00:00:06 ◼ ► You got a lot going on. Do you want to talk about the Vision Pro stuff first, and then we'll get to Hovercraft?
00:00:11 ◼ ► Yeah, just get right into it. You know, there's no warm-up, huh? You've got a lot going on, too.
00:00:33 ◼ ► Yes, I've got a lot going on. Sandwich Vision was a company that I set up to start doing Vision Pro stuff.
00:00:40 ◼ ► The first thing I did was television, which is a cute little novelty app where you can put television in your space, in your immersive space.
00:00:51 ◼ ► I made that with a co-developer, Andy Roth, who is a seasoned Apple developer. He works on Aura Ring full-time.
00:01:05 ◼ ► And then it was about what comes next. This device is obviously so good for entertainment, maybe primarily for entertainment, for certainly a lot of people.
00:01:19 ◼ ► It's not quite the office workhorse that people maybe assumed that it would be, being an Apple product.
00:01:26 ◼ ► But since entertainment was such a strong direction to go in, we started thinking, like, what's another screen?
00:01:41 ◼ ► And that was the dream, is like, let's make people nice movie theater spaces they can sit in and then watch the movies that they want to watch.
00:01:50 ◼ ► The dream was honestly to, like, partner, almost be a movie theater, like, partner with distributors, get access to films, first-run films.
00:02:07 ◼ ► I mean, it's interesting that the content is going in the other direction, that immersive content is going into the legacy movie theater spaces rather than the other way around.
00:02:20 ◼ ► But I think it'll just be a little bit bi-directional until it's got a little bit more parody.
00:02:26 ◼ ► But then I brought my good friend Jose Marquez in to really run the Sandwich Vision business side.
00:02:32 ◼ ► And he had this epiphany that we shouldn't just be showing movies, traditional movies, in our virtual movie theaters.
00:03:04 ◼ ► Even before the Blackmagic immersive camera, the Apple immersive video camera that Blackmagic built existed, we captured your live talk show in 2025 and live streamed it through our app.
00:03:43 ◼ ► The epiphany that Jose had was that there are these giant screens in science museums around the world and a really dedicated, passionate community of filmmakers making films for these screens.
00:03:58 ◼ ► What you would think of as like planetarium kind of stuff, but even like sometimes less sciencey, more artful and that these filmmakers are really hungry for more screens and more scale and like more access to audiences.
00:04:11 ◼ ► So he started identifying all these top tier immersive filmmakers and partnering with them to be able to distribute their films in theater and sell tickets to them.
00:04:23 ◼ ► And now, at this point, we have the largest library of these films to the point that creators want to come to us to distribute their films for the format.
00:04:38 ◼ ► The next phase of that was like that there are now a lot of Apple immersive video filmmakers that are making great stuff.
00:04:49 ◼ ► But we do and maybe like 50 other people out there that we know own them and are making cool stuff and they want to distribute their work.
00:04:57 ◼ ► And so Jose is out there meeting everyone he can in that community and putting Apple immersive video.
00:05:20 ◼ ► Apple is aware of us and we love being able to feature that branded immersive format in our app.
00:05:58 ◼ ► I think before the device came out, I had this feeling that it was going to try to, it was sort of like rewire how we think about interacting with digital stuff.
00:06:34 ◼ ► I think if it had had more penetration, then you would be seeing a lot more software inside and outside of the device like that feels like Vision OS.
00:06:46 ◼ ► It's obvious that Apple wants this to happen in their OS 26 line of everything that they're doing.
00:06:54 ◼ ► But because I fell in love with this, I started and I think it was in the context of being like screen sharing and presenting on a Zoom call or whatever.
00:07:04 ◼ ► I started thinking, why can't I just pick up a window and move it around like I do in Vision OS?
00:07:13 ◼ ► So about a year and a half ago, I started working around with early AI coding tools, like mostly in Cursor and the earlier models.
00:07:23 ◼ ► And seeing if I could take a crack at building a Mac app and I hit a wall so freaking fast.
00:07:29 ◼ ► I hit, I like, as soon as I got up against OS level stuff, like the core media stuff, I could not solve that problem.
00:07:36 ◼ ► I could not make a camera extension and get it working as a virtual camera in any capacity.
00:07:52 ◼ ► And then for six months now, I've been AI building, I've been building software with AI in other ways, building tools for my company for Sandwich and built the backend for theater, the CMS that runs it and all this other stuff, like way bigger stuff than I ever thought I could do as a non-coder.
00:08:16 ◼ ► And so a few weeks ago, I decided I needed a palette cleanser because my head was in hyper AI space all the time, just like scale and intelligence and models and tokens.
00:08:46 ◼ ► And then it was just a matter of using all the tools that I've been using to make it better and polish it and work with AI in the way that I love to work with AI now, which is as a creative partner.
00:09:25 ◼ ► But you, we were in a, we're in a group chat hosted by our mutual friend, Brent Simmons.
00:09:32 ◼ ► And you piped in, chimed in, I don't know, two weeks ago, 10 days ago, like before hovercraft came out and, and we're like, Hey, you guys, you want to try this thing that I built?
00:09:48 ◼ ► It's that I, I made a very incorrect assumption that it was a more ambitious, sprawling app than it is.
00:09:58 ◼ ► And that didn't mean that I didn't want to try it, but it meant that I thought, well, I need, I need to try this when my plate is clear.
00:10:06 ◼ ► You don't, you don't try a, a significant major sprawling new app with a couple of minutes on your hands.
00:10:30 ◼ ► Cause I had some, some quick thoughts, not you, you had notes, but, and I felt bad because you asked and you are, you are my friend.
00:10:50 ◼ ► And I, well, let me caveat that everybody should, every software maker should welcome John Gruber's thoughts on their software.
00:11:14 ◼ ► And then I asked the group again in our, in our Slack and Alan Pike offered, like, I'll do, I'll review it.
00:11:23 ◼ ► He does this thing where he screen records himself using an app onboarding for the first time.
00:11:35 ◼ ► And because he does it with a smile on his face, it's not like, it's not mean spirited at all.
00:11:50 ◼ ► And that's the perspective that we, that who are maybe more, a little bit more technically minded or forgiving of rough edges.
00:12:32 ◼ ► I told, I told Jose, my theater business partner, it was going on and he was like, he slacked like, that's some Shaolin master shit going on right now.
00:12:47 ◼ ► So I just did all your notes in the night and I sent it, I sent the bill to you that night, which was great.
00:13:02 ◼ ► It's, I am behind on the whole vibe coding thing and I can't quite explain it and it's not for lack of interest.
00:13:11 ◼ ► And I know that AI in general, for some number of people is polarizing where they're on both sides, that there are some people who have clearly lost their minds going into it.
00:13:28 ◼ ► But I do think that they are missing, missing a bit of the forest for some of the trees and where the, the, some of the trees are the reasonable objections to some of the aspects of it.
00:13:44 ◼ ► It feels like some of the knee jerk resistance is eroding a bit, but, you know, and I don't think anybody who reads Daring Fireball or listens,
00:13:58 ◼ ► I mean, I think if anything, it's a better criticism that I don't write or talk about it enough.
00:14:21 ◼ ► I don't assume everybody subscribes to Dithering, but my general internal sense of what I'm talking and writing about is sort of all of that combined.
00:14:40 ◼ ► But I have so many friends, and, you know, Brent and others who are developers, and Alan's one of them, right?
00:14:56 ◼ ► But he was one of my first developer friends who, AI trusts Alan's opinion very, very much.
00:15:07 ◼ ► But his enthusiasm for AI, for code generation, was so off the charts that you couldn't help notice it.
00:15:23 ◼ ► Yeah, he's like me, is that he has an agency, and his team runs the agency for him, so he can go off and build AI stuff.
00:15:51 ◼ ► It's not just, this is why I wanted you on the show, but I'm not saying that because it is, because you are on the show.
00:15:58 ◼ ► I just, I mean, I've known you and I have been friends for a very long time, and I know you've never been a programmer.
00:16:03 ◼ ► You said that when you were describing it, and when you first had the idea of trying to build it with AI coding tools 18 months ago or so, you couldn't even get it done.
00:16:13 ◼ ► Whatever mix of actual programming you can do versus where the tools were, you couldn't get it done.
00:16:18 ◼ ► And you didn't bring in any of you, you know, as you mentioned, you've, with previous things that you've built through Sandwich Vision with television and the theater app, you've had, you've brought in developer partners.
00:16:34 ◼ ► And I tried to learn to code Objective-C at the same time Cameron did, and I flopped majorly, and he took off.
00:16:45 ◼ ► Right, Birdhouse was, for people who don't remember, it was in the very early days of Twitter, which are, Twitter came out right before the iPhone, like 2006, and then the iPhone comes out in 2007, and then the App Store comes out in 2008.
00:16:59 ◼ ► And Twitter's exploding, but it was mostly, it was so, nothing like the Twitter of today.
00:17:05 ◼ ► Yeah, and it was that era when things like using at the person's username was just a convention.
