Empowerment Through AI-Assisted Software Development

Hello and welcome to Mac Power Users. My name is Stephen Robles. Joined as always by my friend, David Sparks. How's it going, David? Hello, Mr. Robles. How are you today? I'm doing really well. I'm super excited. It's the first time I'm podcasting with our special guest today, maker of Overcast, the Incredible New Transcription Service. And we're gonna talk about 3D printing a little bit as well. Marco Arment, thanks so much for joining us. Hello, nice to be here again. And Steven, welcome, I guess. I don't know. I've been on the show before, but not with you. And I've loved it ever since you've taken it over. So good job so far. Thank you so much. Thank you for for saying that. Yeah, and you have been posting about 3D printing as well. And so I think for Mac Power for more Power Users, we're going to talk about 3D printing and uh because my middle son he just has three D printed stuff all over and so I'm curious your experience. Absolutely. And and Stephen Robles, you don't even know what kind of Pandora's box you open when you talk to me about three D printing. So I had no idea. Thanks. You were on one of us too. I I I want to change my middle name to Gridfinity. I'm just saying. Oh yeah, yeah, yeah. I haven't even gotten into that yet. Oh no. Oh no. We have a lot to talk about, Margot. We sure do. All right, very good. Oh, and one quick announcement, our second monthly bonus episode is out. We mentioned it last week, but you can hear all about David uh smart home escapades and more tips there. And so you can look for the m pro Max users feed. Uh that link will be below. And we're actually going to do follow-up from ATP and our last episode about backups because we talked about Backblaze last week and accidental tech podcast talked about it a few weeks ago about how iCloud drive files and folders are no longer backed up to Backblaze specifically. And we had a wonderful email from Brian, a Fat Cat Software, talk about it. And we just wanted to be explicitly clear. Uh, listener of MacPower users actually got an email from Backblaze when they asked. But to confirm, iCloud drive files and folders are no longer backed up from Backblaze, even if you have download locally and it's all syncing. And hopefully Marco, is that can you guys confirm that? Because you guys went really deep on it on ATP. Honestly, I have lost track of what the answer is, so I can't confirm or deny. Well it seemed like the support email uh one of our listeners specifically asked and uh we'll put a link. I linked it last week and I will link to it again, but even if the file and folder is downloaded locally, as long as it's uh designated as an iCloud drive folder. This is from Backblaze's support. This is from Janet at Backblaze. It says that it's just cannot include iCloud Drive folders. Anything that macOS controls via cloud sync. So if you have files and folders outside of iCloud Drive, outside of your desktop and documents, if you're syncing those, those will get backed up. And it can still do photo backup, which is what the email from Brian was stating. But he also has an app which is for backing up photos and other photo features, which we'll link down in the video description. But uh we just wanted to be clear, and if anyone else has heard from Backblaze, uh, you know, we'd love to know. Yeah. And Brian's uh from Power Photos, he was a guest on the show a year or two ago, really good guy. And right on top of this stuff, of course, he's very interested in photo backup with his application. You know, it sounds to me like it's it's almost on Apple just the way they're treating this API now. We all need to be aware, because I know a lot of people are relying on off-site backup and they need a good one that works. Uh Steven and I have a new backup show in the uh that we are starting to prep. So we're going to go into this deeper when the time comes. But uh everybody be aware. And Marco, what's your current system? Are you still using Backblaze and do you sync desktop and documents to iCloud? I have tried syncing desktop and documents to iCloud. It basically never does what I want. Uh and it it introduces so much complexity in ways such as the like file path changing of where that actually is on your system, which is part of the complexity that that we see here, that it just it introduces so much complexity that I just don't want that. Also, so I operate with two Macs. I've been doing this for a while, which I'm sure we'll get into. I have like a main desktop and a laptop. They both happen to be physically laptops, but that's a that's for a different time. Um but I actually don't want them to be perfectly the same. If I wanted them to be perfectly the same, I would just take the laptop with me everywhere, the same one. I actually like having two different Macs set up in different ways with different work being done on them and different, you know, like the portable one, it has like less software. It's a more focused device. It's more, you know, things that I would only really need while out and about or while traveling. So I actually don't want that level of sync, not only because I find it unreliable and complex, but also because I actually don't want those things to match up perfectly. But that being said, to answer your question , yes. Backblaze. That's that's so curious. I want to ask about that too, because I'm the opposite. I want exactly every file and every app available everywhere I am. So if I edit a thumbnail on my Mac Studio, I can go to the patio and open the same file. It's in iCloud Drive, has the same changes, and I can just keep working on it. And so I want to get more on your setup in a second. I also want to say, David, you'll be happy to hear I am backed up officially now, even on my iCloud drive. I used Parachute Backup, which you guys mentioned on ATP. We mentioned here on the show. It is backing up both my entire iCloud drive and my iCloud photos. It's actually a dual system. You you decide where you want each to be saved. And they are both uh running weekly and they go to my network attached storage, my Synology, and I just checked it. All the backups have been successful. So David, I'm officially backed up. Bless you. Bless you, C . Thank you. And uh and it worked. And and one parachute is a one-time purchase, just in case anyone wondering, we talked about it a few weeks ago. But it does it, it does the thing. You know, you have to have a NAS or some other hard drive that you back up to, but it it works. It wor Yeah. I use carbon copy cloner for that. Um I used to use super duper, but for some reason I switched to carbon copy cloner at some point. Marco, what do you use an app like that? Do you do like the the mirror drive? I have done that in the I I don't currently. Currently I just use Time Machine for that role. Um but I have used both Carbon Copy Cloner and Super Duper at various times over the years. They're both great apps. They they kind of like trade back and forth who's on top in any given moment, I think, with features and you know OS updates and things. But um I've had great luck with both of them in the past. Yeah, agreed. Well we we got we got off on that tangent uh but I'd like to get back to Marco for a bit . So Marco, now you've been uh for folks who don't know you, uh you are the developer of Overcast. Now uh was it coming up on twelve years, right? Something like that. It's it's a long time. Yeah, twelve yeah, twelve years this year. Seems like just yesterday you were announcing it, and you know, and uh that's pretty cool. And uh you are now also an entrepreneur. Well, I mean, well, I would argue I was the whole time in multiple ways. You you added restaurant . How about you are now a restaurant ur? How's that the word? That even that feels like a lot. Uh so I my wife and I now own a restaurant, I would say. Yeah. I think restaurantur suggests a level of of engagement with the rest of the industry, maybe , or multiple restaurants that that we don't really have. Uh yeah, you know what? I think we just missed it. Uh but we'll see. We'll see for the future. You know, you never know. There's actually a great show on Apple TV about Michelin stars, like the road to Michelin stars. It's uh pretty wild. So highly recommend. I have a couple questions about your restaurant though. Absolutely. Tech wise. And so I've I actually have a a family friend who owns a restaurant. They're just on this journey now. And the whole online ordering and DoorDash and Uber is very complicated. You know, they've decided not to do that. And so I was curious, it it looks like your restaurant does online orders, but you haven't opened up to like third-party apps like DoorDash and Uber. Have you explored that? Is that too much added complication? Why or why not? Well, we actually have a huge amount of complexity running our particular restaurant that reduces the amount of complexity in this area. So the complexity of our restaurant is that what we did was we when our when our favorite restaurant went for sale uh because the owners wanted to retire, we decided to buy it to kind of to preserve it really and to make sure that no nobody else could buy it and ruin it. Um, because we cared about it so much. And it's a really big part of our community on Fire Island. Now, because it's on Fire Island, you can't drive there all summer along. There's no no like , you know, consumer vehicles are allowed, only emergency vehicles really. Uh and so there is no DoorDash and there is no Uber Eats. There's no delivery services whatsoever, unless it's a kid on a bike with one of those big baskets on the back. So we actually don't need to deal with that at all. We we need to deal with a lot of other things, like how do you get all the ingredients there? And the the the answer is on boats, really. Um so there's other complexities we deal with, but we don't do anything w with those services because they don't operate there. Um we do have so the online ordering system that we have is really just a feature of our point of sale system, which is toast, which is a pretty popular one that a lot of restaurants use. Um, Toast just has a thing where you can like, you know, put a link on your website that links directly into your, you know, ticketing system and people can basically submit orders directly and it pops up and, you know, somebody, somebody it prints out and it beeps and people go, you know, start making it. And that's so it's kind of the old the older school way of doing it, but that's how we're doing it. You know, when you bought the restaurant, I was thinking, because you know, I had a fr prior life where I used to represent companies that owned restaurants. And that was the worst business. Those people work so hard. I mean, and like they had they have the most failures in restaurants. It's like , oh my gosh, what's he getting himself into? How is it? How are you feeling now that you've Aaron Powell We are very fortunate that we bought a successful restaurant. So all we had to do was keep it going. Exactly. Look, owning a restaurant was never on my life to-do list. Um I it surprised both of us uh that that we were doing it. And and when we were going through the process of you know, first of all first evaluating whether we should buy it and then going to like the bank and the accountant and the lawyer and figuring all that stuff out. Everybody kept asking, so do you have a lot of experience in the restaurant business? And we're like, well I I worked as a busser when I was 16 uh and that's about it. And they're like, why are you doing this? But and if it wasn't our favorite restaurant, and if we weren't in our minds saving it, and if it wasn't already successful, and critically, if we didn't have all of the existing people on board who were going to stay and manage it, uh we would never have done it. And we would never buy a different restaurant. Uh it is only it was a very special case for this one. And it helps tremendously that the people stayed and that they are doing most of the management. And my wife, Tiff, does, you know, mo for for whatever part falls on us, she does the vast majority of that. So very little of it falls on me, except basically technology issues and handshakes and occasionally filing some permits. Not only does it have killer Wi Fi, um and of course, backup internet service, so we don't go offline. Um but also uh the other restaurants are now asking me for Wi Fi advice for their for their restaurants. So I might heard your story about become like the de facto IT person for the restaurant business there, which maybe I shouldn't, but you know that's that kind of happens. I've heard you talk about you, you know, the whole extended ubiquity setup uh in your restaurant on ATP, which is awesome. Do you have uh iPads and stuff in the restaurant that are like business devices? And if so, are you using like Apple mobile device management, Jamf? How are you managing that that kind of stuff? That's a great question. Um so the point of sale systems, they're they're all like toasts terminals, they're all custom Android tablets with like the cache drawers and stuff like that. So those aren't ours. Um we do, however, use an iMac in the office for like a general kind of office things. Um, and we have an iPad mini as the music player that feeds the mixer and everything else. And so and those are the only Apple devices. Oh, sorry. And we also use Apple TVs to run the two TVs plus a security monitor in the back. So we have three Apple TVs. We actually manage them all as like home devices in a way. Like we have we have like a couple of locks, like we know, keypad locks on a couple of doors for stock and stuff. And those are just we have an Apple home called restaurant. And we have, you know, the we have like users of the home, guests