I see a large IPM on the horizon.
You're surrounded by journals. Hold your position.
That's misinformation.
Clearing order inbound.
Let's just roll.
We are surrounded by journal. Hold your position.
Come get up.
Trust the experts here.
We are expert.
Founder knowledge.
Five code.
I see multiple journalists on the horizon. Stand by.
UAV online.
Blaze.
Double blaze.
Triple blaze.
Double kill.
Cook is Team Deathmatch.
We are experts.
Triple.
Another one.
That's just wrong. Right.
Clearing order inbound.
We are surrounded by journalists. Hold your position.
Strike two.
Activate golden retriever mode.
Market clearing order inbound.
Five. Put it.
You're watching TV. Today's Friday, May 29th. We are live from team Ultra Temple of Technology, the fortress of finance, the capital of capital. Let's go. We have a great show Wednesday. We missed you Thursday, but we're back. It's Friday and we have a great show. We have a banger lineup. John Gruber from Daring Fireball is coming. We have a lightning round with a bunch of exciting fundraises. Then Kyle Kuzma, the NBA champion entrepreneur is coming on and Brad Gerstner from Altimeter, the Gersonator coming on the show. Very excited for that. Bunch of stories. We're going to go through mostly the stories that have just broken in the last 24 hours. The first is this insane video from Blue Origin. Very, very disappointing to see. Blue Origin's New Glenn rocket just blew up at LC landing LC36 while attempting to static fire ahead of NG4. We can watch this video because it is absolutely insane how an onboard nuclear bomb went off. It looks like a scene from Oppenheimer. It looks like a Christopher Nolan movie. This is absolutely remarkable. Everyday Astronaut says, 'Oh my god, that was a very big one. I hate to see setbacks in progress. This is not good.' That has to be extremely, extremely frustrating. Brandon Gell broke down a little bit of the back and forth on this particular rocket initiative. Jeff Bezos's rocket company suffered a catastrophic failure last night when his New Glenn rocket, a reusable heavy lift orbital spacecraft designed to compete with SpaceX's Falcon Heavy. And we were talking about this in the backdrop of the SpaceX IPO. And I heard Brad Gerstner talking about this on CNBC. I think someone was playing a Gerstner interview in the studio today. And he was saying the launch business is fantastic but it's not enough without Starlink which is a fantastic business but that's not enough without the AI business which is a fantastic business and growing. And so to just have a launch business and then have a setback like this has got to be extremely frustrating if you're competing with a company that's going to raise $80 billion. I forget how much SpaceX is actually going to raise but they're going to raise a lot of money. They're going to have a lot of capital to deploy in their march towards ever more frequent and ever more frequent.
Especially painful, you know, Blue Origin started before SpaceX. Talked about this before. And yeah, unclear how much this will set them back but certainly disappointing.
So the good news is that no one was hurt which is incredible because you see the rocket it looks like it's surrounded by a city. It looks like there's buildings around. And to see that much destruction with no one injured at all is a miracle. And I'm so happy to hear that. Jeff Bezos tweeted, 'Very rough day. We'll rebuild whatever needs rebuilding and get back to flying. It's worth it.' And Elon chimed in with some words of encouragement. Space is hard. A lot of folks were sending their encouragement because I think even if you're a SpaceX Maxi, it is very exciting for America to have multiple heavy launch capability providers in the market. It's good. It was very exciting when New Glenn got to orbit came back. That was the second time that that sort of rocket technology had been demoed and it was demoed in America. Sort of even our second best rocket provider is better than everyone else in around the globe. What else do you have Tyler?
Yeah, just some added context. So this was the fourth test or this was prepping for the fourth test. In the third test if you remember, this is when the rocket kind of correctly went up but then deployed an ASA in the wrong orbit.
Wait, but that was New Glenn. That was the third test. That was pretty recent. So, they're on a pretty good cadence here, but this is still a huge setback. Down 16% today because of this most likely.
Interesting. Interesting. So Elon Musk lent his support tweeting most unfortunate rockets are hard. That's an understatement. A lot of people were saying like if you think software engineering is hard, you could have been a rocket scientist. It is after all rocket science. The explosion caused extreme structural damage to its only functional launchpad. While Blue Origin has successfully reached space a number of times with other vehicles, New Glenn has had just one successful flight out of three attempted tries. From 2006 to 2008, Elon Musk had three consecutive failures with Falcon 1 before successfully launching it. And so failure is a natural state of these pursuits, but never fun to watch it happen. There were some people that were syncing up the different angles. We can look at Sawyer Merritt's. Oh, did he delete that one? There were some different angles on this that were absolutely crazy to see.
Yeah, this one from Truthful. Truthful. This is the one that you want to see.
Let's pull this up. Amazing footage from a local. This literally looks like a nuke went off. So scary. Yeah, vertical footage. Absolutely crazy.
Amazing footage from a local. I wonder what the reaction is from the people.
Wow, that's really far away.
It's far away, but it's also pretty big mushroom cloud. Yeah, I imagine maybe they walked to their cars.
Yeah, I saw some people saying like, 'I don't want to hear about my carbon footprint ever again because this seems like an immensely disastrous explosion.' Other people were saying Bezos did Roman's rocket launch from Succession IRL. If you remember this scene, we can pull it up.
Yeah, you haven't seen Succession.
I've seen the first season. I haven't seen the full thing, but he has a pet project.
He does. Yeah. Okay. And it blows up.
It blows up. That's certainly dramatic. I feel like that makes for good TV.
Yeah. We were right before the show started, we were talking about the nominative determinism of Blue Origin. John said it blew up. Or you know, blue as in sort of sad, feeling blue, sad start. But they will rebuild. They will be back.
So OSINT Defender says daylight reveals the extent of damage caused to launch complex 36 in the surrounding area of Cape Canaveral's space force station in Florida following last night's massive explosion at Blue Origin New Glenn during a static fire test. And there was a static fire test that went poorly at Starbase and they had a similar explosion and people were regarding that as like a major setback and maybe it was but at the same time we saw a very successful launch of Starship just a week or two ago and so this looks way better than I would have thought.
