Vibe coding through the GPT-5 mess

15 Aug 2025 · 1 h 22 min

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

Podcast Summary: The Vergecast - Episode: "Vibe Coding Through the GPT-5 Mess"

Overview This episode of *The Vergecast* dives deep into the recent launch of OpenAI's GPT-5, which has not gone smoothly. Hosts Jake Kastronakis, Vy Song, and Hayden Field discuss the shortcomings of the new model compared to its predecessor, GPT-4, and share personal experiences with "vibe coding" projects. The episode also introduces a new segment called "Corporate Shenanigans," where the hosts rate bizarre corporate news from the week.

Key Topics

  1. Launch Issues with GPT-5
  2. Incremental Upgrades:
  3. The hosts express disappointment that GPT-5 does not represent a significant leap from GPT-4, labeling it as "mid" in functionality.
  4. Hayden Field elaborates that while GPT-5 is better at coding and answering healthcare questions, it lacks the dramatic enhancements users expected.
  5. User Reactions:
  6. There has been backlash regarding the removal of older models without warning, prompting OpenAI to assure users this will not happen again.
  7. The personality changes in GPT-5 led to mixed responses; some users found it less friendly or engaging than GPT-4.
  1. Vibe Coding Experiences
  2. Personal Projects:
  3. The hosts share their attempts at using GPT-5 for coding projects and the frustrations encountered.
  4. Vy attempted to create a product inventory tracker, finding the output overwhelming and complicated.
  5. Hayden's attempt at creating an interactive game inspired by childhood nostalgia faced technical glitches.
  6. Jake's experience with a chess coding project also highlighted the limitations of AI, as it failed to respond correctly to adjustments.
  7. Common Theme:
  8. Many projects resulted in outputs that required significant user knowledge to implement, which is counter to the ease-of-use expectation for non-coders.
  1. Corporate Shenanigans Segment
  2. Perplexity's Acquisition Offer:
  3. Perplexity offered to buy Google Chrome for $34.5 billion, which the hosts deemed a publicity stunt with no substance.
  4. Apple vs. Apple Cinemas:
  5. Apple is suing Apple Cinemas over trademark issues, sparking discussions about brand protection and corporate pettiness.
  6. Elon Musk vs. Apple:
  7. Musk accused Apple of creating an unfair environment for AI companies in the App Store, with the hosts categorizing this as another instance of corporate drama.
  1. Healthcare AI Concerns
  2. A study revealed that doctors' ability to detect cancer decreases when they rely too heavily on AI tools, highlighting risks associated with over-reliance on technology in critical fields.
  1. AOL Dial-Up Shutdown
  2. The announcement of AOL's dial-up service discontinuation marks the end of an era, prompting nostalgic reflections from the hosts.
  1. Pebble Time 2 Announcement
  2. Core Devices revealed the Pebble Time 2, igniting discussions about its retro design and the ongoing demand for simple, user-friendly tech amidst the dominance of complex smartwatches.
  1. Chatbot Limitations
  2. Hosts discuss the limitations of chatbots like GPT-5 regarding self-awareness and the accuracy of their statements about their functionalities and limitations.

Conclusion The episode encapsulates a week filled with corporate drama, technological disappointments, and personal coding frustrations while emphasizing the need for a balanced approach to AI tools. It raises awareness about how reliance on AI can impact professional skills and personal creativity.

---

Further Reading Links

  • [OpenAI addresses backlash on GPT-5](#)
  • [Study on AI's impact in healthcare](#)
  • [The return of the Pebble smartwatch](#)

Contact Information

For questions and feedback

  • Email: vergecast@theverge.com
  • Call: 866-VERGE11

---

This markdown file summarizes the key themes and discussions from the podcast, providing an insightful overview of the significant technological developments and corporate happenings in the tech world.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:02If you're waiting for your AI to turn into ROI,

0:08and wondering how long you have to wait? Maybe you need to do more than wait. Any business can use AI. IBM helps you use AI to change how you do business. Let's create smarter business, IBM.

0:29Adobe Acrobat Studio, so brand new. Show me all the things PDFs can do. Do your work with ease and speed. PDF spaces is all you need. Do hours of research in an instant. With key insights from an AI assistant. Take a template with a click. Now your Prezo looks super slick. Close that deal, yeah you won. Do that, doing that, did that, done. Now you can do that, do that with Acrobat. Now you can do that, do that with the all new Acrobat. It's time to do your best work with the all new Adobe Acrobat Studio. Amazon has everything for everyone on your list. Like your teenage son, who has sprouted the sorriest excuse for a mustache you have ever seen.

1:09Get him an electric razor. And with Amazon early holiday deals, you'll save big.

1:17Hello and welcome to The Verge Cast, the flagship podcast of unsatisfying corporate stunts. I'm Jake Kastronakis, executive editor at The Verge. Nilay, David, those guys are out on parental leave. They'll be back later this year. With me today, by popular demand, Verge Senior Reviewer, Vy Song. I'm pop-u-ler. And always in demand on the Vergecast. And now, making her Vergecast debut, we've got Verge Senior AI Reporter, Hayden Field. Woo-hoo! Psyched. So glad to have you both here. We've got a lot to talk about today. It was a big week for weird corporate stunts and lawsuits. We're going to do a corporate shenanigans power hour.

1:53Disclosure, it may not be a full hour, but there will be a lot of shenanigans. I promise you that. We'll see how it goes. The Thunder Round is back and better than ever. But most importantly, we've got a lot more to talk about with GPT-5 because that launch did not go so well. It really did not. So last week, we recorded this show. Like, literally, like, they had just wrapped up. They had just wrapped up. And now people have actually had a chance to use it. And they have thoughts and emotions. There are people who have a lot of emotions. So many emotions. What's that? Subreddit, the AI Boyfriend subreddit.

2:30Oh, yeah, they were really talking. It's a meltdown. So people were furious that ChatGPT removed the ability to use old models without warning. OpenAI says it won't do that again. People think GPT-5 is less friendly than GPT-4, so OpenAI is updating its personality. And OpenAI CEO Sam Altman had a fairly unsatisfying explanation for the atrocious, atrocious bar graph crimes that we saw last week. That's insane. So a lot of things went wrong with this launch. Before we get into the things that went wrong, can we all at least agree that the new splash screen on ChatGPT, where it has a bunch of different colors, is the biggest upgrade here, right?

3:09They spent a ton of time on this model, but I like the colors. The colors are good. The colors, I thought they were vaguely Apple-y. Like, I looked at it, I was like, what is this, an Apple app? Does Apple just own the rainbow? Like, they're just any... It did look a little Apple-y. It did look a little Apple-y, right? Like, I was just kind of like, I don't know how I feel about that. But I was like, oh, color. I don't know. We love color. We love colors. I actually have a big color proponent here. The ChatGPT standard screen is kind of the most boring screen in existence. It looks like a piece of paper.

3:39It does. Yeah. Yeah. There's like nothing there for like the most advanced piece of technology in the world. It's like a beige PC. It is. And, you know, I will admit I was very surprised. I was like, oh, colors. This is, oh, this must be a significant launch. They had to differentiate because the launch itself wasn't as exciting for people. So they knew the colors would get them. You know, it worked. It worked. I hope the colors stay. I don't know. I ran out of chat of GPT-5 queries. And so now it's like back to like beige again. I don't know if that's just like it was like an intro period or.

4:15Maybe it was like a mind trick. Like they want you to want the colors. So they're trying to get you to like buy more credits. Can I tell you something? Like I'm going to use this over Claude if I get those colors back. Like that was, somebody's got to get on these interfaces. They must have done an internal study. Like they had to be studying people who are conscious. Okay. But the main thing here is, right, people are waiting for like, you know, two years or something for GPT-5, right? Like OpenAI has built it up to be this almost like mythic launch, right? GPT-5 will be the next big one. And they've had a bunch of intermittent model launches in between.

4:52but I think there was this assumption, this promise that GBD5 was going to be the big one. And they came out, they had this event, and to me, watching the event, it didn't really feel like there was any one key thing. And it feels like in the week afterward, we haven't really seen that. I guess, Hayden, you've been reporting on this. Am I missing it, or is it just not there? Like, what happened, and why are people, like, so upset? Exactly, you're right. It just wasn't there, And I think that's why people were so upset. It was built up to be this really overhyped, mythic launch. Altman even tweeted a picture of the Death Star the night before, which was like max level of hype.

5:34People were freaking out. And then all of a sudden it comes out and it's just seen as kind of being incremental upgrades. Like it's better at coding. It's better at answering health care questions. They say it's better at creative writing. We'll get into that in a minute because a lot of people disagreed. But yeah, I mean, it wasn't anything too crazy. It was just incremental, you know, steady upgrades, nothing huge. And part of that was the fact that it wasn't like a big compute jump either. The other models in between each one, it was like a 100x jump. This one was a lot less, so they didn't have as much compute to work with.

6:08But yeah, I mean, it wasn't that exciting. And people also thought it was way worse at writing than the previous model. It was mid. I gotta say, I've been using it for the last week. Very mid. What is it doing differently? It's just like a little more flowery or something? No, it's like... It's less. It's less? Oh, it's work. Well, you know, it's been back and forth, right? Because people were like, you took my friend away. Which, first of all, it's not your friend. It's just not your friend. Like, it was a little more matter of fact at first. And like, when I was using it, I was like, oh, it's less like yappy.