of the home are our staff. And when somebody like you, know un,locks the stock room door, I get a notification on my phone saying, you know, Julie, unlock the stock room. You know, like stuff like that. It's all through Apple Home stuff. And home kit locks and and we have, you know , what locks? What lock models can you say Although as it turns out, home key is so slow that we just type in our codes. No one uses it. You should get the ultra wideband ones from ACAR and now. People just approach. Honestly, I I have had a couple of ACAR things in the past and I wasn't too impressed. Uh in general, I have had pretty poor luck with most home kit made devices. I had awful experiences with the Logitech uh secure home video cameras. Um I had, you know, the the Accara lock. I' Ive had one on my house and I had to replace it with the Saga one because it was so unreliable and finicky. Um I've had a couple of the um the Eve like air monitors and stuff and I found those to be pretty finicky. So most of the home stuff I've had, I I've actually just left on separate platforms. Like my favorite light switches are all Lutron Caseda. Has its own entire you know system. Lutron's the best. Yeah, like it connects to homeKit. So I can use homeKit if I want to to control it, but in practice I never do because it's running on its own system and I just use its own stuff. Or I use physical switches. Now you also said you've got a camera monitor at the restaurant, and that's probably using the one of the home kit camera setups. Not even close. Nope, all ubiquity. Ubiquity, yeah, I was gonna say. Gotcha. And you're not using like home bridge to get that into home kit or whatever? Again, like I I I have found that like whenever I put more stuff into home kit, I I mostly don't use those integrations or or they end up breaking eventually and then I just never reset them up again. You know, like you know, light switches again, they're all with Lutron. I have like all of my temperature sensors and everything, that's all Yolink now, which is great, super inexpensive, and it has its own radio protocol that's very, you know, it's super long range, low power, goes through walls and critically metal refrigerator boxes. Are those Zigbee? Is that what they use? No, it's it's LoRa. It's it's it's a different system, uh L O R A, and it's made to be I I think there's only a couple of brands that make stuff that uses it, and I don't even know if they're interchange or if they're um if they interoperate, but it's very inexpensive little sensors that just have very long batteries. You know, like I have these temperature sensors all over the restaurant and all over my house that run on a couple of double A batteries that last years on those batteries. Um it it's great. Okay, let last restaurant question because we got to get to overcast and transcripts. The Apple TVs, I've had to set up like digital signage in venues and it's always a challenge like what software to use, what are you playing? Are you just looping videos? What are you putting on the Apple TVs in the restaurant? We don't really have a great solution. That's kind of up in the air. The challenge is there are very few video streaming services that are licensed for business use. You can't just put on YouTube and call it a day. Well they do, but they shouldn't, you know, like so there's, you know, we we're bouncing between whatever crappy app we can find that season that supports it. Um and it's it's never like right I don't even know what it is right now. We just changed it like last month and I forgot. But like it like But you know, the music service is much simpler. There's there's a whole bunch of those. Um we had Sound Machine for that last year. This year we're doing Pandora's thing. That's the their Pandora's commercial thing. Um but for video services it's an awful landscape out there. Yeah. I think there's an opportunity on Apple TV for an app developer to come up with like like remember when Panic made that great status board app years ago and like stuff like that for businesses and public places as well to really turn the Apple TV into a marketing platform. It's just kind of surprising to me that nobody's ever really jumped on that. Yeah, it's it's an incredibly capable platform that unfortunately, when Apple launched the TVOS platform and the App Store, um they made a couple of implementation decisions to make things simpler for users in theory that ended up making it pretty difficult and cumbersome for developers to develop for it. One of the biggest problems is persistent storage can't really be counted on on Apple TV. And so it makes it very unpleasant to develop apps for it. Then by basically never lowering the pricing or introducing cheaper models that were meaningfully cheaper, I think Apple severely limited their market power as well. And then for games, that's a whole separate story of like requiring the remote to be compatible, not having an official controller that's shipped with every Apple TV, and so that pretty much killed the game market. So in addition to the storage problems I was mentioning earlier. So uh it's it's a tough platform. It basically only makes sense for the the you know the major video apps to develop for it. All right. Last quick ones. You just walked around Manhattan, thirty two miles. That's right. Crazy accomplishment. The Great Saunter. Yep. The Great Saunter. You wore two fitness trackers. You had an Apple Watch Ultra which model? Two or three? Three. Three. Apple Watch Watch or three. And another one. And you said accuracy was actually pretty good between the two of them? Yeah. So you know the the the other one I wore was the Sunto Race S, uh, which is their it's one of their smaller models. Most most like sport, like extreme sport watches are very large. Um, and they don't really they're a little too large for me, especially like most of this training was happening in cooler weather, like you know, in the winter and spring, and trying to fit a big chunky watch under a jacket sleeve is really awkward. Um so I wanted something smaller, so I went to the s the Sumo to Race S. Great platform, great watch, super lightweight. Um basically this the summary of my findings on all of on that whole world, I first tried Garmin, then landed on the Sunto , is basically that they are terrible smart watches, but very good sport watches. And so if what you want is more like extreme sport abilities or very long battery life, very long endurance, you know, extreme condition uh abilities to operate, they're they they can be better than the Apple Watch and in certain ways, even more common stuff is actually better on them in certain ways, certain workout uh preferences and things. But if you want like smart functionality, notifications, responses to notifications, Siri access, things like that, they either can't do those things or they do them very poorly compared to an Apple Watch. So like as computers, the Apple Watch is much better. Um, but as as like sport trackers, some of these can be very good. And I found the accuracy uh to answer your question in a very roundabout way. The accuracy I found it be for like GPS tracking and stuff is good enough between the Sunto and the Ultra. They were they ended up deviating by the end of the 32 miles, they ended up deviating by about, I think about eight or ten percent, something like that. So not nothing, but not enough to really be a problem. Aaron Powell I feel like all that fitness tracking stuff you just have to treat as a relative marker when you get it and don't expect any of it to be actually correct. Yes, that's exactly right. I've been going to you know I've been I went to the doctor and I got a scan body scan. You know, what's percentage of you as muscle and fat and all that. And I thought the DEXA scan? Uh I forget. I think mine was called in body or something. But it it gave me a printout and the doctor said, Oh, you're doing good on this, you need to do better on that. And I thought, that's actually good data to have. And I bought uh off Amazon a home version of that where you pull up a handle and you hold the handle while you weigh yourself and it runs an electric current through you. That's the the technology is kind of amazing. They used to put you in like a tank of water to figure this stuff out. Sounds terrifying. Yeah, but but now uh it's very simple. But it's shocking to me because the the same day I went to the doctor and did the home one, there's like a 5% difference, you know, and like significant results. And I'm sure the doctor one is is the accurate one. But you know, I was telling Daisy, my wife, I'm like, but now we have a measuring point. It's like it's not necessarily accurate, but you know if that number goes down or this number goes up, that you're making progress. And don't worry about, you know, the actual number. Yeah, that's what I found to be the case with almost all consumer wearable or fitness kind of tech is that it almost it's almost always fairly consistent relative to itself. So if you ha if you do a run one day and it says you burn 400 calories and you do another run the next day and it says you burn 500 calories , those probably are not accurate numbers in absolute terms, but in relative terms they can be useful. You can say, oh, I I went about about 20% harder or 25% harder um, you know, today than yesterday. And so and same thing with any kind of, you know, calorie measurement, body measurement, fat evaluation, all those different things those do, they are they are good relative to their own past measurements. Um but not in terms of absolutes, like things like calories This episode of Mac Power Users is brought to you by NetSuite. Businesses everywhere are asking the same question. How do we make AI work for us? The possibilities are endless, and guessing is a bit risky, but sitting on the sidelines is not an option, because one thing is almost certain your competitors are already making their move. No more waiting with Netsuite by Oracle, you can put AI to work today. NetSuite is the number one AI Cloud ERP, trusted by over 43,000 businesses. It's a unified suite that brings your financials, inventory, commerce, HR, and CRM into a single source of truth. That connected data is what makes your AI smarter, so it doesn't just guess, it knows, intelligently automates routine tasks, delivers actionable insights, helps you cut costs, and make fast AI powered decisions with confidence, and you've got total flexibility. From software and IT services to healthcare, equipment manufacturing, financial services, and many other great American industries, Netsuite delivers a customized solution for your business. This isn't another bolted-on tool. It's AI built into the system that runs your business. Whether your company earns millions or even hundreds of millions, NetSuite helps you stay ahead of the pack. I've been on the inside of trying to build these systems out for companies, finding a tool for this and that, and then they don't talk to each other, let alone using one AI to communicate across all that information, and that's exactly what NetSuite can simplify and make easier for your business. If your revenues are at least in the seven figures, you can get a free business guide, demystifying AI at net suite.com slash MPU. The guide is free to you at N-E-T S U I te dot com slash m pu and that link is in the show notes below. But thanks to NetSuite for their support of this show and all of Relay Marco, uh it's been a while since being on the show. Uh you have the latest and greatest, right? You probably got the latest and greatest Mac, right? I think I actually might have the so I'm using the M3 Macs MacBook Pro 16 as my desktop. Is that what I was using last time? I think it might be. It may have been. So you're using the M3. How's it going? Yeah, it's going so far great. Like every time a new you know, M Max Series chip comes out. I'm like, oh, should I upgrade and look at the numbers? I could get a little faster, but so far I'm fine with it. And y I also I spec them up a lot. Like I get the eight terabyte. Um I think this is 128 gigs of RAM. I'll double check that. Yes, it is. So it ends up being a pretty expensive purchase. So if I'm only gonna get like twenty percent more CPU performance, I'm not gonna spend seven thousand dollars to do that. You know, or whatever it is. Like, you know, that's that's a lot. So I I'm gonna wait a few years, a few generations. So my current plan is, you know, this so far the M three Max is working out totally great for me. I don't think I'm ever meaningfully bottlenecked by it. Um, usually my thought is much slower than what the computer can keep up with. So But be honest, if you could have gotten an M5 Max MacBook Pro without Tahoe, would you have done it? Hmm. That's a little tempting. Okay, I'm gonna say no, only because there are certain things in Tahoe I've gotten used to. So I I did uh a few months back, I did switch my browser from Safari to Chrome, and when you're on Tahoe, Chrome will do the autof ill SMS verification message thing. Right. And because they're like they made an API for that in Tahoe. That alone is so convenient that I actually might I might like stick on it just for that. So do you have Tahoe on your M3 Macs MacBook Pro now? Yeah, I I upgrade I ended up upgrading both. So I I've gotten a little bit um you know first of all like I'm I'm developing software and there are certain things that are easier. So for instance, we'll get to Overcast, I'm sure, in a second. Overcast transcription, the transcription app that runs on the transcription servers requires Tahoe because it uses an API that's only on Tahoe. So when I was developing that app all last summer and fall, I was just I was like developing it mostly on my little laptop and then having like shipping it between