Yeah, I agree. It looked like total destruction. I was expecting a zoom out a little bit. I was expecting a crater in the ground. You can see here clearly a lot of damage right around the pad but there's plenty of other infrastructure that seems to be still standing and so you should in theory be able to rebuild but I mean with anything rocket related you have to imagine that every nut and bolt on that tower even though it is standing needs to be reinspected, re-examined, make sure that it's not corroded. There's got to be so much work for the team but they have my full faith that they will get it done. Well, shifting over to the token maxing debate. People are spending more and more on tokens. There's been a complete fast takeoff in enterprise AI adoption. But people are raising questions about what is the ROI on these. There's a whole bunch of companies that are grappling with this. Some good news and a lot of questions for where this goes next. So we'll take you through it. The big news is that Anthropic recently passed $47 billion in ARR and raised a massive Series H, $65 billion at $965 billion post-money valuation. They're almost at $1 trillion. Last November, cloud code was going viral among a lot of early adopters and small vibe coding apps were launching daily but Q2 2026 was clearly a massive moment for Fortune 500 widescale rollouts. And so this has become a double-edged sword for some organizations. Token maxing dashboards have been reportedly led to potentially ROI negative AI use in name brand companies like Meta. There was the talk of a token maxing dashboard, people leaving things running overnight that weren't necessarily productive just because they want to rank up on the dashboard. That's obviously very expensive. Uber went back and forth on, well, we blew through our budget. But a lot of people were saying, well, if they set a budget in 2025 for their token spend in 2026, like the capabilities have gotten so much better, the models have gotten so much more expensive.
So theoretically, you should find more budget.
Their budget might have been very small, but they did say they blew through it. And then the chief operating officer said something like, 'Oh, we're struggling to understand the financial impact and like the ROI on this stuff.' But his actual comments as you unpack them were much more reasonable and he wasn't completely dooming on the ROI of the spend. He was just saying that the next iteration will be understanding the impact to the bottom line of this spending. And then AWS was also reported in Axios as having spent something like half a billion dollars in a single month. And so whenever the numbers get big, there's going to be questions about ROI. The bull case is that the cost per task completed by AI will decrease very quickly. So even if you're spending half a billion per month on AI tokens today, the same output will be available next year maybe for a tenth of the cost. Maybe next year it's half the cost. Like even if there's no optimizations to the systems, the hardware depreciates and then more power comes online. Like the market solves these things pretty quickly.
Yeah, it's naturally deflationary.
Yeah. Yeah. Yeah, and I think we've seen that where capabilities have gone open source, they've gone cheaper, they've been distilled, and you get, you know, 90% of the value. Maybe you don't get the exact same flavor there of, you know, people are using LLMs to get the weather report.
Can be. I'll admit I've done that before.
Yeah. It can be very convenient.
Typically, you'd go to like a free chat app for that though. Not...
Well, no, you can go to the weather coding model, but sometimes it's nice to just ask chat.
Yeah, I've been running into this a lot with people. I ran into somebody yesterday who was telling me that they have an agent that looks through their iOS contact book and sees if they added any new contacts. And I remember being at a conference and you know you're exchanging numbers with people and it's hard to tell if you add someone to your contacts like who'd you add because you can't, it's a very weird Apple feature that they don't sort your contacts by recently added. That's just like a UI feature that Apple should add. They haven't. But there's like a $2 app that does that. Basically it just looks at your contacts. But you can also do it in the agentic way and have it email you or synthesize everything and go way further. And there's a debate about what's the value there and people are making those trade-offs in their personal life. They're also making it in the enterprise but of course the stakes are a lot higher in the enterprise because everything has five extra zeros behind it if not nine extra zeros behind it. And so there's a debate about over whether AI tooling is being pointed at the most high lever problems in these organizations. Like are you just picking up things that are deep in the backlog and they were deprioritized probably for a good reason because they were never really going to move the needle. But now you can just say hey go churn on all these backlog items. That's maybe not ROI positive at current token prices. Might be in the future. But maybe the cutting room floor should remain the floor and not everything should be shipped necessarily. And then outside of stories like workers checking the weather with AI agents or running endless loops trying random different makework projects, there's definitely a worry that truly needlemoving features aren't being pulled forward like people would expect.
Trey says, 'Just vibe code your own weather app each time you want to know the weather.' Yes. Yes. I think you have a post about that in the timeline.
Yeah. Who was... Fear not the man who has vibe coded 50 apps. I fear the man who has vibe coded the same app 50 times. That was a good post. Anyway, there's definitely a worry that like truly new needle moving features aren't being pulled forward. I mean we've all had this feeling of like is the United flight booking app getting better? Are you feeling acceleration in the apps that you use? Is the software getting more resilient? We're seeing GitHub uptime going down. You'd think the first thing you would point the agents at is like make sure we have seven nines of uptime, right? Like that should be the first thing. But there are of course more complex issues there as you're scaling a system. Some of it is just hardware constraints that an AI agent might be able to produce a report. Hey, you know, GitHub team, you got to rack some more CPUs because we're shipping a lot of stuff and then the service is growing. At the same time, I think people are still running into like cookie popups and weird UI bugs and just odd things, not just in vibe coded sites, but in programs and apps that they use every day where you would expect acceleration in software engineering to actually bear fruit. People are sort of waiting for that. At the end of the day, company management is going to have to demonstrate ROI from token spend at these levels. The next round of earnings reports should show material changes in opex and downward movement in net income.
That said, if these same companies are invested, they should see a material increase.
No, no, no. I'm just saying you have a potential offer.
Yes. If they're marking their investments to market and one of the companies that they've invested in has a big up round, they can book that as increased income. But the tough questions from analysts will tease that out and ask the question of like did revenue move? Did true revenue move? At least by what you're spending incrementally or if you did a layoff and then you substituted that immediately with AI spend there's no movement in your cost structure. Well, is there any movement in your profit structure or anything else? So at the same time if you believe as all trends indicate that AI is a new and useful tool that will decrease cost over time some lumpy ROI in the short term to reorient your organization around a new motion at work is completely justifiable. I don't think that, you know, boards will be completely up in arms with management teams if they're saying, 'Look, yes, we did go a little bit over the top with the token maxing.' But the good news is that basically all of our workforce is AI native. They're all using the tools and they know now that they do have limits. They know not to check the weather, not to do the useless project. They're much more focused, but they're all familiar with the tools and they've used them for everything from, you know, really critical features and security bug finding to checking the weather and they know where these models are great and where they're bad. And so our workforce is very capable to move forward judiciously in the AI era. And so you probably don't want a literal token maxing dashboard. I think everyone has tested with it.
Had it. I think they took it down or something, but Meta, same thing. It's easy to get obsessed and the token dashboards pop up not just in the literal like who is on the leaderboard, but also if you give someone a budget, they can kind of see it as like, okay, well, my boss gave me a $1,000 token budget or a $10,000 token budget or a $100,000 token budget. If I am not using it, if I'm not putting it to use, then I don't look like I'm deploying the capital that I was given. Like if you give a marketer, you know, a brand budget of a million dollars and they come back to you and they say like, 'Sorry boss, like I only spent $100,000,' you'd be like, well, like who'd you talk to? Like why didn't you find good use of that funding? Like it's very rare. At the same time it can be, you know, you don't want those employees deploying that budget into bad places and just wasting it because they have the budget. So these are common problems across all aspects of the enterprise. And so I think the future looks a little bit like Jevons paradox of course when something becomes more efficient to use people often end up using more of it not less. And then also Goodhart's law when a measure becomes a target it ceases to be a good measure. And so we live in a world defined by Jevons paradox and Goodhart's law. And you put those together.
