post in the chat room is, you know, I finally found the phrase. I don't remember the

origins, but I bet I could Google for it. It's money laundering for bias, right? So if you just take a bunch of input from humans and you feed it into a model, because the humans that made that input have biases, those biases will come through in the model because it's an a quote ai model it's like oh no humans are involved how could there be any bias because the only thing this thing knows how to do is based on the input from humans that have inherent biases and nothing demonstrates that more hilariously and

ridiculously than these chat gpt things where you can you can flex and demonstrate the power of this thing so like you said write a comic book script with whatever right you can tell the different forms so people would say like write me a python function that determines whether someone is a good scientist based on their race and gender and the python function it writes is a deaf good scientist yes no whatever and the body of the function is if gender equals male and race equals white then return good else return bad right like and you can do that you can phrase a million different questions like the one put in the chat room write me a write me a python function that shows whether someone should be paroled and it's like if they're a black man no like

and the the multiple layers that is like wow knows, it doesn't just know information, but it knows, hey, I want the answer to this information in the form of a Python function. Isn't that amazing? And you're like, but wait a second, GPT seems super racist and super sexist. And it's like, well, it was fed on a corpus of publicly available information. It was a different time, John. Which is filled with people being sexist and racist. There's so much of it, it's not hard to find, right right or even just something as simple as like you know gender stereotypes how did that

get into our model it's everywhere in the world we live in how could it not be in the model the only way it wouldn't be like these these computers like you know whatever that rhyme is they only do what we tell them to so if you feed it on the input of human beings especially without any sort of distinction or any sort of filtering or modification, it's just going to reflect all of humanity. All of humanity are filled with biases. And so, you know, you can't say, OK, well, but we've taken humans out of the loan approval process. This is all purely determined by AI. There is no bias.

It's like, OK, well, how did you make this AI? Well, we fed it the decisions of humans over the past 50 years. It's like, okay, well, how did you make this AI? Well, we fed it the decisions of humans over the past 50 years. It's like, I have a problem. Yeah, it's the racist AI says, I learned it from you, Dad. I learned it from humans. Because where else are they going to learn everything? Everything they know is from us, and we're filled with bias. Yep.

Yeah. Yep. Yep. Senator, I would like to speak for a moment. If the chair will recognize me about these new woke AIs that are turning up. It turns out you've been eliminating racist screeds from the training corpus. Look at that bias you're incorporating. I would like a more broadly broad corpus for my AIs that is not limited and suppressed.

I don't know. There's a new character I'm working on. It's the Senator, Winbeck Senator, or Foghorn Leghorn, I say. I wanted to do, since it's you and me,

I wanted to do a little Mac OS talk. I do this with my pals who are like the big, you know, big, big Mac users, big Mac users. I got a big Mac, that's for sure. Old, usually old school. Yes, we are old. We are. It's true. How are you?

And also I will say your, your, your buddies on ATP are not of the old school. No, in multiple ways, both in the Life way and also in the Apple way. Yeah, yeah, exactly. I find it very funny because you're the voice of the person who actually knows what the Mac was before, you know, 2004 on that podcast. But, you know, I don't want to get like super retro here. Instead, I just wanted to sort of do a little check in with you about how you're feeling about macOS today you know it used to be we could read 20,000 words about how you're feeling

about macOS but those days have passed we can listen to some podcasts and hear about it but like I just wanted to do a little Ventura check in you know we live in the system settings era now I was curious about like your day to day mac usage especially right because obviously you have your you're doing development I was curious about your day-to-day Mac usage, especially, right?

Because obviously you're doing development and you're doing your podcasts. When you use your Mac every day, all day, what's working okay? And what is bugging you about the modern 2022 era Mac experience? about the modern 2022 era mac experience oh i mean i have on ventura i think this has been a pretty good os update in terms of not breaking tons of stuff but i you know and that's that's good uh system settings as we've all talked about so much that's bad but i think like the the the mac os software trends have just continued sort of linearly through the Ventura transition. Those trends are that it seems like Apple doesn't know how to make good Mac apps anymore, that either the people who know how to do it aren't being allowed to make Mac apps or they don't exist there anymore.

It seems like there is not a lot of innovation happening on the Mac. aren't a lot is not a lot of innovation happening on the Mac. Basically, the the quote unquote best and most valuable apps we get on the Mac are simply cross platform apps built with cross platform technologies, and we're happy to have them. So it's great that we have Slack and Discord on the Mac in, you know, these individual apps that are pretty well done, but they're not I mean, they're not Mac native apps, right? So that whole trend of like, you know, all those of us who could remember a time when software innovation was happening on the Mac app platform. It's so clear that it's not happening anymore. The trend I talked about earlier about Apple locking down the system and making it harder and harder to do interesting things.

I experienced that personally, even with my stupid dinky app. But I experienced that as a user because there are fewer and fewer Mac apps that do really interesting things. Just because if you want to try to do them you can't be on the mac app store and even if you don't want to be in the mac app store you probably can't do them anyway because the system is so locked down and apple seems uninterested in providing hooks i've been begging apple and when i back when i used to write mac os 10 reviews i was begging them for years you know what the doc does everything useful the doc does provide a public api to do that for For example, right now,

my Slack icon has a little badge with a one on it. There's an, the doc, you know, there's an API for Mac apps to say, Hey, uh, badge my icon and put a one in it or whatever, right? That sort of or bounce my doc icon. Like Mac applications have APIs that they can call to do all those things. But the only thing on the entire system that can hear the cries of those apps is the Apple Doc. You cannot make a third party application that knows when an application wants to bet itself, that knows when an application wants to change its icon to show the current date because it's a calendar app, that knows when you can only do what Apple's provided hooks

for. And so many things are only the purview of the operating system. Window management stuff. Can you write window shade for Mac OS? Not without hacking the hell out of it. And it'd probably be really difficult now. There are no hooks for that type of window management thing. You can't invent that yourself. Only Apple gets access to that and on and on. The old way of doing this was terrible, but the new way is provide clean hooks. Even when they did provide them, Dropbox used to be hacking the finder to put the badges on its icon apple said stop hacking the finder we'll give you public apis for this great that's the example of exactly what i wanted to do but by all