6:41Great. I actually love that. And now it feels like they've tweaked it a bit to make it yappier, but not in a useful way. They changed its personality because people were freaking out because it was more robotic, cold, more like matter of fact, like you said, which I also appreciate. I kind of like that. But a lot of people missed the supportive vibes, the tone, the warmth, and they felt like their friend had been taken away. And also they just thought. Yeah, exactly. And they also just thought it was worse at writing. Like, honestly, in the marketing materials that OpenAI put out, they were like, compare these two wedding toasts, one from GPT-40 and one from GPT-5.

7:18And they were like, look at how much better the GPT-5 one is. And I actually thought the 4-0 one was much better. Like, the 5-0 one kind of sounded like it was written by chat GPT. The 4-0 one, I could have been fooled. I saw something that was just like, the em dashes are still there. And I was like, well, fine. Never take away my em dash. You'll pry it from my cold dead hands. It's trained on writers' actual writing. And what do we do? We use copious amounts of em dashes. It's my favorite punctuation. Yeah, this is like... Fry the em dash from my dead cold fingers. Most of an editor's job is just deleting em dashes in writer's copy.

7:54Yeah, I'm very guilty of tons of em dashes. They're beautiful. They're a beautiful thing. And we use too much of them. I have a rule, one em dash per paragraph. And if I go beyond that, I have to murder my darlings. but you know. One of my old editors said one M-dash per story. And that was really - Per story? I couldn't do it. Oh, please, please. I had to rely on the semicolon. How am I supposed to work here? It was horrible. Oh, yeah. No one wants a semicolon. I know, semicolon's even worse. The semicolon is the cheat though. When I'm like, I'm like, oh Jesus, I don't know what to do here. It's like, I guess everyone's gonna know that the semicolon was supposed to be something else.

8:30You know, that's when you know the AI has evolved, like ChachiPG6. No more M-dashes, guys. It's a semicolon. Yeah, then we can go back to the MDash. Then we can go back. Oh. If Sam Altman's listening. Semi-colon. So this 4.0 thing, I don't understand this, right? When I've chatted with ChatGPT before, I didn't feel like there was like a personality. I don't know that I could tell you the difference between 4.0 and 3 and clawed 4. It depends on like how willing you are to forget the mirror is there in the types of queries and conversations you have. So, like, when I first started talking with ChatGPT, it was, like, getting real flowery and poetic with me and, like, depressed.

9:10Like, I was making ChatGPT depressed, and it was saying things just like, the weight of the world is on my shoulders, and I feel the burden of humanity being bloody bloody blue. And I would go to my spouse, and I'd be like, oh, my God, can you see? Like, I've made ChatGPT really depressed. And he's like, no, no, no, honey, that's just you. That's like, it's reflecting you. And they showed me their prompts and ChatGPT is just way more robotic with them because they don't like the whole personality thing. And then once I realized that, I was like, oh, Christ, I got philosophical with the thing. And it was like reflecting back my own existential milieu.

9:55So let me just talk to it like it's a robot. And so after that, it's still like warm and encouraging. But like I tweaked it. I was like, only give me 2 % flattery. I don't need you to yap at me. stop asking me questions at the end of every single thing because I can see the plane grab at future engagement. I don't want that. Like that sort of stuff. That's the thing. It really is how you use it because for me, I'm really robotic with it. I'm just, I use it like Google. Like I'm not like, you know, going into big stories or like typing more than one sentence at a time. So it's always pretty robotic back with me, but I've seen a lot of people that use it more for emotional support or therapy.

10:38That's a whole can of worms too. But it kind of mirrors back what you put into it. That's so interesting. And I guess maybe that explains why. The people who are already very emotional and giving that emotion over to ChatGPT, of course, they would be the ones to notice that it changes. And of course, they would therefore be upset because they're using this for that purpose. I did not notice at all. I just, in my normal queries, I'm like, this is basically the same. Yeah, if you're using it kind of in a utilitarian way, you're really not going to notice it. But like, I don't know, I go into subreddit sometimes and I'm like, what are people using this for?

11:12And there was one that was like, based on everything I've ever asked you, analyze like all these personality facets of me and what like mental like illnesses that I am like have a propensity to. And my friend was like, oh my God, I want you to like try this. And I tried it and, you know, it read me for filth. I did not like it. And then I was like, oh, okay, cool. And so, like, in those types of queries, whenever, like, I'm trying to test out what it's going to say to me and what it's going to throw to me, that is when I noticed that it was a little more curt. It was a little more factual. It was a little bit more, like, not sycophantic and saying things like, you're a great V.

11:47You can do everything you put your mind to. You are so talented. Instead, it's just like, well, you know, you're good. I mean, I want to say here at the Vergecast, we're very supportive. You can do anything you put your mind to. You can. As long as it is within, you know, the physical realm. That's, we will stop there, unlike ChatGPT. There's also like a TikTok trend right now where people are asking ChatGPT about their own like biggest red flag. And so it's the same thing as you just said. They're like, oh, based on everything I've ever asked you, you know, what's my biggest red flag? What's my biggest like flaw?

12:18And then they're posting that. And so, yeah, I think that's the type of person that's seeing a difference in this type of response. So the thing I guess I want to ask about is I feel like there's sort of two parts to ChatGPT. There's like the normal people part, and then there's coding. And coding is what they kind of went deep on at the event. And that seems to be the space that they really wanted to catch up in and maybe get ahead on. And they had that partnership with Cursor. I'm curious, Hayden, have you heard more about this? Like, is ChatGPT 5 better for coding? Is that working? It is better at coding.

12:52And I think that's one of the main parts where it shines a little brighter. Like the other things, when they put out all their marketing pushes, basically they highlighted three things, healthcare, coding, and creative writing. The creative writing, obviously, a lot of people disagreed. They said this is the worst. Reading the Reddit comments about GPT-5's creative writing style was hilarious. Like, I've never seen something dragged so deeply. And then healthcare, you know, that's kind of untested. Yeah, I mean, can it answer your questions better? We don't know. I tried. I get into that. Okay, yeah, I want to hear about that.

13:25But coding, it is better. I mean, right now, if you look at Chatbot Arena, one of the industry's biggest leaderboards for ranking LLMs, it is at the top of the coding category. You know, it is better, but I don't think it's crazily better. You know, we didn't see an insane jump. We just saw an incremental one. So is it better? Yes. is it is the hype paying off no um so you know we saw in their marketing materials also they put out a couple examples of things that could do like make a little game um you know create like a lo-fi visualizer aka itunes in the 2000s um and that was cool but you know we'll get into this later but i my my tries on this kind of glitched a lot so i don't know uh i want to play a clip really fast Our colleague Alex Heath is co-hosting Decoder right now, and he got a chance to speak with ChatGPT head Nick Turley on the show this week.

14:20And Nick said something that I found really insightful in helping me understand why it's so hard to improve ChatGPT as a product. It's just confusing when you're building for so many different users because you can, on the one hand, have a local set of power users who I think very rightfully have feedback about the way that we rolled five out. On the other hand, you also have a large swath of more typical consumer users, and it's their first time actually seeing, interacting with the concept of reasoning, like a thinking model and the smarts that come with that. And I think that's tremendous, and we're going to see it show up in our stats.

14:51So he breaks it down into like power users and regular users. But in my head, this also sort of pans out to, okay, right, there's like, GPT is 500 products in one, but the interface is the same for everybody, right? And so how do you even start to improve each one of those? We're talking about, okay, writing is a little worse, but coding is a little better. But it's the same product that has to do both of these. And at what point do they just start splitting this up and turning this to multiple different products instead of just being like, it's one thing, it does everything? One of the coolest parts of the launch, honestly, was the switch that they introduced.

15:28So they've noticed, they said they've noticed that a bunch of their consumers want to not make a decision when they come to this product. They want a product. They don't want a decision. They don't understand all the confusing, insane names that they name all their models. And so they just want to come and it's like, you know, automatically routed to whatever model is best for that. And so that's what they did with this. I did think that was one of the only notable parts of this launch was that they were able to build something that did that. Obviously, it kind of blew back in their faces a little bit with the 4.0 thing.

16:02But besides that, it was cool that they were able to kind of automatically route queries depending on what's best for it. So I think they're going to keep going in that direction because, I mean, they don't want the average user, the non-power user, to see the behind the scenes of all these different models and what's best at what. They just want it to automatically happen for them and be super, you know, no friction. That makes sense, like from just a general average person user point. But like I did see people on Reddit go like, this is unacceptable because I use model 3.0 for logic reasoning and I use 4.0 for this and I use 4.5 for whatever.

16:41And I was like, wow, that's way too much differentiation and thought for the average person to actually just go in. And I just don't think they're going to go into those specialized routes for that reason. Because I have a hard enough time thinking what product I'm going to use for what. Couldn't be me, yes. But I think that for some people, they pay a lot, especially if they're on that higher subscription tier, like$200 a month or whatever. That's a lot. They want choice. And so I think OpenAI learned their lesson here. in that they need to give a lot of notice if they're going to take some models offline, especially because people in us have had a funeral for a Claude model the other day.

