SpaceX’s IPO Triumph, Anthropic’s Fable Fumble, OpenAI’s Price War

Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover: 1) SpaceX's massive IPO 2) What SpaceX's success means for its AI competitors, OpenAI and Anthropic 3) Is SpaceX's outsized valuation a feature, not a bug? 4) Goldman Sachs eats Big Bang burritos 5) Anthropic's flubbed Fable rollout 6) Does the gaffe mean Anthropic really believes in the power of Mythos? 7) A fun Fable conspiracy theory 8) OpenAI may drastically lower prices 9) Are AI models a commodity? 10) Wait, does Siri work now? --- Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. Want a discount for Big Technology on Substack + Discord? Here’s 25% off for the first year: https://www.bigtechnology.com/subscribe?coupon=0843016b Learn more about your ad choices. Visit megaphone.fm/adchoices

• 0:00Speaker 00

  SpaceX IPOs soars and makes Elon Musk the world's first trillionaire. Anthropic's fable rollout doesn't go as planned and OpenAI may drop rates drastically. That's coming up on a big technology podcast Friday edition right after this. This episode is brought to you by Genesis. What does it actually take to put a Gentic AI to work across an entire enterprise? I recently spoke with Genesis chairman and CEO Tony Bates about why orchestration is emerging as a competitive advantage in customer experience and what companies need to get right in the AI era. Then I sat down with Adam Mitchell, head of enterprise business solutions at Voya Financial to go inside a large scale agentic CX transformation, including the business decisions, trade offs,

• 0:45Speaker 00

  and lessons along the way. You can watch both conversations now on my YouTube channel today. It's never been more challenging to get frontier compute. How do you know where to look, who to trust, or how much to pay? Akash is an open cloud that provides easy access, flexibility, and peace of mind. That's how companies like Razor were able to cut their inference cost by 15 times. Visit akash.network slash Alex for a free consultation. That's akash.network slash Alex for a free consultation. Welcome to Big Technology Podcast Friday Edition where we break down the news in our traditional cool headed and nuanced format. This is definitely one of those weeks where I have been waiting to get Rajan's take on the news for days on end.

• 1:29Speaker 00

  I haven't heard his thoughts on the fable fumble, but I'm so excited to get his perspective on what it means now that we've seen a version of Mythos. So we'll do that in the second segment. We'll also talk about OpenAI potentially drastically lowering prices for its GPT models, but first we'll talk about SpaceX and the SpaceX IPO, companies trading above trillion dollars will unpack what it means. Joining us as always, of course, on Fridays to do it is Ranjan Roy of margins. Rajan, great to see you.

• 1:59Speaker 01

  Wow, you know, when the fable fumble is going to be story number three, we've got a, we've got a busy week, but we do have the world's first trillionaire. I think it's time to celebrate, right?

• 2:11Speaker 00

  That would be Elon Musk. You know, this is a very interesting IPO for a number of reasons. Obviously, it's going to factor heavily in the AI world because if you think about it, a year ago, SpaceX didn't really have an AI play. Then in the run up to the IPO, the company signed a $15 billion deal with Anthropic, a $30 billion deal with Google to license out the data centers that it's built for its own AI efforts. And now it's become like one of the biggest clouds in the world. And that has propelled it to a $75 billion IPO.

• 2:51Speaker 00

  It's above $2 trillion. So Ranjan, I just would love to get your perspective on the IPO. Of course, the broader story about what it means. But also, you know, they came out telling us that 26.5 of 28.5 trillion of their total addressable market is AI. So is this the world's first AI IPO? And if it is, what does it mean for OpenAI and Anthropic?

• 3:14Speaker 01

  Is this the world's first, well, it's not the world's first AI IPO. I think it's certainly out of the big three. It is amazing, I have to just say, like, how in the span of, I think, what is it? Six to nine months, Elon Musk has turned SpaceX into an AI company with the acquisition of XAI, his own company, with these incredible Anthropic and Google AI Cloud deals, with the potential kind of a, I don't know if it's an acquisition of Cursor, at least it's the right to buy Cursor, suddenly has gotten in the fold. And again, we talked about this the other last few weeks.

• 3:58Speaker 01

  The numbers are interesting, again, in terms of like the actual revenue, in terms of how much money they're losing, in terms of Starlink, the actual infrastructure business being the cash cow. But I just have to say, like, let's take this moment to actually...

• 4:15Speaker 01

  This is what Elon Musk is incredible at. And like, I have to admit, between the Anthropic and the Google contracts, between the way he rolled up all his different entities, to pull this off in this way, at least for this moment, is pretty spectacular. It's pretty masterful.

• 4:34Speaker 00

  Right. And, you know, there was certainly money that was going to be earmarked towards AI IPOs from investors, right? And certainly, people were waiting to invest in SpaceX. But you could argue that without the AI side, SpaceX is not a very compelling story. It's a communications company with Starlink, and this launch business that's not profitable. Of course, a lot of that money is in R and D. The AI story, you know, gave the market, the market, I think, was looking for something to be enthusiastic about for SpaceX, something to justify putting in the money. And now they got that story and they responded accordingly.

• 5:15Speaker 00

  So the question to you, Ranjan, becomes is, is Elon taking money out of Sam and Dario's pockets? Because maybe some of that money was going to go to the OpenAI's and Anthropic's of the world. Or has he just set the precedent that if you want this, you know, AI company that's sort of spun up in a moment, you want to get behind that story, then you should definitely get behind the other two? What's your thought?

• 5:37Speaker 01

  I think, so, I actually think that the like, masterful pivot that was made, and again, I say this all, not saying that any of this is good for the markets as a whole, or if like retail investors in general. But again, the brilliant kind of pivot he made was, that Elon Musk made here was, it's no longer just kind of like an AI play in the way XAI and Grok was like, where he was positioning those businesses. Data centers became the business du jour. And he turned this into a data center business. Remember, originally, he talked about Colossus, he talked about all of, you know, the, it was

• 6:19Speaker 01

  a major part of the overall infrastructure, but it was to fund Grok to be the leading AI business. But then he just, with two contracts, two phone calls from

• 6:30Speaker 01

  I believe, I know, Google is certainly an investor in SpaceX. I'm not, I'm not anthropic. I'm hoping that would be crazy if they were. That would just take this to another level. But, but again, couple of phone calls and Elon Musk managed to completely shift the, it's crazy. Like, when was SpaceX founded?