00:17:19 ◼ ► It was just a convention that we had to sort of say, hey, I am directing this comment at my friend whose name is at Lonely Sandwich.
00:17:40 ◼ ► The late Dean Allen had for a year, I would say at the peak of this era, what was the name of the site?
00:18:02 ◼ ► It was just sort of a draft repository for material that you were going to post to Twitter, which makes no sense in today's world.
00:18:19 ◼ ► And there were a lot of times where you'd have, it was both that you wanted to make it a sharp joke, and a lot of times, it was 140 character limit at the time.
00:18:58 ◼ ► And then the funny part, too, I'll just say, let me just say, because I think it's important in history, is, I don't even know, what the hell were you doing at the time?
00:19:27 ◼ ► And what happened is, and this is very similar to the business model for Daring Fireball Story, where it was, like, I was trying to sell T-shirts and memberships, and the memberships included access to an RSS feed.
00:19:39 ◼ ► And then Google Reader came out and didn't support the authentication for the RSS feed, but everybody wanted to use Google Reader.
00:19:46 ◼ ► And I thought, well, why don't I just sell weekly sponsorships for the RSS feed, and then I'll make it free for everybody.
00:19:55 ◼ ► And you came out with Birdhouse, and you made a very clever movie, little video, to promote it with you as the pitch man for the product.
00:20:07 ◼ ► And as beloved as Birdhouse was as an app, it wasn't really long for this world or a good business.
00:20:14 ◼ ► But Adam Lisagore-style promotional intro videos to new tech products, often hosted by you, Adam, a delightful screen presence, that was actually a very good product idea.
00:20:43 ◼ ► And Eagle Eyes, a couple of Eagle Eyes, Greg Noss noticed that I was, for this Hovercraft demo video that I put on the landing page, it's the same chair that I'm sitting on, as I sat in the original Birdhouse video.
00:21:05 ◼ ► I think what's notable about this time is, because this is when you and I became friends, and I remember that you were working on Vesper, and you were obviously around software a lot.
00:21:23 ◼ ► Well, I just remember getting into these conversations, these thoughtful conversations with you about the aesthetic and taste matters of software for the first time that I obsessed about.
00:21:35 ◼ ► And I remember that when you did work on Vesper, you called yourself the director of it, I think.
00:21:41 ◼ ► And that notion that you can be the director of software, it's sort of like the arbiter of taste for software.
00:21:47 ◼ ► Not necessarily a designer, because you're not in there maybe with, you're not in Photoshop or Sketch at the time.
00:22:01 ◼ ► And at this point, when people ask, sometimes I describe what I do as creative directing software using other software.
00:22:09 ◼ ► Because working with AI coding tools to make software and then build it and ship it is very much like creative directing.
00:22:22 ◼ ► And the skill set required to be a creative director for little films or commercials or whatever is very much like the skill set required to do it well with AI coding tools.
00:22:40 ◼ ► But I think that in my community, the creative community, it is underappreciated that so many people have that skill set and don't realize that this can be their canvas.
00:22:51 ◼ ► I'm out there banging the drum that software, making your own tools, can be the next creative frontier for you if you're looking for one.
00:23:04 ◼ ► So, like, given that you're sort of, you're calling yourself sort of new to the AI coding realm, I don't know.
00:23:11 ◼ ► Like, yeah, I would love to show you or talk about what is successful, even just like very basic level mindset things that you, that maybe you don't know stepping into it for the first time.
00:23:25 ◼ ► And it's stuff that maybe sounds obvious, but I just think, like, people who have gotten good at it, like probably Alan, they know that there are these, like, sort of unexpected ways to work with the tools that result in a better outcome.
00:24:40 ◼ ► Like, things like sometimes you get the initial tracking number before it's even in the carrier's hands.
00:24:47 ◼ ► It's like you buy a shirt, and they create the tracking number, and you go and check the status.
00:25:12 ◼ ► In addition to just tracking numbers from carriers, and, of course, really smart, like if you just copy the tracking number and you go to Parcel, it reads your clipboard and takes a guess based on the format of the tracking number, who it is.
00:25:30 ◼ ► They also have Amazon integration because Amazon ships a lot of this stuff itself these days, and they don't give you a tracking number like the other guys do.
00:25:41 ◼ ► And so what you do with Amazon is you go into Parcel settings, you sign into Amazon, just sign in in the settings, and then every three months you have to re-up just for security reasons.
00:25:56 ◼ ► But then everything that Amazon ships themselves, you get tracking notifications for that too.
00:26:00 ◼ ► So on your iPhone, on your Mac, it is a great, real, native Mac-asked Mac app with widgets.
00:26:16 ◼ ► I just have a widget there, and every package that's coming to me is there, and it's one of the things I check in the morning.
00:26:37 ◼ ► So if you get like a piece of paper with a tracking number on it, you can just point your camera at it and get it from there instead of typing 20 digits by hand.
00:27:05 ◼ ► You get a premium subscription with all features for just $6.99 per year through the App Store.
00:27:23 ◼ ► Oh, by the way, also, one of my very favorite features, I think it's off by default, but I think it should be on by default.
00:27:39 ◼ ► But if you're like me and you really – I worry when it's the Postal Service because who knows?
00:28:27 ◼ ► Now, I don't know if that's the same Yamazaki that makes the very good Japanese whiskey.
00:28:31 ◼ ► I love the way that there's a couple of these Japanese companies that make everything from, like, heavy machinery for construction to fine whiskey to cookies.
00:28:52 ◼ ► And if you think about it from, like, a 20th century perspective, well, of course it makes sense that wherever the hell they were making these Oreos in North America, what sense would it make to put them on a boat and ship them over to Japan when you could just pay somebody to make Oreos in Japan without – and, you know, and then they're right there.
00:29:11 ◼ ► And I guess the gist of it is that people who knew, people who visited Japan, I have never – Japan is probably number one on my list of countries in the world that I have never visited that I cannot wait to visit as soon as possible.
00:29:33 ◼ ► And it turns out every – apparently back in the day when Yamazaki was making Nabisco Oreos in Japan, they were – people who knew would say these are better Oreos than the ones you get in America.
00:29:45 ◼ ► Then the fucking bean counters step in and somebody looks at what they're paying and says, well, we could just pay some company in China to make these, ship them across the strait, and we'll save, I don't know, a couple of cents per package.
00:29:57 ◼ ► And so they told Yamazaki to get lost and started making the Oreos for Japan in China, and everybody noticed right away.
00:30:13 ◼ ► They said, well, we know how to make really good, delicious chocolate wafer sandwich cookies.
00:30:18 ◼ ► We're going to make them ourselves, and we're going to make them even better than we did when we were making them Oreo branded, and they came out with Noir, which are exactly Oreo-sized.
00:30:27 ◼ ► If you didn't know any better, somebody who wasn't paying attention, you just put them out in a dish, you would think they were Oreos.
00:30:34 ◼ ► And somebody – after I posted about this story, a couple of friends said, hey, I could get you some.
00:30:47 ◼ ► And they might be better than – I've long been a proponent of the Numino's from Paul Newman, which are excellent, excellent Oreo-style cookies that taste far better, made with much better ingredients.
00:30:58 ◼ ► Like, you look at the ingredient list, and it's like cocoa and cane sugar and things that you know, and you look at the ingredient list for Nabisco Oreos here, and it's – I don't know.
00:31:24 ◼ ► If you were hungry and you really just really wanted calories, you'd be better off with Numino's.
00:31:42 ◼ ► There's obviously sugar in them, probably more sugar than you should eat, but just like most things that are sweet, you don't need a lot of sugar.
00:31:55 ◼ ► There is less white filling in a Noir cookie than a Numino or an Oreo, and I think that is correct.
00:32:03 ◼ ► And theirs is – I don't have a strong opinion on it, and if I peel any kind of Oreo-style cookie and do the thing where you pop the two pieces apart, and of course you try to get all the cream on one side and empty one on the other, I actually prefer the side without the cream.
00:32:38 ◼ ► I think that says a lot about a person's personality because I want double stuff but double stuff.
00:32:46 ◼ ► It's like take two double – take two double stuffs, pop the lids off, and then put two of the cream sides together to make a quadruple stuff.
00:33:10 ◼ ► We're going to order from three different levels of pizza place and then do a Pepsi challenge.
00:33:25 ◼ ► I haven't had a Hydrox in a long time, and every time this comes up, people remind me about Hydrox, and apparently Hydrox is the original.
00:33:31 ◼ ► The Oreo was a knockoff of Hydrox originally, which I do find hard to believe from my childhood memory of the brand, but I have to admit I haven't had one in a while.
00:34:05 ◼ ► I have fondness for Oreos just because we watch old commercials in our house a lot and got that ad campaign in the 80s that the jingle, oh, oh, oh, oh, who's with the Oreo cookie?
00:34:24 ◼ ► And it's such a good, catchy jingle in an era when every commercial was a great, catchy jingle.
00:34:45 ◼ ► I never looked it up, but the fact that if you spelled the name aloud and you kind of pause, you get O-R-E-O, and R-E-O sounds like Oreo.