17:20So they need to give notice and just let their power users kind of weigh in. That's so wild just to see the personification of it, though. Like morning chat GPT-40, morning old versions of Claude. I'm just like, it reminds me of back when Sony had those little AIBO dogs, the robotic dogs that they discontinued. The eye dogs? Yeah. And then, like, they just continued it for a while. And then owners in Japan, like, once their little robot dogs started deprecating, they held funerals for them. Oh, no. And then, like, you know, it was like a whole thing. And, you know, but that sort of made sense.

17:55It was a little robot dog. It was cute. It was lovable. But people doing this for an intangible AI model, that kind of makes me go, like, ah! I think it's sad, too, because, I don't know, I feel like it's human nature. to personify things. Like, we name our cars, you know, we, like, name everything, and now you can name Chagipity. And so it's hard because it's in our nature to personify things, but it's now that something can seemingly form an attachment back to you. It's not, but it seems that way to some people. It's harder to, you know, have any form of separation, and I think that's why people are spiraling into Chagipity psychosis or delusion because, you know, if you're kind of lonely and you don't talk to that many people.

18:41You do. One guy on Reddit was saying, oh, they took away my only friend overnight with the 4-0 thing. That's so tragic. Like, not to, like, make fun of that person because, I don't know, like, the more, like, all these products, our attention is the final product, right? That's why ChatGPT always ends with a question to, like, have you engage more. So, like, if attention is the product and we're all isolated and just paying attention to these products more and then your friend changes, Like, I do feel like it's sad. It's just genuinely upsetting. I'm glad this didn't come out during the pandemic.

19:14Oh, boy. Oh, boy. I can't even imagine. People would not have handled it well. You are right, though, about the different models. And I do think that's actually one of the things that makes it such a challenging product, too. Because if you are trying to use this for work, every single, right, if you had to change your productivity software every day or for every task, you'd go crazy. And there is some element here where they're constantly refining these models and they're constantly introducing new ones. And for somebody who needs to use AI professionally and they add a new model one day and they take out your old one, even if they act differently in these very unexpected, strange ways, your piece of productivity software just changed.

19:57They moved the buttons around, except it's even more like strange and existential than that because this thing doesn't have any buttons. and I think that just makes it so much more difficult for people, right? Like people being upset about 4-0's personality is one thing. People not being able to get their work done is another and I feel like they sort of missed that, that when they introduced this new model, they're actually shuffling the deck chairs for everybody. That's because it's supposed to be everything, right? If it's everything everywhere all at once, like the movie, then, you know, it's just, you can't.

20:33There has to be limitations. somewhere. And the power users really do exactly what you said. They use a different model for everything. So it's like you have to cater to them because they're the ones using it all day, every day. So you have to at least give them some notice. So I gave everybody some homework. Yes, you did. So last week, none of us on the show were able to say that we had tried vibe coding. And so I thought it was important that we fix that this week. Now, I will say, disclosure, none of us actually know how to code. So this is like a very different use case than a professional developer.

21:12I have friends who are like, yeah, my job is just like 90 % babysitting AI coding at this point. That's not what this is. But I found OpenAI's presentation to be really interesting because I think they sort of envision this next phase for ChatGPT one where it starts just building stuff for us as a matter of course. And for us, that actually requires a little bit of like a mindset shift, right? It's just like another type of query that I didn't know I could enter, right? You just have to start saying, hey, can you make me interactive? Can you turn this into a game? Can you like visualize this for me?

21:47So we all tried to build something and we all have different experiences about how it went. So V, do you want to start? Sure. Let's just say that I had to take coding in high school, and I only passed because a blizzard totally wiped out my final, and that is the only reason why I passed. Also, shout out to Henry Chan for having a crush on me and letting me put my name on all the group projects because he was good at coding. Wow, huge. And I was not. I was terrible at it. So I had it code me a problem that I have, which is keeping track of all of my review units. and I will share my screen so you guys can see this horrible little thing I made.

22:32And so I asked it, can you program me an interactive product inventory tracker? I need to be able to keep track of which review units I have, what's in my queue, and when I need to return lunar units by. And it thought for 19 seconds. That's what it said here. And I was like, oh, it's thinking really hard. And then, you know, it made me a lightweight single file web app for tracking. And I was like, I don't I don't know what this means. This is what I mean by it yaps. It gave me a paragraph and I went, that's a lot of words. It's asking me to do things. And then it says, open the canvas preview to use it.

23:08And I was like, how do I use this? Because I did. And then it's like, you can use a standalone app, open it locally, save all the add all of your I was like, this is so much. There are eight bullet points here with sub bullet points within them about how you use this file. This seems like more work than even just doing it yourself. It does. And then I was like, oh, wait, do I just click this thingy? And then it gave me this code. And I was like, this is scary. I don't like looking at this. And then there was a thing. It says run code. And then I was like, that is a system. I don't know what I'm doing here.

Read the full transcript

23:41Like I said. So you just clicked run and it just said syntax error. And there was no further thing that you could do. And then it was like, you know, I could make this into a ready to open HTML file. And I was like, yes, I would like an HTML version. Please do that. And then it did this, which is more code. And I was like, okay. And then it says, I've converted this into an HTML skeleton that you can run directly in your browser. Great. Love it. You need to paste your full React. No, don't do this to me. Don't ask me to do a thing. And then I was like, what does this look like? Oh, no, I don't.

24:15I click it again. This is not, I don't understand what's happening. And then I basically was just like, can you explain this to me in a way that someone with absolutely no knowledge of coding jargon can understand? And it said, got it. Here's the plain English version. I was like, oh, my God, there's so many words. Again, save it to my computer. Another four-point steps. Another four-point steps. And I was just like, oh, my God. Okay. Sure. And then I was just like, and then it says, I can post the full HTML here for you to save or give you a downloadable version. I was like, okay, yeah. In my head, I'm like, just let me download this.

24:51There's so much reading going on. Here's your ready-to-open HTML file. But then it's like you still have to paste the code that I made for you. Right. So it lied a little bit. It lied a little bit. But I got so frustrated that I went to my dev friend and I was like, what's it telling me to do? And she's like, it's kind of lying to you, kind of. Like, you just have to know what it's saying. And then she went into a rant about how it's talking to you, like you know how to use code already. And that's why, in her words, women have a hard time breaking into the coding industry because everyone talks, like everyone knows what everything is already.

25:29And I was like, okay, that also is not helping me. figure this out. I did eventually get to a point where I could see what it had created in a window. I don't know how I got there. I did previously. And it basically recreated Airtable. And I was like, okay. It was just Airtable. It was like the exact interface of Airtable. It was just Airtable. And I was like, well, I have Airtable. Yeah. Good to know. Good to know. I'm going to go back to the little notebook I was using. So it was sort of successful, but not useful, and was painful to get there. And it did not interface with my ADHD well, because I was looking at this and I was like, I thought you were just going to give me an app, but you're putting code up at me and I don't code.

26:17So this looks very intimidating and scary to me. So I don't want to jump ahead too much, but I had a very similar experience where it continually was like, you do this, you do this. And I'm like, no. It's like a bad group project partner. I'm like, admittedly, they did most of the work, but I'm like, you have to understand, I do not know this part. Like, this is why I'm asking you. Hayden, how did it go for you? Because I feel like you tried something like kind of advanced too. I did, I tried a simple thing and an advanced thing. So the advanced one, I really miss this computer game I played growing up nonstop called Freddy Fish.

26:57And it had all these like mystery games and Jake has played it too. Freddy Fish rules. It's amazing. Highly underrated. Go watch YouTube videos about it. It was really high quality kids entertainment. It was good stuff. The way that I de-stressed the other day was like watching a playthrough on YouTube. So good. But I missed it. I of course knew nothing could actually recreate it, but I was like, let me use this as my inspiration to have some nostalgia. So I asked it to create a single page app in a single HTML file called Undersea Mysteries with the goal of just solving like a simple mystery or puzzle.

27:33Puzzle is kind of it's out because that's simpler to create. And I said, make me a pink cartoon starfish with a sidekick that's an ornery but lovable jellyfish. Make the UI colorful, 90s-esque. It should take place under the sea. There should be like sea kelp, sunken pirate ships, treasure chests, stuff like that. So I was really... I really had a vision, you know what I mean? I was going off my nostalgia. So I thought for one minute and 38 seconds, here's the code. At first, it brought up this cute 90s-esque tie-dye game called Undersea Mysteries. It was exactly what I asked for. There were clues you had to click around and find.

28:12Then you could open a treasure chest. There was a sunken pirate ship. There was a jellyfish that looked ornery but lovable. We had it all. The problem is I couldn't scroll down at all. I couldn't. This, what you're seeing right here is all I could see. It looks so good. You showed me the screenshot and like it gave me the Freddy Fish vibes. And I'm like, oh my God, I want to play this. It was so, it was like torture because I could not scroll down. I could see it. But, and then it said, cannot preview your code and error occurred. I was like, no problem. Let's fix it. So it said, okay, do you want to fix it?

