Hello and welcome to The Vergecast, the flagship podcast of making apps for your phone on your phone. I'm your friend David Pierce, and today on the show, what it's like to be a vibe coding app developer and also a content creator. But first, here's everything else happening on The Verge today.
I'm Josh Kahas, and this is 90 Seconds on The Verge from Monday, October 19th, 2026. Apple has sent invites to a second product launch this month. According to Bloomberg, the big announcement is expected to be a new, lighter MacBook Pro with an OLED touchscreen and an iPhone-like Dynamic Island. The event is next week, October 28th, at 1 p.m. Eastern. And as always, we'll be covering it here live at The Verge.
Meanwhile, You Can See Everything had the biggest opening weekend for a documentary in 20 years. Nathan Fielder's three-hour doc about Theranos founder Elizabeth Holmes raked in $15 million in its opening weekend, landing in second place at the box office behind Street Fighter. But that's not the only doc making headlines right now. Just ahead of the wide release of the new Elon Musk documentary this past weekend, TikTok banned the film's social media account from its platform. While TikTok has no affiliation with Elon, part owner Larry Ellison appears in the film via text message offering to help Musk fund his Twitter purchase. A TikTok spokesperson confirmed to The Verge that the ad account for the Musk doc has, quote, been restored after an error, unquote. Our TikTok feeds may be blessed by Elon Musk once again.
And finally, in a world full of digital licenses and no guarantee of ownership, there is some hope. New York Congressman Richie Torres introduced the Power to the Players Act last week. If passed, this bill would direct the FTC to establish regulations that guarantee consumers ownership of a game when they pay for it, rather than a licensing agreement. Hopefully, this bill has better luck than that store once PlayStation phases out physical games in 2028. You can read more at theverge.com. That's 90 Seconds on The Verge from Monday, October 19th, 2026.
Support for the show comes from AWS AI. AWS AI is helping the world's leading organizations transform their industries in production at scale right now. From bringing fans inside every swing at the PGA Tour to helping small businesses grow online with GoDaddy to taking vehicles from concept to customer with Audi. AWS AI is how businesses innovate, lead, and revolutionize. AWS AI is how.
Over the last several months, there's been this really interesting trend, at least in a particular corner of the technology world, which is that a bunch of people on YouTube and elsewhere who have spent a long time reviewing apps, recommending apps, talking about apps, have started making apps. People who, especially in the productivity space, have like spent years being mad that no to-do list app did exactly what they wanted, have turned to things like Codex and Claude Code to actually make their own to-do list apps, trying to put their own values and ideas into software into the world. I think this is really interesting and really cool. I don't know if it's leading to better software or just more software, but I think it's a really interesting change in how a bunch of creators and thinkers are actually able to put their ideas into the world.
So, I wanted to get one of them, Stephen Robles, who has been on the show before. He's also the co-host of podcasts like Mac Power Users and Primary Technology. He is also, I believe, the world's foremost expert on Apple Shortcuts, to come on the show and talk about his experience as a vibe coder. He's built a bunch of different apps at all different scales. And I want to know what it's like to become an app developer and make things that are used by other people, that other people are actually starting to pay for, and how that changes your relationship with your community. Let's get into it.
Stephen Robles, welcome back to The Vergecast.
Hey David, thanks for having me. Congrats on baby number seven.
Thank you very much. They all just run around in the background as that's just part of the family now. Um, I just want you to know that I still get emails from people who tell me how right you were and how wrong I was when we fought about Shortcuts. The world does not appear to be on my side, even though history and facts are on my side.
You know, I prepared specifically by wearing a Shortcuts shirt on the show and even...
Made a Shortcuts pillow just to troll you.
That's the best the Shortcuts logo has ever looked in its entire history. That's very good. Um, but I have not brought you here to talk about Shortcuts. I have brought you here because you are, I think, at the tip of a trend I find really interesting, which is a lot of creators, a lot of people, but especially a lot of creators, people who have been making content and thinking about tech stuff for a very long time, are now becoming app developers in a very real way. And I want to start at kind of the beginning of your journey here. You've been a Shortcuts guy forever. You're sort of deeply immersed in how a lot of these platforms work and how a lot of these apps work. You've been covering apps in meaningful ways for forever. When did the turn happen to you becoming a person who made and shipped your own apps?