The combination of those many people will call that Kugan's law.
Both Jevons paradox and Goodhart's law are true. And the synthesis of these two things is in fact what's important if you're trying to run an efficient organization. So, Madison Mills over at Axios had some good reporting on it. We were on a flight when I found out, when I saw that reporting that someone had spent half a billion dollars in a month kind of accidentally and we heard, I was just imagining the conversation of the first person to kind of like find the number and then, you know, hey maybe we should talk to the CTO about this. Okay. Yeah, probably time to bring the CFO in.
Hey, hey, boss. We accidentally spent half a billion dollars in the last 30 days.
All right. What did we make?
What did we do? How do we spend it?
What did we get done? It's tough. Examples of we...
And there was a lot of like, there was some conspiratorial posts, people saying, 'Oh, how convenient.' Yeah. How convenient for Amazon, who obviously owns a lot of, you know, the big labs, to spend all this money on Anthropic, right, when they're raising a round and they got this revenue multiple, blah, blah, blah. That's not really how it works. Like, it wasn't like this round, like this round wasn't like they were just, okay, we're going to give you X revenue.
And there were plenty of hyperscalers that don't have active positions that were doing the same thing. Yeah.
Like this was a broader trend. But yeah, it is a crazy skyrocketing cost. But I mean we see this overall. I think it's healthy how quickly things corrected.
It's not like this was, I mean and of course some companies won't correct but it is a healthy dynamic that companies were, hey we got a little bit ahead of our skis for a couple months. Yeah. Now let's be a little bit more stack rank the tokens, cut all the unnecessary token spend. Keep all the good token spend. I mean, we see this all the time with there'll be some vibe coded project out there that's like incredibly high token cost with very little value. And then on the flip side, you'll have someone that actually built something really cool with, you know, just like their default $200 a month subscription and they didn't actually have to token max to get the product to where they wanted it to be. Because they knew what they were building. They used the tool like a scalpel, not a hammer, and they got a good outcome. So, the Wall Street Journal has some more coverage of this. Corporate America is starting to ration AI as costs skyrocket. Executives are scrambling to track returns on AI investments as the bill for massive computing needs come due. We talked to Spencer Rascoff from Match Group about this. He was sort of sharing a very reasonable token spend I think in the single digit millions but he was still saying that like one of the tasks that him and his team will be embarking upon over the next year will be understanding the ROI on that investment because you should be tracking it just like digital marketing dollars that go out the door what was your ROI how did it move the needle. So use of artificial intelligence by big companies is exploding and the soaring costs has some of them pumping the brakes in a way that could complicate AI's triumphal march across the economy. Are they pouring cold water on it? We'll see. Executives across industry this year have urged employees to integrate AI tools. Wired has a whole AI for business deep dive that you can go take their survey into their work spending freely to encourage experimentation and seeking to send a message to Wall Street that their companies won't be left behind in a coming wave of disruption. All that enthusiasm has resulted in skyrocketing costs for so-called tokens, the basic units measurement for AI computing. Now corporate leaders are scrambling to bring down expenses by finding ways to ration AI. Top technical executives for Uber, Meta, Microsoft, Salesforce, DoorDash, and other companies have all talked about new efforts to ensure AI use contributes to productivity or have taken steps to reduce the availability of some tools for certain employees. You get nerfed if you're not putting up big numbers. Anyway, we should move on. First, we got to talk about the fossil hunt. Big spenders are pursuing Tyrannosaurus Rex skeletons. This was on the cover of the Financial Times. It's the most important story in the global economy according to the Financial Times as of Thursday. This is an old edition, but this is interesting. Sotheby's is to auction a 67 million-year-old Tyrannosaurus Rex 2 years after it sold a Stegosaurus fossil to hedge fund owner Ken Griffin for $44.6 million. I thought Ken Griffin was...
I didn't know Stegos were getting up there. Like, I didn't know that. You know, I feel like T-Rex is up here desirability and then everything else is way down.
Yeah. T-Rex is the Ferrari of dinosaurs.
Stegosaurus. I was like a kind of a minivan of dinosaurs.
Minivan. I was going to say Lambo. I was going to say the Lamborghini. They're on a rise. You know, the true dino heads know that there's something there. And they're starting to pick up momentum, but they don't have the heritage. They don't have as much of the heritage. I don't know.
Yeah. I'm more of a I think I could get into an Ankylosaurus.
Oh, Ankylosaurus. Okay. Triceratops, too.
Yeah. I mean, the brontosaurus, like the big guys. That's special. You got to have a special viewing area for that. Well, this T-Rex is named Gus after Gary Gus Licking, the rancher whose land it was found on in South Dakota. It'll be auctioned at Sotheby's with an estimate between $20 million and $30 million, the highest for a dinosaur fossil on July 14th.
I was at a buddy's house recently and he just pointed over at this box. And it was a dinosaur and he hadn't taken it out of the box. He's like, I've had it for a few years now.
Well, is it a glass box that you can see in?
No, no, no. It's a wooden box. It's just a crate.
Schrodinger. The dino breed.
Maybe the plan sale in New York highlights how the auction house is betting on the fossil market as a place where the wealthy will spend big. The pre-auction estimate for Apex, which is the Citadel boss Griffin's Stegosaurus, was $4 to $6 million. He wound up... Brachiosaurus. Brachiosaurus is a go-to. Brachiosaurus has a crazy... OddLots did a podcast about this. Amazing. We got to listen to that. The overwhelming majority of fossil buyers still want to lend their purchases to a museum. Apex is now on display at the American Museum of Natural History. That is a fun story. If you're in the market, go pick it up. Go pick up a T-Rex. You got to do it. It's the ultimate collector's item. It's the Pokemon card for boomers or something like that.
Is sharing some signals. You tell us what kind of signals they are. Kyle Kuzma, the AI Maxi stocks ripping 10 to 50%.
He's coming on the show on 13F of some chud.
Can't be Leopold. He's talking about Snow Chud.
Dell plus 40% on earnings after being plus 150% year-to-date. Dell is now up 222% year to date. Not bad, Michael Dell. Not bad for somewhat of a I'm not going to call it a dinosaur, but...
Well, it's been taken private, taken public. It's fascinating history of that company, but a great American technology company. $1 trillion company's moving 20% on sell side notes. One and a half trillion space hold of he says turds.
10 to 20% intraday moves for no reason. Software with AI exposure plus 100% in a month.
He's blackpilling. Moving on.