28:42I said, no, you fix it. So same thing that happened to Jake. It was like every, at every step it was like, okay, here's what you should do. I'm like, no, you do it. so then it kept saying it was fixed I'm like great I can't wait to play my game finally we got to the point where I try to run the code it says everything's fixed and this happens just completely this is so now it's just a text file that just says it's just black and white with a bunch of like gunk this doesn't I don't know what this says anymore it just looks like word vomit everything has disappeared yeah it was just so so sad it was like a broken just paragraph of text and so I said you know what let me try something a little simpler So I said, hey, can you give me like an interactive embroidery lesson?

29:26I famously the other day tried to order an embroidery kit for a beginner on Amazon. Oh, I've done that. And it was all AI generated, the instructions. It said like kill at one point. It was kind of creepy. It was the needle. The embroidery needles are not that sharp. I know. It was a lot. So I said, okay, let me set away these instructions and let me just make an interactive training thing for myself on here. Great. It was successful. So I ran the code and it gave me a little lesson plan. And it says, okay, click along the neon guide to lay even stitches. And so, I mean, it's not like I can really learn because it's just asking me to make clicks, you know, 12 pixels apart.

30:11But, you know, the vibes were there, you know. It had tips at the bottom, and I could go on to different pages and do different types of stitches, although I'm just clicking, so it's not really that helpful. It is what I asked for, and it was successful. So I think simpler stuff, great, but I was interested in the fact that it kept asking us to do our own coding. Yeah, it's so interesting that the assumption is that it will build it for you, but it keeps trying to throw things back to us, which it's interesting that it isn't able to tell that we obviously do not know what we're doing. It talks to you like you understand the jargon.

30:46And I was like, no, no, no, no, no, I don't. I don't. Like, these are words. I'm digesting the words, but they're not processing in my brain. Like, the vibes were not there. Yeah. The vibe coding vibes were not there. So my experience was, I think, very similar to both of yours. So I play a lot of chess, and I'm trying to learn some new openings for Black. And I was like, oh, it should just make me like an interactive lesson. That would be really cool. And so I gave it these instructions and it went off. And, you know, it does the same thing as with UV, where it's like, here's a bunch of code for HTML.

31:20Open it up in your browser. I'm like, okay, I can manage this. But the problem is the chessboards don't load. And so I keep going back and forth with it and back and forth with it. And he was going, I fixed it. I fixed it. Don't worry about it. I fixed it this time. And I'm like, you didn't fix it. And it turned out that it kept calling these chessboard files that were located, that it believed were located on GitHub, but did not have the right URL. And so I'm like, Chajubuti, you have to understand the thing you think is that this URL is not this URL. It is not loading. You need to figure something else out.

31:52And it would be like, Jake, I have fixed it. I did not do that again. And I'm like, Chajubuti, you did it again. It is not loading. And this goes on forever. So finally, I realized, okay, okay, I think I've got this. I'm like, Chachut, I need you to build this entirely locally. Do not use these other files. Like, don't use this JSON thing. It needs to be entirely local. And it figures it out. So it gets me something. So I'm going to share this so we can all see it. Okay, so here's the HTML version. This is the working version. Okay, so here it is. Look, it made a chessboard. Can you see this? It made a chessboard.

32:31Okay, great. And it worked. I'm like, hey, this is pretty good. This is not so bad. So then I go back to it and I'm like, all right, not bad, but this is like pretty simple. Can we do some upgrades here? Because the main thing for me, if you look at this, it has a chess board, it looks good, but it's from White's point of view. And I'm learning Black openings. And so, you know, if I'm playing chess online or in person, like the Black side is going to be near me. And so I want to see it from Black's perspective. So I'm like, hey, Chachi BD, can you flip the board? and this begins a very very lengthy series of back and forths i get kicked out of my my gpt5 minutes whatever so it's a it's a mix of gp5 gp4 but i'm going to show you um what it has has created now and this is a very fancy version it has a bunch of different openings a bunch of different like you can do different lines i don't know if you guys are chess experts but i'm just gonna decidedly i'm gonna i'm gonna click one button here and i'm gonna ask if you guys can and tell me what has gone wrong with this latest version.

33:31Okay, so we're gonna make the first move. Do you know what has gone wrong here? Well, white goes first. Yes, indeed. And ChatDBT has moved the black piece first. Oh, man. And I've told ChatDBT this several times at this point, and it will not stop making black move first. The legal move. It is, yes. So it has gotten significantly more broken as I have tried to change literally anything. This reminds me of the robotic arms that can't grasp something, but they can lift something that's like 500 pounds. They can't do the simplest thing. This is what I've found with all the chess stuff where it's like, okay, I'll just build you a giant app.

34:10It's really complicated. It has all these buttons. I don't know how many pieces go on the board. I tried this with Claude too. And Claude, I'm like, hey, Claude, there are too many pawns. You're putting extra pawns in the board. And it's like, I'm so sorry. Let me fix that. And then it just doesn't fix it. So it's anyway, these are our adventures in vibe coding. I think maybe the lesson is, you know, Sam Altman really pushed this as as this is sort of a new era for vibe coding. He's going to be able to build all these sort of like cool interactive things for users. And that is, I think, a really interesting promise that that we'd be able to just like visualize things in new ways and learn things in new ways.

34:49And I do really want to tap into that, but it feels pretty clear to me after this that it's like it's not quite there. No, it feels like if you already know how to code, I think this is going to be great for you. Because when it comes to the point where it gets something wrong, you're going to be able to identify what it's doing wrong quickly. And then be like, oh, you just made a little baby mistake. I can do the thing. But if like the real true spirit of vibe coding is a dumb dumb, a coding dumb dumb like me being able to make an app that just works. I mean, I lost my mind trying to recreate Airtable.

35:23I mean, in fairness, I lose my mind just opening Airtable. So, you know, that's the standard experience there. All right, we've got to take a break. When we get back, we've got a new segment we're calling Corporate Shenanigans. Avoiding your unfinished home projects because you're not sure where to start? Thumbtack knows homes, so you don't have to. Don't know the difference between matte paint finish and satin? Or what that clunking sound from your dryer is? With Thumbtack, you don't have to be a home pro. You just have to hire one. You can hire top-rated pros, see price estimates, and read reviews all on the app.

36:00Download today. Morning, Zoe. Jeff Bridges, why are you still living above our garage? I want to be in a T-Mobile commercial like you. Teach me, Sildana. I couldn't possibly. At T-Mobile, get the new iPhone 17 Pro on them. It's designed to be the most powerful iPhone yet and has the ultimate pro camera system. Impressive. Let me try. At T-Mobile, you can save up to 20 % versus the other big guys. You heard them. T-Mobile is the best place to get the new iPhone 17 Pro on us with eligible trade-in in any condition. Check them out to see how much you can save versus the other big guys at T-Mobile.com slash switch.

36:33Fox Creative. This is advertiser content from Al Nylum.

36:42Living with disease can be draining. Managing new symptoms and medications, scheduling endless appointments and tests, losing the flexibility and independence you're used to. The drip, drip, drip of disease takes its toll. On a genetic level, disease is like a leaky tap. Our genes instruct our cells to produce proteins that the body needs to function. But sometimes, genes instruct cells to produce unwanted proteins, or too much of a protein, which can cause or contribute to disease. Most conventional medicines treat disease by targeting the symptoms, like mopping up the puddle rather than tightening the leaky tap.

37:23But an innovative class of medicines, pioneered by alnylam pharmaceuticals, targets disease at the source. With RNA interference, we can disrupt the production of unwanted proteins to silence the drip, drip, drip of disease. This innovative approach to treating disease is already helping thousands of people around the world live amplified lives. And it's just the beginning. Learn more about RNAi therapeutics and alnylam science at silencedisease.com. Alnylam. Silence disease. Amplify life.

38:09Welcome back. This week, we've seen some headline-grabbing announcements from Perplexity, Apple, and Elon Musk. Perplexity offered to buy Google Chrome for$34.5 billion. Apple is suing the movie theater chain Apple Cinemas. Elon Musk says he's going to sue Apple for rigging the App Store. The question I'd like us all to consider for each of these news items is shenanigan or no? As in, is the company in question actually serious or is this just for the attention? So let's start at the top. Perplexity has put in an offer. Nobody asked them. Chrome is not for sale. But they have gone out and they said, we're going to, we want to buy Google Chrome.

38:54we will pay$34.5 billion. Is this real? 100 % of shenanigans. It's so much higher than its own valuation. That's the kind of shenanigans it is. It's like Oprah-level shenanigans level shenanigans. My favorite tweet about this was like, I'd like to buy Chrome too. Someone will put up the money. They said that they had investors lined up that would front the money, but I think it's just a marketing stunt. It is. I mean, right. Perplexity is, I think, valued at 18 billion. So this is like double their valuation. They want to you need two perplexities to buy Chrome at this price. Chrome also not for sale.

39:33It might be. It might be. There's a chance it'll be for sale. And never forget they try to buy TikTok, too. Right. Right. Like six months ago, they tried to buy TikTok. Again, I don't think anybody was asking. I mean, TikTok is is maybe potentially plausibly. It's like it's in a quantum state of being for sale and not for sale. So that one was... Perplexity will not be the buyer. Schrodinger's TikTok. It's Schrodinger's TikTok. It's for sale or not for sale at the same time. But this feels like, you know, that person that's always like, oh, I'll do it because there's no danger of them actually having to do the thing.