• 6:51Speaker 00

  Two thousand. Two thousand two.

• 6:53Speaker 01

  Two. A 24-year-old business, you completely transform how the market understands it with two phone calls. Couple of lawyers, I'm guessing, maybe a couple of meetings, but that's it.

• 7:07Speaker 00

  All right. But then, Toc, what do you think about the implications for the other two?

• 7:11Speaker 01

  I think it's, it's hard to say. As of recording time, SpaceX trading at 171.3. I think it's at like a 24% increase from the IPO price, 135. We are at a 1.75 trillion dollar value. All that sounds good, but I mean, we got a long couple of months ahead of us. Like, I think, I don't know what, normally the summer months are, especially on this, certainly any IPO activity, they're just quieter. I think we're going to be in for two of the, two to three of the most wild news months in the summer ever in any of these. And again, I think, where is SpaceX

• 7:56Speaker 01

  trading on Monday? A week from now, a month from now, it is impossible. Remember that so much of how this is orchestrated, it's only a 4.3% float. Like, it's tiny. It's like, everything was done and it was done perfectly to actually kind of like bring about this massive valuation and IPO, but. We got a long ways to go and I don't know. Do I, where do you think it's going to end up for our friends Dario and Sam?

• 8:31Speaker 00

  Yeah, okay. So since you're dancing around the question, Ranjan, I'm going to take the answer, okay? I do think that, yes, this is, so it's a small float, which means it's a small percent of the company. But that being said, it's still the biggest IPO ever times two and a half or three, right? So, no, times two and a half, right? Because Saudi Arabia raised 29. So this raised eight, you know, in the realm of 80. We're still not sure because there might be more shares that they, that they sold as this thing went live. So that's a lot of money. And I don't know if it would have been that level of cash if the AI side of this wasn't included.

• 9:12Speaker 00

  So I think that that money would have otherwise gone to OpenAI and Anthropic. And, and I think that because Elon Musk has, you know, with, say what you want about him, he's held the value of Tesla pretty much. If people were looking to make an AI bet, they wanted to make this bet and Elon has become a vehicle for it. Even if, you know, like we said already, XAI hasn't really panned out as an AI foundational model company or an application company. But then again, Elon Musk did build all that infrastructure and he's been able to lease it out. I wouldn't even call him a NeoCloud. He's just a cloud. And remember last week, we talked about how OpenAI has these ambitions to become an AI cloud?

• 9:57Speaker 00

  Well, you know, overnight, pretty much SpaceX has beat them to it. And that is what we're seeing reflected in the IPO.

• 10:05Speaker 01

  Well, from, okay, let's take OpenAI first. I mean, Sam has definitely talked about AI cloud ambitions very loosely at times. Like, there's never been any real concrete plans, I believe, that they've kind of communicated. But he'll talk about it every now and then as like a potential business line. I do agree. I think this is very interesting and how quickly, just on the span of two contracts. They kind of took that, I do think they have taken that away or kind of taken the legs out from OpenAI moving in that direction, which is also funny that OpenAI, sorry, Anthropic kind of helped spur this as well.

• 10:47Speaker 00

  Oh, they knew what they were doing.

• 10:49Speaker 01

  They knew what they were doing.

• 10:50Speaker 00

  And so did Elon.

• 10:51Speaker 01

  Maybe Google as well. And maybe so did Google. Ooh, wait, I like this now.

• 10:55Speaker 00

  World against OpenAI.

• 10:57Speaker 01

  It's Anthropic at the enterprise, Google at the consumer, OpenAI, sorry.

• 11:03Speaker 00

  SpaceX at the infrastructure.

• 11:05Speaker 01

  Sorry, I can't even remember what I'm supposed to call the Elon business right now. SpaceX at the infrastructure layer, one by one, all going at OpenAI. I don't know, but I feel, do you think that even enters the head of any of the dealmakers as they're approaching these kind of contracts?

• 11:28Speaker 00

  Absolutely.

• 11:29Speaker 01

  Absolutely? Okay.

• 11:30Speaker 00

  Yeah, OpenAI is a big threat to all of them. So, and well, where it's not a threat, it wasn't a threat to the space business, certainly was a nemesis when it comes to Elon. Yeah, I do think they all want to hurt OpenAI. But let's talk a little bit about this valuation before we go, before we move on. So the valuation is a little bit like crazy. It's a $2.2 trillion company. It's not going to have earnings that justify that at all for years. And you look at the breakdown in the prospectus and it's like, well, what is going to fund this company, right? And so the space, when it comes to the total addressable market that they put out there, space and connectivity are like $2 trillion to them, the total addressable market.

• 12:11Speaker 00

  The total addressable market for AI is 26.5 trillion. We've talked about it before. It doesn't make any sense. But let me, um, let me sort of make the case for why this crazy valuation is a feature in the system and not a bug. And then I will turn it over to you and you will be very mad. I'm toying about publishing this on big technology. I probably will before this episode comes out. Without these crazy valuations, right, which are just not mapped to reality of the business, you wouldn't have satellite internet or space data centers or these moon shots.

• 12:51Speaker 00

  And so this valuation craziness is actually enabling the risk-taking, similar with AI, right? It's enabling the risk-taking and some of those risks eventually pay off. And that's where you see progress. What do you think about that?

• 13:05Speaker 01

  That is a hot take, Alex. I will give you that if we, for listeners who are not watching this on video, you can, Alex is just doing some air guns here with this after the hot take.

• 13:20Speaker 00

  You're gonna get suspended by the NBA now, but okay.

• 13:24Speaker 01

  Wait, are you not allowed to do that in the NBA?

• 13:26Speaker 00

  Well, Ja Morant can't, but anyway, let's move on.

• 13:28Speaker 01

  I mean, but Ja Morant, he can't, he shouldn't. But, okay, anyways, so I do find this interesting, the idea that for a number of decades in America, one could argue that for these literal moon shots, this is where the innovation, like the US government partnered with private organizations would come in and fund these and kind of like push forward a lot of, whether it's the internet, whether it's the original space program, and kind of together, the public sector, along with the private sector, would kind of get us to the moon. Now, so

• 14:07Speaker 01

  what you're saying is the only way we can do this is to have a highly orchestrated conglomerate of kind of self-interested companies enrich us into the world's first trillionaire at the expense of the mainstream retail investor in order to one day hope as the permanent underclass, we can get some data centers in space. Is that what you're saying, Alex?

