OpenAI is Flailing, Anthropic’s Mythos AI is Too Powerful, Are Vertical Tabs Good?

0:00That's for if things get really hardcore, or if you want to blow up some moons. Welcome to Primary Technology, the show about the tech news that matters. Tons of AI news this week. OpenAI required a podcast, and Sam Altman compared AGI to the pandemic. Anthropica created a scary model that they're not even letting normal people use. Google makes it easier to deepfake yourself. Meta released a superintelligence labs model. I finally got a refund for my XDR Vase amount, and a ton more. This episode is brought to you by CleanMyMac, Granola, and you. All of you who support the show directly, thank you for that.

0:31I'm one of your hosts, Stephen Robles, joined, as always, by my friend Jason Aiton. How's it going, Jason? Much better than last week. Very much better. That's right. Your tooth, you've not been to the dentist, but you got something taken care of. I went to the emergency room instead and just had them punch me in the face really hard. Let me show you a little trick to take your mind off that pain. I had an infected tooth last week. They gave me an antibiotic, which is really the most effective painkiller when you have a toothache, because probably it's infected, which it was. So, yeah. Yeah. There you go. I will be less grumpy, but only slightly, because grumpy is my shtick.

1:02And he's grumpy about the Kindle news, which we'll also get to later. Yeah. All right. We have a bunch of five-star reviews. Shout out. Smashing Sensation from the USA. Appreciates the chemistry. Looks forward to the podcast. Buridi from the UK. I love this guy and the other guy. We're both guys in his review. Awesome podcast with attention to links, chapters, and now videos. That's right. Also, he goes for dog walks in sunny South Wales. Robert is Cocker Spaniel. Loves our voices. Well, that's super fun. Luke the Dev from Australia. Another Australia listener. Battery percentage on phone and back pocket.

1:32Back pocket phone. Leave us a five-star rating review if you put your phone in your back pocket. Let us know. RGB the tech guy from the USA. And now he supports the show. Thank you for that. Cogen 27. He gave us a four-star review. Now, listen. I don't know if that was a mistake or not. He was requesting music during the ad breaks. But listen. I'm letting go the five-star rating goal. It's okay. Thank you all for doing it. But keep the ratings coming. Let's just get to 1,000 and we'll see what we are at then. See you for a five-star review. Because it seems as though people are now trying to hold us hostage. And I want to be clear that the primary tech position is we don't negotiate with terrorists.

2:05So, if you're going to leave us a review, leave us five stars. Leave us whatever stars you want. We're no longer going to be held hostage to the idea that if you break our rating, now suddenly we have to change the entire show. We do appreciate people leaving a rating. And the suggestion is not bad. But I don't think it made it a four-star podcast instead of five. Exactly. So, only mentioning five-star reviews from now on. We'll say that. Shahab from Canada, which he's also in my shortcuts community. Hey, Shahab. He does keep Apple boxes and Apple TVs on all the TVs. He's also Vision Pro M2.

2:36There you go. Vision Pro. OG Vision Pro. OG. Podcast 728 from the USA. They don't know why they're still using Apple Podcasts. That's okay. Listen, Overcast transcripts launched publicly now. If you want transcripts for every show, including member feeds, because that's one of the things where if you do support the show, and we'll put the link down in the show notes, I want chapters. Get it for $25 a year, $2.50 a month. The Overcast app will now transcribe even private feeds, which is wild. Apple Podcasts doesn't even do that. And I just want to say, even if you don't listen to ATP, Marco Armin, the guy who makes Overcast, one of the co-hosts of ATP,

3:09it's like two weeks ago maybe where he talked about this. I actually recommend just go back. Go listen to that episode where Marco talks about setting up 48 Mac minis to do local transcription and his adventure. That's an all-time great podcast segment across any podcast. Spoiler, we might talk to him on Mac Power Users about that. All right, there you go. I'm not on that show, so I wouldn't have known that. I do have some potential guests for this show, though, Jason, I have to talk to you about. It's pretty exciting. One of them just wrote a book about robots. That might come on the show. Oh. Uh-huh, uh-huh.

3:40So, anyway, I'll just tease that. Our listeners might know who that is. And last thing I wanted to mention, we have really cool listeners all across the world. I know we have multiple ex-Apple people that listen to the show. Some text me while they listen in real time, which is fun. But one who has been listening to a long time in the communities, Daniel sent me an email. He worked for Apple from 86 to 94, so not during the Steve Jobs era, but he literally sat across from Phil Schiller in the Apple marketing group years before he was Phil Schiller, he says in the email,

4:14and also worked to get products placed in certain things like movies and TV shows. And he sent a bunch of pictures on, like, projects that he worked on. And I just wanted to share this because I thought it was super cool. Things like in Forrest Gump, when Forrest looks at the letter in the mail and you see it's from Apple Computer, he worked on that. And also in Jurassic Park, there's an Apple Macintosh Quadra 700 that you see in the background behind Samuel L. Jackson during a scene, which is amazing. Worked on that product placement.

4:44And also in Seinfeld, when Jerry uses a Mac, a Macintosh Duo. A Duo, yeah. Yeah, a Duo, which was a laptop with a docking station, which almost seems ahead of its time. So, Daniel, a former Apple, really cool. Thanks for sending that. And, yeah, if you're ex-Apple and you have some cool stories from now or back in the day, especially now, we'd love to hear it. Or if you're current Apple and you'd like to tell some stories, we would love to talk to you. Love. We would love to talk to you. All right, well, before we get into all the big AI news, there was moon news. I should have made the quote. Oh, I didn't ask you the quote.