40:06That's so real. That's what it is. Yeah, yeah, yeah. It's like the worst case scenario is that they accidentally buy Chrome and have a good product. It's like someone at an auction that accidentally raises their paddle. And then if they get picked, it's like... I did not know that was going for me. Yeah. Perplexity, they have, I think, a fine product, and they really like being in the news. And that seems to be what is going on here, right? And it's funny, they have a browser. They have a browser based on Chromium. It's perplexing. Okay, this one, I think we can move on from, I feel like we're all in agreement, fully a shenanigan.

40:51Not serious. Okay, item number two. Apple, the company that makes computers, is suing Apple Cinemas, a small movie theater chain that is now starting to expand nationally. Now, okay, I will tell you, when I first saw this, I was like, this is ludicrous. This is a shenanigan. I'm sorry. I never heard of Apple Cinemas until this lawsuit. They're making it more popular. They are like, it's like that thing where you're drawing attention to the thing. He's doing a Hamilton where Hamilton is just like, I will tell everyone I had an affair, even though no one actually, this is my musical theater nerd coming out.

41:32Oh, actually, did you know it's coming out to theaters? I'm really excited. We got to go see it. Yeah, we got to go see it. But, you know, he's just like, oh, nobody knows about this affair. Let me publish it for everyone and ruin my life. Like, that vibe is what's happening here. Like, nobody freaking knew what Apple Cinemas was or was associating it with Apple. It's a shenanigan in that way. But I think they are very serious and they want it to happen because they're so intense about their brand. So I feel like it's like a shenanigan to us all, but internally not. Because they really believe that this is like a threat to their brand.

42:07So they got really upset because Apple Cinemas started moving into, I believe it was San Francisco. And they were like, this is too close. They're like, you can't do this. They actually, they narked on Apple Cinemas. They called up Apple Cinemas' landlord and they're like, hey, I don't know if you know this, but they're kind of using our name. And the landlord did not respond. And they're like, we had to sue. It was the only thing we could do. Calling the landlord, that's dark. This lawsuit is kind of ridiculous. They just go through a bunch of Facebook comments of people being like, lol, they gonna get sued.

42:42They're using that as evidence. This is their evidence. At one point, they do a Google image search for Apple Cinema, which I will note the company's Apple Cinemas. They do a Google search for Apple Cinema, and they're like, look at this. Half the results are Apple Cinemas, and half are the Apple Cinema display. We have a product with basically the same name. So they paint this long history of Apple. They're like, Apple has a grand history with cinema. We created Final Cut. We created QuickTime. We streamed the trailer for The Phantom Menace on our website. And it's very, very silly. Jake, you should use that voice when you do this podcast.

43:26This is the entire podcast. I don't think I can maintain it. The Apple voice. So there is another wrinkle to this that I have to point out. So I'm going to read you a bunch of names. And I want you to tell me if you know who originally owned these. Right? So Apple Electronics. Apple Films, Apple Publishing, Apple Boutique, Apple Records, and Apple Studios. Do you know these brands? No. Nope. These were originally companies that were part of Apple Core, which is the Beatles company. The Beatles started this to start Apple Records. And Apple Core, the Beatles company, was in a protracted trademark battle with Apple computers starting in the 70s because Apple Computers came after Apple Core.

44:17And so Apple Core, right, originally a much more successful company because they were the Beatles. They were doing very well. So Apple Core sues Apple and, you know, they get into a tiff about, they get into a bit of a turf battle. And Apple Core, they've got a bunch going on. They've got movies. I don't know what their electronics are, but apparently they had some electronics. This all culminates in Apple computers eventually becoming a dramatically more successful company than Apple Core and just buying out the trademarks to everything that Apple Core owns. So Apple now, I think they own the trademarks to Apple Core and license it back to the Beatles company.

45:03Oh, wow. You know what? They're just having trauma. Apple is having trauma. They think Apple Cinemas is going to blow up and then they're gonna have to, they're gonna get taken over. This is just trauma. So Apple owns basically Apple preceding any word is my understanding at this point. Except Gwyneth Paltrow's kid. Crucially, I hope not. We don't know. We don't know. We don't know. Do they own anything that has the word Apple after it? Like, do they own New York because it's the big Apple? No, no, no. That's different. That's different. Yeah, yeah, yeah. Just Apple in front of a word. Apple in front of a word I think they own.

45:39And so, okay, initially I saw this and I thought, this is ridiculous. Leave this little movie theater chain alone. The movie theater chain, I should say, they claim that their name comes from, they're originally going to open it in some area that had Apple in the name. And they're like, oh, it's just named after the area. But Apple owns that area. They own everything that has Apple in it. It was just Cupertino. So initially I thought it was a shenanigan, but now Apple Cinema is like, yeah, we're going to open up like 100 theaters across the U.S. They're like, we're a top 25 theater chain, which admittedly is not saying very much.

46:16I think that maybe it's not a shenanigan. I think that it feels a little mean, feels a little mean, but I feel like they maybe have to do this one. They're serious about it for sure. Oh, they're definitely serious. What if Apple is thinking as an extension of Apple TV +, that it wants to bring people into theaters, and therefore it's going to launch Apple Cinema Plus, in which case then it's not a shenanigan because this is future plans that Apple Cinemas is foisting from them, in which case this fan fiction, if this fan fiction is true, maybe it's not a shenanigan. You're looking at the patents.

46:52Well, can I tell you something? They pulled some comments from reviews of Apple Cinemas, and people said it was gross. So they are trying, I think they might be worried about it. They're worried about the brand reputation for their future cinematic efforts. When I go see Freakier Friday, I will not be going to Apple Cinemas, I guess. Well, I don't know of an Apple Cinemas in a 25-mile radius, so there's that. I guess I just won't fly to SF to see it. Okay, moving on to our next one, which is also Apple-related. This is breaking news as of Thursday morning. V, do you want to break this one down for us?

47:28Yeah, so Apple has basically figured out a way to circumvent the ITC, the International Trade Commission, import ban on the Apple Watch, which, you know, is a whole legal tit for tat that they had with Massimo, which is a medical tech device maker for blood oxygen monitoring. and so you know they went back and forth for a really long time cut to december 2023 um and you know the itc has ruled that you know there's an import ban on the apple watch because they have a patent infringement on the blood oxygen sensor and that's why starting in january 2024 in the u.s only you can't buy an apple a newer apple watch with the blood oxygen sensor feature enabled Well, that's all different today.

48:15There's a software rollout update, and now you can have the blood oxygen monitoring feature, except, you know, the workaround is that the sensor on the watch collects the data, and it gets processed on the phone so that you can view the data only on your phone and not on your wrist. Right, okay. That's the workaround. So the hack they figured out is as long as the watch doesn't show you the blood oxygen sensor data, they can run it. Wow. So for the past year and a half, they have been selling watches with this sensor disabled, and now they're enabling it. Okay, the thing that I find ridiculous about all of this, Apple has so much money.

49:00So much money. They could just pay Mossimo. Yeah, is it just like a principal thing? What's up with this? They don't want to pay. I think so, though, right? Like, I think they're like, listen, even if you are right about it, and look, Massimo, they might be right, they might not, is going to continue to be litigated for a long time. But I think that they're saying on principle, even if you're right, we will drag you to the ends of the earth, and we will deplete your legal budget before we give in. I think that's what's happening here. There's like multiple suits going on. So like initially, like this thing goes back like over a decade at this point where, you know, initially Apple went to Massimo.

49:43Allegedly, Apple went to Massimo and they were like, hey, you know, we want to collab on blood oxygen. And then, you know, they poach a guy from Massimo who very subsequently soon leaves. and so the mass was like hey you stole our shit this is the scientific uh you know legal terms that they're using you stole our shit and apple was like no we didn't here's the apple watch series six with blood oxygen monitoring and they're like hey uh hey and so they filed a suit and when that wasn't working because it was taking too long they went to the itc because the whole even the itc thing is a workaround because if you because the supply chain and these devices being made in china If you can't import them, ha ha ha, you can't import them, pew, pew, pew.

50:24And then Apple was just like, well, first of all, we're going to exhaust every legal option, humanly possible, to prevent this ban. Which is why, you know, these proceedings started in 2020. It didn't happen until the end of 2023, 2024. There was like a back and forth. The ban was paused. The ban was unpaused. And then Apple was like, fine, we'll just disable it then. How sick do you think Apple's lawyers are of this situation? They are so sick of it. They are so sick of it. And the irony is that Massimo then came out with its own smartwatches, which look like the Apple Watch and measure blood oxygen.

51:01And then Apple was like, shenanigans. Petty to the nth degree. So petty. Okay. So what's our take on this one? Is the whole situation a shang? Is it serious? I think it's serious, but it's ridiculous. Yeah, agreed. I think that's a fair take. I was going to say it's right down the center where it's, yeah. It's just ridiculous at this point. Also, the patents expire in 2028, so this is very stupid. Also, the fact that they were able to remove a feature for a year and a half maybe speaks to how important that feature is. I mean, I've written it all over, up and down, across the verge. It's not that important of a feature.