• 14:37Speaker 00

  Exactly right. I'm not saying that it's the only way to do it. I'm saying that like there's another side to see this valuation. But I also think what you hit on here is really important, which is that retail. So everyday investors will be a very big part of this, right? So more than 20% of the IPO allocation went to retail investors, which means, and this is sort of where the negative side of it is, the risks typically borne out by government and venture capitalists are now being borne out by everyday investors, who may end up doing fabulously well, but they may not. And I won't be the one that says they shouldn't be able to invest. But it is something to think about when it comes to, my hot take is, yes, these big bets are being enabled, but who is enabling them?

• 15:23Speaker 00

  And retail is becoming a bigger, bigger part. Your thoughts?

• 15:27Speaker 01

  Well, I think, again, it's part of the whole game. It's been for a number of years, like Tesla was the warm up to this. Tesla was, again, an incredible product, very like, you know, market leading. It was revolutionary. And then that gave enough of that credibility to kind of build this larger story, kind of like, to keep up, you said, maintain this valuation as well. And have that retail energy and to kind of translate that into, there's always, you know, like the, the SolarCity, the Boring Company, like all these different pieces were all, all feel kind of like a warm up to this is like the greatest act of all time.

• 16:12Speaker 01

  And this is like, you know, to then pull it into world's first trillionaire, $2 trillion IPO off $20 billion of revenue. I guess we extrapolate that out after the Google and Anthropic contracts a little bit more, losing $4 billion a year. I mean, it's incredible. It's truly incredible.

• 16:33Speaker 00

  Okay, I'm gonna give you an option of questions to answer as we end this segment. Here's a couple. Is the SpaceX IPO good for society and business? And can SpaceX maintain its valuation perpetually? Or is this eventually gonna come down to earth? No space pun intended.

• 16:55Speaker 01

  I'm gonna, the first one, no. I mean, I think it's very good. And I'm actually, my favorite, my favorite little tidbit I saw online was,

• 17:08Speaker 01

  as the investment banks who led this IPO are making a record, are reaping a record-making $550 million in fees, Goldman Sachs, who is leading the deal, has remade its entire lobby and cafeteria with a space theme. There's a fueling station with lattes and macaroons that looks like, that look like moon rocks. In the cafeteria, the bank is serving a mission control brunch with big, quote unquote, "big bang burritos." So when we're talking about who's benefiting, again, Goldman's got big bang burritos because they're very happy about all the fees as retail investors are coming in, thinking they're excited.

• 17:51Speaker 01

  So if that, does that, in a roundabout way, help you understand whether I think this is good for society or not?

• 17:59Speaker 00

  Oh, it sure does. I mean, of course, myself and all of our listeners and viewers will react in a couple of ways. First of all, yeah, that's a lot of fees. Second of all, this is clearly a perspective of a hater who just wants his own big bang burrito. And I can't blame you. I mean, I would be hating too.

• 18:19Speaker 01

  Do you think the big bang burrito is different than, do you think the big bang burrito? I'm only, I'm here on 34th Street, 200 West Streets, maybe I'll take the subway down. Everyone else is protesting, I'm just gonna have a sign, just one big bang burrito.

• 18:37Speaker 00

  Better be special burrito. Like, I think it will have to have like a Snickers in the middle or something like that.

• 18:43Speaker 01

  That sounds like, that sounds like what, remember when GPT-2 was like giving us stuff like that?

• 18:52Speaker 00

  Gluing glue on the pizza?

• 18:54Speaker 01

  Actually, wait, Google told us to eat rocks. I miss those days. I miss the days of when Gen AI was just telling us to eat rocks. Now we got data centers in space.

• 19:07Speaker 00

  I will say that I'm more bullish on it. I think ultimately I'm sticking with my, let's fund the crazy things. But if you're a retail investor, you have to be careful. Okay, can it hold the revenue or can it hold the valuation?

• 19:20Speaker 01

  Well, certainly, I hope it can hold the revenue. Again, Starlink is very interesting. Like, Starlink is, it's an $11 billion business built over a number of years, we all know. I still, I fly Delta usually, and I wish it had Starlink. So can it grow? Can it hold the valuation? I think it is completely susceptible to whatever the overall market trend is doing. I actually think everyone is talking about what SpaceX is doing to Anthropic and OpenAI. I think what's more interesting is if they come out poorly, I think that's going to dramatically impact SpaceX's valuation. So maybe Elon shouldn't be going at Sam so aggressively. What do you think?

• 20:05Speaker 00

  Yeah, but I don't think they're going to come out poorly. I don't know. It's one of those things that he will have a lot of, a lot of support, so to say. Like, as they say in the financial world, that stock will have support and it will have a floor. But it'll probably be tough to grow much, much more than where it is today.

• 20:27Speaker 01

  No, but the, this is where the floor or the support line comes from one of two areas. One, if there's enough vested interest that is just so widespread and distributed that has capital, that is not getting hit in other areas, they will do whatever they can to kind of hold that valuation because there's been so much money that is just being made off of this. But if that money starts getting hit in other parts of their portfolio, which there hasn't been anything dramatic yet, then... They're not going to be able to hold that. And then the other side is retail as well. And the same thing would hold to them. So both on the institutional retailers, retail side,

• 21:07Speaker 01

  I feel like it, to me, it's just in a happy market, this can continue for at least a little while. Anything, little, anything, uh, kind of, what was the liberation day? Almost forgot the name of it. Anything like that could screw a lot of this up.

• 21:28Speaker 00

  Yeah, I think as long as you keep feeding people the Big Bang burritos, you're going to be happy. But once that supply runs out, it could be trouble for the stock.

• 21:38Speaker 01

  Okay, Big Bang burritos are only for bankers. Let's be real. Not everybody gets a Big Bang burrito. No, no, no, no, no. This day and age, Big Bang burritos only for the bankers.