5:15Do you have any idea what movie that's from? It's pretty obscure. Actually, I vaguely think that it's from one of the movies my children made me take them to the theater to see, which means it's either Pokemon or Guardians of the Galaxy, and I don't know which one. It is the latter, spoken by Rocket. Which Guardians of the Galaxy? I think it's the first one. Okay, I've seen all three of them. This is a thing my boys and I do. We actually went and saw Super Mario Galaxy this weekend. Well, yeah, how was that? It was terrible. But it was a lot of fun to take my kids. That's what I'm seeing.

5:46People are like reviewing the movie poorly, but then people with kids are like, you know, it's whatever. It's a fun watch. I think even my older son was like, they didn't actually have to make this into a movie. It was a fun game, but I'm not sure it's actually a movie, but it's fine. That's not the reason you take your boys to see movies. So that's the thing. That's the thing. So, yes, you were right. You got it again. Guardians of the Galaxy. I wanted to mention the moon because the Artemis 2 mission and the crew of the Orion did their lunar flyby. I'll put links to NASA's image page, which is images.nasa.gov, where they have a bunch of high-res images of the moon and of the mission, which is amazing.

6:22And also, I tried building like 18 different shortcuts this past week because I wanted ways to easily pull the high-res images from the Artemis 2 mission. And NASA uses Flickr, of all services, to host their images in addition to like on their website. And like, why are we still using Flickr, Jason? Like, I understand this is a government agency. Well, listen, a lot of people are still using Flickr. And like Web Summit still uses Flickr. Really? Southwest still uses Flickr, all these places. I don't know why I just convinced an organization to move from Flickr to Google Photos, like, literally this week.

6:57I mean, like. It was actually the easiest thing I've ever tried to convince anyone of. It was like, yeah, let's just do it. I want to feel like an AWS bucket or something. Like, just put the images somewhere where they're easily accessible because everything from NASA is, like, copyright free. Like, anybody can take these photos, use them however you want. So there's no, like, copyright. So I wanted to make a shortcut, and I was like, all right, well, Flickr has an API. You have to pay for it. So I literally bought a year of Flickr Pro. For $80. $80. So I can get access to the API because I wanted to do something cool.

7:29And then I created the shortcut. It pulled the latest images from the Flickr page. And a bunch of people shared it. And then Fernando Silva, very kind of him, actually included it in a 9to5Mac article, and the shortcut died because the Flickr API was, like, too many requests. I was like, I paid for this API. He didn't tell me there was any usage limits. And so it just totally borked. So I had to create a different version that's now linked in the 9to5Mac article. And then Federico Vatici, the true shortcuts king, he created this version where he basically just, like, linked to specific images and just used the raw URLs for the original sizes and made it so you can make them all into wallpapers.

8:12So you know what? Download his shortcut. It's called Lunar Wall. And it's in the Mac Stories article, and it's really good. So you should get that. But I also wanted to mention, after we recorded last week's episode, I was at the Artemis 2 launch, but I also went to, like, this YouTube dinner thing. And I met Brandon Butch, YouTuber. He's got, like, one and a half million subscribers. He's a really cool guy. Really smart business sense. He was telling me how he runs his channel. But he's really cool. And I got to meet him, so I took a selfie there. And a couple other quick follow-ups. Yes, I did actually get my refund for my Studio Display XDR Vase amount.

8:44This is the screenshot of it hitting my Apple card, a refund of $430. So look at that. I feel better about it. You pay a lot of tax down there in Florida. What, $30? On a $400? I mean, that's, like, what is that? Like, almost 8%? Yeah, something like that. But it did happen. And I appreciate that it happened because I got to write an article about you getting a refund. That's right. It was perfect. It was like, Apple fixed a $400 pricing mistake with a four-sentence email. And can I say, now that the moon photos are out there, I've set these moon photos as my wallpaper.

9:18And now the XDR feels really worth it, especially now that it's cheaper. Because those blacks are great. I can understand that that would make a lot of difference. But I think, I just want to say one more time. And I actually heard the ATP guys speculating on this, too, this morning. Because that's what I do on Thursday mornings. I listen to everything that they talk about. I'm just kidding. That's not really true. That is true. That is what you do. But that's fine. I don't listen to it in order to prepare for this show. No, no, no. But you listen to it before we record this show. I do often listen to it in the morning. Yeah, yeah.

9:48But, because I listen to the bootleg feed. I wait until the regular feed comes out. It sounds weird. I don't like the regular feed. It sounds too polished or something. I don't know. Anyway, that doesn't matter. But I think this was just a mistake. I literally think they just goofed. They meant to charge a different price. Because they're charging a different price for the non-XDR one. I think they just biffed it. And then they won't. But they don't want to say that. Is biffed a word? Oh, absolutely. Or did you just make that word up? No, it's definitely a golf term.

10:19Isn't Biff the guy from Back to the Future? I'm not saying it's not a name. Wait, what's the guy's name from Back to the Future? The bad guy. It's not the villain. But, you know, I guess time is the real villain. Yeah, it is Biff. Biff Tannen. And anyway, sorry, this is a fun energy show right here because I don't know what's happening. Last thing before we get to the AI stuff, I wanted to mention I did receive AirPods Max 2 in the mail April 1st. And I posted pictures of them because I got the Starlight version.