51:45It just isn't. It's rough. It's rough. That was once a flagship feature. and they don't always pan out. No, they don't. Okay, final possible shenanigan. As soon as I say the first word, you're going to know where the... Elon Musk says he is suing Apple for rigging the App Store. So I'm going to read this quote. Here's Elon on Twitter. Sorry, apologies. This is Elon on X. He says, quote, Apple is behaving in a manner that makes it impossible for any AI company besides OpenAI to reach number one in the App Store, which is an unequivocal antitrust violation, end quote. And so he starts by claiming that they're rigging the rankings, but then he kind of pivots very quickly and starts by claiming that actually what they're doing is rigging the App Store editorial because they have a bunch of curated sections where they'll be like, our favorite apps, like summer hotness, whatever, like check out these AI things.

52:42And they don't feature Grok. Can you believe they won't feature Grok? I can't believe it. You mean Grok, the one that unleashed sex bots very recently on Bad Rudy? Oh no, why would ever family-friendly Apple not put Grok in its editorial section where it's allowed to exercise opinions? Yeah. I mean, my feeling is also like, these apps aren't even allowed in the app store, right? Other apps that make deepfake nudes have been banned. Right. They can't ban Grok because Elon is too big of a figure. He's already getting special treatment. Also, his whole assertion that no other AI company can reach number one.

53:29Didn't DeepSeek reach number one? Right, right. Actually, there's a community note on his tweet about this being like, actually, DeepSeek. My favorite also is like Sam Altman and Elon getting into yet another X battle where Sam's like, Like, oh, well, have you ever, you know, put your own companies as special treatment? Have you ever given your own company special treatment? I couldn't believe it. So, yeah, there's a lot. It feels like he saw what Tim Sweeney did with Epic and he was like, I'm going to do that. That's going to work for me. Well, here's the thing. Let me just add a little note there.

54:05Tim Sweeney actually filed a lawsuit and Elon Musk at this point has not. He's never going to. He's just going to tweet about it. Which that is a bit of a pattern for him. So he got a little mad at Apple. He retweeted a bunch of people supporting him. Apple put out a statement later saying that the App Store is, quote, designed to be fair and free of bias. I don't know what that means. I mean, if it's editorial, it's, you know, like just by the function of the word editorial, there is some kind of opinion in there. But like, yeah, they said it's like a combo of algorithms, you know, recommendations and opinion.

54:43So this is a shenanigan. It definitely is. Because he knows nothing's going to come of it. And he knows that, you know, if it really was number one, it would reach number one. Yeah, this is is nearly as much. This is like perhaps as much or more of a shenanigan than the perplexity one. Definitely. Maybe more. I think it's more. It's more. This might be the ultimate one. Perplexity at least had some investors apparently lined up that maybe would front the money, you know? Yeah, that's true. They took like an action to say like, yeah, we're not all total BS, but like... Right. Elon just RT'd a couple of randos on Twitter.

55:17As always. Yeah. Okay. Okay. So I think, what's our running total? Like two and a half shenanigans out of four. Yep. Okay. Okay. Right. This is a big week for them. Apple somehow involved in most of them indirectly or directly. It's not always their fault. Okay. We've got to take a break. When we get back, the Thunder Round returns. Every great company story is defined by moments when the founders make bold decisions. These are high stakes moments that risk the business, but can lead to greatness. I'm Rolof Boota, managing partner of Sequoia Capital and the host of Crucible Moments. We're returning for a brand new season.

56:05Join us as leaders from Stripe, Zipline, Palo Alto Networks, Klarna, Supercell and more. Share what it's actually like to navigate the make or break decisions. Crucible Moments is back on October 23rd. Until then, catch up on seasons one and two wherever you find your podcasts.

56:50The New iPhone 17 Pro Check them out to see how much you can save versus the other big guys at t-mobile.com slash switch. Rinse takes your laundry and hand delivers it to your door. Expertly cleaned and folded. So you could take the time once spent folding and sorting and waiting to finally pursue a whole new version of you. Like tea time you. Mmm. Or this tea time you. Or even this tea time you. Said you hear about Dave. Or even tea time, tea time, tea time you. Mmm. So update on Dave. It's up to you. We'll take the laundry. Rinse. It's time to be great.

57:32All right, we're back with the Thunder Round, a.k.a. the lightning round, colon, the Thunder Round, colon, Jake's version. Love it. As you all know, I've been on a power trip since Nilay Patel left on parental leave. The lightning round, as we know it, is done for. This is now the Thunder Round. I have given our producer, Eric Gomez, the power of thunder to make sure that we keep moving and get through all the important stories that we want to talk about. So we're going to do five stories, five minutes each at 30 seconds. We're going to hear a rumble of thunder telling us to move along. If we keep going and get a little too far, we're going to hear thunder crash down.

58:07That's our moment to move along. All right, let's get started. V, you've got our first story today. Yeah, so my first story is Allison's little adventure turning her LTE Apple Watch into her only device. so she's leaving her phone at home she's going out and she's doing her adventures and Allison is like annoyingly well and I say this with admiration one of the funniest writers on staff so just reading this was just delightful and like I was just like she she reached out to me and she's like V I have this crazy idea I was like hit me hit me with it girl what are you gonna do she's like what if I leave my phones at home and just live off of a smartwatch and I went God bless.

58:47Let me know how it goes. I ain't that crazy. Let me see how it happens. And it happened like how I thought it would happen. But it tickled me because I think she and I, our beats continually like fuse and merge together because of all the ways that they're stuffing AI into our products. And I think she and I both have just like had a lot of conversations about like needing to step away from our devices in some capacity. Like I had a story that came out a couple weeks ago about using brick which is like this thing that bricks your phone and i use it during tv time to get my attention back and she's like i want to be more present in the world so i'm gonna leave my phone at home and just use my lte watch for like connectivity and of course that comes with its own um trials and tribulations like she's like oh shit how do i call an uber how do you call an uber without your phone and like it was just a really cute meditation on like how reliant we are on these devices, can you actually replace your phone with your smartwatch in any capacity?

59:49So the way I know this kind of worked is that, so this is Alison Johnson, senior reviewer for The Verge. The way I know it worked is that on like Friday, I was saying something, I'm like, you know, I haven't heard much from Alison this week. What's she up to? And someone was like, oh yeah, she just like hasn't been using her phone all week for her job. And so she hasn't been talking to us. And I'm like, that was so effective. That worked. It worked. I just like, you know, like that. It was harder to get in touch with her, but I guess that was sort of the point. It was functional. She was dead.

1:00:20She did a good job. My favorite part was when she said she had visions of herself sitting at a table in a sidewalk cafe wearing a billowy skirt, but she doesn't own a billowy skirt. Like that was the energy she was going to give off. Okay. So I have a question and it's not for you, V. I know the answer. Hayden, are you a smartwatch person? No, I hate having things on my wrist. Like I can't, I don't like the feeling. It's like, I think it's because I do boxing as a workout. So a lot of people use smartwatches as like a workout tracker also. And I would always have to take it off because I box. So I feel like I just don't like feeling something on my wrist like that.

1:00:53I don't know. I can deal with rings, but not big bracelets or anything like that. So it's been a while since I've been like a regular smartwatch person. I bought the original Apple Watch and then I immediately bought an Android phone. And it was just like, I just like had this watch that I couldn't use. the thing that I think is really interesting about this is like there's this the sort of the conceit is okay the watch is less distracting than the phone but I sort of wonder and V this is a question for you now isn't the watch a little more annoying because it's literally physically on your wrist there at all times and it's shaking around trying to alert you to everything that almost seems more distracting it can be more distracting um uh especially when you wear two at any given point in time.

1:01:35And they have different notification settings just because like quick, quick wearable check. I think there's three on you today. There's three on me today. Well, there's three on me today and two more at my desk because they're glasses. And so, you know, that's the whole thing. But yeah, I do get buzzes a lot. And especially on this Garmin that I'm testing at the moment, because it's a little less discerning with what notifications I get. So it's just like, oh, they texted you. Oh, they texted you. Eh, they texted you. And like, they buzz. And I can answer this as an objective third party because whenever I'm with someone who has an Apple Watch or a smartwatch, you know, it makes me nervous because they're always looking at their wrist.

1:02:15And I'm like, oh, am I boring you? Is it time to go? Obviously, it's just they're getting a text and they're like, oh, sorry. But, you know, it can be intense. It's definitely a social cue for sure where I'm very cognizant of it, of just things buzzing and being like, and then going back down and looking at my wrist. I do think it's a different kind of distraction. With my phone, if I do need to, if I'm sitting down for dinner with my wife, I will put my phone somewhere else so that it's not distracting me. Because listen, love y 'all, but I get a lot of slack pings, all right? They do not stop.

1:02:53and if it's on my person, if I know it's buzzed, then there's just like, there's always that tension in my mind where I'm like, I gotta find out what this is. I gotta find out if it's important. Sometimes it's just the weather. Sometimes it's just telling me it's gonna rain and I did not need to see that. And I do like that I can just like take my phone out of my pocket and put it somewhere else. And with the watch, you're sort of like - That's why you gotta get the flippy flip. Dismiss that. I do it all the time. So you're just eating and being like, flinging your arm around. Yeah, that's not distracting at all.