• 21:52Speaker 00

  Well, that is the seeds of SpaceX's demise. Okay, so let's talk about Fable. I've been waiting to hear your Fable take for a long time. Let me set it up. This is from the Wall Street Journal. Of course, Fable is Mythos. So this is from the Wall Street Journal. Anthropic's new Fable AI model is met with user backlash over restrictions. Anthropic is finally giving the public a taste of its next generation model, but its blunt safety barriers are angering some artificial intelligence developers and users, and feeding a growing debate over who should be the keeper of dangerous AI capabilities. Dubbed Claude Fable 5, the new model that Anthropic released Tuesday is an update to the Mythos model that the company said was too dangerous to release widely. That model

• 22:35Speaker 00

  spooked government officials and cybersecurity experts for its potential to find unknown vulnerabilities in software used around the globe. Fable comes with broad restrictions, Anthropic says, are aimed at kneecapping its ability to assist users with potentially dangerous activities. So Ranjan Mythos is here, but it is a neutered version of the model, and it's called Fable, and it has sparked the biggest backlash to a model release I have seen to date. I've been waiting for this. What is your perspective on the model and the rollout?

• 23:10Speaker 01

  I gotta say, in the same week... And I've been saying this for regular listeners will know that I feel there's like a large element of marketing into the way Anthropic has kind of communicated around Mythos and now Fable 5. It's crazy to me when Gary Marcus, David Sacks, and then even Gurgely Oros.

• 23:36Speaker 00

  Gergay.

• 23:37Speaker 01

  Gergay, the pragmatic engineer. Sorry for butchering the name there. Everyone is saying the same thing. This was never about safety. This was all marketing. I feel when you have three people from that who generally I would say like are just in very different areas in terms of perspective, all coming around, I'm going to feel a little vindicated. Was the, after saying that it was not just marketing and it was a true threat to humanity, do you feel scared today with Fable 5 in the wild?

• 24:15Speaker 00

  Let me read a little bit about the restrictions and then I'm going to ask you a question. When a user touches on sensitive topics, sensitive topics like bioweapons and cybersecurity, Fable pops up a notification and then redirects the conversation to an earlier, less capable model. Fable also degraded the quality of its responses about high-end AI development to be less useful for developers looking to build AI tools that might not have the same safeguard. For those responses, there was no pop-up notification. However, the company cited national security and its own Terms of Service as reasons for the invisible restrictions. It's yet, it's since rolled some of that back. Here's my question to you, Ranjan. If Anthropic

• 24:57Speaker 00

  didn't believe that this model had these capabilities for cyber hacking, for bio, potential bio warfare, for building models with, without the same safeguards, if it didn't believe it, why would it score such an own goal here with the rollout? Like, do you think it's just complete, a complete sense of toe-end deftness that it thought that it would put, let me finish the question, that it thought it would put its company in a better light to do something that ended up becoming so despised by users?

• 25:31Speaker 01

  See, I don't think it actually is that much of an own goal. I think like, This is within, I'm guessing, listeners of the Big Technology Podcast for us. We're all kind of like watching this roll out, but I think still the mythos, better than any other kind of like moment around, like this is the potential end of the world, really had kind of just mainstream people starting to build this kind of, I will say, no pun intended, mythos around Anthropic. Sorry, there's something, somewhere in there, the

• 26:11Speaker 01

  pun would exist. So I, so, so I think they did a good job of it and it worked. And like, this is just all kind of, again, like, think about how ridiculous that is. Like, if you ask for a bio, about a bioweapon, one, shouldn't there be content, like, I don't know, like, I'm guessing most of these have always had some kind of content moderation around if you ask it to help you make a bioweapon. I have not, I will tell listeners right now.

• 26:40Speaker 00

  But it's not, it's not just bioweapons.

• 26:43Speaker 01

  No, but now, but I love.

• 26:44Speaker 00

  Anything close.

• 26:46Speaker 01

  I know, but I love, hold on, it's going, it degrades you to a lower model. That's what I love. So are they saying like, we'll give you the like, you know, like a dollar store bioweapon. You don't get the good one, though, but we'll still give you the Sonic Force 6, or maybe a little haiku. Like, what are they trying to say there?

• 27:06Speaker 00

  I mean, they're trying to say that they drew a line where the model would be able to actually help people with these weapons and they're not going to let you use it. But it, to me, that's the sort of, okay, if that was the line, then I would say, fine. It went so much further. Here, I'm just gonna read again from the story. "Many complained the model was blocking them from discussing ostensibly benign topics like mathematics, biology, and chemistry, or even analyzing Fable's own publicly released system information. One user posted a screenshot of Fable refusing to answer a query on basic cellular autonomy tell me about mitochondria." So, so this goes back to my question, right? Like, If you put, so obviously, like some of these things, it's not going to want to block,

• 27:50Speaker 00

  like asking Fable about its system card. But if you didn't believe that this model is capable of the harm, why, why you would, like, it doesn't make any sense for to do this as a marketing exercise because you're pissing off your users and you're causing a group of people who usually would not have any common ground, the ones you read off before, to rally together against your company.

• 28:13Speaker 01

  Okay, okay, that's fair. May- maybe this is all, again, but they just need to get out an IPO. Like, after that, everything's fine and it all goes away and all, everyone's made their money. So like, no, I mean, but, but like, hold on. How would you explain, why would they do it this way? Like, what, what would be the, I'm going to flip that back on you and ask, like, why would you do it this way? If you really believed it, if you really believed it, like, you wouldn't release this, you would, this neutered way, you wouldn't really, I mean, honestly, actually, I think the, the true own goal here is now with Fable, and I mean, I'm seeing this very firsthand and hearing this, like, working in enterprise AI,

• 28:58Speaker 01

  like, everyone is talking about the data retention right now. Because enterprise agreements, Anthropic and OpenAI, as well as everyone else, like, claim no data retention. Now they're just explicitly saying that they're going to for 30 days and like, that's a huge deal. So why, why, why do you think they would even release it if it's going to cause them this much, this many problems?

• 29:26Speaker 00

  Well, okay, so the argument would be you put it in the hands of people because when they work through, and again, like, let's just talk this through. When you work in the areas you're allowed to use the stuff for, you can, you can actually do a lot of things. Like, there's a thread that they posted of people creating animations, building games. I guess they're doing a simulation of an airplane. Um, all these things created, programmed with Mythos or Fable. So basically, like, the, I think the strategy is they believe in the model. If I'm taking the non cynical view, right? They believe in the model and they want people to be able to use it for productive uses, just not for the ones that they find scary.