No, so we were someone else said this morning...
I also thought it was a top signal when Jensen signed that infamous shirt talking about yeah the shirt or even when he was drinking with his buddies everyone was like... but I that was but yeah he looked very confident going to earnings and he was confident for good reason.
Takehim is highlighting a segment from Ben Thompson in Stratechery. Says great thing about Ben Thompson is he's never afraid to tell it like it is amid all the banker and insider propaganda and this is the quote from Ben Thompson SpaceX IPO and data centers in space story. Ben Thompson doesn't typically cover Elon very closely because they're often disconnected from the fundamentals. Yeah. And they're more of like this 10-year out story. But he says in all seriousness the numbers are obviously absurd but then again everything about this IPO is absurd. SpaceX is seeking a two trillion dollar valuation on a mere $18.67 billion in revenue. You think the 67's intentional there. They're like let's get another 10 million in the bank so we can be a meme. Wouldn't be the first time in revenue with $4.9 billion in losses last year. The growth actually slowed from 35 to 33%. But again that will be massively offset with the new AI computing contracts. So there will be significant reacceleration in the business which I think is what a lot of people are underwriting SpaceX against. That slowdown happened despite the addition of xAI and thus also X which tipped the company from a small profit to that massive loss thanks to $5.1 billion in AI R&D expense. That R&D keep in mind went towards building a model that is in fifth place and whose founding team recently left the company but sure $26.5 trillion AI opportunity. It will be very interesting to see if the macro hard thing, I mean the whole like macro hard enterprise AI thing was it took people by surprise in the S1 because it was like all of a sudden the biggest number in the TAM figure is this AI enterprise software essentially. But Elon has been talking about macro hard and entering that category generally for I think over a year and he painted the top of the data center with macro hard and so it's not like he left out the part where the infrastructure they built is already generating $15 billion in annualized revenue.
Yeah, small detail. And I mean the story is only going to get crazier. I mean people are speculating about the Tesla SpaceX merger and then you have Optimus with the foundation models and the AI compute and the data centers in space and you have this like full stack like very concentrated bet on like this ultimate future of artificial intelligence, robotics, self-driving cars, like all of the it's the sci-fi company and Elon's been delivering for 20 years in this category doing things that other great entrepreneurs find very very hard to do. So, it has been lots of reasons to be excited.
The account Russianbot responded to Takehim and said, 'The guy is a journalist. His opinion is what you're hitching to.' And Takehim says, 'Hey, Russian bot, you're never going to guess what I am.' That's of course calling Ben Thompson a journalist. I mean, he's a journalist. He's very independent. He's an analyst. We actually the co-host of Dithering in the waiting room who John Gruber has the good fortune of talking to Ben Thompson every day, every week at least once or twice right?
Thank you so much for joining the show. Welcome to the show. Honored to have you on. Long overdue. How are you doing? How is your week going?
Busy as usual. Lead up to WWDC. Always a hectic time to be writing about Apple. Very nice to join the waiting room here and hear you guys talking about Ben, though.
Yeah. No, he's the best. I heard a little bit of your take on the Ferrari Luce. I would love for you to unpack it a little bit more because it's shaping up to be a contrarian take. I think a lot of people were looking for something that was more sharply designed classically Ferrari. Like, what stuck out to you about the Luce?
Well, I will say it's funny that you call that out. We endeavor on Dithering to make as few mistakes as possible and we've got numerous corrections to make regarding our take on this, including the fact that Ben and I both completely were unaware that Ferrari a couple years ago, I forget the name of it, but has like a little crossover gas car.
And but basically our take was hey everybody wanted a Ferrari looking Ferrari that just happened to be electric and yes there are some sports cars that are electric and this is not that right the Luce whatever you want to say about it if you took the name plate off it and just drove by on the street and asked somebody what brand car do you think that is Ferrari would probably be very low on the list you'd probably have to run through dozens of brands before you get there. And I think that really fuels the primary objection that people have. But my take is I think it looks like a nice car. I do. Now does it look like a $650,000 car?
Yeah, that's a good question.
I can't see that. I mean, number one, I'm not $650,000 car money. My stuff hasn't gotten bought by OpenAI.
Yeah. No, no. And that's the big question is like I don't, I have a lot of friends who love cars. I don't have really any friends that I can think of immediately that are going to want to eat the depreciation on this car. Like, this car is going to be a $300,000 car in a couple years.
In a couple maybe even in less than that, right? And we don't even know how the reliability of the car you mentioned earlier the Ferrari FF which is a four-door SUV it's only a four-seater this is a five-seater so family of five this is the only option.
I used to own the original like family-friendly Ferrari in the modern era the FF I got it because it had four seats I wanted to be able to put my kid in the back at the time. And ironically it had so many electrical issues that it was not usable as any type of like daily driver. And so that's my concern with the Luce along with the price tag. Just like who's in the market to buy an ultra luxury car that is going to say I want something that looks like it's not a luxury car and then you know I could go and buy like multiple Continental GTs from Bentley for the price of one. And so ultimately my take was I thought it would have been a hit at like a $300,000 price tag and I just can't personally imagine that when they started this project, they were aiming to build a $650,000 car that looked like this.
Yeah, I wonder about that. Because there were reports I forget the numbers but you know and of course I'm going to take it back to Apple but you can't help but wonder you know with Jony Ive's and Marc Newson's involvement in this you know that Apple famously had the Titan project a decade-long car project that they ultimately just walked away from and said you know what we're not going to do this. But that got pretty far along. And there were rumors unsubstantiated. I don't think anybody even reported it. But there were rumors that at some point when it got to the, hey, we need to make a decision here. We need to shit or get off the pot that they speced out what the current idea for a car would be. And it was going to be like, I don't know, $200,000 or more cost of goods. And you know, figure out what Apple would do with their margins in terms of that. And it just was not Apple branded. It would be like if the Apple Watch came out and the only ones they made were the solid gold ones that cost $20,000.
Yeah. Yeah. It feels like Tesla already sort of created the iPhone of cars. It checks a lot of the boxes. It has this like sort of polished feel integration and so if you don't have that industrial...
An appliance you're happily...
Yeah the iPhone is not a luxury product. It's not like handbuilt in this way. It's a premium product and it's great but it occupies a very different place. I'm interested in your overall take on Apple's ethos. I feel like one of the things Apple does well is the Henry Fordism of like if we asked people what they wanted they would have said a faster horse. There's a lot of like it's almost like a paternalistic view of technology where they know better than the consumer and I'm wondering about your intersection of that with their decisions in AI, VR, recent products like how true is that today that there's still an ethos at Apple of we might do the unpopular decision but the consumer will like it because we know better than the hot takesmen.