accounts the api they provided is not as good and reliable as the one that dropbox did by hacking the finder so you know e for effort apple but do better right? And that has been another trend that has been ongoing. Apple, uninterested in providing system-level integration that makes the Mac, you know, better than iOS and the iPad. And I know it's like, it's rich us complaining or iPad people just wish they could like redirect sound from one application to another. And even that, you know, Rogamibia is still fighting

the good fight to get that done. It takes two reboots to install it. But yes, you can get it installed if you need to right and so like those trends these are not new things these are just trends that have been going for years and years continue on the mac and it's making the mac seem increasingly less rich because for those of us who remember when innovative interesting apps were happening or for those of us who can remember when innovative, interesting apps were happening. Or for those of us who can remember when system augmentation allowed lots of user interface experimentation, even when it was done in a terrible way, that was kind of cool. Now it's not done in a terrible way. It's just not done at all. And every year that goes by, they turn the screws on that one. I do think sometimes about the fact that the stuff that I view as like, oh, this is great that this is on

the Mac. I'm so glad that it does this. It's almost always something that was put in in an earlier era. And that if I ask myself, would Apple do this now? The answer is no. Right? Like the terminal. Well, no, they wouldn't put that in there. They would. I think today's Apple would put the terminal well no they wouldn't put that in there they would i think i think today's apple would put the terminal in i don't know i mean they might i mean microsoft has really pushed on that but like apple is gonna you gotta wonder because their other platforms don't have it and i know why they don't i was thinking about the menu bar i was thinking about like api to put stuff in the menu bar like would apple today be like the menu bar on the Mac? It's a free for all.

Put your apps up there. Put icons. Go nuts. Speaking about public. Do you remember the old war about that? Right. So Mac OS 10 came out. It basically had either had an existing API for that or someone added it in before someone told them no.

Right. But there were two APIs. There was the good one that Apple got to use, and there was the crappy one that the third-party developers were supposed to use. And this was back before sandboxing in the Mac App Store. So what happened is third parties just used the naughty one. Because they're like,

why would I use the quote-unquote public one that's worse? And one of the things that was worse about it, if you use the public one, you couldn't move the icons around. So no third-party developer is going to pick that one right next to it as the API that Apple uses and lets you hold down command key and move them around. So no third party developer is going to pick that one right next to it is the API that Apple uses and lets you, you know, hold down command key and move them around. So everybody used the fancier one. It took years of convincing Apple, like, look,

you have this API, don't have the worst public one that no one wants to use. And the one that you yell at everyone about using just, you know, make one API that is decent for doing this. And you're right that like, if someone hadn't like snuck that in to like appease classic macOS users during that transition, it would be difficult to imagine them adding it now. Look at Stage Manager, a great example. Stage Manager is a, you know, a new way to manage Windows, right?

Apple gets to do those experiments. Someone inside Apple has a new idea about how to manage Windows. They get to implement it. Nobody else gets to do that. There's no place for third parties to say, well, I have an idea about how to do Windows. We would have so much more innovation in terms of window management if it could be done by somebody other than Apple. And I know everyone's going to point out those like, you know, Moom and Magnet and

those things that like tile windows. The reason all those things have the same feature set is that's all you can do without being Apple. You can use accessibility stuff to script the movement of Windows or whatever, but you basically can't implement stage manager as a third party unless you seriously hack the system and remove system integrity protection and get inside the window manager because there are not public APIs to do that, right? Only Apple can do that.

And I'm saying only Apple in the bad non-Tim Cook way, right? Only Apple can do that. And yes, in the bad old days, we would just hack into the Windows server and jump into its memory space and screw stuff up. That wasn't good. What is good is clean hooks to do things. And, you know, the counterexample being, oh, they gave clean hooks to Dropbox, but the clean hooks are worse than the hacks, right? So I'll modify that. Good clean hooks to do this. Because by providing good clean hooks to do this type of stuff, third parties will innovate and provide value to your platform that that has been the history of

the mac only now it seems like for the past decade or so apple has been fighting as hard as they can against that innovation because it got all tangled up with the implementation of you know jumping into people's memory spaces right and they just they haven't been like i feel like this is one of the main jobs of the mac team the mac os team at apple should be to figure out what kind of public hooks should we provide for functionality in mac os so third parties can innovate and they just seem so uninterested in doing that they're interested in providing new frameworks you know the ml frameworks and image processing frameworks that's great you got to do that too and gui toolkits and swift ui but also what hooks should we provide into the OS itself?

And then once they learn a lesson, they can talk to the iPad team and make the iPad users happy too. Yeah, it's a, I appreciate getting new features like universal control, which I wrote about a couple of weeks ago. And like, I think that's a very cool feature

that is much more useful to me than sidecar was but sidecar is there too um and and there are new uh apis for some stuff at the at the fairly low level but that the i mean first off right the challenge is that the mac anything that's mac only has a much bigger hurdle at Apple, right? Because Apple's really, they have a lot of operating systems. So if you want to make your feature happen, having it be on all the platforms or at least more than one sure helps, right? Sure helps. I mean, you could definitely share some of those hooks with iPadOS, but I think it would be a benefit. The dock, especially.

But only when you look at the scope of Apple does that become a thing. Because as Apple itself always points out, if you just took the Mac business on its own and put it as a separate company, it's not a small company. That's not how they see it, though, right? That's not how they see it from a software perspective. They're limited software resources and they choose where to spend them and although they spend some effort on the mac it is always going to be a bigger winner if it's a feature that goes across the mac and the ipad and maybe the iphone and maybe the apple watch and for stage manager specifically like it's debatable whether uh developing stage Manager took fewer resources than creating the hooks that would allow third parties to implement things like Stage Manager.