• 30:09Speaker 00

  And if they actually didn't believe those things were scary, they would have just released it, you know, wholesale. But I can give you my, my conspiracy theory take here. Um, if you'd like.

• 30:18Speaker 01

  All right. I like both.

• 30:20Speaker 00

  Here's my conspiracy theory take. You know, even though it might cost them a little bit in public perception, there is like some story you could tell about anthropic releasing this nerfed fable, getting everybody talking about how annoyed they are with their restrictions. And then adding more and more companies with a lot of money to spend into this glass wing coalition where they have access to mythos without the same boundaries. So maybe fable, maybe fable acts as a, uh, as a sort of, look what we've given the public. If you want the thing that works and you've seen some of these examples, you're going to have to pay up.

• 30:59Speaker 01

  I, I did see somewhere, uh, it was kind of like around, you know, two track AI or a K shaped AI. And now this is kind of like, kind of increase at both like the token costs. And we're going to get into that conversation just a little bit, but even the access. Okay, wait, wait, I like this. I like this that, so maybe let's, let's give them, it's not even conspiracy, recognizing the genius of their marketing and communications over the last six to 12 months, especially. Could they score known gold this badly? Or that's an interesting one. It's, we're going to give you this. If you want the good stuff, gotta pay up and only if we invite you.

• 31:41Speaker 00

  Yes. I mean, could you imagine being a salesperson and, you know, walking in, you got the deck up and it's like, well, normal users can't talk about mitochondria with mythos. For the right price, you can talk about anything you want.

• 31:57Speaker 01

  It's not even a price. It's only if I let you. Only if you're invited.

• 32:01Speaker 00

  And if I let you for a lot of money. Yes. Yeah. So I don't think that's what's happening, but you know, it's, we have to air it out there as a possibility. But, um,

• 32:12Speaker 01

  Wait, does mitochondria have something to do with, like, is that how you make bioweapons?

• 32:17Speaker 00

  No, it's a simple biological thing that was.

• 32:19Speaker 01

  I know, but manipulating mitochondria?

• 32:23Speaker 01

  Claude won't tell me, so I don't know.

• 32:25Speaker 00

  I have not tried.

• 32:26Speaker 01

  Okay, good. So both of us are

• 32:29Speaker 00

  are clear.

• 32:29Speaker 01

  bioweapon free. This is a bioweapon free zone. Neither of us have ever gone down that path.

• 32:36Speaker 00

  Yes, if you're going to come to the summit next week, Rondon and I will not release altered mitochondria built with Mythos during our session. That is the one thing we can guarantee you about the summit next week.

• 32:47Speaker 01

  Can't promise you much, but can promise you that.

• 32:50Speaker 00

  We can promise, that we can promise. And some coffee. Okay, let me, let me sort of ask you this other question, which is, you know, people really got on Anthropic for the idea that they wouldn't allow people to use this, this model, fable model, in the pursuit of building competing AI models. And on one hand, and I think both of these things are true. On one hand, it showed just how much power a frontier lab has. If it makes a decision about something, you know, you're sort of at a loss if you're reliant on its capabilities. And it's rare that you're going to see a tech company use its power that way.

• 33:31Speaker 00

  The thing you could say on Anthropic's behalf is,

• 33:38Speaker 00

  you could say that they will have their model distilled. If they, or they will have, they will have, they've had people distill their model, use the technology to create competitors and, and they're, they're getting wise to it. And you don't want to enable it. So I'm actually curious to hear your reaction to those two statements. And it would be great to just hear your, hear what you think and whether there's any credibility to it.

• 34:04Speaker 01

  No, no, I think on the latter, there definitely is. Well, hold on. Walk me through exactly before I, before I fully agree with you, the second point.

• 34:17Speaker 00

  Basically, the point is, you, if you're a company like Anthropic, you've seen other companies take your models and glean some of the IP off of it through, you know, distillation, things like that. So why would you make your model available to people building AI models, knowing that that's a possibility? I think that's probably more the same way for me to say.

• 34:37Speaker 01

  No, no, I think, I think that, okay, and I do think it is a tough one because on one hand, you could imagine it would factor in, but on the other, I don't know, I've been thinking about that a lot because like, how do you try to control for it? Do you add like, because even right now, that's like only a mild safeguard, I feel, but it's not really going to solve the problem. And it's funny because for as complex and like advanced and revolutionary as this technology is, it's amazing to me that it's like dealing with the problem that like the apparel industry and most others dealt with in the 2000s, like, like, yeah, someone can come steal your model and that's your high margin product and that's like where

• 35:22Speaker 01

  you will make most of your money. And so they have to, but I don't know, I can't imagine that is going to dramatically, that's going to be like a major reason behind this because it can't be that relevant. Like, I mean, at that point, a simple like throttle, throttling mechanism would be, I think, more, more useful than just trying to like catch you in a prompt if you're making some AI stuff.

• 35:52Speaker 00

  Yep. Well, they did, they did roll it back.

• 35:54Speaker 01

  Do you think they're being defensive here?

• 35:56Speaker 00

  Of course they're being defensive. Yeah. They rolled it back to some degree, but, but it is interesting. And as these things become more capable, if one company takes a lead and wants to protect that lead, we might see more of this, which is, I mean, we know Anthropic restricts the use of Claude code and its competitors. So we may end up seeing more of it. All right.

• 36:18Speaker 01

  Let's go to one, one quick, one quick question. Even for eight, when is the last time you felt something magical? What was like the model release that where you like typed some stuff in, had a couple of turns with your agent and were like, holy shit, this is incredible. Was it, when was it?

• 36:41Speaker 00

  I'm gonna answer in a way that's going to, I was talking with a listener this week and I was like, Ron Johnson's been stacking W's on me and I'm gonna get another L today.

• 36:53Speaker 00

  To be honest, the places where I've had the more holy crap moments have been using the medium powered models like Claude Sonnet in, in the harness, in something like Claude Co-Work.

• 37:12Speaker 01

  Oh, yes.