laptop and desktop to keep running things and then like eventually shipping it to the servers and it was a it was a big pain in the butt. So now having everything on Tahoe, despite me not liking Tahoe, you know, interface-wise, um, I think it's a mess, but it is more convenient for me to use it. So I tolerate the parts I don't like in order to get the productivity benefits from it. Now I know you're using the actual Apple APIs for the transcriptions and the overcast transcripts, but are you using AI anywhere in your workflow like clock cowork or codecs aside from all that transcription work? Um and first of all, I would even say like it's questionable whether transcription using Apple 's framework counts as AI. Right. Sure. But Apple actually gave up that fight. They've really worked hard to not say AI for a long time. Everything was machine learning, and then they were like, forget it. It's we gotta do it. Yeah, exactly. Do they say artificial intelligence or do they say Apple intelligence? Well, that's a whole thing. Right, right. But so where I use AI in in my work so far is first of all, it has largely replaced or at least precluded a lot of Google searches for me. Um you know a lot of web searches I where I'm just I'm looking for a piece of information, I usually will go to my AI of choice first for that because usually it will find it for me faster. And then if I need to verify it, they they are now much better at providing links to verify. Um so for me, I use AI a lot every day for basically what used to be web searches. And then I use it earlier. So I'm not using it to write huge amounts of code. I am using it though to help me write the code that I'm already choosing to write. So for instance, earlier today I was working on affix to the way Overcast tries to match up transcripts after dynamic ad insertion has happened and messed up the alignment uh with your copy of the transcript, which the you know the current shipping version in the app store doesn't always get it right. And I was I've been trying to line this up, I've been trying to work on that algorithm for months. Um, and I was just hitting walls . And the other day I had Claude Code take a look at it. I said, Hey, go to this in this file, look at this function. What do you see? And it found a couple of little edge case bugs that I haven't found for months. And you know, earlier I was doing a different part of that, and I had and I said something like write me a function that tries to find this chunk within this array by the center and then expand outward and find the best match and find and like I just described that in a paragraph and it wrote the function for me. And it's you know a fifty line function. It's not a very it's not a lot of code, but it wrote it and it was right. And for me to write that function would have taken half a day because I would have gotten a lot more wrong along the way, I would have had to go back and fix. And then I said, gener ate tests so you can verify this answer. And it generated a handful of tests and it it made a couple of tweaks. And then I said, generate a lot more tests. And it did, it ended up making like 45 different test conditions, tried to different like insertion and deletion patterns and different overlaps and things like that. I never would have done that part. So it took the part that I could have done and would have done and made it take a lot less time and made it have fewer bugs. And then it took the part I would never have done and did it for me with no effort and no complaining. And I just kept saying more, more, more it. did And didn't In't have to feel bad about that. Am I making am I making somebody do this for me? Like, you know, I'm I'm so demanding, it's never enough. You know, avoid all that. I didn't have to do it, so that's great. And then it helped me find a super obscure bug in the code I did write. And so overall I'm finding a lot of uses for it like that. Now I am also working on a second complete app, um, tentatively called Reminder, that I I can't imagine Apple will ever let me ship that with that name. Um but I will try. But but I'm making an app that basically is like snooze like better snoozes and management of your reminder's data. And I've been using it myself to manage my own stuff. And that app I completely vibe coded the first version of so I could start using it myself. And then once I kind of got it to a place where I want where I realized like this, this is something that I like, I I've gotten it where I need it to be, but now I actually want to make it good so I can tolerate it more and maybe someday ship it as a product. When I got it to that point, I took over. And again, I'm still using AI sometimes for things like, hey, can you you know write me a quick function to do this? But I found that for me to make an app to my quality standards, things like you know, I would ask the AI before I took it over, I would say, like, hey, this animation is is messed up when you you know move this thing to this thing, um, and it would it would just just be totally unable to fix that um no matter what I did. So certain things like that, you know, I I found it is better to do it myself. But I think, you know, if you ask me again, if you ask me that same question in six months, it might be a different answer. It's moving so quickly and AI coding is so good. Most of the time I'm not limited by its abilities, but rather I'm limited by me not thinking to ask it to do something because I I don't even consider the possibility, or I would assume it wouldn't be able to. Yeah, this is a thing I've been talking about a lot lately, is I think we all have mindsets about work and how we work with computers and AI turns it on its head. You know, the analogy I use is is uh uh motion pictures. Like in the old days, you'd go to a play, you watch a play when the first film cameras came out, they'd put the camera in the audience to watch the stage. And it took them time to understand you could put the camera on the stage. And we have the same problems with the way we work with computers as this thing becomes more of a tool. Absolutely. I mean ear,lier on in a lot of my development career, and a little occasionally today, but usually not today, but I I used to have a problem where I would I would not even consider attempting a certain feature or algorithm in my code because I would just assume the hardware won't be able to do that fast enough. One thing about that was when I w back in a million years ago when I was working on instapaper, um certain algorithms about like detecting where the page break should be for pagination to work right. Um, my first few attempts at that, I had an idea of how to do it, but I'm like, there's no way the iPhone will be able to do that quickly enough. And eventually I tried it anyway, and it was fast enough. And then with Overcast, you know, when I was prototyping that 12 years ago, 13 years ago actually was the prototype. Um, when I was doing that, I thought about smart speed and voice boost processing on the audio. And I thought there's no way the iPhone's gonna be fast enough to do that. And then I tried it, and sure enough, it did it with like you know five percent CPU usage at the time. And of course now it's even lower. And so I was I was limited by by me underestimating the performance of the hardware. Well now I feel like I'm limited by underestimating the capabilities of AI tools. And I think many of us will be for a while. That right now, like oftentimes I I I will be doing a task that AI could do for me, but I didn't even think to ask it. And then other times I will I will finally have that insight and be like, I wonder. Like recently I um I was looking at um a big table worth of numbers, there's like a like a huge you know budget list and everything and I'm like I paste into one of the AIs, I'm like, give me the story about this. What do you see? Look around for errors. What do you and it like it was able to find correlations between different sections of the document. It was able to find like, you know, it was like, oh well this this looks like this was reallocated over to this side over here to like it it it found not only did it did it do everything I have to do, but it found connections I never would have found on my own. And then when I went through and did a human review afterwards, the story it was telling, I found all the evidence to support it. It was correct. Um similarly, I like I, you know, I recently I I uh I took over some of my like financial stuff that used to be managed and now I'm doing it myself. And I I had like, you know, an account statement from a financial product that was complicated. And I just uploaded the PDF to one of the AIs and I'm like, explain this to me. What are the what this does term mean? What is the value of this over time? And it broke it all down. I didn't have to like manually copy in text or like of course it read the PDF perfectly, even though it was a scan of a piece of paper, but it read it perfectly. Like ever ything, it just whatever I think to do with AI, when I think more ambitiously and I try it, so far, it works almost every time. And I'm always blown away. Like, wow, I didn't think it would be able to do that, and it just did it. The however, this is the where you're at the scary point. Because I'm teaching a chorus on this right now. I'm seeing a lot of people go through it and they like first they're skeptical, then they become believers. But at that point, you have to be careful because what it still doesn't do and frankly shouldn't do is judgment and taste, right? And I think a lot of people, as you start to have faith in it, start to give too much value to its judgment and taste , which is you. That's the part that's the human part of the equation. So I I that you know I agree with you on everything you're saying, but I also think uh for folks listening, be careful as it starts getting more powerful that you don't let it take over the agency part that's most important. Yeah, and I think it's worth pointing out like, you know, it I mean, first of all, a lot of judgment and taste it can do for you. Um but where I find AI is is best in like you know more creative or abstract endeavors is basically serving the role of brainstorming. Because what you know, when you think about you know what are the way these models work at a fundamental level , especially once you throw in randomness and uncertainty with things that are not easily answered? Um, they're basically BS generators. Yeah. And what is brainstorming? BS gener ation. Basically, you're like, give me a bunch of ideas in this area. Or where could this go? And AIs are really good at giving you an answer of what could be that answer? Like, here's something plausible, here's something that follows these patterns with these restrictions and has these outputs and everything. So it's an amazing tool for brainstorming. You know, give me theoreticals, give me some ideas, and then you the human can pick up from there and direct things further. You know, I'm I when I was trying to figure out a name for my reminder app that's not reminder that Apple would actually use, um, or that Apple would actually permit, rather, I went straight to LLMs like, give me uh give me some help with ideas and you know, give me some name. And LLMs are horrific at coming up with product names. Yes. They are awful. I've used them a few times to start this process and they're horrific. However, they did give me direction. And I and like then I, you know, I fed them like here's give me some give me some starting points. Here, and then it gave me, and then I was able to take the the BS that it generated, and that might jog my memory to think of something better or that might kinda send me on a path. And then I was able to say, All right, how about this this name, this name, and this name? Do any of them have any obvious trademark conflicts that you can find? Any competing apps in the app store for Mac or iOS and and you know any any weird negative associations in different cultures I might not be aware of. And it's able to break all that down for me with you know far less work than me coming through the US PTO site looking for trademark conflicts. You know, I can I can do that first and I can rule out a bunch of stuff before I've put in all that work and send me down directions I never would have thought of. I will say on the brainstorming part, I do think the basis or the prompt for whatever brainstorming task you're giving it also has a lot to do with the output. Because if I just ask it, give me some video ideas about iPhone, that's gonna be trash. But when I give it a transcript of a video I've already made and I ask it for ideas about titles that might be clickworthy or curiosity generating. It does a much better job at that. And I'll still tweak those and then come up with my own, but you do have to like give it something. You know, to try and do something out of whole cloth, I find it is doesn't give as valuable information. Absolutely. I found the same that that you know part of my process for for trying to have it generate names, which again it's terrible at um was I I wrote in it a big Apple note first like here here are some products I have named in the past. Here's why I named them these. Here's why I like these names and here's how they have worked or not worked so well in the market . So based on that, you know, here's you know, here's what I'm working on now, give me some name ideas. And and that was dramatically better than than it being undirected. However, it is still pretty bad. Well, that's why I always laugh when people like say, you know, I'm gonna pick this engine, this model, or that model, it's two percent better, whatever. But they don't spend any time thinking about context and it's all about the