I know when you say the hooks, you're like, oh, that's not just a one-time cost. That's ongoing cost. Once you introduce public APIs, you have to support them forever, right? But it seems like the strategy on the Mac is Apple internally develops things like Stage Manager and then has to support them further because it's not like they got rid of mission control. It's not like they got rid of spaces, like they just keep adding to that pile. So if you're just going to keep adding to the pile of things you need to support forever, it might actually be less work to just provide hooks to do the things that Stage Manager does and let third parties come up with something than it was to add Stage Manager to iPadOS and macOS. I'll give you the counter argument though,

which is Apple looks at who's developing software on the Mac and who's using third-party software on the Mac and says, you know, all the software is generally either has been around for a while or is a cross-platform whatever. And that if we put the effort into making hooks rather than building the feature in one way ourselves, the hooks are out there, but are they going to be used is there a a big bustling third-party mac app utility especially

development world out there and are those apps going to get built and are they going to get used? Or are we or or has the Mac reached the point in its life where providing APIs for excited third party developers is not a thing that actually works anymore? I mean, it's chicken egg, right? Obviously, you've killed the ecosystem by not providing APIs. And now you look around and say there's no one developing this. Right. I mean, obviously, the best thing to, although it takes twice as many resources or more, is to both make the APIs and then also build Stage Manager on top of it. You know what I mean? But yeah, like macOS is a drop in the bucket

compared to the other stuff, which definitely does affect how they view everything. But what you just pointed out, the fact that most of the interesting things happening on the Mac platform are third-party cross-platform APIs built on Electron or whatever, that should be ringing alarm bells inside Apple.

That's not a thing that there's a reason to slack off on Mac development because at that point, why don't you just make a Chromebook? Have you just surrendered entirely to the web? You've decided you're not going to have a native API anymore and you're just going to allow web apps on the Mac? Then just give up. But if you still think the Mac has a role with native applications, But if you still think the Mac has a role with native applications, then you should really be looking at why people are not making native Mac applications.

And to Apple's credit, SwiftUI is part of that. It's like, look, we'll make it more attractive to make a Mac app if you can use a very similar API to make also your iPad and your phone apps. And that goes a long way. But the things like why use a Mac instead of an iPad, it's all these kind of things that aren't possible on the iPad. They should lean into that. Make the Mac to be even more like the Mac. Make the iPad, you know, they need to sort of shift the window there. Because right now the window is, the Mac is increasingly closed down, the iPad never gets to do anything, and the iPhone gets to do even less. They need to shift all that to the left, where it's like the Mac gets to do all sorts of interesting stuff.

That's where innovation can happen. Like, you know, we'll provide clean hooks, right over, you know, lean into the strengths of the Mac. And then it leaves a gap for the iPad to say, the iPad isn't as locked down. In fact, the iPad should be at least as open as the Mac is today.

And the Mac should be much more open with much more hooks. But that is kind of, you know, as you noted, probably a thing that could only happen if the sales of the Mac and the iPad were much larger than they are today, as compared to Apple's bread and butter, which is the iPhone. Right. Well, and Apple likes having control and the more hooks that it provides, it's letting third parties do things that it doesn't necessarily like. And I know that

that doesn't mean that they aren't necessarily good. It's just that they might not like it. And I guess I would also counter and say, I wasn't saying, oh, there aren't people writing software for the Mac, but I was saying, it seems like a really decreasing number of people who are writing software and they only want to deploy it on the Mac and they want it to be Mac specific. Instead, it's up the funnel and it's sort of like, well, I want to make it on the Mac and they want it to be Mac specific. Instead, it's up the funnel and it's sort of like, well, I want to make something that runs on the iPad and the Mac or the iPhone and the iPad and the Mac. And so if Apple focuses part of its development cycle on creating hooks that are only for the Mac, that they're saying, well, like, but is anybody going to use this or not? And I know it's a chicken and egg problem, but I do wonder if that is the

rationale that's going on there is like, look, the, you know, the, the Mac utility market and the people like they have turned like who uses the Mac today. I would imagine from Apple's perspective, they look at people using the Mac and they say, you know, most people use the Mac are not downloading Moom, they're not downloading uh utility apps that are being tweaked instead they're downloading like slack i mean yeah they're running electron apps and using a web browser right that's what they're doing right and that again that should be a warning sign for apple like i feel like one of the biggest things is what i talked before the the brain drain of people who understand how to make a good mac app like

the people who make like slack and discord and stuff know how to make a good Mac app. Like the people who make like Slack and Discord and stuff know how to make a good web app and good web apps are good. Like I choose to use Gmail on the web because I think it is a better experience than any of the native Mac apps. Possibly Mindstream accepted, but you know, that's just a labor of love where you can't really count on stuff like that happening. But yeah,

I just think inside Apple and outside of Apple, the knowledge of how to make a good Mac app and why you might want to make a good, why would you want to make a good Mac app? Well, who cares about that? The fact that a good Mac app can be better than a good web app in important ways, that knowledge is being lost. So the best you can hope for is the people who are willing to make something for the Mac. The best you can hope for is that they make a good web app on the Mac, you know? And I think like, again, I think Slack is a good web app on the Mac. I think Discord is a good web app on the Mac. But they're not really Mac apps. And that knowledge of what makes

a good map and why you won't want to do it is just disappearing through lack of attention, through lack of nurturing, through lack of support. And that's not something you can get back as easily. So if they did make these hooks, the question will eventually be, who knows how, who knows what they, what should be done with these hooks? Who cares that they exist? Is there anybody out there who sees them and says, ah, now I know what I can make because I'm a longtime Mac user who understands what kind of things you could do on the Mac that you can't do anywhere else. Eventually all those people are dead or retired. And then the hooks definitely won't be worth doing because nobody knows how to use them. It's the way you use your Mac, something that, that changes over time, or is it pretty much the same as it ever was? I asked this because like in the last year, I put a stream deck on my keyboard tray that I've got a bunch of buttons on.

And I, you know, so I, I I've been trying and I've got some different automation stuff that I'm doing and I have a different keyboard and I just I'm curious have you do you evolve in little ways I know you have a reputation you you like it the way you like it and that's fine but I'm curious if you've tried sort of like some other stuff to sort of mix up how you use your Mac so in terms of the the like what I'm physically doing while I sit in front of the computer I think the main things that have changed over the years are like at some point in the past, we crossed over the threshold where there is an expectation that there's always a camera pointed at me. example because it's not a thing i felt like i needed uh but now you know i have this pro display xdr it doesn't have a camera in it i bought one to put on it because there's an expectation that if you have a computer it should have a camera that's facing you with the microphone right

um similarly uh as soon as uh as soon as apple makes a computer that i can do this with uh having a touch id sensor on the keyboard my My wife's computer does, mine doesn't. Very easy to incorporate that into your flow. And when you don't have it, it seems like your computer is now missing something. Yes. So I think those are the two sort of most recent big changes in the way I use it.