• 37:14Speaker 00

  And I have really just become Co-Work-pilled recently. I'm using it for everything. I'm having it like draft, draft emails. I can't even begin to tell you, like, I'm having it like register people for the summit. Like, when I get a list of people that are gonna be, that are coming to the summit, I'll just give the list to Co-Work and be like, "Log in and out of our event platform." Because we don't have like the backend tools to do it like batch and add each one of these people with this, this coupon code, you know, and, and it's doing it. So this is for one specific group, like press, right?

• 37:50Speaker 01

  Welcome to the Harness Hive. Welcome to the Harness Hive.

• 37:55Speaker 00

  I'm in the Harness Hive now. I'm in the Harness Hive now. I'm just like, wow, I wouldn't even want to use these high-powered models for it. It works so well with the medium-powered model. It doesn't need like the biggest brain in the world to accomplish this task.

• 38:10Speaker 01

  No, I felt that. I've honestly felt that with 48 even, like, I saw it, like, where you're just like, yo, calm down. It's okay. Let's just, we're trying to do something basic here. So, so yeah, welcome to the Harness Hive. I'll pass it back to you if you want to break. Now that I've gotten that, now that I've gotten that, I might just walk away from this microphone.

• 38:32Speaker 00

  Don't leave the podcast. Look, I'm not saying the model doesn't matter. Like, clearly you need good pot, you need good models in order to be able to create these experiences, especially if you create a lower-powered version of them. But, but yeah, I'm in the Harness Hive now. I think I fully have, I've joined. It's pretty amazing what this stuff can do. It just saves so much time when you start to like, think, instead of how do I use my computer for these things, how do I just turn it over to something like a Co-Work?

• 38:58Speaker 01

  It is funny, though, that still travel situations, I feel I've been, and we're going to get into a little bit of Siri later, I do love that, the thing that makes me the happiest about Agentic and where it's gotten to in the last six to eight months is now most people, other than even Apple Marketing still, have recognized that it extends far beyond just travel planning. Remember, everyone's like, what are we going to do with the Gentik? It's going to buy your plane tickets for you. It's going to book your hotels for you. But we're in a good place. We're in a good place.

• 39:31Speaker 00

  It's incredible. Handing it a spreadsheet and asking it to do stuff is pretty amazing. And you can see the things kind of will their way through stuff that they don't understand at first, or they can't accomplish at first and get better. And that's definitely been like a totally new revelation to me, even though I've been using things like Claude Code to build websites, like the Summit website is built with Claude Code. But I just think this cowork stuff is, I'm starting to use it more and more, and it really, really is super productive for me. All right, let's take a break. We're going to talk about a potential pricing war between OpenAI and Anthropic coming up. And then Ron John will

• 40:11Speaker 00

  tell us all that he apologizes to Tim Cook because he finally fixed Siri. Or he won't. We'll talk about it when we come back, right after this. This episode is brought to you by AppPoint. Everyone's racing to roll out AI right now. Co-pilots, chatbots, agents doing real work. But here's the part nobody loves talking about. All that AI runs on your data, and most teams have no single way to see it, secure it, and prove it's under control. That's exactly what AppPoint does. For 25 years, they've been the trusted layer beneath the world's most demanding data, now extended across your entire AI estate. Your data, your cloud, and the agents acting on your

• 40:52Speaker 00

  behalf. It's how more than 28,000 organizations deploy AI with confidence. So innovation scales without scaling risk. AppPoint, the unifying trust layer for AI. Head to apppoint.com to see how enterprises deploy AI with confidence. Learn more at avpt.co/bigtechnologypodcast. That's avpt.co/bigtechnologypodcast.

• 41:20Speaker 00

  Violent threats against executives are growing at an alarming rate. When a company experiences backlash, executives are the first people blamed. When their address and other personal details are sitting online, this backlash can lead right to their front door. The team at Ironwell knows this better than anyone. They've protected some of the most targeted executives and individuals on the planet for almost two decades. As a wide-glove enterprise security service, they protect your people with continuous personal data removal, proactive prevention tools, and human-led emergency support. So when someone goes looking for your executives, Ironwell ensures they hit a dead end. So here's what to do. Go to ironwall.com/bigtechnology, fill in the quick form, and request your free risk assessment. Their team will show you just how exposed your executives are

• 42:05Speaker 00

  and how to lock it down before a threat reaches their front door. That's ironwall.com/bigtechnology. Learn how to stop online threats before they become real-world attacks. The link's in the show notes.

• 42:19Speaker 00

  You know those pieces in your closet that you find yourself wearing on repeat? The sweater that goes with everything? The perfect pair of pants? Or the tee you somehow always reach for first? Those are the pieces that make getting dressed easy, and that's what Quince does best. Quince focuses on high-quality wardrobe staples made with premium materials like 100% Mongolian cashmere, organic cotton, and merino wool. The styles are timeless, versatile, and designed to become the dependable pieces you reach for day after day. I am kind of obsessed with my black zip-down sweatshirt from Quince. It's great for planes and malls. It's easy to tie around my waist whenever it's warm out, and then, when the breeze comes in, back on it goes. Find your next fall favorites at Quince.

• 43:04Speaker 00

  Download the Quince app for app-exclusive offers or go to quince.com/bigtech. Get free shipping on your order and 365-day returns. Now available in Canada and the UK, too. That's Q U I N C E dot com slash bigtech. And we're back here on Big Technology Podcast Friday Edition. Next week, before we get into this OpenAI story, folks, next week, we're gonna have a little bit of a different show. It will still be me and Ranjan, but it will be with audience participation from the Big Technology AI Summit. So we are gonna go through one story next week. And then we'll take questions live from the audience and and it should be a good listen. So stay tuned for that. All right, let's talk about this price war. So this is from the Wall Street Journal.

• 43:48Speaker 00

  OpenAI considers drastic price cuts anticipating war for users with Anthropic. OpenAI is considering drastically lowering its prices, the prices it charges users as it seeks to win customers from its rival Anthropic. The customer, the company is weighing significant cuts toward it charges for tokens, the unit of measurement, artificial intelligence firms used to build for their products. The move would be in anticipation of similar similar cuts the company expects at Anthropic. Business executives have begun to balk at high prices for AI usage. OpenAI chief executive Sam Altman said at a recent event that costs have become a huge issue. I think we'll have a lot of ways we can help people get more value for less spend. Is this the beginning of a price war between OpenAI and Anthropic?