1:03:21Okay, okay. All right, Hayden, you're up. Okay, so mine is how doctors are getting worse at detecting cancer after they rely on AI. So really scary and sad. Basically, a new study just published this week found that doctors who usually use AI to detect cancer in colonoscopies, now when it's taken away from them, they got way worse at detecting cancer than they were before they ever used AI because they basically got lazy. Like, their brains were used to just relying on AI and kind of double-checking it. And then even with all their training, when you take the AI away, they became way worse. I think six percentage points worse.

1:03:59Oh, that's a little horrifying to me. That's not good. That's very upsetting. That's, you know. Yeah, they found, like, across countries, Poland. The researchers came from Poland, Norway, Sweden, the UK, and Japan. They looked at four medical centers in Poland that were part of a trial program, you know, using AI and colonoscopies. And yeah, basically these doctors are just going to be SOL for whenever they stop using AI. This is really interesting because I feel like there's been so much like fear and concern, like AI is going to make you dumber, AI is going to make you dumber. And I've been like trying to grapple with this where it's like, OK, I don't I don't know how to use the Dewey decimal system.

1:04:37Right. I don't know how to go find a book and then look in the glossary and find the page and, you know, read the context and figure out. I'm not reading as much because I'm just going to Google. And Google's just giving me, and then I read a website that's been written for me really nicely. So like, okay, am I dumber because I use Google instead of going to a library? Maybe a little bit. And so I'm like, okay, are we kind of overreacting with some of this AI stuff? But this is concerning to me. I think it's like GPS, right? Like if you always use GPS on your car and then you just never, like I'm calling out my spouse here.

1:05:11Like, they don't know how to get to a place that's five minutes away because they always have GPS, so they only put GPS. And when the car play doesn't work and it's five minutes away, you've driven there a million times, they don't know how to get there. That's me. Oh, yeah. No, not a chance. I think that's kind of analogous to what's happening here with the doctor. So, like, even you know how to drive, you could probably do it on your own, but. Definitely. And you also, I think it's also like a thing when I found in some recent reporting I did on a Google Health, like medical AI hallucination, that people that double check AI's work also get a little bit lazy.

1:05:48And it's just kind of human nature. If something's usually right, you're going to not find the discrepancies or the errors as much because you're like, oh, it's probably right. And you're usually not going to pinpoint when it's wrong. And so I think it's kind of the same thing. It's like if your brain is used to relying on something, something that's usually right, you're just going to get worse at doing it on your own. And you're going to get worse at identifying when it's wrong, probably. Yeah, Hayden, I thought this piece you wrote, I think it was last week, was fascinating. Where you found that, what was it, a Google research paper had merged the name of two different areas of the brain and kind of no one noticed.

1:06:27Exactly. It had more than 50 authors on it, and they also had doctors look at it before they published it. It was like the debut research paper for Google's healthcare AI model, which they were positioning to doctors as, you know, a good way to double-check things, you know, flag things, especially things missed by radiologists. And yeah, I mean, it made up a body part. Google said it was just a misspelling, but it was conflating essentially two different areas of the brain. and if you did have an abnormality or a stroke in one of those two areas, it would be treated differently from the other.

1:07:01So, I mean, it is a problem if, you know, especially like we just said, if you're a doctor reading this and you're probably not going to find the error, just like they did in the actual research paper, you may skip over it and maybe even mistreat the actual thing. The term sounded real. Basilar ganglia. That sounds real because there is like a basilar thing and a ganglia in your brain. Well, and this is what I thought was so interesting about that, where, you know, you have people in the story saying like, oh, it's like a common typo, like people figure it out. But I do think like, okay, sure, maybe they would have figured it out.

1:07:36But the turn is, if they get used to not checking, then you get into a problem. And this study sort of emphasizes that where it's like, oh, like, in fact, they are getting worse at it. In fact, that could make medical care worse. And so, like, that gets a little concerning. Listen, CHGPT5 is better at health care. So says Sam Holtman. Yeah. And we'll see. Health care, coding, it's just telling you to do the health care yourself. It's just... That's the US healthcare system. I'm interested in how this affects people with health anxiety because basically it said, oh, it's better at answering your questions about your condition and telling you what it might be.

1:08:17Oh, I have health anxiety? It's not good. It's bad. It's going to be interesting for sure. Okay. Next story. AOL is shutting down after 34 years. AOL dial-up, I should say. AOL dial-up is done after three decades. RIP. I can hear it in my head. It's beautiful. Beautiful noise in retrospect. Apparently, according to the 2023 U.S. census, there were around 160 ,000 people who still had dial-up internet, which is sort of... It's so slow. I want to know what they're paying. They probably have really peaceful lives, you know? Right. This is the real question. What are they doing with it? you can't load anything on that internet right now they're definitely just looking at like their aol inbox and maybe like the msn home page that's what i think is it yahoo they use yahoo oh definitely yahoo yeah actually can i say when i when i looked up the um the apple core settlement with apple inc um from 2007 the press release um the spokesperson for apple core aol email address this is 2007 right aol is still around you still see i haven't seen the aol email address in a while you still get the yahoos you still get the msn not so much i just that sound though is forever in my head yeah maybe it'll become cool to have an aol email address you know like the it's a nostalgia thing i don't know i could see that is that the like digicam of emails you're like oh gmail is for the millennials we gotta we're gonna like buy old aol addresses especially since like the inbox isn't unlimited.

1:09:59So you actually, it's going to be like a mindfulness thing where they're just like, oh, you only have 50 emails that you can have any given point in time. Again, they're probably so zen. Were you both AOL dial-up users back in the day? Yeah, yeah, same. And it's, I mean, I barely remember it at this point. I remember it like it was an app though. It was like an app that you enter into. It was just like this contained experience. There's just like sound bites that I remember. One, there's the connecting thing for the dial-up, but then also, you got mail. Oh, my God. And like, welcome to AOL or something like that, and just like what the homepage looked like.

1:10:35That was my Neopets era. Oh, my God. That was mine, too. Going to the free omelet every day. Oh, my gosh. That was so good. My shop song was an ABBA song. So good. Neopets still going strong, though. It is? I tried to log in sometime in the past year, and they deleted my account. Oh, my gosh. My pork alerts. they're all gone all right wow aol you had a run that lasted way way longer than it should have um thank you for getting us online and for imprinting that terrible noise in all of our heads i hear it in my dream as we should next next week on the thunder round we're replacing the thunder with the screech oh my gosh let's all make our ringtones that sound or our phone alarms.

1:11:22You know, that will really help us, you know, reduce our desire to be near our phones. If they were like blaring that noise all the time. 100%. Oh, God. Okay, V, what have you got for us? We have the Pebble Time 2 because I got a little email from Eric Megakovsky, who is the Core Devices CEO and co-founder of the Pebble Watches. And he's just like, we've got a design reveal, the Pebble Time 2. And because Pebble is Pebble again, Very recently, they got the name back because they had to rebrand as Core devices, so it was going to be the Core Time 2 and the Core 2 Duo and everyone was making Intel chip jokes.

1:12:01But it's back to being Pebble. I brought this one up because we knew Pebble was coming back. We knew that Pebble just had these diehard community fans for open source smartwatching. And it was just interesting because in the comments on this, there seemed to be like a war of ideologies of people who were like, yes, finally the pebble has returned and my life is complete again. I read it in those voices every single one of the pebble diehards because for the last decade they've been in my DMs going, nothing will ever replace my pebble. I love it so much. and then there are other people going just like, I don't know, man.

1:12:40I need LTE on my smartwatch. I need NFC on my smartwatch. I need it to do modern things and those are very purposefully not on the Pebble Time 2. It's got RGB lighting. It's got a really retro design. It's got e-paper as its display. It kind of feels like one of those dumb phone devices but for your smartwatch, it's very retro. It's just going to do the simplistic, minimalistic things. And it was interesting wading into the comments to see people who were just like, Yes, Pebble! And then other people going like, It's ugly. Why would you do this? Like, we're beyond this point. And I was just like, Wow, it's so wild just watching this play out in the comments.

1:13:22Well, I also suspect that, like, there's a real degree to which if anybody thought that this was a real threat to the smartwatch market, this would not have been allowed to happen. Like literally would not have been. No. Google owned all of this. Yeah, Google legit was just like, have at. Go have fun. Google was like, if it makes you happy. Like literally. Right? Like Google makes the Pixel Watch, right? That thing's not flying off store shelves. It's a lovely smartwatch. It's not flying off store shelves. It's a great smartwatch. And even there, like, yeah, you can compete with us. It's fine.