10:53Come on. These look silver. Aren't these silver? I'll link my post in the show notes, too, if you're listening. Ooh, that one looks dirty. That's the picture. That's what I was thinking of. Well, the earcups are slightly off-white, but the backs of them, like, that's just straight-up silver. Well, that's definitely just silver. But if you go back to the other one. The one with the dirty earcups? Yeah. I didn't even wear them when I took these. I haven't even worn them. All I'm saying is I definitely had a teacher in middle school who drank a lot of coffee, and that's the color of their teeth.

11:23Thank you. Now, Jason's just thinking about teeth now because of it. But look at that. I mean, that looks better. Anyway, I did get an engraving on it, and I kind of like it. It's very small and discreet. Which is how you know Steven can't return them. No, well, you can return an engraved item in two weeks, can't you? I don't plan to return these, but. What would they do with them? I don't know. Buff it out? Sell it refurb? Someone get my SSR AirPods Max? And Basic Apple Guy was taking guesses on my middle name, and actually, it's Sir Steven Robles. Okay. That's the SSR. I've been knighted. I just haven't told anybody.

11:55Anyway, AirPods Max 2, guess what? They sound good. The ANC's good. Transparency's good. They still clamp your head pretty hard. Yeah. At least if you have a big head like me. Because being knighted is the kind of thing you'd want to keep a secret. If I was knighted, I would be so loud about it. That's what I'm saying. I know, I know, I know. I would literally be posting about it nonstop. Who is knighted? There's Sir Ian Kellen, Sir Patrick Stewart, Sir Elton John. Isn't Sir Johnny Ive? Wasn't Johnny Ive knighted? Yes. Yeah, Johnny Ive was knighted. How do I get knighted?

12:25I'm not going to become an EGOT. I've given up on that dream, and of being a five-star podcast. But can I be knighted? If only that was all it took to become knighted, King Charles. If you could hold on to your five stars for a month, you could be knighted. Dame Judy Dench. She can't knight people. You have to be knighted by, like, anyway, I don't even know. You have to be knighted by the king, right? Like, that's the only person who can do it. I think so. I think so. Anyway, so the AirPods Max 2 might have been. Imagine the email she would be just getting. Could you knight me, please? Do you think that if that was happening, she'd just be like, sure, you're knighted?

12:57Yeah, why not? Speaking of silly purchases, I bought this MagSafe e-reader, and I made a little short about it yesterday. This is the jankiest e-reader ever. It's MagSafe, but the battery is placed poorly. And so the only way to attach it to your 17 Pro Max is perpendicular, unless you have the iPhone Air, and then it'll fit. And so I posted about this. I was like, don't buy this. This is $70. You have to put EPUBs on here via the microSD card in the side. You can't wirelessly transfer.

13:28That doesn't work yet. And so I was like, don't buy this. This is a silly purchase. And then someone posted on Threads, and they were like, actually, if you flash this other operating system that you could download off GitHub, this thing's really cool. And I was like, I didn't want to tell him, but like, listen, I'm not doing that. It's still not really cool. It just has a better operating system. But I wanted to... I saw your message, or I saw your video, and I thought, I hadn't watched it yet. But if this is one of those videos where Stephen's like, definitely don't buy this thing, then I'm not going to tell you why you should buy it, because it's super cool.

13:58I was going to be really mad at you. So I'm glad you stuck with it. This is going to be the first video you've ever said don't buy a thing. So I don't usually do don't buy a thing videos, but I saw, you know, Snazzy Labs has talked not about this one, but other e-readers that attach to the back of your phone. It's also not a touchscreen. You have to use the physical buttons, which I get it. Like, it's small. But anyway, I wanted to talk about dubious purchases, because I bought the AirPods Max 2. I bought the silly e-reader. And our bonus episode today, I bought something much more expensive that might be even more dubious. And I have not told Jason yet, and he's wondering what it is.

14:29I've been thinking about it. I've been asking Claude, what might Steven have bought?

14:35Okay, we're going to talk about AI right now. And can I just say, I don't like how much AI is telling me it knows about me. Like, I asked it, I asked Gemini, this was a question that the family was asking in the car as we were driving home yesterday. And I was asking, my wife was wondering, do fish know what direction is up? Because in the water, if you're underwater your entire life, do you know which direction is up? Because, you know, it's kind of...

15:06Why would you not know which way is up? Well, if you've only ever swam in water, weightless... You're not weightless in water. You literally are. That's why space astronauts train in pools. You're not weightless. You're just somewhat buoyant. And so you... Yeah, you're buoyant. It's reduced effect of gravity. But you're not weightless. So do you think fish know what direction is up? Yes. Do you know why? Because the sun is up. Well, that's actually one of the reasons. Yeah. So Gemini... That just seems super obvious.

15:37No, I didn't think it was obvious. I thought they knew what direction is up. I knew that you were forever trapped in a sinking ship and you didn't know what direction is up. You follow the bubbles. Sure. The bubbles go up. But apparently fish have like these little, like, things in their ear canals that are weighted. And so gravity pulls on those. And so fish know that. And also because light. They know light is up. And so if you shine a flashlight in an aquarium, sometimes fish will swim at a 45 degree angle because they'll think the light is like the fake light is the sun.