context. If you give it enough context to any of the big models now, you're gonna get a good result. But it's way more about the context in my opinion than it is the actual model at this point. Absolutely. You know, the a the models now when you are using them when they are relying solely on their training data, or relying mostly on their training data, they have significantly worse output, like in terms of accuracy and things like that, than when you are giving them a whole bunch of context or they're finding it themselves on the web, for instance, um, and they are then sifting through that. So the more you can give them or the more you can direct them to go out and find, the better your results will be compared to just based on your training data, give me some BS. This episode of the Mac Power Users is brought to you by Mercury Weather. Go to Mercury Weather.app you for forecasts beautifully done. Download now for free. Mercury Weather is a thoughtfully designed weather app that shows all the essential weather details at a glance. It has a gorgeous colorful interface that dynamic ally adapts to the conditions with a warm orange palette on a sunny day, icy tones on a cold day, or a deep blue on a rainy night. Mercury weather uses a glanceable chart layout to present the hourly and daily forecasts in a way that feels intuitive right away. And this is a really cool feature for frequent travelers. Mercury's trip forecast feature automatically shows the weather at your destination right in your daily forecast time. So you'll always see the weather for where you'll be, not just where you are. What I really like is the UI . It's not only nice to look at, it's very gl anceable. You can get all the weather information you need very quickly. They've spent a lot of time thinking about it, and you can tell. When the weather gets serious, Mercury offers storm and hurricane tracking with maps, live positions for,ecast paths, cones, and intensity, plus widgets so you can keep tabs on a specific storm or the closest one right from your home screen. Mercury Weather's gorgeous interface makes it a delight to check the weather every day, even on gray and rainy ones. And the app's business model is simple: no ads and no selling of user data. Mercury is available for iPhone, iPad, Apple Watch, and Mac. You can download it and use the standard features for free and upgrade to Mercury Premium to unlock all features. Head over to MercuryWeather.app slash MPU to download Mercury Weather right now. Use that link to let them know you heard about it here on the Mac Power Users. Once again, that is Mercury Weather.app slash MPU. Try it out and get all the standard features for free. Thank you, Mercury Weather, for all of your support of the Mac Power users . What what do you think about the future of all of this? And I feel like the local models are gonna get really good soon for the mo most of the stuff we do. And just imagine where you don't have to pay Anthropic or OpenAI, but your Mac has local models strong enough to do a lot of the AI stuff we talk about. Um, do you think Apple's gonna go there? I think it's too early to really know that. Um obviously there's a huge finan cial incentive to try to shrink models down into more specialized smaller versions that can run on smaller resources. Because as we are seeing right now, you know, the entire industry is being constrained by the amount of massive data center capacity they can build to run massive chips in massive arrays, use being powered by massive amounts of power. Meanwhile, we're all walking around with supercomputers in our pockets that are mostly sitting around doing nothing or waiting to show us ads. So obviously, I there's there's all this power that we can take advantage of in our pockets. Um this is you know part of why I built overcast transcription the way I did, uh where it just runs on a bunch of Mac minis and it's the base model Mac minis, like the 16 gig M4 base model Mac minis. It's 48 of those. And uh, which now looks like a great investment, actually, in retrospect, but because now you can't get them anymore. But i i i intentionally wanted to build this feature only with what could run locally, especially also run on an iPhone. So while they are Mac minis, they are they are us ing APIs that are also available on the iPhone, and one of Overcast's features is you c if I don't have a transcript for your episode or your version of your episode and I can't figure out the alignment, you can just transcribe it locally on your phone. And it's not a super fast process, which is why I'm not just doing them all this way. You know, an iPhone takes a good few minutes at least to transcribe most podcast episodes. And there are certainly limits on how much it'll let me use, how much it'll let me execute in the background and stuff like that. Uh so it's it's a little bit cumbersome to do it locally, but it can be done locally. So my plan with that is over time, as the phones get faster, maybe I won't even need to run these servers anymore. Um, you know, i will I still be running them in 10 years? Probably not. I think the phones will be fast enough then that everybody could just, you know, one person could do it locally and it'll just send it to everybody else, and we wouldn't necessarily need me to be running 48 10 year old Mac minis. Will I So then I'm glad you I'm glad you said this because we're in an overcast transcripts now. And I just want to say I'm gonna link to uh the ATP episode where you go into this. I don't want to I didn't melt my rug is the name of that uh episode, which was an amazing story. And so we I won't rehash it here, but we'll link that in the show notes and also the Reddit post in your overcast subreddit because you have a bunch of pictures here of the entire server rack, the 48 Mac minis. It's incredible. And so , one, like, can you explain when an episode hits a feed. So when this Mac Power Users episode hits our RSS feed, what is happening in Overcast? Like, where is it seeing that episode? Where is it going? The Mac minis and then apps devices Aaron Powell Sure. So Overcast has always been server-side crawling for feeds, which is now how almost every podcast app works. So you know the the your app on your phone does not directly access your RSS feed. It's syncing back and forth with overcast servers, and then overcast servers are crawling the feed. And when updates happen, they pick them up within a few minutes of the episode being published, uh, and then they push out updates to everyone's phone that contain that episode. And then your phones then go and download the episode directly from you. So it's kind of this triangle. But so w when transcripts were added, it really just became part of that process . When a new episode is uh detected , my servers add it to the transcript queue. And the transcript queue, like all those Mac minis, are just queue consumers. So they are just constantly checking in with the servers. Hey, what do you got? What do you got? What do you got? And the servers, every time a new episode comes in, they add it to the queue. Most of the time, the queue is empty because the servers are keeping up with it. Um, occasionally it'll like if there's like a really high traffic period or, if I am doing a big backlog uh of a big podcast, sometimes that queue might you know have a bit of a weight, but it's it's sorted by popularity of shows. So your show being pretty popular is gonna get transedcrib almost immediately after the episode's released. So then I transcribe it, like one of those Mac Minis claims it, it transcribes it, it puts the transcript up on Overcast storage, and then it sends an update to all the phones that had that episode saying, Hey, we got this. Here's a transcript. And then when you either download or play that episode, it downloads a transcript and shows it. So if I as the podcast creator , like if we were to upload a different MP3 because something had to be edited out or changed, does that whole chain get triggered again once the RSS feed gets updated? Or how is the transcript, and then that transcript that lives, I think, on Cloudflare , is that then updated? Like how does that work? No. So Overcast, the scale of how many podcasts and episodes out there there are would be way too it would be way too much for overcast to check for updates to the audio files after they've been published. So it doesn't. So if you if you upload an edited version of the audio file, the transcript will no longer be accurate to it. However, you would be then running into the same logic I use for dynamically inserted ads. So what happens is before the phone shows a transcript for the episode, it compares the signature of the file that it got from when it downloaded the episode from the publisher. And then the transcript knows the signature that it got that it transcribed. And the phone compares those. This is what I was talking about with my alignment earlier. The phone compares those and says and it can figure out, okay, well, it looks like audio was inserted at this time stamp and deleted at this timestamp and changed in this time range here. And so it just basically zeroes out those parts of the transcript that don't match up. It displays as the ellipsis in Overcast when it when it hits one of those. Um, which I believe, I think is actually the same thing Apple Podcast does with their transcripts. Um and so any part that does not line up with what the server's transcribed will just be an ellipsis and it will adjust the timing of everything else. If my code is right, it will adjust the timing of everything else to line up with the version that you actually downloaded. And if a creator remind me, if a creator at taches a transcript in the RSS feed, are you doing anything with that? Not yet. I'm actually in the process of um of building so I I ship transcripts with you know one version of the alignment code that I thought covered all bases, but it turns out it didn't. And the way that the text is rendered in the interface is using Swift UI, using a giant series of text, basically, of of text instances. And the there are a couple of issues a couple of issues with that. Number one was the performance is not amazing. That's why right now when you if you tap the mini player to expand it up on a very long episode you get, a little bit of a delay before it expands upward. That's because it's rendering all those text nodes in Swift UI and it's it's not efficient enough. Um the other problem is of course, the very first feature people ask for is the ability to copy and paste text out of the transcript. And as it turns out, doing that when you're rendering a whole bunch of Swift UI text instances is not good. It's it makes it it's it's either impossible or it certainly does it in a way that nobody wants. So what I'm in the process of doing now, in addition to fixing those alignment bugs, is one thing I was working on yesterday was converting the rendering of the transcript to a web view. Um now web views have a bad rap, I know, but Overcast uses them already and most people don't even realize it. Like the show notes, for instance. That's a web view. I just make it a good web view. It is a good web view, I can attest. Thank you. So that's a w and so I'm using the same logic now. I'm putting transcripts in a web view. That way text selection will be much faster and the rendering of them and the tracking of them as they scroll also is far faster when using a web view than it is using Swift UI. Okay, so I recently did a video about the best podcast apps, and I did put you as the best audio because of this. Oh, thank you. The transcript feature is awesome. Even the little music notes when you know music is playing behind like an ad or something. Very nice touch. No other app does that. For private feeds, which again, I think yours is the only app that will do transcripts for private feeds. It is. Which is awesome. So that's transcrib ing locally on device if that transcript does not already exist. Is that accurate? No. Oh it is not. Okay. That's also, you know, private feeds, they're still being server-side crawled. So what I'm doing is trying to identify clusters of private feeds where you know most private podcasts, they are not, you know, they're they're private, but they're still serving the same content to a whole bunch of people. Yeah. So I'm basically doing like duplicate detection there of like if I'm seeing the same audio file being published to in different feeds, I can just transcribe it once and serve it to all those people. So it's not doing everything because there there is still like a minimum number of duplicates that I will, you know, cue for transcription. Like if if you're the only person listening to a to a feed, if you have if you have something served just to you, Overcast's not going to transcribe that. But once it reaches a few people or or more, then I will add that to the queue. And then it, you know, once I transcribe that version, if I see it's being served to somebody else, it's the exact same file, they'll get the transcript too. So do you are you ever doing on device transcribing right now? Never without user uh requests. But if the user says do it, then that's doing it. Yes. How do you because I have a lot of shortcuts they use the built-in transcription action . And what I have found is that action specifically on a Mac can transcribe three hour plus audio files no problem. It seems different on the iPad and the iPhone and its ability to transcribe longer files. Like it just won't. It'll fail usually if the audio file is like twenty or thirty minutes or or longer. Is there like w the way you're doing it, is there any limit to how that on device model can do it? Like is the way you're doing with API somehow different than a shortcuts action that it's doing it? Um