00:35:05 ◼ ► If you listen to the way they did it, they clearly coached or, as you would say, directed the people singing the jingle to emphasize spelling it, saying the letters R-E-O in a way that sounds the most like Oreo.
00:35:38 ◼ ► Apparently, and other people have told me that if you have a local, and you in Los Angeles surely have.
00:36:01 ◼ ► But they are truly excellent, and I expected nothing less, but they even exceeded my expectations.
00:36:22 ◼ ► Okay, so here's an interesting subtopic, is that a lot of our friends who are Apple-adjacent developers maybe would be the later ones to sort of consider using.
00:36:38 ◼ ► Yeah, there's a very helpful developer ecosystem and tools for Apple development, but also this sense that the coding tools that are really good at web coding, at like React, Next.js kind of code, are not going to be able to nail anything Swift, anything that Xcode can do.
00:37:01 ◼ ► Or you might want to just like embed a coding agent into Xcode, but even then, it's so fussy that it's going to get a lot wrong.
00:37:09 ◼ ► But that was, what I found is that when making a Mac app is that it is fussy in a lot of different ways, but as long as you have time and patience, you're going to push through and solve a lot of problems.
00:37:27 ◼ ► You're using a lesser model and it's going to take a lot of guesses and it's going to cause a lot of headaches and get it wrong.
00:37:43 ◼ ► I've always been a Cursor fan or probably going on two years now I've been using Cursor.
00:37:49 ◼ ► Because when I was working with a couple of engineers trying to build a platform pre the AI coding era, I had to just by working with this team, if you ever want to like install a dev build yourself, you have to know your way around an IDE somewhat.
00:38:09 ◼ ► You have to type in a terminal, you have to know a little bit about adding keys to a local ENV file, just so you can authenticate with your APIs.
00:38:30 ◼ ► And especially if you get into like building for iOS, like with Electron or something, or Expo rather, like building React apps for iOS, you have to know a lot of technical stuff that's just mind numbing and gross.
00:38:53 ◼ ► And then when Cursor offered that same VS code style of tool with an AI agent on the side of it, and then when it got to the point where the AI agent was actually able to agent stuff and do stuff to your code, write code for you, even talk to GitHub,
00:39:10 ◼ ► even do commits and pushes for you, even like use CLIs to get into different third party services, work with your Vercel, like re-query your Supabase, Postgres database.
00:39:24 ◼ ► When the AI agent was able to do all that stuff, then it became really, really powerful.
00:39:29 ◼ ► Because that's pre that, before that, you were just chatting with ChatGPT and it was telling you kind of like what to do without any visibility into your own user experience or knowing the tools you're using.
00:39:40 ◼ ► And then it's just taking a lot of blind guesses and you're getting it wrong and it's so frustrating.
00:39:57 ◼ ► And so therefore, the sort of things that I am working on where it's like a single page script, it works.
00:40:04 ◼ ► That's just going back and forth with ChatGPT scales to the level of the coding that I have asked for help with.
00:40:17 ◼ ► But that's why I've never even tried making an entire app because I can see how it wouldn't.
00:40:25 ◼ ► But there's like kind of an unlock that happens when you're using something like cursor in your file system.
00:40:31 ◼ ► Because what you realize is that all of these like really complex Apple scripty type of tasks that you might do that you might spend a lot of time building a script around.
00:40:42 ◼ ► You can just give cursor access to that folder in your file system and it'll be like, got it.
00:40:49 ◼ ► And you'll say like, please organize this completely chaotic three-year deep folder of project files for like a domestic thing that, you know, like or a financial thing or something.
00:41:11 ◼ ► And for somebody who's like kind of endlessly curious about how stuff works, you can read every one of those lines of its thinking process.
00:41:36 ◼ ► And so the way to sort of learn by using these tools a little bit around the periphery of how code works is to really just read every line that it thinks and every instruction that it does.
00:41:49 ◼ ► And you don't necessarily have to read the code that it's writing because that's just like that's a foreign language.
00:41:58 ◼ ► But the beautiful thing about a model like Claude is that it communicates like a human as it's doing the code work.
00:42:10 ◼ ► And if you're somebody who likes to direct or creative direct, that's the name of the game is you're talking through the whole process of work.
00:42:20 ◼ ► And you know that the better you communicate with whoever or whatever you're collaborating with, the better the outcome is going to be.
00:42:28 ◼ ► So in its ideal state, working with a coding agent is this wonderful, fluid, creative, communicative process that feels like working with the best creative partner.
00:42:39 ◼ ► Now, I have to admit, I know I've, of course, heard of Cursor, but I've never tried it.
00:43:12 ◼ ► So, like, when I started power using Cursor, when I started just working solo, it was when Sonnet 4 was – Sonnet and Opus 4 were the latest models.
00:43:53 ◼ ► But then, as I started building more seriously, and then Sonnet 4.5 was the first one where it felt like, oh, wow, it's able to solve a lot of problems in advance that it didn't used to be able to.
00:44:05 ◼ ► Like, hiring a really good engineer who has the experience to identify problems that will come and solve for them proactively instead of, like, oops, something happened.
00:44:22 ◼ ► So when Opus 4.5 came around, and I realized that the money that I would spend on Opus tokens would more than make up for the time solving problems with Sonnet, then I just started going pedal to the metal on Opus tokens and spending way too much money building this stuff just because I needed to build a lot.
00:44:45 ◼ ► And so now I'm basically like Opus 4.7 all the time, which – so my kids are not going to go to college.
00:45:00 ◼ ► I mean, and I know that this is more than just – this is more than just for Hovercraft, though.
00:45:05 ◼ ► You're using this to build, as you said at the outset of the show, a lot of other internal tools for Sandwich.
00:45:14 ◼ ► I'm sort of building like a big ecosystem of platforms for collaborative work across a lot of different verticals with embedded AI intelligence.
00:45:38 ◼ ► But it's more than the – it's more than the $20 most people are used to spending on chat, GPT, you know, pro or yeah, or whatever.
00:45:46 ◼ ► Like, you really have to get – you have to get acclimated to that idea of, like, you are hiring an AI and you're paying sort of labor wages for it.
00:45:56 ◼ ► But that's of the scale of what I have heard from a lot of AI-forward companies that want their engineers to be spending tokens and have a token budget per day and that $200 a day in tokens is like, hey, that's like a productive day.
00:46:19 ◼ ► And that they – you know, but it just – at this point, it is, though, the orders of magnitude matter, right?
00:46:27 ◼ ► Like, normal people – well, frankly, the mass market people don't want to spend anything on software, period.
00:46:33 ◼ ► And, you know, and that's why ChatGPT and everybody is going towards building an ad business as they want to scale whatever.
00:46:48 ◼ ► But I pay $20 a month for ChatGPT and I don't think – I've never gotten a warning that I'm close to running out.
00:46:57 ◼ ► But I think I don't really use – I think that I use it like a couple times a day for things to ask questions and get answers and I don't really have a seamless workflow for it.
00:47:07 ◼ ► But the longer an article that I write is, the more likely it is that I'll send it through ChatGPT for a proofreading pass first.
00:47:18 ◼ ► But I – you know, I've got a good little prompt that it gives me – it only tells me the type of shit that I want to hear from ChatGPT.
00:47:32 ◼ ► I do use it for programming, but it's just the sort of little baby scripts that I tend to write, like little one-page scripts.
00:47:47 ◼ ► Do you remember – do you ever hear of the app Yojimbo from Bare Bones, the people who make BBA?
00:48:31 ◼ ► And even before, I was sort of like, eh, I think we're in trouble because we're not going to –
00:48:36 ◼ ► We really needed to get a Mac client for Vesper, and we kind of ran out of time before we had both
00:48:49 ◼ ► And maybe if we tried it again today, we would have been able to move faster with a small team
00:49:01 ◼ ► And I was talking to a friend about this, but I said the way that I change Notes apps or bookmarking systems
00:49:28 ◼ ► But then every once in a while, I would fall in love with a Notes app, and I would use it,
00:49:32 ◼ ► or a bookmarking system, and use it for years and have hundreds or thousands of notes or thousands of bookmarks.
00:50:01 ◼ ► But for the most part, I just leave the old family behind and move to a new city and start a new family.
00:50:48 ◼ ► A problem with Yojimbo is it's a Mac app that doesn't have a proper iOS or iPad client.
00:50:57 ◼ ► But what I wanted to do is write a script that would go through all 9,500 items and chart them by year.
00:51:06 ◼ ► And it came out exactly like I thought, which was to say, lots of Yojimbo came out in 2006.
00:51:16 ◼ ► But I started using it in 2005 because I was a former Bare Bones employee and friends with the company.
00:51:28 ◼ ► And then by 2010, my usage started declining because guess what I had in my pocket all the time by 2010?
00:51:38 ◼ ► And then as the 2010s turned to 2020, every year I had fewer and fewer new Yojimbo items.
00:52:00 ◼ ► And it was because I forget all the shit that I used to know about how dates work in Apple script.