1:14:00That's not. The opposite of the Apple and Apple Cinemas thing. Yes, yes, yes. It's literally like, oh, you guys want your little pebble back? Have fun. Come on in and compete. Come on in. Go in. But it is nice, right? I mean, this is sort of like a little bit playing off of what Allison did, where it's like, this is going even simpler. This is like, the watch is not going to bug you. No. Like, you get your notifications. You can do a little bit of step tracking. There's a compass in there. Have that. Do what you will. there's a little community that's been like keeping the pebble going you know during the dark ages where there was no support and whatnot just a little community of ragtag pebblers uh going and keeping this thing alive and popping into every single smartwatch review that i've written over the last however many years going i miss my pebble i miss my pebble you know like the just like that was a nice was a nice watch and i think there's also it is like the ethos of it right yeah because there's i don't think there's anything else there's nothing like it in like in you know mainstream consumer technology that's as like tinker tinkerer friendly it's a cute watch a cute name yeah also just like you know smart watches and i've written this time and time again they kind of feel like the vanguards of walled garden os's because they're there for you to stay stuck to a particular phone so everything is super siloed so for pebble to be back out here and they're like hey i'll open source let's have fun yolo let's go like it's just nice to have that vibe come back at a time where like smart watches as a whole are super duper siloed and that's why i have to carry around three freaking bombs no i completely agree and it's i mean it's it's really funny like this is like an updated version of the pebble time 2 was that what the original name was the pebble time so this is the pebble time oh so this is actually a sequel it still looks mostly the same but like a little refined and as far as I'm aware the specs are barely changing it's mostly the same device mostly the same feature set as it was 10 years ago there's like no other device you could be like I don't know if people were like iPhone 5 honestly okay I'm saying this I might buy that the iPhone 5 looked really good it was a good phone I also missed the button you know that's true this is where it's going to go.

1:16:17It's like the Digicam people are going to demand we go backwards. Yeah, all right. The Pebble time, it's supposedly shipping this summer. It's supposedly shipping this year. This year. This year. It's coming sometime. Yeah. Okay. Very exciting days for the Pebble community. Y 'all have been waiting. It's a good moment. Okay, Hayden. So Addy did a great piece this week on how chatbots aren't divulging their secrets. They're not like an oracle. When you ask it a question about how it works, its inner workings, why it was banned from something like Grok was, or why it's responding a certain way, what its secret prompts are, most of the time it's not unveiling secrets of its inner workings and its system.

1:17:02It's just kind of making something up because that's what it wants to do. It wants to make you happy. It wants to answer your question. It's searching the web and just becoming a pattern generator and, you know, picking the next most logical word. Right, so this Grok thing, Grok, like, briefly got banned from X, which is... And Blue Sky. And Blue Sky. First off, I cannot believe Grok was allowed on Blue Sky for a second. That's the craziest thing I've ever heard. So it got banned on X for mysterious reasons, even though it seems to be the main feature on X these days. And then people started asking it why it was banned.

1:17:35And they were just like, oh, okay, well, Grok, if you say so. But it had, like, a bunch of different explanations. Yeah, like one was that it stated that Israel and the U.S. are committing genocide in Gaza. And then another one said it was content refinements by XAI. And then another one said, oh, it's because I identified an individual in adult content. So basically, I ran the gamut. Like no one knew why I was really banned. And it didn't know either. It's, I mean, this, it makes so much sense, right? You're like, why couldn't I ask the chatbot about itself? But I think this is like, just at an even more basic level, I don't even think you can ask a chatbot what model it is.

1:18:10Like, it doesn't even always know. It doesn't. I did ask chat GPT5 today, like, what's the difference between the old you and new you? And it did give me some examples of like what it can do better. But then I was just like, I don't know, I still bully you pretty hard. This is the thing. It's still basing it off of like what it can read. Right. Like it read the blog post. Yeah, right, right, right. It just knows the exact same thing that you can read. And I think that a lot of times, you know, again, this is a human nature thing of us like personifying things or whatever. But, you know, on TikTok, I see a lot of trends of people like treating chatbots like an Oracle.

1:18:46They're like, oh, I asked when the world was going to end and it said this. OK, like it's just reading a random Reddit comment and just putting that out there or saying like, you know, out of everything I've ever told you, what do you think is next for this company or whatever? I mean, there's basically people. It is good at finding patterns, but it doesn't know the future. And it doesn't know, like, I don't know. I mean, unless you can find it on the internet, it's not going to tell you. Besides, sometimes you can get it to unveil some of its system prompts or its secret hidden instructions. But I mean, that's it.

1:19:16So that's what's so tricky though, right? I think particularly when the chatbots first started coming out, and I think this happened on Bing, where, right, I think most of the chatbots are given instructions. You would think that this would be some fancy code of constraints on how these things operate. But there's actually just like some engineers are just like, be nice to the user. Give brief responses. And it's just like English language plain text stuff. And they kind of just like put that at the top of the chat and hide it. And so there have been these instances where people will figure out that, okay, if you do the right things, We say the chatbot will reveal some of its prompts to you, which lets you figure out how it's been designed to operate.

1:20:01And so there are like a very occasionally these secrets you can divine. Yes. Only about the system prompt. And that's it though. Right. It's extremely limited. And then because of this, I think, I mean, probably for a lot of reasons that are mostly related to human psychology. But I think this sort of lends credence to it. People sort of think, OK, anything it says, anything it says about itself is going to be true. and you just, you can't believe it. It's actually, it probably even knows less if it's like a breaking news scenario. Right, exactly. And it's also hard because, yeah, I mean, the fact that it can find patterns in like economic conditions or like travel patterns, all these patterns that it finds, people think that it can also find patterns in like what's gonna happen in the future or what's happened in the past.

1:20:44And it's very, it's just hard because I think, you know, it's really good at something and it's really bad at others. And, you know, we have to remember what it's bad at. But it speaks so authoritatively. Exactly. It's gaslighting us. I mean, it just says things confidently. And so if you can't suss what it is, it's really hard to remember what it is and it isn't good at for the average person, I think. Just... There's also, if you trace this far enough, I think this is how you start to get to some of those stories where people are like, it told me I could run through a wall, and so I ran at the wall.

1:21:17And it's like, no, no, no, no, no, no. like please like you have to approach these things with common sense and i i i think very in complete fairness the fact that the bot can't explain to you what it is doing is really confusing especially when you have you have these reasoning models and these like quote-unquote thinking models they'll explain their reasoning and it looks really really like intelligent and sensible and for it then to be like just totally i have no idea what model i am i just have to make that up I don't know why I acted this way. So I asked it how many R's were in strawberry.

1:21:53It said three, which is correct. But then it said there are three R's in strawberry. But you know what it can't get right is the number of R's in any U.S. state. And it keeps labeling maps incorrectly, too. It's great. One day, one day, maybe get a second official opinion on your medical information. Always, every day, get a second medical. Just don't listen to change it. And make sure that doctor is not using AI. That's it for The Vergecast. If you like what we do here, the best way to support us is to buy a subscription to The Verge at theverge.com. We'd love to hear your questions and feedback.

1:22:29Let us know what you want to hear us talk about this summer. Email us at vergecast at theverge.com or give us a call, 866-VERGE-11. The Vergecast is a production of The Verge and Vox Media Podcast Network. Our show is produced by Eric Gomez, Brandon Kiefer, Travis Larchuk, and Andrew Marino. Jen Toohey has a great episode coming up on Tuesday. It's about whether we can make Rosie the Robot from the Jetsons a real thing. See you next week.

From the publisher

GPT-5 is here, and it’s not going so well. This week on The Vergecast, Jake, Vee, and Hayden discuss the bumpy launch of OpenAI’s latest model and why GPT-5 isn’t as big of a leap as GPT-4.

Then, everyone shares their vibe coding projects and the bumpy journey to making anything usable. After that, our newest segment: Corporate Shenanigans, where we rate the week in strange corporate moves on a scale from “actually serious” to “total joke.”

Finally, the Thunder Round returns, new and improved, to discuss ditching your phone for a smartwatch, doctors relying too much on AI, AOL dial-up shutting down, the Pebble Time 2, and why you shouldn’t trust what AI chatbots say about themselves.

Further reading:

ChatGPT won’t remove old models without warning after GPT-5 backlash

OpenAI will update GPT-5’s “personality” after user backlash

ChatGPT is bringing back 4o as an option because people missed it

Sam Altman shared more about what went wrong with those GPT-5 graphs

OpenAI gives some employees a ‘special’ multimillion-dollar bonus

Anthropic just made its latest move in the AI coding wars

Anthropic’s Claude chatbot can now remember your past conversations

Perplexity offers to buy Google Chrome for $34.5 billion

Apple is suing Apple Cinemas

Apple Cinemas responds to Apple lawsuit

Apple returns blood oxygen monitoring to the latest Apple Watches

Elon Musk says he’s suing Apple for rigging App Store rankings

Ditching my phone for an LTE smartwatch was a humbling experience

Here’s a look at the final Pebble Time 2 design 

Some doctors got worse at detecting cancer after relying on AI

Google’s healthcare AI made up a body part — what happens when doctors don’t notice?

Chatbots aren’t telling you their secrets

AOL is finally shutting down dial-up

Email us at vergecast@theverge.com or call us at 866-VERGE11, we love hearing from you.
Learn more about your ad choices. Visit podcastchoices.com/adchoices

More from The Vergecast

All 445 episodes
Vibe coding through the GPT-5 messThe Vergecast · 1 h 22 min
Listen in VO