I actually I don't know how the shortcuts action is doing it, but I'm yeah, I'm just using Apple's built-in um speech transcription API that they launched with with the twenty six series OS. It it there actually was an API earlier than that, but it was much worse and much more limited. Uh it was dramatically re remade and improved in twenty six. M I do have limits, but they are much higher. It's like, you know, I think it's four or six hours. Okay. And even then , that's more of a limit of like the audio fingerprinting code that I use for alignment and stuff like that. But there is no inherent limitation that would prevent that, except when you are doing local transcription on an iOS device, you have to compete- you have to contend with the limits the system is going to put on you for total processing time. You know, so last year they introduced that API where you where that like Final Cut can use where it makes the live activity for when you're doing an export. Yes. Um so I'm using that API uh during my during my transcriptions on on the phone. That allows you to use much more background processing time than anything previous. However, it still isn't enough for a very long podcast on current phone hardware. Over time, that again, that will that will start to become less important as the phones get faster. But we're not talking about fifteen percent faster. We're talking about like once once they're five times faster. Then so you know, this this might be you know ten years from now or more, you know, it might be a while. But at that point , you know, you'll be able to do a lot more things before hitting those limits. But right now, those limits get in the way. That probably explains the difference between what's going on with shortcuts and what's going on with your app. Probably. Yeah. I'm guessing shortcuts is hitting one of those processing time limits. Aaron Powell Marco, what what model are you using to do the transcription? It's Apple's built-in model. So the the the speech transcriber API um it has its own model. It's the same model that Apple uses for all the places on the system that transcribe speech to text. So that's you know the the dictation button on the keyboard, it's Siri, obviously, it's dictation of voice memos, um yeah, Apple Notes. It's all the same model. So what's your experience with it? How's it how's it do ing? It has a couple of quirks. Um but it's overall it's pretty good and it's very fast. What I needed I needed for this to be fast because for me to have any kind of scale whatsoever, you know, obvious it takes f Pretty much. Yeah, I just kept buying 'em. Like I would I bought I would buy like 'cause m the racks that I the little enclosures I put them in hold six each. Um and so I just kept buying like six and then six more and then six more. And yeah. Yeah, basically you know I would I was I was like scrounging around like whenever there was a sale, I would piece them together. So there'd be this is you know back in the olden days when Mac minis were on sale, um they they would you know I I Black Friday came around and I and I like, oh B and H has them for you know four seventy five and Amazon has them for 482. And then but there's a limit of two here and three here. So I would I would piece together different sales and and just like you know and I I learned little tricks like you know when Amazon has a limit per customer, it's like a hard limit. You you can't just you can't just place another order tomorrow, whereas B and H doesn't care at all. So you can place three orders in a row at B and H for two each if the limit is two, and they'll you'll get three orders the next day. Like it's great. Um so much orders sell this on eBay for right now? Could you get a million dollars for this? You can pay the college tuition off if you uh sold them actually. Yeah, exactly. Have you with accuracy uh this whole system probably wouldn't work or wouldn't be the system you would use if you wanted to use Whisper API or a different, you know, transcription AI. Have you thought about like would you get better accuracy elsewhere? Or have you thought about not using Apples for any reason ? I I'm looking I I've looked at things like um NVIDIA's parakeet model, for instance. That's probably the the next uh the next best competitor for this because it also runs locally and is also very fast. It's better in some ways, worse in others, and I wouldn't be able to practically ship it in overcast on the phone. And so if I if I switch to that, local transcription on the phone would be different than what the servers are doing. Right now it's the same because it's using the same APIs. And that's I consider that a good feature. So I'm I'd like to try to stick with Apples over time if I can. But yeah, Parakeet would be the next one to try. Now if I was doing like a remote API, like what you're saying, if I was just using the Whisper API, for instance, um, usually those have limits for audio duration that are too short for most podcasts or at least podcasts at scale, not to mention the fact they would cost significant We're talking 10 times or more additional cost compared to what I'm paying now. Aaron Powell And honestly, Apple uh the Apple one is as bad as it's ever going to be right now, and it's pretty good. And like in June , my guess is it'll get a nice step up with the new software. So you're fine. Yeah, that's that's kind of what I'm banking on is like right now, it I mean look, it's really imperfect. There's lots of errors. But what I'm using it for is not a replacement to listening to the audio. I don't think any of the machine transcriptions are good enough for that yet, except possibly the like the largest whisper models that that are very expensive and slow to run um that would never scale to this kind of usage. But right now, it is not like reading the text is not a good replacement for listening to the audio. Eventually, hopefully it is good enough that it can be. We're not there yet, but what it is good at now is skimming. And so if you if for instance your podcast is that you're listening to your tech podcast is talking about baseball, um and you want to be able to skim right past that, you can do that really easily. Or suppose they start an ad read for an ad you've heard five times in the last hour. You can just skim right to the end of that really easily. To the point where once you get accustomed to navigating a podcast via transcript, I think using the seek controls feels barbaric by comparison. Uh and that's why like I'm hoping, you know, over time I I I I mean this opens up tons of of great feature ideas and and interface and navigation ideas that of course I'm I'm working on. Um I think over time we're not gonna be using those seek controls very much at all. Uh I I think it's gonna be we're gonna start navigating much more via transcripts or automatic topic selection and things like that. Another example of things that we just have to rethink how we use . This episode of Mac Power Users is brought to you by OnePassword. Our special guest today, Marco, actually uses OnePassword. He's talked about it by using OnePassword with his teams at his restaurant, giving passwords across his family in different vaults. I use it personally with my family as well. And whether you're trying to simplify creating secure logins, finding a way to sync your pass keys across all your devices, maybe you need to be multi-platform or multi-ecosystem, Android and iPhone, or maybe you just want a great place to store all your secure information like account numbers, travel rewards, even passport numbers. That's actually how I use one password. Like I've said recently, since going independent, I've had to do a lot of tax stuff, figure out how to set up accounts for different retirement stuff. And where do I put all those account numbers? I want them in one place so they're easily searchable. And then I can fill in my EIN if I need to. Well, I actually put all that information in one password as well. I know it's secure. I know it's going to be available on all my Apple devices because I have one password on all of them and I know it's safe. So I put my password numbers. I even put social security numbers in there. This way, whenever I need to find sensitive information or fill out a form, and it's pieces of information I don't have readily available elsewhere. One password is where I store it all. And I can also use OnePassword for my autof ill, create those secure logins for all your websites, and a ton more. And there's even a new travel mode for OnePassword where you can disable certain vaults from your dev ices while you're traveling, but all your information is safe, and when you get home, you can re-enable them and you get access to all those logins again. So here's what you do: go to onepassword.com slash MPU, the number one password.com slash MPU and you can get twenty percent off either individual or family plans. That's one password dot com slash mpu for twenty percent off. Thanks to one password for supporting MacPower users and all of Rela y . With the RAM and storage constraints today, in retrospect, would you have gotten different models than the base model for your server? No, only because the reason I chose the base model. Apple's model I don't think uses the GPU at all. I think it's all CPU based. At least on the M4s, it is. And when I I did I did the math of like, you know, what's the best what what's the most CPU performance per dollar that you can get? And the base model blows away any other configuration in that in that metric. It's not even close. Like if you want like the very first M4 Pro model costs I think a little bit more than twice as much and is nowhere near twice as much the CPU performance. So bang for the buck wise, it was no contest. Now where I am limited is none of them have more RAM than the 16 gigs. And so if I wanted to, you know, for instance, like one way I could do one way I could architect this is I could have most of the servers do that first pass and then once they have a transcript, I could send them to like a bigger server and say, use a more complex model to maybe clean this up, find and fix errors, do like topic detection and stuff like that . And I don't have any computers to do that right now. They're all the same low spec. And that's something I'll look at over time as I as I investigate those kind of features. Maybe that'll be necessary. Maybe it won't. But the the price per the performance per dollar of those base models is just so far ahead of every other option. Nothing else was compelling, not even close. Aaron Powell Yeah I I really feel like you're going to uh not need to upgrade those computers. I think the the local models are going to get good, the hardware is going to get so much faster. In a year or two, you'll be shocked how much of this stuff can get done on the phone. I think you're right. Um how are you remote managing them though? I mean that's a lot of computers. Um poorly. So one of so because I am not just a a podcast app maker, but because I'm also a podcaster, it was important to me to have a really fun story out of this. So w as I was buying and installing 48 Mac minis and setting up this ridiculous system and learning how data centers work and then learn and getting my own rack cabinet and learning all that stuff. As all that was happening, I didn't mention it in public at all. Because I wanted to make a really great episode of ATP when I finally told Casey and John live on the air and of course all of the audience as well. They all learn at the same time. Like when when I'm doing something like this, I don't tell Casey and John in advance. Because it makes for a better show. It's more fun because it's like 'cause they're also my friends and I want to like surprise my friends. Yes. You know, it's a better story, I think. It makes for a better podcast. And I just like having that like to surprise my friends. So I did the same thing with buying the restaurant. When John surprised you all and Merlin when he went independent, that was a also wonderful surprise. Exactly. He's like, Yeah, I quit my job. Like, what ? Yeah, it was great. Like, I love doing those stories. And so as a result of me kind of keeping this quiet for you know almost a year as I was as I was de ploying all these. Um I didn't ever want to ask on the show, hey, anybody have any suggestions for remote management and and fleet management of devices for app because I I knew about stuff like MDM. I'd initially looked at some of the products out there and the pricing was just not gonna make sense for what I was doing. When you're looking at like, you know, eight or ten bucks per month per device, that's five hundred dollars a month at least. Um and that's and like I'm only paying a thousand a month for the whole rack cabinet and the internet and the power and everything to it. So I'm like that's that's a pretty big jump for do I really need that? So what I'm doing what I did was manually go through like I I would when I would get my six pack of Mac minis, um I would you know you get one or two of those and and pile up in my office. And one night I would put on some you know podcasts and play it on my phone and I would just one by one plug in a monitor, keyboard, and mouse and connect each one one by one, go through the setup process one by one. I had a list that I was following. I did I did it manually. I have seen that stupid liquid glass welcome screen 48 times. Hello. Yes. And it's it says welcome and it's so slow. It is and you have to wait for the animation. Because they they really want to show you how nice liquid glass is. And did you know image playgrounds are here? I did not. It turns out they are. And that's something that you learn 48 times. Image playgrounds are here. Um and so I went through all that and I did