00:52:07 ◼ ► And it's like ChatGPT just gave me, it took as long as I always used to think it should take to write an Apple script like that.
00:52:20 ◼ ► I think what you might try is just like, because you've already done that task and you know the old way of doing it,
00:52:31 ◼ ► And it'll say, do you want me to write this script and put it on your desktop so you can run the script?
00:53:03 ◼ ► And I asked him to show me how Obsidian, how you can power use Obsidian, because I know he does everything with it.
00:53:21 ◼ ► And what I did was I built an MCP that accesses all three of them that sits locally on my computer.
00:53:28 ◼ ► And then my platform, my intelligence platform, accesses that MCP so I can chat with any of those notes, those various note systems.
00:53:41 ◼ ► It'll actually just sync all those notes and put them in my database for me so that I can use it remotely as well.
00:54:08 ◼ ► This is the kind of stuff that you realize, oh, I could just roll my own version of whatever that is.
00:54:42 ◼ ► So I'm on a different machine than the one where I made the, and then I generated the numbers from AppleScript, put them into numbers, and I made a little chart so I could chart them.
00:54:58 ◼ ► But I was in numbers, and I copied the chart as a PDF, and what I would like to do is paste it into Hovercraft.
00:55:11 ◼ ► But instead, what I think I need to do is I need to go to preview and make a new image and then save it to the desktop and then drag it over to Hovercraft to share it with you.
00:55:32 ◼ ► So I think that when there's an image on the pasteboard, or like a PDF, like a PDF or an image, I should be able to – Hovercraft should be able to –
00:56:00 ◼ ► That's my Contacts app because it's – and it uses the system Contacts database that Apple provides an API for.
00:56:06 ◼ ► It's a hundred times – however much better Fantastical is than Apple Calendar, CardHop is ten times more better than Contacts.
00:56:38 ◼ ► For example, like one of the companions in this – it's like Apple Notes, like these sort of staple apps, right?
00:56:44 ◼ ► Like you've got a Notes app built into the system, a calendar app built into the system.
00:56:52 ◼ ► I'm not a huge Reminders guy, but I can see how they've kind of made it into a really nice to-do app that a lot of people use as their main thing to track to-dos and tasks and the way that it integrates with your calendar.
00:57:15 ◼ ► Contacts are a thing that every OS should be aware of because the people in your life are like the crucial substrate on a hierarchical level.
00:57:31 ◼ ► Maybe that's because they want it to remain a primitive and they don't want to build out a whole beautiful UX around it for whatever reason.
00:57:41 ◼ ► Anyway, the thing that you can do – I'll just finish this – is let's say you want to add a picture to a contact.
00:57:48 ◼ ► Instead of having them be a generic icon, you would like to have the face of a friend and it has them prefilled.
00:58:18 ◼ ► But anyway, ChatGPT wrote that script almost in no – or helped me write the script in no time.
00:58:24 ◼ ► The pasting thing, CardHop, if you have an image on the pasteboard, lets you paste an image onto a contact.
00:58:42 ◼ ► This interruption here seems like as good a time as any for me to thank our next sponsor.
00:59:02 ◼ ► The biggest problem they face is because critical institutional knowledge isn't documented, or if it is documented, no one can find it.
00:59:10 ◼ ► And creating new documentation manually is too painful, so it doesn't get done, even though everyone wants it to be done.
00:59:17 ◼ ► Without good documentation, teammates work inconsistently, shared tools don't get adopted by everyone, and perhaps worst of all, knowledge walks out the door when someone leaves.
00:59:33 ◼ ► Scribe is a workflow AI platform that captures any sort of workflow or process in real time and turns it into documentation automatically.
00:59:43 ◼ ► No manual writing, no manual screenshots, and no starting over from scratch every time someone new joins the team.
00:59:50 ◼ ► Scribe is already trusted by over 80,000 companies and enterprises, including nearly half of the Fortune 500.
00:59:58 ◼ ► To capture new documentation, you just turn on the Scribe extension and do the process as you normally would.
01:00:05 ◼ ► Scribe creates the guide as you go, taking screenshots, noting where you click every step of the way automatically.
01:00:12 ◼ ► Scribe redacts sensitive or personal information, things like usernames, account numbers, email addresses, automatically.
01:00:22 ◼ ► Then, when someone else needs to follow that documentation, they can launch real-time on-screen guidance that shows them exactly where to click and what to do step-by-step, right inside the actual tool.
01:00:35 ◼ ► Scribe will even suggest improvements to the underlying workflow, not just documented, identifying steps that are redundant or complicated.
01:00:43 ◼ ► It's not just capturing how work is done, it's not just capturing how work is done, it helps you do it better.
01:01:26 ◼ ► You just change the backend and cursor, and it's not like—so this is where the analogy to directing or hiring humans breaks down.
01:01:37 ◼ ► Because if you were, again, like me with my notes app, sociopathically just picking up and moving to a new town and starting a new family of notes, if you were just like, I'm sick of you as a developer, I'm going to switch to somebody new.
01:01:52 ◼ ► So they would have to come in and get used to the project and figure out the lay of the land, and no, that's not how it works.
01:02:14 ◼ ► So if you swap in GPT, it's still process—it's still process—like including the last however many tokens in its—in the next turn, right?
01:02:30 ◼ ► But not only that, there are just ways of including more context in every exchange, like with stuff like cursor skills or rules, that it's basically, as an agent, it's thinking like a human.
01:02:52 ◼ ► Should I go and read the rules because the user asked me some factual information about something that's in the database, so I need to know actually how to query the database.
01:03:05 ◼ ► So when you're beginning this stuff, you find yourself telling it the same instructions over and over and over and over again because it keeps forgetting.
01:03:20 ◼ ► But then you start to understand, okay, all I have to do is save the instructions for the robot, and then the robot will find the instructions every time I ask it to do this thing.
01:03:33 ◼ ► And this is—I mean, it's just like the parallels of working with a person are like they're never-ending.
01:03:40 ◼ ► But that idea of context as continuity and the real power feature of something like Cursor is that every—so every time I'm working on a different feature, I don't stay in the same agent because the context windows fill up.
01:04:26 ◼ ► It's like imagine if the world's most productive team was recording every meeting and every part of their process that they ever did.
01:04:42 ◼ ► And then you can use that institutional memory to your advantage by turning it into a set of core processes that you can then use.
01:04:50 ◼ ► And I did this, I showed you what I, what I, like a little glimpse of what I had built using this idea of historical artifact, like historical memory.
01:05:00 ◼ ► Which is, I realized maybe like a month ago, I realized that I have six months of building history in my cursor transcripts.
01:05:08 ◼ ► And I started to recognize that there's a certain way that I work as a builder that I think I should probably try to like capture and encapsulate as a tool that I can then use myself.
01:05:20 ◼ ► So I spent like two days extracting all the transcripts from six months of building in cursor and then turning it into an intelligence that represents me and how I build.
01:05:34 ◼ ► And then I made essentially an MCP where I could access Lissagor and then bring Lissagor into any future development task with me.
01:05:45 ◼ ► So now as a matter of practice, anything that I'm building, I consult with my Lissagor expert before I start building.
01:05:54 ◼ ► And then along the way, many times, and what I find is that that process ends up catching errors that will happen in advance and actually putting me on the right path to save even more time.
01:06:07 ◼ ► That is like figuring out that what you have done as a matter of professional practice for years and years and years is actually value that you shouldn't be losing.
01:06:22 ◼ ► You should be capturing that and turning that into a machine that then you can use or other people can.
01:06:29 ◼ ► This is the part where if I didn't know you, I'd be rolling my eyes and I would think that you lost me.
01:06:37 ◼ ► And but I do know you and I know you're not a bullshitter and I know and I know that hovercraft is real and it is a very it's it is.
01:06:55 ◼ ► It does not feel like two weeks of work for a Mac developer who's been making indie software for a while.
01:07:04 ◼ ► It feels like it would take longer to build this and it for somebody like you who doesn't actually write any code.
01:07:26 ◼ ► It's it's my fault, but I had the V mode on where it it the visibility of the attachment.
01:07:36 ◼ ► But the way that hovercraft works is you open the app and it's sort of like photo booth where it's a single window that shows you a preview of your webcam.
01:07:46 ◼ ► If you're just this is safe for the sake of argument like me right now, I'm on a MacBook Air and I'm using the built in camera and there's a couple of controls.
01:07:54 ◼ ► But then there's a drop zone underneath where you can and it says drop an image or a PDF here.
01:08:02 ◼ ► So if you want to show a slide deck, somebody can you drop a keynote file in there or or you can there's not native keynote control yet, but that's right horizon.
01:08:22 ◼ ► But you drag like a JPEG or a ping or in this case, I have a PDF of this chart that I've been talking about.
01:08:29 ◼ ► I dragged it into there and then you hit a button to make it go live and then it's superimposed on my picture.
01:08:35 ◼ ► I see it in the Hovercraft app and then you see it because what I do and I'm using we're using a web app called StreamYard to record this show.