16:08So, yeah. But have you ever seen an aquarium full of fish swimming at not the correct way? I mean, I understand. Sometimes you see them, you know, make maneuvers. Well, that was... I've never seen fish not swimming. That's what begged the question. Like, it was interesting that... It would seem like then the answer would be obviously yes because they all swim the correct way. But then like whales and stuff, which aren't actually fish. But, you know, they'll swim upside down. They'll turn around. But they don't... Yes, they do that. But it's an intentional maneuver is what I'm trying to... It's like, do the F-35 pilots know which way is up? Because sometimes they do a barrel roll.

16:39Yes, they still know which way is up. So all of that... So I just want to read you the beginning of Gemini's answer. It says, yes, fish absolutely know which way is up. They don't guess. They rely on sophisticated blah, blah, blah. Second paragraph. Since you're a content creator who uses high-end cameras like the Sony A7IV, you can think of a fish like a stabilized gimbal. I don't like that. I don't like that it like throws random context about me in a question about fish. So I don't know. I didn't feel good about that.

17:10I was wondering how... I was really trying to figure out how it like made the leap to the camera. That's what I'll say. But I guess that works. Like a gimbal. That's fine. I think that's fine. It's fine. But like I feel like it feels a little thirsty to like throw in just personal details about myself. So the AI is trying to signal to me that it knows about me and it's personalizing its answers based on our conversations. I don't know. It feels a little weird. I mean, if it would have answered the question with, you know, you could charge $200,000 for five reels about the subject if you get connected with aquariums.com.

17:42I mean, I wouldn't be mad about that, actually, if it could land that deal with the Florida Aquarium. I'm down for that. Speaking of deals, great transition, by the way. OpenAI acquired a podcast after we recorded last week. They acquired TBPN, which apparently originally was like the Tech Bros podcast, which they've since distanced themselves from that name. But if you're not familiar... Only in name. They've only distanced themselves in name. Correct. Correct. TBPN is a live streaming podcast.

18:14They have like... It's like a two to three hour show. It's a video show primarily. They've had lots of like tech founders and people on the show. Well, they made this deal. OpenAI has acquired the show. They average about 70,000 viewers per episode. And the deal is like a nine figure deal. I don't know if the exact number was... The guess is 150, I think. $150 million. And it's going to be part of OpenAI. OpenAI has supposedly stated that they're going to remain not neutral, but like that they'll still be able to operate how they operate.

18:48But it's like, listen, you have a lot of news that you want to get out there and a lot of positioning that you want to have OpenAI. Like, surely TBPN is going to be used as like a channel for that. Propaganda is the word you're looking for. Yeah, exactly. So it's like, hey, listen, TBPN guys, get your bag, I guess. Like, fine. That's great. If someone offers us nine figures for this podcast, we'll probably take it. And then we'll probably start a podcast called like Primary-ish Technology or something.

19:21Primary Technology-ish. Yeah, that'll be independent. But yeah, okay. But you thought this was silly. Well, I mean, good for the guys who just pocketed $150 million. Yeah, yeah, yeah. And the only downside is they have to go work at OpenAI. But they'll probably get some stock and that probably will turn out fine for them. I think this is so interesting because the mythology of Sam Altman is that he's like the greatest fundraiser in Silicon Valley, which seems to have borne itself out. I mean, the guy's raising like insane amounts of money for something that is not making any money.

19:55So good for him. But there are two pieces to this. One is he doesn't seem like he's very good at making deals because I don't know how this is worth $150 million. Their most recent year that says they made $5 million in advertising revenue and they only have 70,000 viewers. Now, I'm not saying that's like nothing. Now, they do a show every day, I think, for like three hours, I guess. But so they have a lot of ad inventory they can sell. I get all that. But I'm like 70. You paid a lot. You paid $150 million for 70,000 viewers, though.

20:26Like, Stephen, like we're not that far off. I mean, we have a ways to go, but we know people who have podcasts with more viewers or listeners than that. Absolutely. And I think a couple of interesting points. One, a large part of their distribution is on X because they stream to X. And that is going to be very curious how Elon Musk, if he does anything to like adjust the algorithm. He literally has a Sam Altman dial and he'll just turn it down. And, you know, they've had a storied history.

20:58The trial between the two companies starts like very soon. And one of the things he's asking for, Elon Musk, is that Sam Altman be removed as CEO. Like, that's actually part of the case. Right. So that's curious. Two, like if the tone of the show changes, like the reason it has as many viewers as it does is because people like the show as it is. And so if they try to change the show, like that's inherently going to change the viewership and listenership. So that's another part of it. Also, this deal was brokered by the CEO of apps or whatever, or CEI of AGI development now, which is Fiji Simo.

21:34And she's been behind a lot of the deals and acquisitions recently. Unfortunately, she is suffering from health issues. And she's about to take a leave of absence for an unknown period of time. If you listen to Dithering, I know John Gruber has some contacts within Love From, which is connected to OpenAI. And people were saying, like, even in recent months, Fiji Simo has not been physically present at the OpenAI offices. She's just in Slack a lot. But she also brokered this deal with TBPN and lots of others, but also just feels like a difficult time to then step away after making these deals and trying to make all these changes.

22:13So it feels weird, man. Well, everything feels weird. One, I think that what was revealed in that conversation that you were referring to is that she's never worked at OpenAI's headquarters, which is in San Francisco. And she lives in L.A. And she's only ever been working that way. And the deal is she's available basically at all hours in Slack, which is a terrible way to live your life. I just have to be honest. It's just terrible. But all of the things that she was hired, I think, from Instacart maybe as the CEO of Applications. And now she's the CEO of AGI, which what does that even mean?