• 44:32Speaker 00

  That just drives the price of frontier intelligence down to not zero, but you know, way lower than it is today. And if so, what does that mean for the AI industry?

• 44:42Speaker 01

  So to me, this is a really interesting story. And I can tell you like the amount, again, how fast this stuff is moving. Three to four months ago, every executive, no executive I spoke with was really concerned with token consumption and costs. And now suddenly that's all anyone wants to talk about. And I do think like this is a really dangerous game. OpenAI would be playing if they move in this direction. Because the way you figure it is like, they are going to lose costs. They're going to lose on costs at some point to a deep seek or others. So they have built themselves as premium products that should win at the kind of frontier model layer.

• 45:27Speaker 01

  And it's a luxury product essentially. So if, again, like any luxury industry knows, like if you cheapen that, that can actually fundamentally threaten your overall like standing brand, business, everything. And it's almost funny to me too, because I saw like a lot of kind of

• 45:45Speaker 01

  OpenAI boosters trying to argue that the actual like individual API call or the individual token that is like distributed or inferred is profitable, is a very high margin business if you strip out all model training costs. So like, you know, you send out that token, it's like a 70, 80% margin. Uh, transaction, if you remove the hundreds of billions of dollars you've spent, and that's why OpenAI loses $9 billion a year. So I think this is like if them going aggressively to Anthropic and trying to undercut them, I think could be short term beneficial for them.

• 46:25Speaker 01

  But I think long term just dramatically hurts them and Anthropic because right now they are premium products.

• 46:36Speaker 00

  Is this just a natural progression of where this business is going to go? Which is that the intelligence layer becomes commoditized because there's so many people building it?

• 46:44Speaker 01

  Well, I mean, and it becomes more about the model and the harness and the product? Yeah.

• 46:51Speaker 01

  No, no, I think, I mean, I do think, I think this is actually one of the bigger threats to both of these companies right now, even more so with OpenAI. Like, and we, hold on, we've talked about this for years now, that no one knows what the sustainable economics of these businesses are, especially like a very high investment frontier model driven research lab turned consumer or enterprise business. No one knows what the economics should look like. We do know they are not software margins, traditional software margins. Like the marginal cost of distributing a product is not effectively

• 47:33Speaker 01

  zero. So like, we know it's going to look different, but them chasing and driving this in the wrong direction, I think, could be really bad. Do they, do you think they're gonna do it, Juan? And do you think it would be good or bad for them if they do?

• 47:49Speaker 00

  It'll be fun to ask Greg Brockman about it next week at the summit. So I will do that. I think they will. I mean, I think that there's, if you think about it strategically, right? Anthropic has been winning, it's also, it's like, part of this is about lifetime value, right? Anthropic has been winning over a lot of these enterprise customers with its offering. And once you win these people over and you get them set on, on, you know, your system, they're not going to want to go anywhere else for a while because they're going to build that habit, right? And because if it's working for them, then why would they go somewhere else? And and so I think that OpenAI probably sees it as, well, if we drop our prices, we're going to lean on our strength, right? And what's their strength?

• 48:32Speaker 00

  They have all this infrastructure. They've invested in more infrastructure than Anthropic. Dario called it YOLO-ing, right? What they're doing on the infrastructure side. If you've built the infrastructure, why not lean on that as a strength and try to win these companies over? And get that lifetime value that will be difficult for Anthropic to get over time.

• 48:52Speaker 01

  But that assumes switching costs are low. Like what you just said there in terms of like, once they're in your system. But remember, in most of these cases, it's just an API call. Like, again, or it's we've seen how dramatically Claude code users could switch over to Codex. And when you're in the command line, it all looks kind of the same anyways. So like, like, it's, they're not, it's not like the average end user has ChatGPT open. Now, maybe like memory and context become part of the mode or the switching cost. But like, overall, from the pure model delivery perspective, I don't think that is is really the case.

• 49:32Speaker 00

  Interesting. Wait, let me ask you something, because we're going to talk about it next week, too. Are you a believer in this, like, in the argument that this becoming more and more prominent, more and more prevalent, that model orchestration is where like the real

• 49:46Speaker 00

  value will be found as opposed to just the raw intelligence of the models you produce?

• 49:51Speaker 01

  I can say that that is at the core of what we sell at Writer, and that's what I've been working on the last 12 months. And yeah, 100%. I mean, I got to say, being like orchestrate, being the orchestration layer, being across and actually like being interoperable across different models, even though we have our own foundation model. This is what we've been building for a long time now. And I'm telling you, I've seen the switch in the last few months where people really, executives really did not think about token cost. And now it's all anyone is talking about. So, yeah, I'm an orchestration guy through and through, beyond hardest, beyond the product.

• 50:30Speaker 00

  But isn't that very bearish for the big model companies? I mean, they can build the most intelligent system they want. But ultimately, if it goes orchestration, then what are they going to do?

• 50:42Speaker 01

  It is.

• 50:43Speaker 00

  Because the use will be much less.

• 50:45Speaker 01

  If you already saw like, I mean, like model routing, even finding, and the thing is, they're trying to push it even themselves, but within their own family. Again, when you are searching about how to alter your own mitochondria, to make yourself a bioweapon, Fable 5 is downgrading you and doing some model routing to, I think, Haiku or something like that. So, so everyone is recognizing it's where things are going. That's just the model routing side now orchestrating across different systems. And this is where there's more value. And cowork is kind of like that entry point into feeling the power of,

• 51:26Speaker 01

  what if I could connect some documents and files and some model? What could I do? And like some logins? I think that becomes the entry point for an individual. But yeah, I'm an orchestration guy through and through.

• 51:40Speaker 00

  Okay. All right. Interesting stuff. We'll obviously be speaking about this over the next month on the show because a lot of these conversations that we're going to have next week, which we'll, we'll try to air as many as we can on the podcast, we'll, we'll feature this, including a conversation with Arvind Srinivas, the CEO of Flexity. Okay. Before we go, Apple, WWDC, it happened this week. You've been begging the company for years to fix Siri. Do you think Siri's fixed?