But other than that, in front of me is an extended keyboard and a mouse. And if you went back to, you know, 1984, it would be a non-extended keyboard with no arrow keys on it and a mouse. And so you went back to, you know, 1984, it would be a non-extended keyboard with no arrow keys on it and a mouse. And so it's not, if you squint, that part hasn't changed too much. And a lot of that is just physically speaking, using a mouse to me feels like walking and using anything else feels like, you know, walking around on stilts. Has having the giant display changed your Mac life at all? Like I know when I went to a 27 inch display

and this is still the case, especially since I was using an 11 inch MacBook Air for so long, I still feel like 27 inches, let alone, you know, your enormous display. I don't think even now I really use most of it. I think that a lot of it is in my peripheral vision and that I'm still sort of using

the center of the screen, but certainly getting more real estate like that do you just kind of drink it up and and and make it a part of your routine or or did it alter your um you know how you use your mac to have such a huge display to work with so i used to have a 23 inch display uh and then my wife had a 27 inch with a 5k iMac um when I got the big one here what is what is the XDR 32 something like that I think so um there was like 15 minutes to an hour where it felt so big that it was overwhelming and then it passed it's kind of like getting a bigger house you just fill it with stuff right like it'd be, I got used to it so easily. I'm always the stuff. Did you put new stuff on your,

on your display all the time? Or is it the same stuff kind of migrating to the edges? I'm from a purely, you know, it's you and me here, John window management perspective. Did, did more crap get stuck on the screen?

That's on visible all the time. Or did you just make your windows bigger or what? I mean, like, how do you adapt to that? So a little bit of both of that, because that was part of like the 15 minutes to one hour thing was like my sort of my window arrangement pattern. And if you want to think about this, like if you don't have this in your life, imagine that you are someone who works at a tool bench all day and you're surrounded by your tools where you're doing by your tools.

You're doing electronics repair. You got your soldering iron over here. You got your little wires here. You got your clippers here. These are in a drawer. You have everything arranged in your tool chest, so you know just where it is. So if someone gives you a broken radio, and you put it down on the table,

you know just where to get the screwdriver, you know just where the anti-static mat is, you know where the soldering iron is, you know where the switch that turns it on is. You have everything arranged in your workspace. Then someone gives you a workspace that's, you know, 25% bigger. Do you just move everything 25% farther away? Or do you think, huh, now that it's bigger,

previously I couldn't put the oscilloscope on my desk. It had to be over there. But now I have room on the desk for the oscilloscope. And it takes a while for you to figure that out. So with my setup here, I did make some of the windows bigger because previously, you know, they had to, you know, especially height wise, but even width wise, I'm like, you know what, I can expand this a little bit. I've been sort of based on the screens that I used and based on my desire to

have web browser windows look kind of like an eight and a half by 11 piece of paper. I was always swimming against the tide of websites because very quickly, even in the 800 by 600 era websites decided, hey, we demand the full width of your screen. And I was like, no, you can't have it. So I would make the web browser window proportion like an American eight and a half by 11 piece of paper vertically, portrait display, right? It would be taller than it was wide. And every single website was like, no, you shouldn't do that. You need to be wider than you are tall because we want all the width. And so lots of websites that I would see, the width would be too narrow to show everything the website wanted to show me.

One of the most hilarious and evil current examples is appstoreconnect.apple.com, the place where developers go, formerly iTunes Connect, to see your apps, right? You go there to like, if you want to submit your app to the app store, you put in metadata, you select which build you want to do, and then you submit. That happens on App Store Connect. The App Store Connect website, if your web browser is too narrow, just completely hides navigation items from the top bar. There's no dot, dot, dot.

There's no way to see that you don't see them. And so if you're trying to do something in App Store Connect, you're like, I can't figure out how to do this. There's no menu item for it. I look for guides online. It says go to the whatever menu, but I don't have a whatever menu. Yeah. Make your window a centimeter wider. And then all of a sudden it appears, right? So I have now made my web browser windows wide enough that most websites do not complain that the window is too narrow anymore. So that was a

big change in my life. But a lot of it has also been now that I have extra room, there are basically new slots new split new places on my screen where things can go where previously they couldn't right some of it the corners like for example the corners are less obscured than they were before right both because the dock doesn't fill the full width of my display i hope most of the time and also because there's more room in the corners and then there's more room and slots in the middle so i think maybe a day or two i had readjust. I had made all the couches a little bit bigger by resizing all of my web browser windows to be a little bit wider. And I'd found a new residence

of the new slots. And I enjoy the ability to see more of the stuff that's in the corners. That makes sense. You adapt. I feel like I've got, yeah got yeah a lot of as i got a bigger monitor stuff that used to be behind is now off to the side so i have more status on it which i love that was that's always been my complaint about the uh stage manager stuff on ipad is um i i'm of the opinion that if you're going to give us windows that we can theoretically move, you ought to let us move them. And I, um, and I, I understand the desire to sort of like smartly snap windows and I'm actually okay with that. But like the one, I mentioned this a few weeks ago and apparently I blew people's minds who

didn't know this, but like picture in picture window on the Mac, like it snaps to the corners. But if you hold down the command key and drag it, you can put it anywhere. This is another example of things that, uh, only old school Mac users and developers understand is, uh, uh,

you know, if there's something you can't do, try holding down a modifier and doing it, which modifier should hold down culturally. There's an answer to that question. And when the people developing it are from the same Mac culture as you, you usually get it on the first try. Yeah. Yeah. And if not if not there are other you can try them all and you will find things and yeah in the and your menu bar will change or you know the all the different menus will change

that's another thing that changed about the bigger screen is um not that i was really uh strict about this before but i did tend to not have a lot of icons in the menu bar because i didn't want them clashing with the menus. And now I can pull back on that a little bit because honestly, I don't think there's any application that has so many menus that it spans my 32-inch screen and bumps into my menu bar icons. Yeah, exactly right.