I always wanted to, and years ago I knew enough Xcode to ship one app that was like all web views and got it in the App Store, like it actually worked, and then just did not get into actually learning the code after a long time. But then once ChatGPT was coming out and coding was better, I was like, let me just see, is this even possible? Like, can I get an app in the App Store? And I made something very silly. It's still in the App Store called Coffee Finder. And I spent the Saturday afternoon.
Is it what it sounds like?
It is exactly what it sounds like. You open the app, you give it access to your location, and it just shows you coffee shops using the built-in Apple Maps plugin or whatever. So, pretty good. You know, some people like it, but I just wanted to know like, is this possible to vibe code an app and get it through App Store review and get it out there just for fun. And so I did that and I was like, that's cool. Never going to do it again, but it's out there and I left it out there.
But as a creator, and a lot of creators do this, you know, we talk about apps all the time. I do lots of videos with the top 30 apps for this. And sometimes you just have an idea that's like, I wish I could try to make this into something, like manifest this into a real thing. Or there's an app that I just want to tweak a little bit, and like Reminders, which is what I was using before, I just want to see can I actually now make something with the tools available. And turns out you can.
So what's your AI coding setup like? When you sit down to make something, how does it all go?
So I used Codex before, but for some reason me and Claude just get along better. Uh, not the same... well...
I feel the same way. I could not describe why I like Claude Code better than Codex at this point, but I do.
It's verbose. Well, I can describe because I started using Claude more and more, and when Opus came out, I realized it is like oneshotting a lot of my prompts into just real things and they work. And so I've stuck with Opus 4.5 for everything, and now I've given it API key to my App Store Connect. I've given it access to my Cloudflare account. So all I do now is just go into Claude Code and type something and everything else just happens. I never have to leave that app.
And I did try back when it was 5.6. I downloaded Codex. I put it on 5.6 ultra, which was apparently a bad idea because then it like disperses multiple agents. But I had it work on an iCloud sync issue for four days, and I had my iPhone and my iPad connected to Xcode with full access. I saw weird code things running on my iPhone screen. I didn't know what was happening, but I found that I did not prefer the UI and design decisions that ChatGPT and Codex made. And I would have to like tell it more like, 'Bro, put more padding around this radio button' or whatever. And Claude just gives me better UI. And now because I have so many conversations there, I've made so much with Claude, I can have a very small prompt and it kind of already has a bunch of context for what I want.
That makes a lot of sense. The real working relationship with these tools is a real thing. It feels strange to say it that way, but it is a real thing. Um, I wanted to talk to you in particular because you've made two different kinds of apps, and the first I think is the most sort of straightforwardly understandable vibe coding tool, which is like, I have a single thing that I need to be able to do. I'm going to build the tool to be able to do it. Uh, you've built a bunch of these for things like pulling transcripts out of a video or for quickly pulling the audio out of a video that you download. Like I can sort of see you going through your day-to-day workflow and being like, 'This is a thing I don't have. Can I just make one of these?' Are these things you're building in like an afternoon and just sort of two prompts, ship the thing, move on with your life. You solved your problem.
Yeah. I mean, a couple of them were like two hours from first prompt to shipped on the website. You know, I wanted an app called Leveler because I do a bunch of video podcasts, and when I download that video file from Riverside, I had to tweak the audio a bunch just to get it to sound leveled, and I didn't want to have to edit the audio and then re-export it from Final Cut. So, I just wanted to make an app that leveled the audio between me and my co-host and it sounded good. And so, I gave Claude a raw file from Riverside, I gave it an audio file that I edited how I like, and I said, 'Make it sound like that.' And it built an audio engine in like, what, 20 to 30 minutes. I tried it. It worked. I said, 'You know what? Put this on the website. Let's just ship it.' And so now anybody can download that for free.
So why ship it either to the App Store or just to the website rather than just keep it for yourself?
Well, the Leveler one, I figured if I have this issue, other people have this issue. And turns out a lot of people, especially beginning or mid podcasters, they were like, 'Yeah, I struggle with audio levels.' And so this just lets me drop something in and get it out. And for another, like I made Captioner, and I was using the Edits app from Meta for adding animated captions to my reels, but I had to do it one at a time. I had to AirDrop stuff to my phone to do it. And I was like, man, if I could just make an app that's free, I could just do this in batches. And so I made Captioner where I can give it 10 reels because I make them in batches, and then it adds the same captions on all of them. It exports the files in the source folder and I'm done. And I'm like, I think other people would benefit from that too. And so it's free and open source and anybody can download it.