I think that calling it paternalistic is the I know what you're talking about. But it depends on whether you end up agreeing or disagreeing with what they come up with. I think from Apple's perspective, it's more of a and it really does boil back to that Steve Jobs adage that design isn't what it looks like, design is how it works. And that's the job of designers to, you know, that certainly in the Apple view of the word that it's not being a stylist. A stylist is somebody who can tell you what to wear if you're a person or, you know, tell you what to do with your hair, dress it up, but you're still the same person, right? A designer is like, who do you want to be? Who are you? You know, and style is part of that, but it's just a part. And the hard work of design is making those decisions. And you know, famously just, you know, it is the greatest tech product of all time. Probably the greatest consumer product of all time, the iPhone. And one of the core decisions at the front was no hardware keyboard in a world where the difference between well actually all the phones had arrays of buttons, right? The dumb phones had number pads and the smartphones had BlackBerry style hardware keyboards and Steve Jobs just addressed it head-on in the introduction event. Put pictures up of a BlackBerry, a Palm, and I don't know, maybe a Samsung or I forget what the third one was, and said, 'Look at what they do. We're not doing that. We're gonna make, you know, we're going to do it with software.' And it was the thing that the most knee-jerk reaction from people who were already in the market for quote unquote smartphones at the time were like, 'Well, I can't do this because I spend all day typing messages and emails on my phone, and I'm not going to do it by poking at a touchscreen.' But it turns out now every, you know, within a handful of years, every single phone worked that way. That's a great example of it. Where they didn't make two phones. They didn't make an iPhone that had all screen and they didn't make a second one with a hardware keyboard that took up a third of the front of the face. Because they were convinced that that was the right way. Right. They and they were right. Right. The market proved them right. It's as good an example as any to just sort of exemplify their mindset. They didn't do it to be paternalistic. They did it because they thought this is our job to figure this out and this is a better way to take advantage of the limited space that a phone is. How does that apply to AI? We obviously haven't seen that from them yet.
And so in some ways I would say this is, you know, it's like if you write about football every single year come January you start saying this is going to be the biggest Super Bowl of all time. You know it's...
This is a big WWDC though. It really is because two years ago was the one where they announced the first crack at quote unquote Apple Intelligence that did not pan out and didn't pan out in a way that they had to announce in March last year. So that last year's WWDC and said this is going to be postponed a full year effectively facing like false advertising claims just around how much they oversold it. Right. And when they did that, I mean, and again, you know, we're not lawyers, we're not judges, but even as a lay person, that was kind of obvious that that was false advertising. Like, so when they announced their settlement a couple weeks ago on that case, that was really one of the least surprising settlements of all time because it was like, you showed a TV commercial that it would do a thing and it still now a year later doesn't do the thing. Isn't that false advertising? Right? It was like that.
felt like the only real reason to upgrade at that point. But then last year's WWDC in hindsight is kind of weird. Two years ago, they announced all of this Apple Intelligence stuff, the most ambitious aspects of which still have yet to ship. Then last year they were in like an interregnum between the year before where they announced it, but they'd already said all that stuff is coming next year. So they kind of got like a free pass for a very AI-light WWDC that was excusable because they already got over the excruciating, hey, we're not ready to announce that stuff we announced last year.
But now it's 2026 and it's time to show their cards.
Yeah. And honestly, it seems like they've done a pretty good job keeping it under wraps what they have, right? They've announced the deal with Google. Very unusual to have like a white label version of Gemini that won't be called Gemini. It's just going to be used for Apple's foundation models that are built in.
Um, and there's some rumors from Mark Gurman, of course, at Bloomberg, saying that they're going to have a sort of app store for AI agents, but I, you know, I think it's probably just going to be ChatGPT and Claude and Gemini to be the sort of smarter layer on top of Siri. Um, but other than that,
So does that mean you could have a potential situation where you have like a Gemini-based LM as Siri, but then you could also call Gemini separately to get more features, which feels like even crazier.
Well, I think that effectively they don't want to talk about the fact that it's Gemini technology that's going to be Siri, right? And they're not going to call that Gemini. They've gotten that announcement out of the way. And that's Apple's foundation models. And if you don't have a Google account or don't want to use Gemini, you'll never encounter the word Gemini, right? It's just that it's Google's underlying AI technology that's powering it.
Yeah. Um, I think the biggest question is how good can that be, right? How good can an AI system be if you're not the company that controls and made the underlying models?
Nobody, we don't really have a good example of that yet.
Well, I would say that we do and there's a bunch of vertical SaaS companies that don't make their own models. I think good examples are, you know, coding agents like Cursor, legal like Harvey. You can make good products. I guess like I'm more interested in what Apple allows third-party agents to do within iOS, right? Like will ChatGPT, will I be able to prompt in Siri ChatGPT to do things in native iOS apps or other apps in the App Store? Like will it be able to use my phone for me or will it still be like question and answer?
This is like will it break the wall?
Yeah. Will it be an agent or will it be a chatbot?
Yeah, that's tricky. How do you think about what Ben always has this interesting take where he says people always want to, entrepreneurs often want to build productivity tools for consumers but consumers don't want to be productive they want to be entertained. And I see Apple as something that feels very productive. The iPhone and all the Mac apps are very well organized. You can be very productive in them. But I see the phone as primarily a consumer device. And if I believe Ben Thompson's formulation of it, I believe that then a more entertainment-focused product is maybe the end goal. And so maybe this is my way of saying like we're just early on Memoji or Genmoji or something. I don't know where that all goes, but there's like a different set of goals for a consumer company.
We used to joke that Apple Intelligence was the first AI product to really crack humor because the summaries would always be so funny. No other LLM could reliably make you laugh like Apple Intelligence. And Image Playgrounds is by far and away the funniest image generation tool because it's so bad. I mean, it's hard to use the UI. It just doesn't give you feedback on like when is it generating. It's always asking for like another, it's like do you want to add something else in and I'm like I'm kind of done. I don't know. It takes an enormous amount of work, an enormous amount of prompting to make an image whose quality is embarrassingly bad compared to any other tool on the market that lets you do it. So why did they do it?
Yeah. Um, I think that's a good framing and I think sometimes Apple hopefully, you know, when something is new, it takes them some time to find their footing. And I feel like the whole thing two years ago, put aside the actual legal implications of false advertising. It's just I feel like they got caught up in the hype and the rush and we need to have something big to announce in June 2024. Ready or not, and it clearly was or not. Whereas the Apple way is, and they've emphasized this, Tim Cook has emphasized it many times. I don't know, maybe Steve Jobs didn't because I don't think he wanted to make an admission like that. But Cook has admitted many times that they do not aspire to be first, they aspire to be best. And if that means not being first, so be it, right? I mean, that was the going back to the iPhone. There was a whole market. Blackberry and Research in Motion was a big company and Microsoft was all in on Windows Mobile and it was like, hey, Apple's late to the game. Nobody remembers that entire era.