them all manually. And then I what I did was I set up tail scale on each one as part of that process manually. Uh but but once it's set up, and once you go disable key exper iene on tail scale for all of those. Learn that lesson. Once it's all set up, then now I just have like a bash shell script to whenever I have an update to the software they're running, which is just a Mac app, I just basically SCP it to all of those in a row and use and use SSH to like remotely terminate and relaunch it. Launch D is managing it on the service for me. So the actual operation of them now is fully automated through just you know cra ppy ways like that, with the sole exception of software updates. Those are not managed. So I haven't yet really had to face that yet. I have them doing automatic updates for security, but you know, nothing else. Um so I'm gonna have to at some point deploy something that's gonna manage that for me. And maybe that'll be a project for this summer. We'll see. Have you ever had an issue where you couldn't solve it remotely and you had to go to the data center physically? What was that? And is it like a severance style experience? Like do you have a key card? You like go in the elevator and all that? How's that go? Yeah. There's a pretty a pretty significant part missing. But uh but uh yeah, I have a key card and I gotta like, you know, swipe the key card, type it in and then like all the lights turn on in front of you 'cause it's all motion sensor 'cause data centers wanna save power wherever possible and most of the time there's no one in most rooms. You know, the rooms are filled with cages and computers and everything. And so, you know, you walk around, the lights light up in front of you, and eventually the ones behind you start turning off and and you're just kind of alone hearing a bunch of fan noise looking around like down these these perfectly pristine like white aisles, lots of air conditioning blowing through there. And it's it is it's a weirdly uh lonely but cool experience. I before this I had never been in a data center. Like I've I've always run leas ored virtual servers for all my stuff I've ever done. So I've never actually been in a data center. And it was a really cool process to learn. Thank goodness for AI for a lot of that stuff. Um and the staff there was very helpful. But it's you know, they're not always there's it's like you know one guy at a desk in the front, you know. But you know, they're all nerds like us and they want to help as much as possible. But a lot of it I you know I wanted to figure out on my own. So even just stuff like all right, they they give me a fiber drop. What is that? How do I plug that in? And what do I plug it into? And then how do I configure the thing I'm plugging it into to use whatever the drop is that they're giving me? And there's a bunch of different answers to that and and a bunch of different standards and you know different pieces of equipment, different terms that I had to use. And AI was very helpful in navigating me through all of that. But for the most part, most of it's remotely controlled now to answer your question, um, I hardly ever need to go there physically. I really only if I'm gonna like you know move some servers around or whatever. But what I did uh I I I have like a remote uh power switch thing that I if I need to like remotely power cycle a computer, I can go I can do that now. As long as that computer is not the router. I can I can now do that. If it ever is the router, that'll be an instance where I probably call the data center and have them go do it for me. And that there is a service where they they will do that for some you know hourly rate. Um they'll go do whatever you need to do. Network uh DoorDash. Yeah pretty much that's amazing. All right. Well, um, I have some over cast feature requests. Is now a good time? Absolutely. As we do, as we do on the show. Give you a list of work. Yeah. Sorry, so I dove in and uh so I love the the now playing screen is great, especially being able to swipe over to the transcript. Show notes, you know, across many podcast apps are not formatted well, but yours are formatted really well. My biggest issue with is the Q. And so on the now playing screen. Ah, yes. You know, you could you got all the good all the controls I could ever want. I love the swiping over to show notes and transcript. Can I get access to my up next cue from the now playing screen? So let's talk about cues. This is this is great. I didn't ask about this, but so it turns out uh okay, so when I first made Overcast, I based everything around playlists. Because my playlists, you know, they're they're they're smart playlists in in old iTunes terms. So they don't have to be, but they are they are by default. So they are basically, you know, just filters, like smart filters that it constantly updates. So you can say, you know, I maintain one, like I maintain like a main one that has most of my podcast in it. And then I have one called shower. It's like stuff I listen to in the shower. I have one called car, like stuff I like to listen to with the family on car trips. And so I'll have rules that like exclude these shows from the main playlist so I can save them for shower or car. And I thought this is how everyone does podcasts, obviously. And as it turns out, that's not the case. As it turns out, people have lots of different ways they can organize things. Now, one of the very first features, uh feature requests, rather, was people wanted a queue. And I thought, well, I've already built this entire system of playlists, and it's what it's what everybody wants in a queue. You can like drag them to reorder manually, and you can have stuff automatically in serted as it comes in. And I thought, well, that's great. So I'll just make a special playlist that just called Q, but just it's just the title of it. And when and I'll have you know shortcuts in the interface like add to Q that just create or find a playlist with that title and just insert it into that playlist. And as it turns out, and and this is what I've been doing for the last, you know, probably twelve of these fourteen years, whatever, or 10 of these, 12 years, as it turns out, everyone hates that except me. And so what I and if you look at my my like app store reviews, my star reviews, there's a couple of consistent problems that people report that I haven't solved yet. One of them is storage limits. I know everybody. I'm I'll I'm trying it the the background download system is is difficult to work with, especially when people have like a hundred podcast subscriptions. Um so I'm working on that. The other big issue people have is nobody wants the queue the way I do it. Everybody wants the queue to be the way most apps do it, which is the way Apple Podcasts does it, and I think PocketCast as well, which is just basically like an up next list maintained right in the app, kind of almost like an in-memory thing, like not even like a persistent thing necessarily. That's what people want. And so I've recently decided I need to do that. Like I don't know that I'm guessing I don't know when I'm going to do that, but I have decided I am going to do that. Um it's a big change. It's gonna be a a big effort. Uh but that's obviously what everybody wants and enough is enough, so it's coming. Okay, well that that is very exciting to me. I will say please do not look at Apple Podcasts as an example for the up next queue because I have found above all the apps, and I've told them directly like it is not good. One of the main reasons is because they have both a up next area and a continue playing area and I cannot drag an episode from one to the other, which is just an insane UI paradigm. Like these are all episodes that I want to listen to soon. Why can't I drag something from the continue playing to the queue? That's wild to me. And they'll also have bad, like if I the episode I'm currently playing, if I want to jump to Dithering because it came out this morning and it's a quick listen, if I tap it in my queue, whatever episode was currently playing disappears. And I have to go to that episode or that podcast and click play again. Whereas PocketCast, it moves whatever you are currently listening to back down to the top of the queue, which is how I and I think a lot of people use up next cues. Like whatever I'm currently listening to, unless I explicitly dismiss it or say I've listened to it, never let it go. Don't let me go. You know, whatever the Titanic is. You know, don't do it. So that would be my personal request. All right. That honestly, that's really helpful because one of the challenges of developing you know many apps, but in in this case podcast apps, is there's all these little like micro behaviors that when somebody makes a feature request, say I want an up next queue, okay. Well, but then you start thinking when you have to go implement, okay, well what what should happen if this condition of or this series of events occurs? What sh what should the behavior be? And what people expect does not is not always the same, like between different people. And whatever decisions you make on all of those little microbehaviors, you're gonna confuse and anger some group of people and please another group of people. And there is no set of decisions you can make that will please everybody. And it's very, it's very difficult to try that. Like, you know, one of the one of the biggest problems I have in Overcast um is the episode limit feature. Uh the idea here is, you know, keep only the last three new episodes in this feed. Okay . Um what should happen if you're listening to the fourth one? Right. Uh what about if you go down the old episode list on of that show and you add like numbers eight, nine, and ten, you know, you you want to listen to those? Well then do those immediately get deleted? Do you somehow keep the ones you've added manually, which is what Overcast does? Um, but then people look at the screen and they have six episodes downloaded and you said limit three. This feature's broken. Like the it there is no way to implement an episode limit feature that doesn't seem broken or behave counterintuitively at least a big part of the time. But there's no amount of those decisions that you can make about how that should work that to make that feature actually sensible and consistent and match everyone's expectations. So I would love to just not have that feature. But because I've had it since the beginning and a lot of people use it, I can't get rid of it. Sure, for sure. Well, I I think pocket cast of next queue is probably the standard. That's how my mind works. I know it's not for everybody, but uh I would say just rip theirs. Uh but my my last two feature requests, these are strictly about the widget, only because I'm one who doesn't use the podcast app icon. I have the widget on my home screen, and that's just how I play the next episode or launch the app and the overcast widget, it looks great. Is it too orange for you? Well, because so my I get that one a lot. I got two requests, that's one of them. The first one is the actual play button, you have it centered, whereas most other podcast apps, like Pocketcast, has it left aligned, Apple Podcasts has it left aligned, and also has the remaining time in whatever podcast you're currently listening to, like next to the play button. So one , I would I prefer a left-aligned play button. That's just me. And two, the color coordination. Pocket cast used to do this where when you would play a podcast, the widget would change color to match the show art. Apple Podcast widget still does this. So when you press play, it matches it. And I don't know if this is still possible because PocketCast stopped doing it. So now no third party app matches the show art. But if that is possible, that would be my request. Left the line play button, color matching widget. I I just have one question. Steven, do you ha always run three podcast widgets on your phone? No. No, I literally just added those to the stack right before I was asking the question. Good. Okay. I feel better now. Yeah, yeah. Yeah. I mean that's that's a lot even for me. Um Yeah, I you'll notice right now there's nowhere in Overcast that uses the artwork coloring in the UI. Sure. I've actually one of the things I had my Mac minis doing last summer was calculating the dominant color for each podcast. Um I was I was testing out different parts of the UI that could use that. And you kind of need to know that before the podcast is before the artwork is downloaded to the phone. So I needed something that ran server side to detect that. Um so I was I was trying all that, but I have just no matter what code I've used, I've used a lot of different algorithms. Even you know, Apple actually published a sample code that did exactly this years ago. I tried theirs, I tried a couple of different ideas. I have never found one that I thought looked good in a pla in places I wanted to use it. And I'm certainly open to continuing to try that. I just like when you when you see how the different apps use them, for instance, like Apple Podcasts, its now playing screen basically is is a dark theme. And it uses like a dark version of the episode art colors behind everything. And it the controls, the tint color of the controls is just white. They like all the Apple Podcast controls are just white. And the reason why is if you are trying to layer a UI on top of a translucent content area with unpredictable content behind it, um, it's really hard to pick colors and contrasts and layouts and control designs that work with that um in in all cases. And I have just so far I have not yet found anything in Overcast that I thought was better when I put the episode artwork colors behind it. Sometimes I can I can get it to look great like in one test, but