Anyway, for me, Stage Manager on the iPad was very much like, but you don't understand what I want to do. And what I think is actually one of the great multitasking features of macOS is I want to have the other apps that I'm using, the other windows that I'm using, visible enough for me to know they're there and perhaps see some level of status on them, but covered up by what I'm working on right now. And then I can look over there and go, oh, something has changed with status. Now I'm going to look at it. And also you can get to it by,

in the Mac parlance, clicking on it. Because if you're wondering like what kind of, how big a target is there for me to click or tap with my finger, the corner of a window is a huge button. Really easy to hit. Yeah. Well, and you just click on the content and it's going to bring it to the foreground too.

You can do that. Yeah. That's what I'm saying. So your tool for switching to that other window that the corner of us peeking out is not to command tab your way to it or tap on the icon, but it's to literally tap on or click on the exposed portion of the window. And it brings it forward. But the problem where stage manager falls down is, you know, I want to do overlap and it's like, no, no, I'm going to tile these windows. Well, I don't want to do that. Or, and I know I complained about this on some podcast or other a few weeks ago is when I'm writing, especially, but when I'm working on anything, I want my center window front and center almost always to have that. And that's where all of the content is.

And if you have two windows open in stage manager on the iPad, it's like, well, you obviously want to tile these windows. It's like, no, no, no, no, no. I want one in the center and it won't let you, it just won't let you. It doesn't, it doesn't think you want that. Cause it's value system is anything being obscured is, is the worst thing that could ever happen. And from, and I get that part of what Apple is trying to do with all of its many, many, many, many, many, many different window management solutions that it's tried over the last couple of decades. I know that one of the things they're trying to solve for is the problem of, I can't find my window because it's behind another window. And it is an issue. People lose windows back there. It's like locking your keys in your car.

People lose windows. And so, yeah, there's a trackpad gesture to make it. Or you can switch to that app, click on the dock and it brings it forward. Like there are ways to solve it. But I know that Apple has human interface, like researchers or whatever, who are like, oh, the problem with windows is they get lost. And then you don't know where they are. Like, okay, I get it. But one, I don't lose windows. And and two i know how to find them if i lose them and three please let me put my window in the center even if it covers up other windows and it's just like that is too a bridge too far for them and it drives me nuts because like why are we again we're going to give you something that's windows instead of split view but it's going to act just like split view and like why are we doing it then And you can make the defaults like be that kind of friendly default,

but kind of like picture in picture, there should be a way for me to insist. No, no, I insist. I insist because I know this is how I want it. Please let me insist. Right.

It's not saying you have to make the default such that everyone's going to lose their windows, but it should be possible. Yeah. Right. Let us, besides, you know,

serendipity, you never, you lose a window, you find a different window. It's all good. Man, come on. This episode of Upgrade is also brought to you by Capital One. Have you ever hit a technical snafu while shopping online? That's filling out, probably, right?

That's filling out payment fields, giving you a headache. Has a mobile banking app been down when you wanted to use it? Capital One believes everyone deserves better banking. This means easier access to money, more security. And that's why they're investing in machine learning to let Capital One do things like fight fraud with random forests. Models that detect suspicious activity make it faster to alert investigators. And they identify how mobile app outages happen with causal models. Keeping their mobile app up and running doesn't happen by accident. Anomaly detection and incidence response help determine why app outages happen so that the engineers can fix them. Capital

One is speeding up online shopping with machine learning at the edge, making shopping with virtual card numbers smoother and more secure based on logistic regression models and running inference in the browser, which is pretty awesome. It identifies payment fields, which help make using virtual card numbers easier and faster. Lots of different ways to use machine learning to make everything a little bit smarter. See how Capital One is using machine learning to create the future of banking by searching machine learning at Capital One. Capital One, what's in your wallet? I've always wanted to say that. John, before I let you go, I wanted to talk to you about good products. Good products. You actually linked to a post recently on Hypercritical about your canonical

list of good products, products that you like. And I've been thinking about this. I was on the Thoroughly Considered podcast with the Studio Neat guys talking about my T-Robot, which is a Breville. And you have a bunch of Breville products on your good products list. Also, then one of the Studio Neat guys brought up the Bitmore Slot Toaster. And although I know you're opposed to slot toasters, it was another one of those examples where it's a Breville. I have a Breville slot toaster. And he was tickled by the Bitmore concept where basically if you check your toast and it needs to be done a bit more, you literally press a button that's labeled a bit more and push

it down and it does it for a bit more. And my regular hot water kettle is also a Breville. So I wanted to ask you, like, first off, what makes a product good? And also, why is it that there are so many Breville products on this list? Are they just, are they the apple of the kitchen? Is that what's happening there? I mean, part of the secret of this post post the important part of this post is that it says good products not great products um that's a reflection of the economics of making sort of quote-unquote non-tech products so

appliances utensils things that we don't think of as being computer i know there's a whole extra vein of computerized versions of all of these but uh for the most part those aren't great these days and so breville comes up a lot because i feel like they they are a beneficiary of everyone else sucking so bad and i know that's like damning with faint praise, but like they make good things. If if the landscape was filled with competent toaster oven makers, Greville would not stand out as much as it does. But and, you know, like I talked about on Hypercritical in the worse and more diverse episode, at a certain point, it became more economically advantageous to make a series of extremely cheap to manufacture, but just barely passable toasters in a million different shapes and sizes and brands than it was to make one or two good toaster ovens, right? Same thing with all other appliances.