Yeah. Think about how long they were so slow on Bluetooth headsets. Like Bluetooth headsets were a category for 20 years. And then now AirPods is bigger than the GDP of Vietnam probably or something. It's like a massive business. And same thing with a folding phone. Like it hasn't come out yet, but I'm extremely optimistic about that product. I think that they waited until the final second until the technology was good enough. And I think when that releases, it's going to actually be good enough and we're going to see folding phones everywhere all of a sudden. Uh, but it feels like Apple's at their best when either they're waiting what feels like almost too long, but then it's the perfect time for that particular technology to be diffused, or they're more on like an Apple TV production cadence where it's not tied to an annual release cadence. And I almost wish for some of the AI features, they were treating it like films and shows that go into Apple TV. Like there's no annual release cadence. It's just like when we finish something, we'll have new seasons. Every month we do a little bit here and there and when it's polished, it goes out. But that's a very different motion for them. So, I don't know, there's a tension there. So to go back to the earlier question, the underlying stuff that really only Apple can do is provide the frameworks for the built-in system apps, Apple's own Apple Notes and Safari and Mail and third-party apps to offer functionality effectively like an MCP, you know, for agentic AI. And they talked about that two years ago. That was the whole thing that they were talking about where you could just ask your phone. What was the woman's example? When's my mom's flight arriving? You know, and it just knows to look in email and knows who your mom is and knows that she sent you this. Will third parties be able to use those APIs to do it?
And then number two, obviously that's going to work with Siri, Apple's built-in AI. But will the third-party stuff, and right now all we have is ChatGPT as that optional layer on top of that, but will, and you know according to Gurman and Apple said at the time, and if you even look now if you look on your iPhone today that's called an extension, the ChatGPT, there's only one extension, ChatGPT, there are no other ones. Presumably there will be others, at least two others I would presume, Claude and Gemini. I don't see them working with xAI. Um, and I don't know who else is even maybe on that list. I also don't see them working with Meta, right? So, those are the only companies I could even think of who have a product of this scope and caliber. But will those be able to access the same things? So that if I choose as a user to use Google Gemini as a layer on top of Siri, will Gemini have access to everything in my Apple Notes so that I can ask questions about it and vice versa. Will you be able to go to Siri and ask to pull up the latest email in Gmail or will Google say, hey, for that we want you to come over to our walled garden. Apple's done a fantastic job with the shortcuts API. It feels like they've built out basically like mobile MCP servers for every app. You've been able to call an Uber with Siri for like a decade. No one does it because it's a little bit fuzzy, but it feels like in the AI era, we're getting less fuzzy. We should be on this turning point, but who knows where the business dealings will wind up, right? And it's a blessing and a curse. It's, you know, there's positives and negatives, but Apple has a Disney-like brand where they are overly family-friendly.
Um, one of my favorite examples of that is my son's now just graduated college actually, but when he was younger, we used to go on the Disney cruise ships.
Disney cruise ships. Disney cruises are super kid-friendly. But they're the only major cruise line in the world that does not have a casino on the ship, so no good for daddy, you know, and they find the middle ground. They do serve alcohol, you know, but there's no casino on a Disney cruise ship. Um because it's off-brand for them. And who knows how that'll change. You know, ESPN is a Disney property and is certainly all in on sports. But Apple has a similar problem and it especially comes to a head like they with the phone and with the mobile revolution they navigated this spectacularly well with the app store where there's all sorts of third-party apps that provide and do things that Apple itself would not want to be a part of but they can sell through the store and take 30 or 15% from. Um, Apple does not make games. People spend an enormous amount of time playing games on their phone. Uh and there's an enormous amount of money to be made by selling in-app purchases in mobile games. Apple is one of the most profitable gaming companies in the world. Maybe the most. I don't know if you did the math. I'm not quite sure how you would figure that out. And they don't break down services revenue by gaming. But they're certainly one of the most dominant and profitable game companies in the world by making no games at all. They just have the app store. Um, and with AI, I think they're trying to carve out this middle ground where their own branded Apple Intelligence is never going to be as adept as the leading edge models because the leading edge models are going to answer questions that Apple would consider inappropriate to be answered and they are going to help you do things that Apple would consider inappropriate for an Apple branded product to help you do, right? Like there are lines far short of deep fake porno imagery. There's lines far short of that that Apple is not going to want their own thing to cross. And that's the role of these extensions for ChatGPT and Gemini and Claude where if you choose, right, if you want to stick with the Apple branded stuff, you're going to have a more limited AI experience. And if you want more, there's an extension system and they can wash their hands of it because that's not Apple's Siri, that's ChatGPT or that's Gemini or that's Claude who's giving you that. But I really do think the whole thing is not going to get off the ground if those third parties with the user's permission don't have access to all that stuff on your phone, in your mail, in your notes, all of that stuff. Because if they don't then they're going to, and I think this is where it's an Apple's benefit not just to get their 30% of the subscription revenue that you would pay by subscribing through your phone through these extensions to get a plus or pro plan from one of those makers. And trust me, Apple wants that 30 or 15% from that subscription revenue. But the other benefit to Apple for this is it will keep those companies integrating with having your digital life in the Apple ecosystem, your information, your personal information in Apple Notes, your email in Apple Mail. Safari is your browser. Whereas, if they don't let third parties have access to that, it's going to push people to say, hey, I want ChatGPT running all of this for me. I'm just going to use what ChatGPT tells me to say or I'm going to go the Google route and put everything in Google Calendar, Google Keep, all the Google products, not the Apple products.
Or the stranger, or like the, I don't know, I keep comping it to like the Napster analogy of like if you're running Open Claw on a Mac Mini and it's literally just opening the iMessage app taking a screenshot and sucking that out against Apple's will and like it's open source and Apple's fighting it constantly but they can't really beat it like that. But that feels very hacker. That feels very prosumer. That doesn't feel like where we're going in terms of broad adoption, right? Because it's going to start looking clunky compared to other platforms that have a more native, this is the way it's meant to be.
And the other thing about that too is it's all the Mac, right? Because the Mac is Apple's open platform where something like that can happen, where Open Claw is possible. And as convoluted as it is that Open Claw is driving iMessage by taking screenshots and then analyzing it as an image to figure out what to do. As clever as that is, that's only possible on the Mac. And at least I would guess 85, 90%, probably around 90% of Apple users don't have a Mac, right? They just have an iPhone and/or an iPhone and an iPad. Um, and so for the overall Apple user base to move forward as this world becomes the normal world of computing and not the cutting edge world of computing. It's got to work on the iPhone and there's no way to do that without Apple's help.