01:08:54 ◼ ► And the way that you pick cameras in those apps that you can pick, like if you have multiple webcams connected or if you want to use your iPhone via the camera continuity feature as your camera.
01:09:15 ◼ ► Then Hovercraft itself shows up as a virtual camera, which includes the sort of picture in picture of the image attachment.
01:09:29 ◼ ► There is a very strong chance that now that people see this, that this is the basic concept here is going to get Sherlocked.
01:09:46 ◼ ► So this is the part that seems, okay, up to the point where you're saying I'm using Cursor and Cursor.
01:10:01 ◼ ► Cursor, Cursor has, because Xcode is just a front end for the same kind of code base as anything.
01:10:20 ◼ ► It's able to like, I don't know, when it's able to create a DMG of my build and notarize it for me and, and then open it up in the finder and says, and then say, text this to John right now, because it fixed the thing that he was complaining about.
01:10:41 ◼ ► I think for anybody that's trying to work with AI coding tools, the secret sauce, the secret magic is talk to it like a human and share your emotional experiences going through building.
01:11:10 ◼ ► And I, now I, I see some scribbles appearing on your camera as, as we're, um, and I, I want to do that.
01:11:19 ◼ ► I want to draw on the frame and I want to ship it by the time I'm on my friend, John Gruber's show tomorrow.
01:11:25 ◼ ► Well, this is the kind of stuff that John Gruber cares about because obviously Opus knows who John Gruber is right.
01:11:34 ◼ ► So then context for every part of that discussion is these are interesting choices that you can make before you're on John's show.
01:11:45 ◼ ► It impacts the outcome, the, like the work product, but it just makes the process more meaningful than talking to it like a robot.
01:11:53 ◼ ► And I, and I think that that's the thing that if there's one unlock that I would love for people to start thinking about, it's like share, share yourself as a human, as a builder.
01:12:48 ◼ ► It's also, this was an accessibility thing that I thought of the AI didn't have to think of it.
01:12:58 ◼ ► If you're a right hander, if you're a left hander, you're holding D with your right hand, then you're like, you're basically blocking your trackpad on your power.
01:13:28 ◼ ► This is your, this is your portable export though, is the thing that you're, you've created.
01:13:40 ◼ ► One, one friend, a woodworker in Chicago named Joshua is a power user, but it's really like it's, it's pre-launch, but how did you switch?
01:13:49 ◼ ► How did you switch here to put your picture in the background and your video in the foreground?
01:14:33 ◼ ► So just to be clear though, to the audience though, right now, portable expert is not out.
01:14:37 ◼ ► I could use it cause I know you, but, and you'd let me in probably, but like people listening can't, but you've got.
01:15:07 ◼ ► So I have a browser extension where I can chat with any of my experts in, in the browser window about what's on the page.
01:15:20 ◼ ► And the, the one that I was describing, the Lissagor expert is an MCP that I can call into any of my cursor projects, my cursor chip chats.
01:15:45 ◼ ► And with this protocol that I basically created, the portable expert protocol, it, it calls that material.
01:16:04 ◼ ► And an MCP interface is effectively, like for modern AI, loosely analogous to like, in AppleScript terms, like having a scripting dictionary in the first place.
01:16:16 ◼ ► Like you could make an app that doesn't have a scripting dictionary, and then you can't AppleScript it.
01:16:20 ◼ ► But if it does, then there's a dictionary you can query, and then you can write a script that does things.
01:16:26 ◼ ► And an MCP interface is a way for cursor, you can say, here's, you, you add the access in cursor.
01:17:02 ◼ ► And then anything that speaks to it, you can give access to your Fantastical MCP, and then
01:17:07 ◼ ► whatever it is, it could just be the ChatGPT app, it could be whatever, it has access to everything
01:17:16 ◼ ► So if you need, if you, if it would be helpful to have your personal schedule in your AI system,
01:17:22 ◼ ► Fantastical supports that in a way that I don't know that anything else on the Mac does.
01:17:34 ◼ ► An API is a set of endpoints that you publish so that one piece of software can talk to another,
01:17:44 ◼ ► but it's very brittle because if the endpoint isn't exactly correct, the communication isn't going to happen.
01:17:50 ◼ ► And MCP is more like it takes an LLM into account and says, if we publish sort of like the endpoints as they're sort of like semantic, semantically exist, then the LLM is going to be able to figure out what endpoints it needs just from natural language inputs.
01:18:09 ◼ ► Do you feel like, would you have been able to create hovercraft without Lissagor just with you and Cursor?
01:18:20 ◼ ► I would get, I would get pretty far, but like what, again, like what I, the value that it brings is like,
01:18:38 ◼ ► And then in my experience, when you set Opus loose and, and it starts building, it's nailing it.
01:18:45 ◼ ► It's like your, your debugging kind of mostly goes away, but well, that's not entirely true.
01:18:51 ◼ ► And this is another nuance to AI coding is that a lot of what you see in like the public with like Twitter builders, you know, Twitter AI influencers and stuff is like,
01:19:08 ◼ ► They start up seven agents and then they walk away and they come back and they have a product.
01:19:28 ◼ ► And in my experience, you set an agent loose and I've got five agents going at the same time.
01:19:45 ◼ ► And then it asks you like three questions and it's a lot to process, but then you give it your answers and that's the creative directing part coming into play.
01:19:54 ◼ ► And it's really, really joyful because you get to see the cause and effect of your choices turning into product.
01:20:07 ◼ ► Yeah, and it's in the same way that being the director of a film or a commercial is not, okay, guys, here's the script.
01:20:24 ◼ ► Exactly, which is what I've heard is that a lot of directors in that, especially commercial directors later in their careers, that's how they kind of treat the process is that they just hire their team around them that knows kind of what they would do.
01:20:44 ◼ ► And then while every other decision has been made for them, they show up on the day and then they just kind of like their name is on the director's chair and then they call it a day.
01:21:01 ◼ ► I think you would talk to any creative person or artist and they would tell you the same.
01:21:06 ◼ ► The way that my, you mentioned that my credit for Vesper was director and the way I thought about it and I've been working on this theory on and off for a long time.
01:21:14 ◼ ► I gave a couple times, somebody linked to it recently and I was like, hey, that was actually, that came out all right.
01:21:20 ◼ ► Like in 2010, I think at Webstock, I gave the auteur theory of design where I compare the design of anything to filmmaking.
01:21:29 ◼ ► Yeah, you were there at Webstock, but that it is a curious role and nobody questions it because by the end of the 20th century, when everybody in our audience was already a child, at least the idea that there are these people called movie directors and they, they are the ones who are the film by Steven Spielberg.
01:21:51 ◼ ► You know, and I think a lot of times the people's best, most beloved film directors, whether they get credited or not, they have a lot of input into the screenplay in particular.
01:22:01 ◼ ► There's footage that his daughter shot of Kubrick pecking out pages of The Shining in the kitchen, although he doesn't get a credit, a co-writing credit for the screenplay.
01:22:37 ◼ ► There's somebody whose title is the editor who edits the footage, but that the director is there every step of the way.
01:22:42 ◼ ► And that from the director's perspective, it feels like they're making a movie or making a commercial or making whatever it is that the thing is.
01:22:51 ◼ ► They're making it, even though they're not doing any of those things because they're there and the making part is just the collaboration and having the vision for what they know they want it to be to come out and to be a receptive input for the ideas that are coming from the people around them.
01:23:13 ◼ ► And so, like, the director has an idea for how this looks, but the cinematographer says, yeah, but it would be really cool if we did this.
01:23:25 ◼ ► Or the actor could say, I think I should say this if my character would say this in response to that.
01:23:40 ◼ ► You've never, you're not acting, not filming, not editing, not didn't write it, but you're directing, you're making it, right?
01:23:47 ◼ ► And that's, that could apply, you know, and we do, like you said, creative director, right?
01:23:51 ◼ ► There's, there's filmmaking isn't the only art or industry where the word director is used, but I think at its best, it is a very, very active role.
01:24:03 ◼ ► And that doesn't mean that you're micromanaging it, hopefully not, but you're not taking off for Mexico for a week and coming back while your team made it.
01:24:15 ◼ ► And if you're directing the creation of a cool new app, it is not, here's, here's a single markdown file that describes what I want.
01:24:36 ◼ ► And that ends up being that historical record of where you started and where you got to is like the product.
01:24:47 ◼ ► The final product, as you know, should feel like there were hard choices made at every step along the way.
01:24:58 ◼ ► The, the, I'm, the, the reason that AI coding is going to get a bad rap for a while is because it feels like it is a replacement for human collaboration.
01:25:12 ◼ ► I don't think it has to be, it's like, obviously you can do a lot more solo, but actually some of the more gratifying experiences that I've had have been working with people on my team who are also using coding tools to build tools for sandwich.
01:25:28 ◼ ► And when we can sort of like work together to make a thing that other people are going to use, everybody feels, I don't know, it's scale, right?
01:26:07 ◼ ► Because it effectively, when you're using it, you get this, now it's not coming up, but you get that whatever it is you're showing, it shows up, it like hovers in front of you.