22:46You are the CEO of Digital God. Does that make you Digital God? Like, if you are the boss of Digital God. She's the boss. Oh, that's true. If you're God's boss. How can you be the boss of Digital God? I don't understand that. Like, that is. Wow. That's too many layers. There's a lot of layers happening there. But what that just tells you is it's just words. It is just words that mean absolutely nothing, which just goes back to the second point I was going to make about the deal with TBPN, which is that this just shows you what happens when you have too much money.

23:17Like, you have a product that doesn't make money, but people are willing to give you money anyway. And then you have to figure out things to spend that money on. It's like people who win the lottery, right? And they just buy stupid stuff. It's like, now I have 700 pairs of shoes, 13 houses. I'm trying to get a jet. Yeah. I don't even know how to drive a Ferrari, but I have two of them in the garage. And if I ever take one out, it'll be the last thing I do. And also, I'm going to file for bankruptcy because I have no money left. It's like, I don't understand how this story is going to end, but it doesn't seem good.

23:49Yes. So there's that. There was also this very odd clip of Sam Altman. I guess they also kind of do an internal video podcast sometimes. And John Gruber linked to it, but he links to this post. It's the OpenAI Newsroom X account, which, again, just wild to me that X is still where all these companies send, like, full-on votes. Video and, like, Sam Altman and Elon Musk just hate each other. It's just very strange. But anyway, in this clip, Sam Altman is talking, which, can I just say, if Sam Altman's going to force himself onto a podcast like TVPN regularly, it's not going to be an exciting podcast.

24:23I'm just saying. I'm not trying to, like, down. He's just not a great, like, podcast personality, I feel like. And in this clip, which it feels like they're trying to do something podcast-y, but they're just sitting around a table talking, Sam Altman compares this moment, quote-unquote, namely the cusp of AG. And he basically goes on this long story of what it felt like right before the pandemic hit in 2020 and how he wanted to go for a long walk before he knew he was going to be shut in his house for months. And he's like, that's what this moment feels like before AGI.

24:54He's like, is that supposed to be comforting? Is that supposed to be like, we want this? That this is good? What does this even mean? Anyway. I don't think it's good. Also, I feel like, again, if we're going to have someone whose job is to bring forth this society-changing thing, we need a better spokesperson. We need someone who's a little bit better at explaining it than saying, we are in the days just before the pandemic. Maybe he doesn't remember the pandemic because, like.

25:25He told a very detailed story about it. But what I'm saying is it probably affected him different than it did me, who the day before we knew they were shutting schools down for what they said was two weeks. And I looked at my wife and I said, it will not be two weeks. They're not going back to school for a long time. And she said, then you have to go to the grocery store and buy all the toilet paper, which is what I did. And you know what? There were two things left. The day before they shut stuff down, there was only two things left to toilet paper. So I'm saying, not a good period of time. Not a good period of time. I don't think anybody remembers it fondly. I don't think so. Although my kids, the only thing that we think of that is fun is, like, we got to all be together for a long time.

26:00And then we were all still together for a very long time. And then there was another long time. There was a point of time where my wife and I looked at each other and realized our daughters, who one was a gymnast and one was a soccer player, the soccer player especially was able to go outside because they were outdoors, but they had to wear a mask. And our gymnast, they came up with other ways of doing practice because if you're a gymnast and you don't continue to train, you're no longer a gymnast. But our boys, who were very young at the time, we looked at each other one day in probably, like, June, and we're like, they've literally not left this house.

26:31Not for a thing in, like, four months. We have to go do something. We have to go. Yeah, go somewhere. Park or whatever. So anyway, it's weird. It feels a little bit like OpenAI is flailing a tad. And I understand there's still tons of money funneling into the company. It's not like they're flailing financially, but it does feel like directionally, maybe leadership-wise, they just bought a podcast. I don't know. It feels a little bit like flailing. Does it feel that way to you?

27:01Yeah. Okay. So here's an interesting thing. ChatGPT was an accident, right? It was an accident. They didn't release ChatGPT thinking it was going to be the fastest growing consumer tech product in the history of the universe. They just were like, here's the thing we made. And it was like, oh, wow, this is amazing. And so they were not prepared for what was going to happen next. And this is actually a recurring theme. We're going to talk about Anthropocure soon. But even Claude, the fact that Claude code is so good was an accident. They did not build those models thinking, this will be good at coding. It just happened to be really good. So what we are learning is that we don't always understand.

27:34Well, not don't always. We don't understand any of the outcomes when they set forth to do them, which means you can't build a predictable business model because you have no idea what's going to happen next. The next model might be better. It might be worse. It might hack into the Pentagon. It might do like who knows what's going to happen. So when you say it's flailing, it looks like it's flailing because literally that's what they're doing. They don't know what is happening and they don't know how do we harness this or how do we. And by the time they figure out how to harness it, they will have built something that can no longer be harnessed. That's pretty good. And speaking of flailing, they released an integration with Tubi.

28:08So that's the big new Chagipiti feature. Okay, I take it all back. I take it all back.

28:13Yeah. I mean, anyway, I don't use, do you use Chagipiti? Like your AI usage currently, you know, you have the big four or five or whatever. I mean, you have Claude, Chagipiti, Gemini, Perplexity. Perplexity is not in a, it's a search engine. It's different. It's not. And I actually did. I mean, I don't use Perplexity anymore because I use Claude more. But like percentage wise, like what's your top three? Like what are you using the most? I would say most of the time it's right now split between Chagipiti and Claude.