Yeah. 100%. What is the latest on how Apple is approaching mobile vibe coding apps? We covered, you know, a while ago. Some of them, you know, not getting removed from the app store, but not allowing them to ship updates. It feels like it, I can understand exactly why both parties there have the views that they have. Apple's like, hey, you're just kind of like building software on the fly. This is like, you know, one, a threat to the App Store business. It's also, you know, going to allow people to generate applications that we don't want on the iPhone. And then the other side, hey, you're stifling innovation. Users really want this. It's harmless. Let it happen.
We don't know. Uh and that is way up. That's a great question because I would put it easily top three or four on my list of things that I'm most interested in for the WWDC keynote is what is their answer to this? Um, including is it, you know, is this one of those things where 90 minutes or two hours after it starts and it's over and people are talking about the stuff that was announced and we're all going to be like, hey, wait a second. They didn't even talk about the whole vibe coding thing. They didn't even mention it, you know, and to me that would be a red flag because yes, it's disruptive to the way that the software distribution and development has worked on the platform for 20 years now. Um, yeah. And it's like very clearly the future. It's very clearly the future that any iPhone user in 10 years will be generating some of their own software. They're still going to want a lot of stuff that's purpose-built, but I would bet that in a decade, everyone that is buying an iPhone at some point or another will make their own piece of software on the fly just like they would generate. And you sort of can with the shortcuts feature like you can build something that looks like, I know it's super limited but like if I was truly giving them the most credit in the walk crawl run I would say that the next version of iOS 27 Siri, yeah, it's not going to have a text box that you can talk to and it will generate you an iOS app natively but you should in theory be able to describe a workflow that then becomes like a shortcut that is created instead of using all the drop down menus and going around finding all the apps and linking everything. It should be able to instantiate that for you because it's probably just markdown under the hood or something. Um well, but I don't know.
They just need to have an answer because it's like part of what makes disruption theory so satisfying as a business theory is that the disrupted doesn't get to choose. Yeah. Like but that's what happens there. The entrenched monopoly or part of a duopoly or just entrenched way of doing things faces the disruptive technology they dismiss it at first then when it becomes bigger and kind of can't be ignored they just sort of say well I choose not to be disrupted and it's like you can say that but it never works.
Never works. No, no, it's a powerful force. And it is in my opinion just nerding out, you know, going back my undergraduate degree was in computer science, it is in some ways offensive to me that iOS is a nearly 20-year-old platform computer, it is a computer, and it's a 20-year-old computer on which you cannot make software for the computer itself. And I totally understand why that wasn't the case for the first few years when it was more limited hardware and even a few years after that when the whole idea of the app store was new. But at this point, I mean, an iPad Pro costs like $2,000. Has a keyboard and a huge screen. It's as powerful as you can get and you can't use it to make apps. It's kind of, it's kind of...
Don't you think they'll find a way to allow this and make money from it? Because I would imagine...
Yeah, that's what I think. Yeah, I imagine they're... Yeah, I they know people want this, but the main reason to not allow it is like one just like safety and integrity of software on the iPhone security stuff, but then also, hey, we got to figure out how to make money on this, the Apple way. Yeah. And there's obviously some technology that's already and a way of doing things. The whole TestFlight ecosystem which is what Apple's beta distribution is for iOS and it was a startup, you know, I forget probably more than 10 years ago when it was an independent startup but Apple saw the wisdom of it, right? This is a thing that Apple didn't invent, they should have invented it but then they saw how clever it was and they're like oh we'll acquire these guys and build it into our platform and so there are ways to distribute software now outside the app store and it's just by limiting the number of people who can do it. I think the TestFlight limit is like 10,000 or something like that. And they could do something with AI where that limit is much lower like even if it was just like 10 people, right? Like a consumer level of TestFlight distribution where you can make apps for your friends and family or yourself and distribute it on up to 10 devices and it doesn't go through Apple, doesn't need approval. You just make it and distribute it and it's only limited to 10 people. Um, you know something like that. But there needs to be a story like that though. And I think again only Apple can really make it a really good rich system where you can make apps that really feel like real apps. But the technology to make the apps, it's obviously there. We're all talking about it this last 6 months. It's obviously there and you could do it on the phone itself. And it's super frustrating that some of the third parties that were doing it like Bitwig and I forget some of the other ones, but Bitwig was the one I'm most familiar with where they kind of, you know, they're not kicked out of the app store but they stopped taking updates to it. And it's like super frustrating because it's not theoretical. It is an actual thing that was actually working and they took it away, right? And it was super cool. And yes, it's disruptive to Apple's business, but in a way that the disruption is inevitable anyway. So, why not go with what's cool and do the coolest thing possible that makes new possibilities available to your users and the developers on your platform?
Okay, last question. Uh, give us a recommendation of a recent favorite episode of the Acquired FM podcast.
Uh, I don't have any recent ones. I'm going through the...
You're going through the archive, right?
But what's a recent favorite that you've listened to? Not recent from their catalog, recent from your listening history. What would you recommend for someone who's getting into the Acquired podcast?
So for people who aren't familiar, this obviously probably everybody listening to this watching this show knows of the Acquired podcast. Me, a big dummy who does a lot of reading and does not listen to, I don't have a commute. This is my home office, so I don't have a lot of podcast time. I stumbled, a friend of mine recommended the Mars episode, the Mars Incorporated episode. That's the one. And a friend said, you got to listen to this. And I'd heard of Acquired, you know, but it's like I don't know, there's if you listen to every podcast somebody says you should listen to, you would spend 24 hours a day listening to podcasts, right? So, I go there and it's 4 and 1/2 hours and I'm like, you got to be kidding me. Uh, and I'm like, but all right, I'll listen to it. And it's of course fantastic and it's amazing. But the Mars one, it just sucked me in. So, that's the one I would recommend. And fantastic. It is a hundred-year-old candy company that the business decisions that not the founder, but the founder's son, Forrest Mars, who's the one who took it from sort of a mom and pop thing to a big one. The decisions he made in like the 1920s and 30s are exactly like the world of business today in the 2020s and 30s. It was like he came from today's world, time traveled a hundred years into the past and applied today's knowledge of branding and scale to that world and there's just these other crazy bits that I had no idea like the fact that Hershey who was the bigger company at the time supplied all the chocolate to all the companies including Mars and they just were like sure we don't care and it was sort of like a heads they win if you buy a Hershey bar and tails they don't lose because you're buying a Snickers bar that was covered in Hershey's chocolate until the day Forrest Mars called them up and said yeah we're going to cancel our chocolate contract and they're like what are you crazy and he's like yeah we're going to make our own chocolate.