01:26:32 ◼ ► I think the name was so obvious that it's you, every once in a while you have those ideas that they just come and you're like, yep, that's it.
01:26:39 ◼ ► When I was thinking about what is a Mac app that I can drag windows around, like they're hovering, hovercraft, I'll call it that.
01:26:48 ◼ ► What's interesting about it is how closely it aligns with the whole thesis of what we're already building with theater.
01:27:08 ◼ ► The future of spatial computing is that we can put a lot of digital stuff in our space and it feels like it's supposed to be there physically.
01:27:17 ◼ ► That also aligns with sandwich video, what we've been doing since the very beginning of the company, which is putting digital stuff in the physical space.
01:27:26 ◼ ► So when you start to connect all these pieces together, you realize that hovercraft is an extension of that inner expression of that idea of that same idea of like, let me show you a thing that's digital.
01:27:45 ◼ ► And behind the scenes, like as of three days ago, the core tech that layers digital stuff in hovercraft is actually the same core tech that runs layers of information in theater.
01:27:58 ◼ ► It's actually the same core tech that I it's like a protocol with an actual file format.
01:28:04 ◼ ► And it's really freaking interesting, because if you believe that the future of media is in layers, then you need sort of a codified way of managing those layers.
01:28:16 ◼ ► I think every time there's a democratic leap forward in technology, it is it's more than two steps forward, but there are some steps back.
01:28:27 ◼ ► And like when desktop publishing first became a thing, people who are already trained, the only people who really did graphic design work were trained graphic designers.
01:28:37 ◼ ► And there or if you just look at signs and it is it's it's it's a shame that there's that hand painted signs for like mom and pop retail stores isn't as big a thing as it used to be.
01:28:54 ◼ ► But you could go back and you look at pictures from any time before the 80s and all of the signs look pretty cool because all of the signs were made by professional sign makers.
01:29:08 ◼ ► Old license plates invariably are terrific, just terrific, industrial, serious looking things.
01:29:19 ◼ ► And then we we made these computers where everybody could pick any font they want and pick gradient color blends and put images and mix fonts together that were never meant to go together.
01:29:31 ◼ ► And they used to call it ransom letter typography where you weren't intending it to look like a ransom letter.
01:29:37 ◼ ► But that's what it looked like because you don't know anything about graphic design, but you know how to use the font menu and you've never picked fonts.
01:29:47 ◼ ► Look at all these fonts that my computer came with and look, I can download more fonts.
01:30:03 ◼ ► And you could say, well, this sucks because the world never had anything this ugly with a cursive script title and impact font used for the body copy.
01:30:13 ◼ ► And it's, yeah, but a lot of people who were never going to learn it in the first place when it was a craft that you had to do in a manual analog way are learning to do it now.
01:30:25 ◼ ► And the world is full now of more people who know how to do graphic design are actually pretty good at it than who ever were before.
01:30:32 ◼ ► And the people who were already good at graphic design could suddenly become way more productive than they were before.
01:30:38 ◼ ► And when I was young and when I got out of college in the 90s and I was doing graphic design work professionally as a 20-something, the older people who were still around, people who are now my age, who were in the industry in the late 90s, were all the people who embraced computers in the 10 to 15 years before.
01:30:58 ◼ ► And maybe I met some people who, like, were still doing it the old way, but there weren't many.
01:31:04 ◼ ► And they were like, and they're still doing work, but they were like, do you know how long this used to take me?
01:31:12 ◼ ► And it was like, oh, my God, you'd have to, like, throw the film out and do it all over again.
01:31:57 ◼ ► I don't know if it's on other social media, Pessimists Archive, where it just finds like 100 to 150-year-old newspaper articles complaining about the rise of technology.
01:32:08 ◼ ► I remember one time where it was like, all these women who are riding bicycles, there's going to be a fertility crisis soon because they're going to make themselves sterile because women's bodies aren't meant to sit on seats like that.
01:32:23 ◼ ► And today's world, everybody, you could never, you'd be very hard-pressed to find an adult who does not think children should read more novels, right?
01:32:33 ◼ ► But when novels were a new mass market product, and there were like dime store novels in around the year 19, late 1800s into the 1900s, there were articles saying everybody's wasting all their time reading these pulpy novels.
01:33:13 ◼ ► But how can it not be an awesome thing that people who never were able to make apps before are now making the apps, pulling them from their imagination and putting them out?
01:33:28 ◼ ► It's an app for if you have a watch collection to track your watch collection on your iPhone.
01:33:35 ◼ ► And then every day when you wear a watch, you can just say, hey, I'm wearing this watch today.
01:33:49 ◼ ► And Stu linked to my review on social media because I said, this is a really interesting app.
01:34:06 ◼ ► And it turns out that even though I like to track all sorts of really nerdy things about
01:34:10 ◼ ► my life, which watch I and I have a watch collection, which watch I wear, I don't care to track.
01:35:03 ◼ ► And it has been the one of the dreams of the humane side of software for 40 years, 50 years.
01:35:18 ◼ ► Of, hey, part of the computer for the rest of us of the Macintosh was wasn't just that the rest of us would use computers, but that they would use them to make software for computers.
01:35:40 ◼ ► But the idea was that by making it something that non-programmers could read, that it was something non-programmers could do.
01:35:54 ◼ ► But it did allow more people to automate more things and to create software of some sort, if not full apps, little partial apps that whatever you want to consider the scope of a script versus an app.
01:36:09 ◼ ► More people were writing things that were software based on what they wanted to create than they were before because it was a lower level of complexity.
01:36:22 ◼ ► It meant there was less need for people who could write C code to do the same thing in a full fudge program because people could write scripts to do it.
01:36:33 ◼ ► It's finally happening with the AI moment where it's for full applications like the, hey, let's let.
01:36:39 ◼ ► People who don't know how to make apps the old fashioned way, writing the code, make apps a new way by using the language that they already, their brains already know, which is English.
01:36:52 ◼ ► It took a lot more, and it just took a lot more time, but almost from the perspective of like when Apple script, I think Apple script debuted in 1991 around there, early 90s, almost infinite more computing.
01:37:07 ◼ ► Like the amount of actual like ones and zeros that go through logic gates for one command with you directing cursor to make one change to hovercraft is like infinite computing from the perspective of 1991.
01:37:24 ◼ ► It would be like on a, I mean, look at all the layers of tech that's going on right now in order for us to just have a conversation, just pixels on our screens being captured by, in my case, my phone, you have with a digital layer in front of you with many digital layers in front of us over IP.
01:37:59 ◼ ► And in the same way that people use graphic design tools and the access to a fonts menu to make absolutely atrocious flyers or reports and people who just have absolutely no taste in software are going to make software, but it might make them happy, right?
01:38:18 ◼ ► And if you don't see the problem graphic design wise of mixing Times New Roman with, and you don't see the difference between Arial and Helvetica and you think chalkboard is a good typeface, more power to you.
01:38:34 ◼ ► If you're making something that you're not going to put in front of me, but if it's for you, that's great.
01:38:39 ◼ ► And if you make an app that is really part of what I've made our chat the other night with Hovercraft so fun for me and hopefully for you, too, was that it was already so good and I could just pick nits, right?
01:38:53 ◼ ► Little tiny details like, oh, those things that you're putting on top of the video, I think they should be off the video and underneath the video and only there.
01:39:00 ◼ ► And there shouldn't be two ways, two things to click to toggle the hand tool, just one.
01:39:08 ◼ ► Whereas sometimes looking at a really sloppy app that is just a mess, like the thing I just linked to and wrote about yesterday, the NextPad++ for the bizarre AI driven.
01:39:24 ◼ ► That's why I wrote about NextPad yesterday is some guy took the open source Windows text editor notepad, which is a GPL open source project.
01:39:39 ◼ ► I mean, I'm sure his directions were more than that, but he did it in a month and it's crazy.
01:39:55 ◼ ► My favorite thing I wrote about it was that if using Electron apps and back in the day Java apps, or there are like newfangled apps, like there's a text editor called Zed that is really interesting.
01:40:14 ◼ ► It is native code compiled and in some ways it's closer to being a Mac app, but it doesn't do things the Mac way, but it's kind of interesting.
01:40:27 ◼ ► It's like, oh, most of the characters look like humans, like Luke and Obi-Wan and Han Solo, but then there's Chewbacca and there's the guys playing the music in the band and there's the hammerhead guy in the corner and everybody's welcome.
01:40:40 ◼ ► And yes, they're just different and they're aliens, but this next pad app said looks like Vincent Tenofrio's bug in human skin.
01:40:50 ◼ ► But if you look at GitHub, it feels unholy to me and it makes me, you know, I don't feel like it, I'm not unhappy that it exists.
01:41:07 ◼ ► There are Windows people who are so used to Windows and have used Notepad on Windows for so many years and now for whatever reason, I don't know why they're not still using Windows, but they're on a Mac and they really want Notepad.
01:41:21 ◼ ► And there's like one of the issues is somebody wants, they called it, the request is for a quote, normal menu bar.