then I try it like on a bunch of other shows and there's always enough cases where it looks bad or doesn't provide enough contrast or just looks messy or dirty. Um I just never found a a version of that that I like. Um that being said, in the case of the wid get, there's no technical reason why I couldn't do that. There's just, I think, a a a taste reason why like I just don't like it. Um that being said, I do know that a lot of people would prefer it not to be bright orange. So that's also on my to-do list is like, you know, give more widget color options. And as for the left lining of the play button, that was mostly just pride. You know, when I I obviously I saw when Apple um you know did their did their widget I I obviously you know started with that as as inspiration, let's say I didn't copy it. It was inspiring. But I always try whenever I am adding a feature or a UI situation that other apps have already added, I try not to just directly copy them. And is that that is often to a fault. Like in some cases, I 'm just being contrarian. Like Apple's one button mice. You know, like at some point it's like, okay, the the solution here is to just copy what everyone else is doing, because that's what everybody expects. And it turns out it might even be a better design. But I'm just so stubborn that sometimes it takes me a long time to get there. So in that particular case, I think you are probably right that left aligning it is probably the move. In my defense, I designed that before iOS 26, where everything became left a ligned. Um, but yes, I probably should update it to be left aligned. And putting the the remaining time in there, yeah, there's no reason why I can't do that. I have a request for the audience. If you haven't uh checked out Overcast lately, you should. Um uh Marco has not been sitting on his hands over there. Not only did you build out your 48 Mac mini server to give us all transcripts, which is a remarkable feat. I just want to say. I mean, I I don't know anybody else that would have I I'm sure somebody maybe had the idea, but nobody would execute it the way you did. And that that is really impressive. But the app itself is just, you know, you've really done a lot of work on this app the last several years. And if you're listening, you haven't checked out Overcast lately, you should. And Marco's one of us. So, you know, support one of us. Thank you very much. It is an incredible app. And um uh one last question about your app, because we are now in a video killed the radio star world, it seems like, uh or v you know video is now going everywhere. How do you think about your app with the incre asing video pressures. I know me as a creator, like it is complicated and it's complicated for listeners. Like I have paid subscribers in Apple Podcasts message me at least once a day. Hey, where's my video? I just supported your show. And I have to tell them Apple Podcast subscriptions doesn't give me a way to deliver video to paying subscribers. So thanks for supporting the show. But you don't get video and it's just it's complicated. So how are you thinking? How are you feeling about the current landscape with video and all that? I it's also it's complicated. You know, I I gotta say I I I don't love it, but I also recognize like the so just like uh you know, bigger picture than smaller pictures. Bigger picture , it is not use ful to try to police how people use the term podcast. It it just isn't it doesn't matter. You know, the same way like we used to have arguments over what's a blog. And it doesn't matter what anybody says. People are going to use it, how they're going to use it. And podcast does not anymore, it it no longer means an audio show delivered via RSS feed. Podcast is now just a format. And like a con like not not to say a file format, but it's like a discussion format. So you i a people who like it wouldn't surprise me to see like you know people who like have like an in person event like in an auditorium with a stage, if they have four chairs up there and people are gonna sit there and talk to each other casually, they're gonna call that a podcast. Because podcast just means a f it just means a a a mostly unscripted conversation, kind of more casual, maybe a little long, you know, compared to the kind of videos that you normally make. Like, you know, the like you you're writing scripts and you know, doing like probably writing every word you say or at least most of them by doing research and doing stuff ahead of time, and you're making a polished, edited, you know, v highly produced final copy of something. Um that's not a podcast. That's like a video. Podcast no longer means audio. Podcast just means conversation. And so I have to recognize that's the reality of the world. That's never going to change. Rather, it's it's never going to go And that's just what the word means now. So to some degree, it it presents pretty big problems for Overcast because not only is Overcast not a video app, and I don't really intend for it to become one, but what people actually mean when they say video podcast, uh usually that they mean YouTube . And YouTube has no way for apps like Overcast to legally access its content and play it. Now, there is this new world developing. We're still in the early days of it. Apple Podcasts kind of work with some hosts to create this new HLS-based video standard. The downside is that unless the publishers explicitly add it to their feeds, which not many are doing, it's a totally private proprietary system that is not being exposed in public RSS feeds for apps like Overcast to use. So that means that the way most video podcasts are being published and made available today is not available to players like Overcast. So even if I wanted to build video support, it would support very few shows. Uh and and I think over time that number will go up, but I don't see it ever being most shows. So if I say offer video, I think I'm only going to set people up for disappointment. Uh when when inevitably they try to play the video in Overcast and oh, it only works for two of my five shows that I that I would expect it to work on, say. Um that's just a bad experience, and and I don't think that's a great product uh decision to make to go into that when it when it's just gonna disappoint everybody. So my plan for video and overcast so far is not to support it, to continue to not support video, to just be the best audio player I can be. And that gives me a lot of flexibility. Obviously things like transcripts and lining them up and everything, all of that becomes either impossible or much more complex once you add video. Um you might have problems like for instance, video doesn't really work great with dynamic ad insertion without doing some proprietary stuff right now. And and you know and that's part of the reason they they don't want to make it available to players like us, because it would be super easy to just skip the ads or or not load them not send the data they want because we're not creepy. Um so there's all sorts of reasons why I I think that world is going to remain mostly separate on a technical side. The downside is, you know, it obviously does create problems for perception and expectations. But I think again, I'm just going to focus on being the best audio app I can for now and over time I'll reevaluate. If the video landscape becomes more broadly accessible to public apps like Overcast. Aaron Powell for me, I don't script my videos or podcasts. So technically every video I make is a podcast, and every podcast is a video. So the only difference is microphone in the shot or microphone out of the shot. Yeah, right. That's the only thing. It's a prop to indicate podcast. Video. That's it. But I will say but two things. One, when Apple introduced video, there is an open standard that like you said, some hosts are supporting transistor.fm supports it, alternate enclosure, which lets you actually link to a video file as an alternate enclosure to an MP3. And third-party podcast apps could have supported that. Fountain is a podcast app where you can actually go and switch between video and audio from an RSS feed. Both files are in the RSS feed. And it's unfortunate that Apple didn't go the alternate enclosure route, but instead they went through the dynamic ad insertion purpose route. And that kind of I think a lot of hosts would have suddenly supported the alternate enclosure if Apple had gone that way, but they didn't. Uh but I the second thing is I'm actually optimistic for apps like Overcast because I think those who enjoy listening to podcasts listening specifically, which is still me. I still listen to podcasts primarily like ninety-nine percent of the time. I think having the best audio experience can now be a differentiator for Overcast. And the Apple Podcast experience for shows who have added video like my other show, it's actually gotten worse. Uh they no more chapters with video. The transcript doesn't follow along with the video. And these are things Apple might fix, but because it's now become a worst audio experience, it is very hard, like you said, like to do great both ways, great video and great audio. And so I think for those who are diehard podcast fans and it might not be the majority of podcast consumers, but I think it's going to be a large part and a passionate part of the podcast community, they're gonna love apps like Overcast because it's the best audio listening experience A.aron Powell That's what I'm hoping. And I think it will work out that way for so many practical reasons, production reasons, technical reasons, and business reasons. The reason why the transcript won't line up is because they're they're almost certainly going to use dynamic ad insertion on the audio version. But then that again, we know that messes up transcripts. And that and how and it's going to work differently in video because of the way they implemented things and everything. So that's that's going to be complex there. Um and there's all sorts of even just as you're producing the content, when you are producing an audio podcast, like you know, my podcasts that I make are all audio only. We don't do video. Uh and maybe maybe in the future that'll change. Um, but you know, we know people who do both. And you know, and typically the way like I will make so many little edits in audio that I would never make in video because it would make a weird jump cut or you would notice it in a weird way. And my style of editing for audio is if if I I think you're gonna notice an edit, I won't make it. I want all my edits to be like totally seamless in the audio. So it sounds like the conversation actually happened this way. Not that somebody jumped in too soon and interrupted somebody or not that somebody had to stop in the middle to cough or whatever. You know, those I I edit I make all these little tiny like micro edits as I go along. If if it's video, you really can't do that to that same degree. And you might make different decisions. So for instance, like you might have a different choice of headphones or microphones, or you might skip them entirely because they don't look good on video, but they sound worse then, or like the like the resulting product sounds worse. So there's like if you lean too hard into video, I think, both in the software and in the content part, it does actually sacrifice some of the quality of the audio product. Um I'm very happy to stay in the audio world for almost everything I do . I think that's that's still how I prefer to to listen to podcasts is an audio um the the vast majority of of the time. And I also like watching YouTube videos. And to me, those formats have not merged. Those are two different things that I watch in different contexts and I'm looking for different things out of them. All right. Uh moving on. We've got a new president coming in or a new CEO coming into Apple, John Turnus. Uh, what are you thinking some leading indicators when he starts that's gonna show he's on the right track? Um, it's hard to set a timeline expectation because the company is so big that if John Turnus wants to change certain things that I'm looking for, we're not going to see it anytime soon. kinds of changes, you know, over the next five years or something. Or you know, for for like, you know, practical or PR or power reasons, some things might require, you know, other executives to also retire a few years from now before they can change and things like that. But kind of broad strokes, what I what I'm looking for is I think Apple has gone a little bit astray in a few areas that not only are like you know little pet peeves of mine, but actually make their products worse. One of them is the relentless push into services has given them strong incentives to bother the user with upsells and ads and nags in in a lot of these system apps and and kind of the experiences in using their products that that I think make the products worse. One of those they're they're just launching soon or which or now, which is the ads in Apple Maps. Great example of that. They've recently increased the ad load in the App Store app for when you search for search ads, because they make a bunch of money from those and now they can make even more. And they've they've set up their their incentives such that for them to continue to increase profits, uh, they now have strong financial incentives to continue to ratchet up annoyances and upsells and ad load in various parts of their system. Meanwhile, their brand is premium. Uh they they are selling these products at premium prices to customers who want that, uh, who want premium experiences . And it that's fundamentally in conflict with loading up the experience with more and more ads. It also goes against many of their privacy goals. Uh and certainly it can get them into trouble with