• 52:05Speaker 01

  Oh, man. I know I've been a bit of a hater on today's show, but I am going to say it. I have not used it yet. I have been going back and forth because apparently if you upgrade to the developer beta on your Mac, you can already install Siri via command line. But I have not done it yet. I think I might be, but I don't know. Like I saw Joanna Stern has not been the biggest fan of Siri in the past. And she had some pretty good videos out about her using it. I've seen a lot of people. And then even MG Siegler and his argument, like,

• 52:44Speaker 01

  it just all started kind of hitting. I'm like, oh, my God, if they actually do this. Everyday AI and like, what does that mean for OpenAI? What is that? I mean, it's just, it changes the look. By December, we are having such a different conversation about the landscape. And my God, if Apple pulls this off, but it feels like they have a shot now more than ever. Do you think they're gonna do it? Does it feel real?

• 53:14Speaker 00

  I think they did it. I think they did it. I mean, I was, so again, I was banned from WWDC. You're not allowed into WWDC this year. Fair enough. But I was still on the day of or the day after I was on CNBC and said, look, you know, the stock's down 4%. I think investors are wrong here. I think they're being too negative. Because consumer AI is wide open. And what Apple's done is it's used its operating system to first provide a lot of context for people on their phones, right? Like, your phone is a screen where you're like tackling so many different things. And, and if you could just say, hey Siri, what does this mean? Or how do I get there? Or can you copy this and paste it into my notes?

• 53:55Speaker 00

  These are simple actions that you apply a generative AI layer and all of a sudden the phone becomes more, more powerful than it was previously. I haven't used it yet, but I've seen enough videos to put my early, early analysis in. I think it's gonna work. I think that the bar was also so low that they have to just show a little incremental progress. And everyone's gonna be like, wow, they did it. And I think that's absolutely what's gonna happen.

• 54:21Speaker 01

  I would like, I would like John Turnus to come on this show to thank us. Because thanks to just ripping on Siri for a few years now, we have set the bar so incredibly low that just answering a basic like GPT-4O voice mode query will make everyone just marvel. Because honestly, I saw someone was like asking, they're like, I have a 7:00 AM flight. What time should I get to the airport? And it was like, based on your location and a typical wait for an Uber at this time. And I'm like, I mean, that is such basic Gemini GPT stuff. But everyone was like commenting on Twitter like, this is incredible. So, yeah.

• 55:05Speaker 01

  We have set a very low bar and that's the best possible place you can start. So you're welcome.

• 55:10Speaker 00

  But also, let's just say it's it's not the most advanced stuff. It's not the most mind blowing stuff, but because of where it sits, it doesn't need to be, right? As long as it sits in the operating system and it really is like Biggest Loser, if this thing works is meta, right? Because their personal super intelligence becomes that much harder for people to access if they have a working intelligence on their phone, even if it's not super, just intelligence.

• 55:35Speaker 01

  I'm not, I'm already having like visions of all this stuff they could do if it works, but I'm not gonna let you hurt me, Apple. I'm not gonna let you get me that excited. If it works, my God, it just opens up just like an entire universe of interesting, fun opportunity, but I won't let myself get hurt again, Apple. I'm gonna wait. I'm gonna wait till sometime this year because you haven't even told me when, but sometime this year, sometime later this year, I believe was the exact phrasing.

• 56:07Speaker 00

  They're gonna do it eventually, right? They're a technology company. They're gonna eventually, they've basically taken it off the shelf from Google to a degree and use that to help train their own models, but they're gonna figure it out. Stakes are too high and when they do, it's gonna be cool. So I think this will be the first step toward it. And personally, I, my expectations were low coming into the week and they were, they were exceeded. So I'm excited to use this new Siri on my phone.

• 56:37Speaker 01

  You haven't used it yet. We, neither of us have used it. Do you think it'll let you alter your mitochondria?

• 56:46Speaker 00

  I sure as hell hope so. I have, yeah, three days, five days to bring, to bring that out.

• 56:51Speaker 01

  Imagine if Siri ended up being the AI platform that helped you with that project.

• 56:59Speaker 01

  Fable five, what it's the true danger that it's so good. It is a cybersecurity risk that can threaten humanity and the entire global economic infrastructure.

• 57:11Speaker 00

  I have to say, if we end up like getting wiped out by a superbug that's been generated on Siri, I'm going to be really mad. It'll be ironic.

• 57:19Speaker 01

  Actually, for the irony, I'll be happy, so.

• 57:22Speaker 00

  They did it as we melt away into the abyss.

• 57:26Speaker 01

  You were right. I'm sorry.

• 57:28Speaker 00

  Sorry, Tim. Tim's like, uh, only I had.

• 57:31Speaker 01

  No, no, turnus, king turnus.

• 57:33Speaker 00

  Oh, turnus.

• 57:34Speaker 01

  He's gonna, he's gonna melt us all.

• 57:36Speaker 00

  Right. All right, well, in which case I will apologize to Apple for all the things I've said about their AI. All right, Ranjan, we'll see you next week in San Francisco. Looking forward to it.

• 57:46Speaker 01

  I will see you on stage in San Francisco.

• 57:49Speaker 00

  And we will see you all the day later here in the feed. So thank you again, Ranjan. Great to talk to you as always.

• 57:56Speaker 01

  See you next week.

• 57:57Speaker 00

  See you next week and thanks to all of you, listeners and viewers. Pleasure to do this show for you guys every week and we'll see you next time on Big Technology Podcast.

• 58:07Speaker 02

  Now, more than ever, we need someone on our side. But Rob Whitman? Whitman's votes drove up costs, cut Medicaid and risked 350,000 Virginians healthcare to pay for tax breaks for himself and his donors. He's not on our side. Shannon Taylor will fight corruption, cut taxes and bring down costs. This November, vote for Shannon Taylor. She's for us. Paid for by HMP, the House Majority PAC dot com, not authorized by any candidate or candidate's committee.

• 58:39Speaker 03

  Every mile matters, whether you're driving your dream car or your daily commuter. The Mobile One team knows that. It's why we created Mobile One Advanced Clean Motor Oil. It cleans your engine in one oil change and helps extend engine life, delivering performance, protection and cleanliness for vehicles of all ages. All to help keep your engine running like new, whatever you're driving. Visit mobileone.com/advancedclean to learn more.