How does the release process for those go? Like, are you, when you make one of these, are you making them thinking, 'I'm going to ship this to the world. I need to make the UI understandable. I need to have X, Y, and Z feature that other people might want that I've seen on Reddit.' Or are you like, 'I've done this one thing. If it works for you, terrific. If not, terrific. Here it is.'
Basically the latter. Like with Captioner, I said, hey, I want it to do this. I said, hey, you know the Edits app, Claude? Like kind of like that but different. And uh, you know, kind of gave it a one-paragraph prompt and it basically did it one shot. And I said, you know, just upload this as a DMG file to my Cloudflare website, and it just ships it, all packages it. And that one I thought, well, people might want this as an iPhone app, and I kind of gave it that prompt at the beginning. I said, I'm going to ship this on the website as a Mac app DMG, but also keep in mind I might ask for an iPhone app. Claude was like, sure, bro. And then I did. And so now that one's in the App Store because I just told it, 'Hey, try to make an iPhone version. Let's see if it works.' It did after a couple prompts. And now that one's live in the App Store, too.
What responsibility do you feel for those apps? For other people who download them and have experiences with them? We're going to get to your other app, which I think you have a really different set of responsibilities about, and I want to talk about them. But for something like this, which you're fairly clear about, I made this very quickly. It's very simple. It is designed to do this one thing. It seems to... do you feel responsible for it and for other people's experience of it?
Because these are free and open source, I don't feel particularly responsible, but I like being responsible. And so people will message me, like James Cridland of Podnews was like, 'I love this Captioner app. I would love for it to do blank lines and let me edit the timing.' So I will, as I see that reply on social media, take a screenshot. I have a Shortcuts automation that takes that screenshot, puts it on my desktop, and renames it. And then when I'm back on my Mac, I will take all the screenshots that I've gathered from people just asking for stuff, give all those screenshots to Claude, and be like, 'Hey, can you add and fix this stuff?'
And that's it. It just does it all, which is crazy. And it's...
So your bug report system is you take a bunch of screenshots...
And you say, 'Claude, here's all the things everybody's mad at.' And then it goes out and fixes them.
Literally that. And I have one guy, he sends me an email every week for CanDo, which we'll get to, with like a long list of bugs and requests that he has found, which is very helpful. And I command-A, select all, command-C, paste, and Claude to be like, Claude, analyze this, let me know what's actually fixable. It gives me a report that's very verbose that I don't read at all, and I say, yeah, go ahead, do that. And then it ships those updates.
One of my favorite things is dealing with Claude Code and it's like it has a very specific technical question and I'm just, I'm always like, buddy, if I knew the answer to this, we would not be here. Like, what are we doing, Claude Code? Just go.
A lot of times I say like, give it to me in like two sentences. Or like, give me three options, and even then I don't even read all that.
Yeah, treat me like I'm the idiot we both know that I am, Claude Code. Like, come on.
Yeah, and it's getting better at interacting with me, I think, because I give these instructions periodically. And now like I have a bunch of conversations in my Claude Code sidebar, but it has all the context. So sometimes I'll be in my Captioner conversation and I'll be like, oh, someone asked for Swedish translation, can you put that here? And in the transcriber app it's like, sure buddy, I'll do that. And then it just does it, and I don't even have to like switch. It just...
All that context. Yeah, it's really interesting. And I think part of the reason I like these apps is because they feel like the perfect sort of personal software outcome of vibe coding, where you've just described a bunch of things that some people have, but like there's no company in building any of the apps you just described, right? People are not going to pay for it. There aren't even enough people who need it to actually have this thing work. But it doesn't matter. You can build it in a couple of hours. People who want it can access it. They can build it for themselves. They can fork it and build it into something different. Like if you look at a world in which everybody has access to coding tools, it's tools like this that come out of it that make me the most excited. I think it gets messy in like, should these things be approved by the App Store and live next to things that are run very differently and built very differently and managed very differently. All that gets very messy, but this feels like an AI victory to me in a lot of ways.
I think so. And I figure if it solves even a small problem for me, there's some small group, even if it's just single digits, but it's usually not, that have the same kind of problem. And like when I released Leveler, a bunch of people said, man, Levelator was a great app, and some users have tried to perpetuate it. You know, the people who made it kind of left it aside. People are like, 'This is a great replacement.' And that was a free app. Mine is a free app. And so, I think it's kind of a win-win.