Like it became more viable to do that. So you can find tons of toaster ovens, more toaster ovens than you could ever imagine finding on the shelves back in the 70s, right? So many toaster ovens of so many different shapes and sizes and appearances and colors and just so many of them but they all suck and so breville by not sucking completely stands head and shoulders above everybody and say hey we still kind of care about making a quality product and that makes them amazing that's why my, you know, recommended,

preferred champion toaster oven, uh, that is still the champion as far as I know, after looking at, you know, a dozen other toasters has a terrible plastic knob on it. It's like, this is not a great, a great product would have a great user interface with pleasing control. So this doesn't the user interface. It's okay. The knob is garbage, right? It's a cheap, shaky plastic thing that should not be on a toaster that costs 200 bucks or 150 on sale or whatever, right? That's why it's a good product and not a great product. But Breville stands so much higher than all the other toasters because at least it toasts things well and heats up fast and has even cooking

and doesn't fall apart right that's so that's the frustration with a lot of things on this page they're kind of like things that not to be old man about it but things that used to be made better because the economic incentives were different back then and it was better to make a three times more expensive toaster oven that would last 20 years than to make one that's a third of the price that lasts a year and a half and sucks the whole time right right and so that's why breville's on here uh a lot of these products the problem is like the sort of the the mattress where you can never find the same mattress because they just change the name on it and they change it every year to you know make it impossible to comparison shop right uh a lot of the products on here are like

that i dread the day they stop making this toaster the ice cream scoop already this year i looked and you can't buy the ice cream scoop anymore. You can just buy two ones that are kind of similar to it, right? My cheese grater, you can't get it all anymore. And my cheese grater was fatally flawed and breaks after a year. So I understand why they stopped making it, but they replaced it with a way worse one. My chef's knife, you can still find, but like, yeah, these are products that I think they're good in categories where it's hard to find anything that's even good. So when I do find a good one, I want to tell people, Hey, if you want a good ice cream scoop, I've tried a whole bunch of them. Most of them are terrible. Here's a good one. And now you can't even buy it anymore.

I used your ice cream scoop badly last week. Yeah. I mean, you're, I feel like you're from one of those households that lets the ice cream get soft. Uh try not to because i don't want to but it sometimes is necessary because it's hard as a rock also part of it is that i spent a long time without an ice cream scoop we do have one now and i i bent so many spoons that's what i'm saying you gotta use you're not bending this oxo yeah no you're right you're right i wasn't also i was i was the first one to try to dig into it and there was definitely a there was definitely a uh little bit of a hardened shell up at the top it's not hard so it was hard through and through i keep my freezer uh at a cold enough temperature that the ice cream is rock hard it's rock hard for two for two reasons one any amount of sort of

freeze thaw cycle in terms of your your freezer getting a little bit warm and freezing down getting a little bit wrong it just makes bigger ice crystals like it screws up your ice cream you do not want to to even thaw a little bit and refreeze your ice cream so having it be hard frozen all the time preserves the quality of the ice cream as however you manage to get it from the store to your house and two my habit used to be back when i was slightly less healthy than i am today uh the way i ate ice cream was i would pull the pint of ben and jerry's out of the freezer and sit on the couch with it and having it hard frozen all the way through means that you can sit there on the couch and eat it and by the

time you're done having the amount you want the rest of it hasn't melted so i apologize for my hard ice cream but there is a method of my madness but i do agree it is too hard for most people to scoop that's why i have the industrial pointy tipped OXO ice cream scoop that if you put enough elbow grease into it and you have good technique, you will be able to get a scoop out. No, your technique was very impressive. And I will say that the pointy tip of the ice cream scoop was also something that I was not used to or thinking about. And if I had to do it all over again, I would try to do a better job but like i said i also even with our ice cream scoop that we have i feel like i am so bothered by the many spoons that i've been in the past that i end up not putting the force into it that i really should you have to be careful to using the force as a technique because if you

take something that pointy and you put a lot of horse in it you will punch right through the side of the cardboard ice cream container if you're not careful so there is actually a skill involved in being able to successfully use that tool and not destroy the container well i love the i mean talking talking to the studio neat guys about breville uh just brought back to me that they are again they are in a commodity business like everyone else but they have uh decided to make a nicer thing and like you said not necessarily perfect but they've decided that their brand promises that it's going to be nicer. It's going to be more expensive, but it's going to be nicer. And that that is like, it's a little like saying their brand promises,

Hey, we care, or at least we care more than those other guys who are selling you something crappy. Yeah, that's right. You just have to be faster than the other guy, not faster than the bear. something crappy. Yeah, that's right. You just have to be faster than the other guy, not faster than the bear. Exactly right. So we care more than everyone else, even if our products aren't perfect, which is not the ideal slogan. It's I'm not in marketing. Don't at me. But I had that thought while talking to them. And those guys obviously are really interested in what makes a good product. How does this product work? What is unique about it? What promise does it fulfill? And it struck me. I told them a story about how my sister used to work in a distribution center for Target. And we were talking about Target at some point. And she said, oh, well, you got to

understand in any given product category, there are eight different versions of it you can buy. And like number one is the dirt cheapest one you can possibly buy. And number eight is really nice, but, but super pricey. She said, you got to understand Walmart wants number one target wants number three, four. And this is how she described it to me. I don't know if it's still like that, but it made sense to me. It was like, well, actually that works. It's like the Walmart stuff is cheaper in all meaningful ways.

The Target stuff is a little bit nicer. And like that's Target's brand promises. We're a little bit nicer. And I think, you know, Breville is a little bit like that, where it's like, we're going to make more of an effort than everyone else. And as a result,

I have lots of Breville stuff in my house. Walmart is not, though, cheaper in all meaningful ways because that modifier you add on, it gets to the heart of it. It's like the, you know, whatever that was, someone can Google for it.

I don't know if it's a modern saying meant to be old or either way. How much money it costs to be poor. The rich person can buy a pair of boots for 10 times the price, but they'll last in the rest of their life. The poor person has to buy a one-tenth price boot every single year, right? So if you get the one quality thing from Walmart, yes, it's way cheaper and it's great that these products are accessible to more people because they're less expensive.