01:41:28 ◼ ► And what they want is a menu bar in the window, like on, so if you have two windows, you'd have two menus, each one in a window, like on Windows, right?
01:41:50 ◼ ► It's absolutely a monumental programming effort to make something that is so deeply intertwined with Windows as Notepad++ is to port it to the Mac.
01:42:14 ◼ ► And I think that hopefully one of the effects of this is that a lot more Swift native apps are going to be made just because there's not that barrier to entry that used to rely on the sort of like React transformation solutions for the Electron, etc.
01:42:35 ◼ ► I've got for my expert platform, I made a rudimentary Swift version of it and it works.
01:42:56 ◼ ► And it's just for me, you know, it's like not on the App Store or anything, but like these little proofs of concept.
01:43:03 ◼ ► And that's the other fun thing about AI coding is you don't have to build it for a million people.
01:43:16 ◼ ► And anybody who's installed any of these virtual cameras, forget some of the other ones.
01:43:35 ◼ ► And most webcams come with their own software so that they can have software that knows the actual optics of the camera.
01:43:56 ◼ ► And then the software can do things like put a filter over the image or something like that or make it black and white.
01:44:22 ◼ ► It's so scary that Apple calls it screen recording every time you want to share a window.
01:44:41 ◼ ► Because it always makes it seem like whatever it is that's asking for the permission is looking at every single thing everywhere on your screen in every layer.
01:44:57 ◼ ► And I kind of just have to, you know, with Alan's help, I kind of just created a setup flow.
01:45:08 ◼ ► And then something just within the last couple of days as I was building the drawing on the frame and typing on the frame.
01:45:24 ◼ ► So it can basically just run its own whole set of internal smoke tests for me before I even build the app and futz around with it.
01:46:03 ◼ ► And I just remember when I was using autopilot in the Tesla, where it sort of drives within the contours of the lanes and everything.
01:46:37 ◼ ► And then the flip side of it is that you only get that feeling on a continuous basis for a year or two or a couple of years while the overall technology is very rough.
01:47:06 ◼ ► It could be like streaming video or Wi-Fi speed, like internet speeds or anything like that, where you stop noticing the errors.
01:47:15 ◼ ► Like that retina screens felt retina displays with the iPhone 4 felt like, holy shit, this is incredible.
01:47:22 ◼ ► And Apple had been hinting at it and sort of making baby steps towards scalable interfaces.
01:47:30 ◼ ► But it felt like what – and I think maybe – I think I even talked to people at Apple who thought, yeah, that's what we were thinking too, that they would just start slowly increasing the pixel density of displays.
01:47:46 ◼ ► And instead they were like, you know, it would be easier if we just went to double the resolution in both dimensions all in one step.
01:47:53 ◼ ► And that just felt like Steve Jobs snapped his fingers and we went from – I can't – I remember thinking that the 3GS was so much faster than the other phones.
01:48:04 ◼ ► And as soon as I saw the iPhone 4 in person – and, you know, the rumors said – I mean, that was the one that leaked with Gizmodo too.
01:48:12 ◼ ► But everybody kind of knew the retina screen thing was – we didn't know the name retina screen.
01:48:19 ◼ ► And I've never at any other event so wanted to throw my existing iPhone in the garbage immediately than when I saw a retina screen.
01:48:44 ◼ ► And even when we were measuring in megabytes, I mean, every single app on your system, shit's bigger than a megabyte now.
01:48:54 ◼ ► It's just – but it's ridiculous that I spent years using a Mac LC that only had four megabytes of RAM.
01:49:04 ◼ ► And for filmmakers, it was the digital video transition that was so frustrating because we knew we wanted our digital films to look a certain way and they just looked like garbage still.
01:49:25 ◼ ► The awkward phases of technology are the most fun because you just get those feedback loops of huge –
01:49:47 ◼ ► And everybody was so excited and the footage looked so cool compared to all previous digital video cameras.
01:50:21 ◼ ► And we're there again with immersive video where we have – we do have a camera with an enormous amount of resolution, but it's so slow to work with.
01:50:37 ◼ ► And like everybody who works in that format is just waiting for the tools to make it easier.
01:50:46 ◼ ► I want to take one last break and talk about our next sponsor, our good friends at Squarespace.
01:50:51 ◼ ► This podcast is sponsored by Squarespace, our good friends, all in one website platform designed to help you stand out and succeed online, whether you're just starting out or scaling your business.
01:51:03 ◼ ► Squarespace gives you everything you need to claim your domain, showcase your offerings with a professional website, grow your brand, and get paid all in one place.
01:51:13 ◼ ► They have – they're offering services is one of the things that I think is the most interesting.
01:51:19 ◼ ► And it's – if you are like a consultant or if something where you book time or a trainer or something like that, you can do it, build your own website and sell your own time and or services through your Squarespace website that you own and control at Squarespace.
01:51:41 ◼ ► So every aspect of getting paid, creating invoices, sending invoices, designing what the invoice looks like, sending invoices and then having your clients or customers, whatever you call them depending on your business, pay you through Squarespace because they accept payment on Squarespace.
01:51:59 ◼ ► All of it, all of it, all in one, and you're not beholden to any other platform or sending people through a system.
01:52:15 ◼ ► And Squarespace has been sort of doing the, hey, you don't have to be a professional web developer at all.
01:52:25 ◼ ► You just know they are parts of a website and you can create and design your own website.
01:52:33 ◼ ► Squarespace has moved forward with their design tools for non-professional, non-expert designers.
01:52:40 ◼ ► But just like making software or making apps, being able to write code or know your way around Xcode can make you even better at directing it.
01:52:49 ◼ ► If you do know CSS and HTML and JavaScript, you can get in there and mess with it, do whatever you want to edit it in Squarespace.
01:53:05 ◼ ► And by going to that URL and using that code talkshow, when your 30-day free trial that every new customer gets, 30 days, no limits, no watermarks on your website, you save 10% off your first purchase.
01:53:21 ◼ ► So obviously the theme of the episode was mostly the AI coding creation and how you made Hovercraft and how you're making other things going forward, really.
01:53:37 ◼ ► You've got a whole team of people making – and you've – not that you're uninvolved in that, but you've sort of shifted towards pursuing your goal of making these tools and stuff for the team, internal stuff, the stuff like Hovercraft that you're sharing.
01:54:11 ◼ ► Something that we're doing right now is actually making a software tool for one of our clients to use, which is a whole new frontier for us.
01:54:19 ◼ ► But, like, the reason I started – that I shifted my focus over to software a couple of years ago is because I saw AI coming fast, and I wanted to be able to harness it to, A, help us do the work that we do, but also just turn our creative processes into something that others could use.
01:54:41 ◼ ► So, you know, I always kind of encapsulate it as turning sandwich into software and letting it scale, and, yeah, just, like, it's proven to me over and over and over that creating tools for creative work is a creative enterprise, and it's also one with a business model, maybe after a creative studio doesn't have a business model.
01:55:09 ◼ ► And the intersection – or at the intersection of all of this, your enthusiasm, to say the least, for the Vision hardware platform, Vision Pro and Vision OS, remains as enthusiastic as it was before.
01:55:24 ◼ ► We talked earlier in the episode that it obviously has not set the world on fire sales-wise, but I keep bringing it up, and few people seem to remember it, but it was out there and repeated multiple times before it even shipped.
01:55:37 ◼ ► I think that the screens that the screens that it uses for each of the two eyeballs, at least originally, and we're only two years into it, were supposedly limited from Sony to 900,000 screens a year, which would be 450,000 units a year.
01:56:21 ◼ ► Now, maybe if they had sold more and there was more enthusiasm for it and more people just talking like,
01:56:30 ◼ ► I can't get enough of watching these Lakers games on my Vision Pro, or if you – every time you go on an airplane, you see more and more people wearing Vision Pro,
01:56:43 ◼ ► If you saw more of that, maybe Apple would have more of the pedal to the metal and there'd be more rumors like,
01:56:50 ◼ ► oh, the Vision Pro 3 is coming out soon, or we expect the Vision Air by Christmas this year, by the end of 2026.
01:56:59 ◼ ► But I don't think – you know, there was this weird Mac Rumors story like two weeks ago that, hey, Mac Rumors has learned that Apple's given up on the Vision platform.
01:57:09 ◼ ► And it didn't say – it didn't say how they learned it, didn't say – didn't even say, like, sources familiar with the matter,
01:57:19 ◼ ► And I was like, I know people who work on that team, and I would think they would have reached out to me.
01:57:25 ◼ ► And I checked with them, and they were like, yeah, we were surprised by that story as you were.
01:57:33 ◼ ► Yeah, it's crazy to think that because what are they – does that just mean they're chucking the IP, they're going to liquidate it, maybe sell it to Samsung or something?
01:57:41 ◼ ► Like, they've got a lot at stake, and it's kind of indisputable that it still is the future of computing, whatever that's going to look like.