regulators around the world as they apply different rules to themselves as they would apply to others and exercise privileges that they have that others don't have. So the what I would like to see from Eternus era in that in that area is finding ways to grow revenue that don't sacrifice the quality or user experience of the products and don't put them in direct conflict with each other. Um, that would be great. And the more ads I see in the system and in the system apps, um the that's telling me they're going in the wrong direction on that. So maybe a course correction on that would be nice. And then the the other big area I I think would be um continuing to reduce the reliance on China and on TSMC, uh both of which I think are tenuous for various reasons and just not a great place to to be so reliant. Obviously that's not a small task, but that's going to be, you know, hopefully the Turnus era, we can start to see some diversification. You know, they have some now, but like they need a lot more. Um, so start to s start to see some of those shifts. And then finally , I would love to see Apple take their relationship with developers a lot more seriously than they do. They the the Apple under the outgoing leadership and under Steve Jobs as well, Apple's not good at negotiating. Apple's not good at power dynamics where they don't have all of the power. And that's not really negoti ation. Apple's really good at dictating terms. That's true across everything. Developers, their component suppliers, their manufacturing partners, governments. They're really good at dictating terms. But negotiation is needed when you don't have all of the power on the relationship. And in many of those areas, I think they really need to develop those skills because they have had pretty poor skills in that area go uh so far. Um but with developers , there is there's not as strong of a of a need to force them to be a little more humble and a little more warm in that relationship because they still have so much power over us. But that does harm the products. Look at we mentioned earlier the Vision Pro. Vision Pro has very little developer support. It also has very few install very small install base of customers. So why would anybody make software for it? If developers loved Apple more and were more excited about Apple's platforms, we would make software for platforms even if there wasn't as much of a financial incentive to do so, just because we use them ourselves, or we love them and we see maybe a bright future, or just we want it to exist. You know, back in the old days of the Mac before Apple was today's Apple, the Mac was not a big software platform by market share, but it had a lot of great software. Why? Because people who made great software loved the Mac and used it themselves and wanted to support it and wanted to app to exist themselves. And where Apple has gotten today is very different. Today, their incentives are all just to maximize their own money in the short term, but that sacrifices long term developer sentiment in a bunch of ways that I think are unnecessary . Um if Apple would loosen its iron grip a little tiny bit, I think it would cost them very little revenue percentage-wise and would dramatically improve the relationship with developers, which would dramatically increase the amount of good quality software that's made for their platforms, and would reduce a lot of the regulatory pressure they face around the world that is threatening much bigger harm and much more limitations to their to their hardware products. So ultimately I think I think the Tim Cook era was very short-sighted in how it treated developers. And I would like to see the the the John Turnus era start to realize the longer term picture and and the bigger picture of like if we treat developers better, if we value them more, and not just, you know, in in two weeks, they're gonna tell us how much they're gonna value us. They're just telling us. We want them to show us. And once they show us how much they value us, if that increases, that will be a sign that the turt the turtles era is going in the right direction. To quote Eliza, doolittle, don't give me words, show me. I I would add one more. I think they need a shake up in software. I think Apple software itself, you know, the UI thing of the last year is kind of a fiasco, but I I think just in general the software needs a new focus. I am most ly on board with that. Um I I do think I I think the I mean, I don't know like, you know, who who along the line would need to change or what goal would need to change along the line. Yeah, I'm not calling for people to get fired. I just think they 're But but I I do think uh I think the Craig Federic era is closer to the end than most people think it is. Um I could be wrong. That's just kind of a hunch I have. I don't have any inside info on that, but just kind of a hunch. Um and you know, be you know, between the the Allen Dye shakeup to um Stephen LeMay, you know, if last uh last winter and you know the now the CEO shakeup, I think we're gonna see a a lot of change in that area. But again, this is gonna take time to really see results from. So I hope to see results in that area. But the design choices are pretty spot I definitely want to see the next the 27 OS is obviously we definitely need a lot of help on the Mac. And even on iOS, I think anywhere that you see a blurred bar with text flowing under it, that's a design failure. Whenever you have blurry text, something is wrong. And I would love to see them reconsider parts of the liquid glass design language that require that to be the trade-off. Um because that's just bad design. Text should always be clear. Aaron Powell Uh my litmus test will be when uh John Turner is the CEO, this will know that uh they're looking more for a premium user experience if they take the badge off the settings app for Apple Care. If that goes away, services revenue good sign. That will be the good sign that we're we're in the right direction there. Right. But again, like they they've set it up that they 've created a strong financial incentive to push those add-ons, push the services, push the upsells to everyone possible. That's where you get push notifications now for everything, every possible upsell. You get Apple opting themselves in to certain things that other people have to ask permission for. And you get things like that big red badge to tell you to finish setting up your Mac, you need to give us more money. So Marco, we've gone a little long, but I just wanted to ask any new apps or services that are bringing you joy and delight these days people may want to check out? Uh I mean this is gonna be the boring answer, but for me it's clawed code. Uh it's really it's it's dramatically improving my life in in ways that I did not think were possible. Um that in ways that I'm also just not I'm not seeing any other apps um besides that that really stand out to that degree. I mean, there's lots of great stuff being made, but that's that's the big one. And it's also allowing me, like one of my favorite new apps that I'm using constantly is what is now currently called Reminder, my stupid task management app because I was able to make exactly what I wanted. And I made it largely with the help of Claude Code at first. And and so that to me, like I think we're entering we're in, we're already in an era where it's so easy to make custom software that's exactly what you want it to be, even if you're the only user of it. So I think a lot of the benefit and joy that people are going to get out of computers over the next few years is just stuff they made themselves to it to scratch exactly their itch. I can absolutely attest. I'm teaching a course on artificial intelligence now, largely focused on Claude Co-work, which is more user-friendly for non-developers. And we have a meetup every Monday, we call it the Builders Club, and senior citizens are writing apps to track the political candidates in Florida. And like just every person that gets on camera, none of these people have a lick of knowledge about programming anything, are suddenly creating custom solutions to their compute problems, things that we've struggled with for years. I think the way we use computers is about to significantly change. And um and this really shows the way. I mean, I absol Iutely agree.. Yeah You know, i when you look at like in the past, people people were able to harness the power of computers either if they were a programmer or there were tools that that caught on that gave people some version of that. Like the spreadsheet is a great example. You know, the spreadsheet brought some power of the computer to a lot of people, and they were able to learn that more easily than programming. And they were able to get a lot of work done and harnessed some of that immense power that computers have But they still needed a little bit of knowledge and it was and it was still kind of limited to what spreadsheets could be wedged into doing. I think now the the AI coding era is almost in that same vein. It's almost like the spreadsheet of like, we're now giving people another way to harness the amazing power of their computer without them having to be programmers. And that's very empowering. Think about all the immense value and independence that was inspired that was enabled by things like spreadsheets and databases. To have that now for lots of custom software, that's where that's where we are now and and this is the worst it's ever gonna be. So it's gonna that's gonna get better over time too. I think the the area of like custom bespoke apps made for just you or just your business is gonna end up being very similar to custom-made spreadsheets that you made for just you or your business. It's going to empower all sorts of things for all sorts of people. And we're hurtling toward the Star Trek moment where you just yell out something and you don't think about apps. The computer figures this stuff out and gives you the answer. And computers are so much more forgiving now. Like you know, with AI, it's it's very forgiving of input. So you can you know w whether you're doing it via voice or or text or or image, you can just kind of just slur your answers. You can repeat yourself. You can stutter. You can you can not know the terms for something. Like I had to I had to replace um a charger for an e-bike at the beach. And but I wasn't at the beach. I was I was, you know and I was like, I have an orange e-bike from a company that went out of business a few years ago after they pivoted to a motorcycle and then died. I need a new charger for it. What is it? And it got it. First try, it's like the company was Sondors. Here's what happened, and here's the charger you need. Search for this thing on Amazon or use this brand name. I was like, Yeah. That's amazing. Like I gave it the worst information possible, and it figured it out perfectly and gave me the answer I needed. I don't even fix typos anymore. Like if I'm typing the clot or whatever and there's a bunch of typos like, ah, whatever, it's gonna figure it out. Yeah. Like that's one of the best things about AI is it is so much better at dealing with fuzzy, impressive Yeah. My last question for you, as the creator of InstaPaper, what are you using right now as your Read It Later service, save saving links type service or app? You're gonna hate this answer. I'm not using one. I mostly don't read stuff on the web anymore. Well, but like just saving links, whether it's like book, like a video or a podcast episode, are you saving stuff anywhere? I mean, if it's podcast, I'm saving it in Overcast. Sure. If it's video, I'm saving it to my YouTube Watch Later playlist. Sure. Um and I I sometimes even watch the things on that on that playlist. What about when you're saving links for ATP, like something you want to talk about on the show? Uh I keep the tab open. For the Google Doc or whatever? Yeah, or I'll put it in the doc if it's if I want to share it with Casey and John, or if I want to just surprise them with it, I'll just keep the tab open or I'll make I might put it in an Apple Note . Okay. It's terrible. You should vibe code an app for that. Same stuff. Everybody already has. That's the thing. Well, yeah, that's true. That's the kind of thing people are just doing themselves now. Oh yeah. Like this is like part of why I like I might never release my reminder app because why? Who else is going to use it? I'm making an app that's perfect for me. Is anyone else ever going to want it? Maybe? Maybe not. Maybe they'll want to vibe code their own. Yeah, I mean in my in my little AI community, uh two guys in there are making their own recipe tra ckers and they're sharing notes about and one of them said, Why would anybody make a recipe app now ? Right. You know, why would you? When it knows you personally, it knows your kids' dietary preferences, you can vibe coat it up in an afternoon. Why would anybody make and sell one? Yeah. It it's just a weird time. Yeah. It's it's gonna radically reshape so many markets. It's it's yeah, it's agree. Margo, so great to have you back on the show. I'm in it, gang. Go check out Overcast. It's the best. Hopefully you're listening to this show on it. If not, maybe listen to the next episode on it. Also, ATP, we've been mentioned, but we never really called it out. Accidental tech podcast, one of the premier Apple related tech shows. Uh Steven's wearing his shirt today. Nice. Um and then of course uh you can always go get have Marco make you a burger if you're in Fire Island. I don't make the burgers. Thanks. And we're gonna go talk to Marco about 3D printing real quick and more power users. You could sign up at the link below to become a member. Thank you to our sponsors, NetSuite, Mercury Weather, and OnePassword, and we'll see you next week.