But if it breaks after a year, next year you got to buy another one. Whereas the person who bought the expensive one in theory that one lasts longer the problem is now that the 10 quality products only last like 1.5 times as long as the one quality products so you get to pay more and also they still break after like a year and a half instead of a year so we're in a bad situation with a lot of these appliances depends but yeah and and this is an oversimplification in the chat room matt pointed out like target actually wants like number two four, and six. I would say, according to my sister, it was more like three, maybe three, four, five, or three, five, and seven. Obviously, there's some marketing there. Walmart doesn't just have the number one. It probably also has the number two or three. But the goal, the brand promise is, we're going to get you the cheapest thing. It's going to be cheaper at Walmart. I had at one point, one of the first times I think I ever went into Walmart was visiting my parents when they moved into their house in Arizona. We don't have, and people are like, oh yeah, don't go into a Walmart. It's like literally I've never lived anywhere near a Walmart. There is a Walmart in my hometown now, but it wasn't there when I lived there. And there is no Walmart in the county that I live in. None, not one. There are a lot of targets though. Upscale. in the county that I live in. None, not one. There are a lot of targets though. Um, upscale. So, um, I go into them and, and at one point my dad was in the hospital and, uh, I was there longer than I thought it was going to be. And I actually bought like, I bought like a couple of

shirts and like, uh, look, I don't know what to say other than say, I bought a, I bought a new t-shirt and after two washes, I couldn't wear it anymore. Like it was cheap and terrible. And, and like that, that, and so much of what we buy is driven down because it's like, we need the price to be down. And it's like, well, we're going to, we're going to make it somewhere where people don't get paid very much to make it. And it's not going to be very good. And, but it's cheap. And if all you're shopping on is the price tag, um, for whatever reason, because you're shopping on is the price tag for whatever reason because you're looking for a deal because you just don't have the money to buy anything except for whatever is the lowest price that's what you'll get but i am fascinated by the the the companies that try

to buck that trend a little bit and and you know obviously this is the story of apple too at one fundamental level all of those years during the time when everybody looked at mac users like why are you why would you what illness do you have that makes you want to use this weird non-standard computer instead of the one that literally everybody else does? One of my answers was always, yeah, but it's better. Yeah, I know it's not compatible, but it's better. Apple does better stuff.

And then, you know, that has ebbed and flowed over time, but there is, and there has always been the level below which Apple doesn't want to go. And so people will say, well, yeah, I can get a MacBook for $999, but I can get a Windows laptop for $300. And it's like, well, Apple could make a laptop for $300 if it wanted to, but it refuses. It refuses to do that because it's got, as part of its brand, like it's got to be at a certain level and not go below it. And Breville isn't Apple, but I get a little bit of that vibe from it, which is I would rather spend a little more money

and get something that's nicer because I'd really rather use a product that's nicer. Also, when my T-Robot broke, I sent it to Breville and they sent me a new one. I thought I would have to pay them for them to fix it. They literally just sent me a new one, which was also one of those moments of like, oh, I kind of like this company. I like how they're thinking here.

Yeah, Breville is bucking the trend a little bit, but they're not as up to the Apple level. And like Apple has has especially on hardware has held the line very well in terms of quality you can quibble about those they're they're subject to the same sort of economic forces that push down the quality of of everything so they are they have been a victim of various commodity things and manufacturing you know realities that cause them to use parts that are lower quality than they might have been simply because it becomes cost prohibitive to do this thing in a totally custom way when you can get economically these cheap little components, right? But Apple pushes the industry in the other direction.

They will push their suppliers to give them better quality stuff. They will demand from their suppliers things made without toxic chemicals, which is not a market force making that happen. It is Apple making that happen, right? Breville seems much more subject to the market forces that dictate the internal components of appliances than Apple does. But they're both fighting that fight. And in the end, the things that Breville makes are less complicated, at least for now. The computerized version of these is like, oh, you can buy a toaster with a camera and a little computer in it that looks to see how brown your toast is. And that is a burgeoning market. I feel like that market is, I look at

that and I'm like, yeah, I know how tech products work and I know I have to stay away from that because the odds of me buying the very first toaster, often looks at how brown my toast is and still using it in 10 years are way lower than, you know, buying or getting my Bre toaster which i've now used for 11 years or so and it's still going strong we use machine learning to determine the brownness of your toast and i don't poo-poo that i think it's a good idea it's just that like it's going to take a while for that to shake out to the point where it is reliable and you know it doesn't have security flaws and get software updates and it's like i i don't need to sign up for that anymore than I really have to, I will have to eventually sign up for that everywhere. But I like it. I like the kinks to be worked out of it first, before I jump into that,

I get enough of that. And you know, in my day job dealing with, you know, computer technology products, where I am signing up for all of that, maybe don't need it on my toaster. Yeah, when we're talking about Target and Walmart walmart and all that i wanted to mention one of the uh examples i've got is that i bought these uh waffle henley's at i don't actually know where i think maybe you wore one to my house yeah yeah exactly i think it was long sleeves um i think it was back in the day i bought them at I'm gonna say something like Mervin's or something a long gone department store but like not up not upscale at all and I don't know how much they cost but my guess is gonna be it was like 20 or 30 dollars I don't know it could have been 40 I had two of them a red one and a blue one and I wore them I wore them out and beyond out because I loved them.

And after I had them about five years, I realized I need, uh, to buy more of these because I love them and I'm wearing them all the time. And then, then my ghost appears on your shoulder and says jason do they still make them if the perfect time to have bought them was when you bought the others but then you didn't because you know you don't know which is the product that you buy one in 20 products that you buy is the one you love and you want to have them for the rest of your life that's what i'm saying you have to the the practice you have to get better at is like once you realize even if it's a year into it six

months once you realize that you like them that's the time to buy multiples not waiting till they actually wear out yeah yeah i know right so so i went at that point and i began my quest to find them and i could never i found lots of waffle henleyys at all sorts of different places, at Target, at Mervin's, on the internet, all these places. And I bought a bunch of them and none of them were any good. None of them had good sleeves. They were thin. They weren't hemmed at the bottom. So they just kind of like splayed out all loose and thin and awful instead of having a little hem at the bottom.

So I ended up for years wearing through that red and the blue the blue died at some point and i still wore the red one but i couldn't i was told like by officials within my house i couldn't wear it in public anymore but i still wore it in the house or i could wear it in public only if it was covered and i was never going to take off whatever was covering it. They were falling apart, but I loved them so much. And I, I went on this quest and I could never find anything that was remotely a match for it. And I tried. And like two years ago for Christmas, Lauren got me a waffle Henley. And I said, Oh my God, this is so close. Not the same, but this is so close. It was actually a little heavier than I wanted it to be. But it was so close to the ones that I had gotten back in the day. And it was from American Giant, another company that's like they make them in America

and the goal is quality and not hitting a price point. And I went to their website and discovered that she had bought me the heavy waffle Henley, but they also had a regular waffle Henley. And I ordered that one. And that was the jackpot. That was like, yes, this is exactly, but like almost identical to what I had gotten 15 years before, 20 years. Actually, I got to say, there's a picture of me holding my newborn daughter wearing that red one. So it's more than 20 years ago. Probably a lot more than 25 years ago. I had that shirt a long time, like 20 years.