28:48So it's probably 40-40, those two, and then maybe the other 20% is Gemini. And it kind of just depends on the thing that I need to do. Or if I ask one of those two something and I'm like, you're so dumb, then I'll go to Gemini and be like, what's the real answer? And Gemini will often give me that. So, but Gemini, I would actually, to be honest, I would use more often if it had a Mac app. I would use it more if that was the case. It is just too cumbersome to just only be using it in a browser window. Correct. Which do you use on your phone? Do you have any of them on the home screen?

29:19I typically still use Chagipiti on the phone. Interesting. Yeah, I would say when it comes to shortcuts usage, that's where I use Chagipiti via its API because I have a bunch of shortcuts built in with those requests and it works well. And so I just keep doing that. When I'm on the phone, I'm using the Gemini app for like, do fish know what direction is up? Like I'm using the Gemini for that or shopping. Like if I'm trying to find, I needed to find a trampoline replacement. I asked Gemini. I asked for that. On the Mac, it is mostly Claude.

29:51I'm mostly doing Claude for shortcuts building, for, I'm going to talk about a little later, like if re-did a page of my website and it was like really good and it did really fast. So yeah, it is Claude. Well, my thing about Claude is that the Mac app is terrible. It's absolutely the worst Mac app. You don't like the Mac app? It's just an Electron app. You might as well be using it in Chrome. It is such a bad Mac app compared especially to ChatGPT and it is completely and utterly unpredictable. If you say, give me this thing, sometimes it will just give you text in a window.

30:24Sometimes it will open a side pane with markdown. Sometimes it makes you a Word document and then it says, open it in pages. And I'm like, why don't you just- Sometimes it gives you the menu of options and you won't do anything until you select an option. Just give me a chat like thing. Just let me copy and paste. It's fine. I'll figure out where to put it. I don't want you ever to, by the way, never make me a Word document. Like, come on. You never do that. You don't know me at all. I also find the copy to clipboard icon in Claude doesn't work a lot of the time and I have to literally select the text, at least when it comes to like coding and regex.

30:55I have to like select it manually in the code block. Well, and the one thing ChatGPT has started doing a couple of months ago is you actually have to copy something twice. You can select it all and it doesn't enable copying. You have to do it a second time. And they're like, no, we don't want you to leave. So yeah, copy and paste is a struggle, but they bought a podcast for 150 million dollars. I'm saying you've been able to copy and paste on the iPhone since 2008. Surely the ChatGPT app should have it figured out by now. That is true. Anyway, we're going to talk about the other end of this because Anthropic has made a model that is apparently so powerful they don't want to give it to normies, like to

31:30regular humans. But before we do, I want to thank our friends. And I'm actually excited because I've reinstalled this application on my MacBook Air and have been using it a lot, which is clean my Mac. I was using my MacBook Air as like the testing bed for all the AI stuff. I had, you know, installed OpenClaw for a minute and I did all this and I was like, you know what? I need to clean this Mac. I want to get everything off of it. And I knew the one app that I needed to install, which was clean my Mac. I'm not even reading script right now. I'm just telling you, I downloaded clean my Mac. I had it scan my whole Mac for different things, cleared up a bunch of space, removed

32:03stuff. And then I was able to go tell it like, hey, these are some apps I want to uninstall. Can you look at all the places these apps might have put something and just delete it? Totally does all of that. It walks you through the process. There's an assistant that can say like, do you want to do this? Do you want to do that? Clean my Mac is amazing. And so I'm running that on my MacBook Air and it helped me just have peace of mind knowing I didn't have any of that cruft on there. It can also help you free up space for things like messages. So if you have a bunch of big attachments or you're finding the storage for your Mac is being taken up, maybe by the message conversations, like it can identify all of that.

32:38And also cloud storage, whether it's iCloud, Dropbox or Google Drive, if some of those folders are taking up space and iCloud is notoriously difficult to manage the storage on your Mac because you just have to tell it like optimize storage, then you get no other settings. Well, clean my Mac can help with all of that as well. So get clean my Mac if you either want to just delete apps cleanly, free up some storage space, work on your cloud storage and messages, clean my Mac, it's the way to go. So get tidy today. Try seven days for free and use a promo code primary tech for 20% off at clnmy.com slash

33:11primary tech. I literally used our own promo code on it because they didn't like give us a free license. Like I literally paid for it and used our promo code. So seven days for free, use code primary tech for 20% off at clnmy.com slash primary tech. The link is also in the show notes. Our thanks to clean my Mac for sponsoring this episode and our friends at Granola. Listen, you probably use lots of different virtual meeting things. You've got to do Zoom over here. You do Google Meet for some other things. Maybe use Riverside. Maybe you use WebEx. That's what Apple uses when you do things there.

33:41Well, Granola can work with all of that because it is a Mac app and it will transcribe your meetings and do all the things so you can have those summaries. You can get the bullet points. You can do all of it using Granola on the Mac. It's an AI powered notepad built for the way real people actually want to meet. You take rough notes like you normally would in the background. Granola is going to securely transcribe the meeting, turns everything into a clean, structured, actually useful notes when the meeting ends. And the best part is Granola works through your device's audio, which means it integrates seamlessly into the video conferencing tools you already use.