01:57:56 ◼ ► Right, and I wonder where that – you know, that if there is a – somebody said something to somebody about something,
01:58:02 ◼ ► and there's a kernel of truth that got whispered down the alley or played the game of telephone,
01:58:06 ◼ ► that it's like there's not going to be another $3,500 Vision Pro like the one that we have with the M6 or M7 or something like that,
01:58:18 ◼ ► But that's just like saying there's not going to be another classic Mac that looks like the Mac SE30.
01:58:32 ◼ ► a heavy thing that sticks out that far in front of your face that requires a battery pack connected to a cord that you keep in your back pocket.
01:58:43 ◼ ► And whose ergonomic problems they're solving with a different head strap, you know, sort of midway through.
01:58:50 ◼ ► There's differing opinions of how, well, this is good enough for now, how far along the scale of good enough it is.
01:59:00 ◼ ► And I think the thing I would be happier about and have more confidence in the future of the platform is if there were – if I were seeing more indie and big company apps coming out for it, right?
01:59:15 ◼ ► It's – but there was a story I linked to last week or the week before where there's an eye doctor who's performing – using it in a – performing cataract surgery for the first time.
01:59:34 ◼ ► And, of course, if you're a surgeon, a $3,500 piece of equipment is one of the lower-priced pieces of equipment that your business has to obtain.
01:59:44 ◼ ► That as much as a consumer product, yes, a $3,500 entertainment headset is pretty expensive.
02:00:22 ◼ ► Like, wow, this was something that even the crude, low-RAM, black-and-white screen Macintoshes of 1987 and 1988 were really good for.
02:00:32 ◼ ► It was already better in some ways to do graphic design on those primitive computers than it was without a computer.
02:00:59 ◼ ► And apps sort of being the middle artifact worked really, really well on our phones because there are these units of commerce for how to package a tool and wrap a tool that they made it so easy for us to grab and like buy like we were just at the corner shop.
02:01:26 ◼ ► And that's a meeting of the content that you're experiencing and the core tech that makes it possible.
02:02:18 ◼ ► And yes, nobody's leaving the iPhone or even the Mac because what are they going to go to when they need something that's like the Mac and that iPhone has all the users.
02:02:28 ◼ ► But when it came to, hey, I'm going to spend all this time and effort to develop something that it and developing for Vision OS, there is nothing like React where you can build the same app and it's going to work with minor changes on Meta's headsets or something like that.
02:02:53 ◼ ► But you're right, though, that it might be when the Apple Watch is a good example where like the first Apple Watch, they were like apps, apps, apps, apps, apps.
02:03:12 ◼ ► I linked to David Smith's Pedometer Plus Plus app that puts he commissioned a guy to even make custom maps just to get maps on the watch.
02:03:19 ◼ ► And I think Vision could be like that, too, where apps are never going to be not part of the story.
02:03:43 ◼ ► And in a way that there's never going to be a probably never going to be a product as big a hit as the iPhone again, there's never going to be anything as app centric as the iPhone again, either.
02:03:55 ◼ ► It's and I do think I think something something I'm optimistic there will be the next time Apple comes out with new hardware that actually does look different and almost certainly at least has a version that reaches a significantly lower price point.
02:04:10 ◼ ► Simultaneously, you know, having better screens because tech, that's how technology works.
02:04:16 ◼ ► And all of a sudden, a lot of people who when the initial Vision Pro came out and they're like, why isn't why doesn't Apple have this whole library of immersive 3D videos and experiences?
02:04:30 ◼ ► And when they pay attention the next time, because there's new hardware, they're going to say, where did all this immersive content come from?
02:04:41 ◼ ► It is quiet and it is not a hit, but that doesn't mean that they're not moving ahead on it.
02:04:50 ◼ ► You can go to Apple's job listing site and they're hiring hardware and software engineers to work on the Vision platform.
02:04:56 ◼ ► That's not the type of job offerings people, companies typically offer for platforms they've, quote, given up on.
02:05:05 ◼ ► I mean, I get the, hmm, I wonder when this is going to be interesting for the mass market or even if it's going to be interesting for me or whatever.
02:05:13 ◼ ► But I think it's foolish to think that Apple and I don't I think it it it goes against all the evidence that we see.
02:05:33 ◼ ► They are releasing new immersive content and there is they added Laker games and it is kind of amazing to watch.
02:05:48 ◼ ► Yeah, it's kind of it's kind of it's kind of there might be a way that you could if you poke around, I think there might be a way you could watch some clips for free.
02:06:01 ◼ ► I watched they have a series called Elevated Now, which is just flyovers of magnificent parts of the world that are narrated.
02:06:11 ◼ ► They partnered with the BBC or they distributed something that the BBC made, which was a pops concert, like a classical philharmonic style concert.
02:06:31 ◼ ► Yeah, Jose is always putting new things in the on the in the in the in the app that I'm like, I'll show up and I'll just look at something.
02:06:44 ◼ ► And I think that that is the future that Apple is betting on is immersive media, immersive content.
02:07:00 ◼ ► And I every time I talk about it, I keep banging the point that it is a lot like the original Macintosh in 1984.
02:07:36 ◼ ► You could carry 10 MacBook Neos and it would weigh less than an original Macintosh, I think.
02:07:49 ◼ ► Overall, I would never want to trade my glasses for a Vision Pro to wear all day, every day.
02:07:59 ◼ ► I mean, I wish more people knew about going without the light seal because I haven't had a light seal on my Vision Pro for at least a year.
02:08:13 ◼ ► Because when I'm working, I'm working with the Mac virtual display, sitting on that couch over there with a 30-foot screen in front of me.
02:08:30 ◼ ► The people who use it the most for the longest are very, very likely to be no light seal people.
02:08:48 ◼ ► Like, hey, we're going to, you know, your Vision Pro comes with a light seal and you scan your face to get the right size for it.
02:08:58 ◼ ► It is a very well-named component, but it's as optional as, you know, using an external display.
02:09:19 ◼ ► I think, I mean, yeah, it's so funny how the intention is for it to be a VR device, but it's so much more useful if it's not.
02:09:36 ◼ ► Hopefully, I feel like if there's an advantage to me being behind the eight ball on the overall AI, certainly the AI coding to stuff like Cursor and Codex and Cloud Code, it's that while I'm catching up, I can bring my audience with me.
02:09:53 ◼ ► And I really do think there are a lot of people out there who listen to shows like mine who you've never written an app, but there are ideas for apps in your head that you've always wanted to make, but you don't know how to get it out there.
02:10:07 ◼ ► And you could get started now, but you probably soon, you're going to be able to do this.
02:10:18 ◼ ► I can't wait to see the first app that you make, you specifically, but also your audience.
02:10:42 ◼ ► Holy cow, the models can do these things that they didn't used to be able to do this last month.
02:10:52 ◼ ► You don't have to go to Lovable or Replit or any of these, like, no-code sort of cheaty kind of things for normies.
02:11:00 ◼ ► You can just go right to the real stuff, and you're going to have a power tool in your hand that you can use.
02:11:08 ◼ ► Hovercraft is a real Mac app that does the real thing as well or, honestly, better than if it had been made by most developers the old-fashioned way.
02:11:37 ◼ ► So, in this world of incredibly bloated 100 megabyte downloads for an app that doesn't seem to do much, which is the Electron way, here's an app that's 4 megabytes, which is, like, the first sniff test of, like, hey, is this a piece of junk or not?
02:12:01 ◼ ► So, everybody should, anybody who does any kind of screen sharing and video conferencing type stuff, you should absolutely check out Hovercraft.
02:12:30 ◼ ► A friend who said maybe before I linked to it on my website, you should have a free trial.
02:12:47 ◼ ► Some of my favorite features now came from people on Twitter reacting to it on that first day, one of which is just, like, disable hand gestures so you're not accidentally throwing your window all over the place.
02:13:05 ◼ ► If you want to do anything that's more like layering stuff on your live camera, like drawing on it, typing on it, sharing a window in a very light-touch way, like, think of it as layers or real-time compositing, then that's what Hovercraft is.
02:13:21 ◼ ► And I'm sure it's going to extend itself in other ways, too, like, with different features that are a variation on the theme.
02:13:32 ◼ ► And I will, I promise, swear to God, no joke, we'll have a link in the show notes, sandwich.vision slash Hovercraft.
02:13:49 ◼ ► We had Parcel, the great package tracking app, Scribe, the AI workflow system, and, of course, our good friends at Squarespace.
02:14:01 ◼ ► I swear it is a bit of a, like, a double take after I tried Hovercraft and realized how much, how you did this yourself effectively.
02:14:13 ◼ ► And it's sort of a moment, like, this week sort of stands out, I think, as a sort of marker on the timeline of my understanding of where AI is today and where the tooling and the system is.
02:14:28 ◼ ► Like, oh, like, for my timeline of, it's like the week where Adam came out with Hovercraft and then sort of explained to me how he made it, it's just sort of like, oh, I've been underestimating vastly how much, how many people could be doing so much.
02:14:47 ◼ ► And that anybody out there who just thinks, oh, when you use AI, all you get out of it is sprawling, loosely glued together slop, try Hovercraft.