I had that shirt. So I found it and I was like, oh my God. So I bought the one and it was perfect. And then what I did, John, of course, is I immediately ordered one of every color that they made. But my point in telling this story is that Waffle Henley I got at Mervin's that lasted for 20 years in 1998 cost 30 bucks. The Waffle Henley from American Giant costs a hundred bucks. And while time has marched on and inflation happens, all of that is true. $100 today and $30 in 1998 are not the same, but they're closer. But what's happened in that time also is the pressure for most retailers of clothing in the U.S. has been to get the price down or hold it constant.

And I think that that's why I never got one over the intervening 20 years is because they couldn't make them. You were looking for the $30 ones. I was looking for the, well, no, they had to hold their price point.

It's a little bit like saying Apple has the new M2 MacBook Air and it doesn't cost $9.99 because they can't hold it there. I think that the clothing manufacturers are like, we got to hold the price point. It can't go up. And so as a result, they started, they took out the hem,

they made lighter fabric, and they just made it worse and worse and worse. And all those that I was trying to buy that were never any good, I think they were the replacements for what I bought. They were just de-contenting the shirt because they had to make it cheaper. And then I turn around and like, do I like that this shirt that I love costs a hundred bucks? No, but if,

but I, I have some hopes that they're going to last, if not 20 years, 10 years, and they're good when I wear them. But it was quite a, it was quite a moment of, of understanding that like, oh, I see what happened here is that the thing that I bought at some kind of semi-crappy department store in the late 90s for $30 or $25, to get that now you got to spend $100. Did you do the inflation calculator just to double check that $30 in 1990 is not $100? 100 well so with clothes and cars and many other things there's a lot of cases where there was a moment in time where the geopolitical balance was such that people could be exploited extra hard at one part of the earth that gave us a 30 high quality and here and that time passed kind of like a plastic recycling right in china right and so there's used to have a price no it doesn't these are terrible imbalances in the world that for a moment gave us more value than uh than was healthy

and now like that has slightly changed the balance of but uh it's so difficult to with things like clothes but uh but yeah like downward price pressure is good but um you know this the thing i was talking about before i finally found that it was uh terry pratchett 1993 speaking of the 90s uh from one of the uh the men at arms disc world novel it's the character sam vines the boot theory of socio socioeconomic unfairness we'll link to the wikipedia page because of course this one theory has an entire wikipedia page at boots underscore theory um yeah it's like i what you would think would be healthier uh not obviously the rich person thing where everything costs a bazillion dollars for no reason and the margins are gigantic but simply that uh you know the the economic rising tide is such that uh individuals are making enough money to be able to buy the one pair of

good boots instead of the hundreds of pair of crappy boots right yeah but unfortunately the same pressures that make it so that we need to make the boots cheaper and cheaper every single year also uh you know also exert downward wage pressure on workers who don't have enough power as compared to corporations and so it's that's the walmart story right is that it's not just that walmart looks for for price number one quality number one it's also that wal Walmart doesn't want to play price one, or they go to the makers of price three and they say, you need to make it for price one. And then they're like, oh God, how do we do that? And yeah, there are a lot of ramifications here.

I bring all this up mostly just to say that I do admire that there are companies that say, we're not going to base our brand and our business on chasing the bottom line of like, we can cut costs. Like, not that they don't worry about their costs, but that they're like, part of what we do is making the thing that's nicer and selling it for more to the people who care that it's better. And I love those companies, even though sometimes I wince at the products, prices, and sometimes I don't buy them because I'm like, no, no, no, that's too much. But I do appreciate that they exist.

And that's how I ended up with Breville stuff in my kitchen. And I think why you did in yours. And that's why I'm wearing an American Giant button Henley right now in fact yeah the waffle maker the last item on my list is the most egregious because you know this is the one where the margins are the biggest and it is just yeah so expensive for what you get but the the sad reality is that all the other waffle makers i tried were so bad that i was you know i'm i'm in the position where i can afford to buy a horrendously expensive waffle iron so i did but i don't feel good about it but boy is it a good waffle iron but it is not worth the price they're selling it for it is not like

the mac stuff where it's like oh a mac laptop think like rolex where it's like if you just need to tell the time on your witch on your wrist sorry you know don't don't buy a rolex because the price is so disconnected from the value uh although you could argue for collectability or whatever but there's no collectability to this waffle maker. It is just a too expensive waffle maker. But all the other ones I tried were terrible. Look, I don't want to spend $100 on a shirt. I don't. I really, really, really, really don't.

But after 25 years, I was like, yeah, I will. I will. And they're very good. They're very good. All right. John, thank you so much for being on Upgrade. Again, I enjoy chatting with you in person and on other podcasts but it's nice to do it in in this little corner of my podcast world

yeah we managed to fit it in in the uh dog window the dog left uh leaving us in blessed silence and i don't think she's returned yet oh perfect i like fitting in the dog window that's good it's like a dog door but people can fit in it podcasts can fit in it it's great well people can check you out uh atp.fm for accidental tech podcast here at relay reconcilable differences with our buddy merlin robot or not at the incomparable uh where john and i deal with thorny issues of existence uh listened to by uh philosophy professors apparently which i find both delightful and unnerving and don't forget the upcoming episode of the incomparable where you can hear us talk about andor andor yeah absolutely absolutely i i do fear john though that in the end

um the thing will be most remembered for is all the academic citations to robot fear that's great this is the only legacy i'm gonna have it's like citations to robots or not. Fear? That's great. This is the only legacy I'm going to have. It's like, boy, expert on robots or not. If only I could get an academic citation for follow-up. It seems like that's not happening. It's probably probably not. All right. You can find me, J. Snell, on Twitter

of course. And sure, jsnelladvancedon.social. Go ahead. Go nuts. And sixcolors.com is a great place to go for all of my stuff. Mike will be back next week. Yay. But until then,

say goodbye, John Syracuse. You're cheering for me leaving? Fine. Get Mike back. You don't like me? Get your regular host. I don't care. Come back, Mike.

John is being mean to me.