34:14No setup, no awkward tools. Like I said, it'll work with any meeting app, even if it's in the web browser that you're already using. And you can actually listen instead of frantically typing every word. Granola, I want an app like this. If I'm going to be doing transcribing on my Mac, I want a Mac app like this, doing it in the background based on the audio input, not dependent on what app is running in the browser. And so Granola is the way to go. So if meetings are eating up your day, Granola is a no-brainer. You can try it totally for free for three months. That's a long free trial. Three months.

34:44Just head to granola.ai slash primary. That's granola.ai slash primary to get your time back. And you get three months for free at granola.ai slash primary. Thanks to Granola sponsoring this episode. Now we talk about OpenAI and whether or not they're flailing or not, but who's not flailing is Anthropic as far as the kind of models they're building. And so they built Mythos AI, which let's just say, epic name.

35:14If you're going to call something Mythos, this thing better be like legit. This thing better be serious. But this model, Mythos, can apparently identify and exploit software vulnerabilities, apparently with unprecedented accuracy. And so there's some concern that this model could be used by nefarious actors. And so Anthropic is only allowing like big companies to be able to use this model and actually run this AI. And so this is like, I guess, exciting and scary.

35:45Yeah, and I think it's interesting, like I sent you the NBC News link because that just tells you this is broken out of just the AI tech bubble. This is NBC News that is reporting on this new model. It is interesting. And this is why this is also difficult, because if what you wanted to do was show that you are by far the leader in making a model that is smarter than everything else in the universe, the best case scenario is that you created a model so dangerous you can't give it to the

36:16public, right? But you're going to give it to these other companies and let them use it to poke around in their software and all that kind of stuff. And so this is like a marketing coup for Anthropic. I don't think that they set out to do this. I think what I said before is true, that they have no idea what the outcomes are going to be when they start building some of these models. And they're like, wow, this one is really good at identifying vulnerabilities. So we probably shouldn't release it to the public because suddenly every bank and every whatever is going to be vulnerable to hackers who now have the strongest, most powerful tool

36:48in the universe. Also, like it does seem to beg the question, like, should we keep doing this? They never asked if they should. Yeah, it's like Jurassic Park. Why are we doing this? Like, why do we keep doing this? And why does no one think that this is a problem? I mean, because this is the the promise of AI, like this is like they're going to keep pushing because they want these tools to find a use or vice versa for people to say like,

37:19OK, here's an unequivocal use for this model that is valuable enough for companies to pay a lot of money for. And now suddenly it doesn't seem like a bubble because they've created a product that is actually valuable. And if this is really that good and companies like NVIDIA, Google, Microsoft and even Apple are going to use this model to identify vulnerabilities and it does better than like, you know, I don't know if people know Apple will have what do they call it? The bounties, right?

37:49Like if people actually just... Bug bounties, yeah. Yeah, there's like bug bounties. Like Apple will literally actually know someone personally, not in my family, but like very close that actually found a vulnerability in iCloud, reached out to Apple, they confirmed the vulnerability and they gave them a bunch of money. Like Apple literally does that behind the scenes. And so if this model can actually find all the vulnerabilities faster and find more vulnerabilities than even the bug bounties of paying actual humans, well then yeah, that is a legitimate

38:23use case. Like that's something that companies like Microsoft and Apple and Google would probably pay a lot of money for. But yeah, man, it seems like if it's too scary to release to the public, but I also get that because some dude with a bunch of Mac studio in his basement is going to try and get into Wells Fargo. And so yeah, don't release it to the public, I guess. Well, and the reason is like it is better because it doesn't need an app and it doesn't require Red Bull, like it can just keep doing this, right? No, I mean, like there is a, there is a point in which it reduces all of the friction associated with trying to hack into things, right?

38:54It is better at finding these things than even humans, but one of the, it doesn't need any human help to do it. So yes, you don't want to just release that freely into the world where the people who might be inclined to do bad things with it can do that super easily. But I also just feel like we, I don't think we are prepared though for a world where this becomes the norm, right? We are the, many of the things that keep us safe today are predicated on the idea that

39:24it would simply take too long for computers to find the way in, right? Encryption. Encryption is just based on the fact that the math is just hard enough that it will take too long for a computer to get through. That's why quantum computing is a problem, right? And so that's why there's post-quantum cryptography to like develop cryptographic algorithms that can't be broken in a matter of hours or days instead of a thousand years or whatever. So, and if you're able, like many of the vulnerabilities that exist in all of the stuff that we're using on a regular basis is simply, we're just kept safe because like, it's just

39:57too hard to find them, right? It's like security by obscurity, right? Like we just don't know that these things are there. And so now suddenly we're going to know that they're there. Now it is good for companies like Apple and Microsoft, especially to have the ability to find those bugs more easily because if they exist, they're a vulnerability and you can't just count on the fact that someone won't find them to keep us safe. So that is good. But I also just think there is a part of this where maybe it was better. Maybe the trade-off was better when it was harder for us to find the bugs to patch them,

40:29but it was also really hard for the bad guys to find the bugs. So they weren't exploiting them. Now we're just making it easier for everyone. It's kind of like if you put the back door and the encryption, then the good guys and the bad guys can use it. I'm wondering, it's like at some point we will have crossed the line and we're like, maybe we should have stopped. Maybe we just shouldn't have done this. Yeah. So I'm going to put this link in the show notes. Apple's article about quantum secure cryptography.

40:55One, like I've talked about the quantum computing video that I've watched with Hannah Fry. She goes to like IBM and like there's like two quantum computers in the US and it takes like an entire facility to run it because they have to like cool the chips down to like zero Kelvin, like colder than space. Like it is crazy stuff you have to do to be able to get quantum computing to just work. So this is not something some bad actor can do in his basement. And so if you're on the quantum secure level of cryptography, you should be good.

41:27But that's also assuming there are no vulnerabilities around those safeguards. And that's what maybe this model would find. But yeah, let me say one thing about that. The risk right now is not that the guys in the basement are going to be able to get into the stuff because they can't because there's, you're right. Quantum computing is not ubiquitous. You don't have access to it. Like you have to go to like the Mayo Clinic or somewhere else that IBM has, like Google has one, whatever. The risk is they're just scooping up all the data right now. They can't do anything with it now.

41:57But eventually everything that is not protected by quantum secure encryption will be decodable, right? They will be able to get all of that data. So that's why the race is on now, because you need to make algorithms that will resist quantum computing, even though those things don't exist, even though the quantum computing capabilities are not there now, because all China has to do is scoop up everything on the internet. And someday in the future, whether it's two years or 10 years, they're like, sweet, now we can just get, we can, we can decode it all. And so that is like where the real risk is.

42:29It's not that someone can hack into your bank account today. It's that everything on your hard drive or everything that you sent on the internet, right? Because everything you send back and forth on the little S in HTTPS, right? It means it's encoded. The traffic back and forth between your computer and a server is encoded, but if they just scoop it all up anyway, fine. Someday they'll know. Sheesh. Well, if you just got some, some bad feelings like I just did, just try to ask Claude or Gemini what today's date is, and you might feel a little better. No, you won't. You will not feel better.

42:59Because they can't, they don't know what today's date is. So we're still, hopefully a little ways off. But anyway, so that's Mythos AI, but also some anthropic news. Apparently, a court has declined. I hate how these headlines are phrased. I understand why these courts. There's like a triple negative in here. It's a triple negative. It's so annoying. So a U.S. court declines to block Pentagon's anthropic blacklisting for now. Okay. I'm going to explain this. I'll explain this for you. Okay.

43:29You may remember, listeners, that the Pentagon, because they were real, real mad at anthropic, because they wouldn't let them like, whatever. I'm not even going to get into that part, but they were real mad. So they said, fine, you won't let us do anything we want with it. We're going to say that your supply chain risk. Again, the logic there doesn't make any sense. So anthropic sued and said, we would like you to not be considered a supply chain risk. They filed a case in a federal court and asked the court in the meantime to issue a preliminary injunction that would pause that.

44:01They would say, until we decide the case, please, we'd like to keep operating the way that we were. And what the court did is, we're going to hear the case. We haven't decided the case yet, but we are not going to issue that injunction that blocks that for now. So what that means is that the Pentagon can continue to just remove anthropic from its systems. It could still be the case that in the future, that designation gets reversed. I'm not a lawyer, so I don't know a ton about this. But typically, I believe that if a court will issue a preliminary injunction when it thinks

44:33that it's just like sort of eyeball test more likely than not that you might prevail, like we don't want to have irreversible damage now if you're probably going to win later, right? So the fact that they didn't do that means that the court is not convinced that either of those things are true, right? That there's irreversible damage now or that there doesn't mean the court is saying you don't have a case and you're going to lose. It just means that balance of equities was not high enough to be like, eh, we should pause this right now. So that's what that headline means. Right. Thank you. Thank you for that.

45:05So still unknown, like in process of whether that happens. Last thing on Anthropic, they actually have another product that handles the part of building AI agents. And so they're trying to create products to build AI agents to do things. So like robots building robots, basically. And I'm going to throw in here. I actually interviewed Microsoft's VP of like core AI stuff for SaneBox. I know I just said a lot of words there, but actually. And most of them I didn't expect to be in the same sentence.

45:36I know. But I just talked to this guy yesterday and it's interesting. He was like, oh yeah, I got agents running everywhere. I got this agent running Excel over here. I got this agent that books my travel over there. And so like, you know, internally how these like AI teams are using AI agents is quite interesting. But anyway, that conversation will come out probably maybe a month or two from now. But yeah, agents building agents. It is like maybe you could build something useful, but I don't know.

46:06Also, you know, what if the agents are mad at you and they build agents you don't want to have around? If they don't know what day is, can they even do anything bad to you? I don't even know. That's true. They have to chase you through the future. It's like, what was that? What's the Justin Timberlake movie or whatever? Oh, out of time? Out of time. Yeah. Listen, that is not a great movie. No, that is a good movie. But it is an excellent. Sorry. Out of time is the Denzel Washington movie. The Justin Timberlake time movie is. Please don't tell me it's just in time.

46:38No, it's just in time. Okay. In time is it. But it's actually. I think Nate and I talked about it on movies on the side. Like the premise and like the. It is. It's. It is a good watch. It is a very entertaining movie. And I think Cillian Murphy's in it. And I just always think of him as scarecrow. Yes, he is in it. Amanda Seyfried. Olivia Wilde. They're all. It's actually really good. Basically, the premise of the movie is like in the future, the currency won't be like dollars and money. The currency is literally time. And when your time hits zero, you die. And you have to like work for hours and weeks and minutes.

47:11And anyway, it's. And the rich people have just infinite time. Correct. And so it is a really interesting movie. But anyway. Okay. Anyway, I'm sorry. No, no. I don't know how we went there. Listen to our movies on the side episode on that in time, which is now video in the last couple episodes. And movies on the side still featured in Apple Podcasts for video. So YouTube is going to make it easier to, has made it easier to deepfake yourself.