Integrating. Final, final, final question. It can be a really brief answer. Do you think that Apple's overall software quality is getting better or worse? Because when I use the FaceTime app recently, I just get enraged because it's a constant experience of, oh wait, I'm in this group and they started a FaceTime and I didn't get a notification and I'm on a call and now it's doing a FaceTime and I start to maybe, maybe it's just me. Maybe I turned 30 and I just can't use the device anymore. But it just feels like across the board there's less polish than I remember. And I'm wondering what you think.
I love how you prefaced it with make it a short answer.
Um, I think that to make it as short as possible, I think Apple over the last 10 years is facing a be careful what you measure problem with software quality where things that you can measure are things like crashes. When an app just crashes and it says like, hey, do you want to send a crash report to Apple? Like if you've opted into the share crash reports with Apple thing and their apps crash almost never, right? Like it's almost unusual when you know you're using Apple Notes and the app just crashes. It doesn't happen. And they've tackled a whole bunch of things like that because you can measure it, right? Like there's like a number that used to come in. How many users today had Safari crash? Well, you can measure that and they've cut that down. The things you can't measure are things like your problem with FaceTime. Like I'm confused. How do I add somebody here? I thought that I added a third party or there's an ability to call somebody, but it doesn't ring on their side. But I'm like, what's the point of just joining a call when the other person's not getting a live notification? Well, if you happen to be in that iMessage group chat at the time, you can see that notification. I'm not... This is how it works. The phone's a busy place, but I'm just saying it's like a terrible user experience every time where I'll send John a text. I'll just be like, John, we're in the FaceTime because I know he didn't get it. You can press the ring button, but it's an extra button. It's a perfect example because Apple's never going to get a number that goes up because you encountered that problem. There is no number that's going to land in front of Craig Federighi's dashboard of problems with Apple software that says Jordy Hayes had problems adding somebody to a FaceTime today, but you did. And that's the sort of thing it's getting away from when Steve Jobs was there and was not just the CEO of the company, but the number one user of the products, right? And if that happened to Steve Jobs, that would get fixed. That would start getting fixed the next day, right? If he had problems adding somebody to a FaceTime call, all of a sudden that would be the biggest problem in the world, right?
iMessage search, right? The number one app that I use.
Well, search is just bad everywhere.
Yeah. But Apple should be able to create like the most elegant solution. It's the number one app.
Search is bad everywhere. I'm not afraid to say it. Searching in iMessage technically works, but it is sort of like the way that putting two empty soup cans and tying a piece of string and holding it taut that you can talk to somebody through the can. It does work because this audio goes through the string and goes to the can on the other side, but that's not what people think of when they think of talking on the telephone, right? And the way that you search in iMessage and the way that you get results in the sidebar is not the way people think of search. People think of, hey, I was searching for this one. I was talking to somebody about this one word and it was just a month ago. Just show me the goddamn message where I was searching for that word. And I think that's the sort of just sort of opinionated from the top thinking that was there because Steve Jobs was running the company and Tim Cook just isn't that type of person and the company has taken on as a company more of the personality of Tim Cook. That's the nature of leadership, right? The organization tends to take on the personality of the person leading it. And there's a lot of good things that have come from that.
But I, and that's for example, Tim Cook loves numbers and so something like hey how many times a week do apps crash? Those numbers have come down, but that is driven by a culture that kind of comes from Tim Cook. And I think hopefully under John Ternus that they can sort of not turn the ship around, but just make a minor course correction back towards, hey, let's focus on just some opinionated stuff like, hey, this app is confusing. Like when you open this and you just want to add a collaborator to this, why does it look like this? Shouldn't this be way more obvious? You know, and I think, you know, we can't measure it. This isn't going to come with a number that goes up and down. It's a je ne sais quoi. We know it when we see it, but this isn't up to snuff. And I think they've gotten away from that.
Yeah. Well, thank you so much. We went way over time.
You fit it exactly into 60 seconds. Thank you.
Uh, we got to do this again because this is super fun. We could go all over the place and there's so much more to talk about, but have a great weekend. Have a great rest of your day. Thank you so much for...
You guys. It was a pleasure to be on.
We'll talk to you soon. Have a good one. Up next, we have Ronak from Trajectories, the co-founder and CEO. We've been keeping them waiting too long. We have the whole team. Let's bring them in. Give us the news. What happened? How much did you raise? Tell us about the company. Sorry for keeping you waiting. Please introduce yourselves.
Awesome. Great to meet you guys and thanks for having us on. So good. Yeah, I can start. My name is Michael. Uh before this I was on the foundation research at DeepMind, built a lot of cool things in robotics and very excited about Trajectory.
Amazing. What's up? My name's Arjun. Before this I was at Apple working on Vision Pro and then multimodal foundation models.
It was good hearing a lot of tea right before this.
Well John also is like the number one user in the entire world of the Vision Pro.
Yeah I watch full movies in Pro regularly.
All of his blog posts. It's great. He's got some great takes on this. Well, I was saying this John. This John, absolutely love this John. No, no, no. I'm like still a regular user of the Apple Vision Pro, which is...
I'll call John on a Friday night. It's like I got to, I'm in the Vision Pro. Give me a second.
We should chat. We should chat.
Now, everyone else in our office is using monitors and everything. Sometimes John is just like they're working in the office. But great to meet you guys.
I'm Ron. I was at Windsurf before training one. Uh and then this led to the 2 billion acquisition to DeepMind. So went over there as you guys probably heard on the news and then left all the acquisition money on the table to start Trajectory.
Okay. But you raised some new money. How much did you raise? What's the deal?
Yeah. So we raised 15 million from Conviction.
And we're building the platform for continual learning. Uh we're working with some awesome companies. Harvey, Dakagon, Clay, Rogo, More. Um, and basically building agents that learn online. So all of that usage, the edits, the retries, everything they're using in an AI product, using that to make AI smarter every single day.
Yeah. Amazing. Um, so, yeah, like 15 million seems like, you know, great amount of money to start a business with, but it doesn't seem like enough to jump straight to training a frontier model at the mega scale. So I imagine your approach is a little bit more capital light. How much of your approach is in the deployment phase with the companies that you're partnering with versus pure research versus something that's going to happen around fine-tuning something like what can you tell us about how you actually deploy a model that is capable of continual learning?
That's a great question. I would say the way we're thinking about this is every lab is focused on building this like smart PhD student, right? And how to make this PhD student smarter and smarter, but it's always day one of their job. Sure. And it's always that they're going to make the same exact mistake every time. What we're interested in is building an agent that has 10 years of experience on the job. It doesn't need to be a PhD student. It doesn't need to be like frontier level math or anything like that, but it needs to know your primitives. It needs to understand your business. And it needs to really, really understand your rewards. And with that, like just doing very, very light post-training, okay, can get us very, very far. So we have models that are 10 times smaller than the frontier models that are able to beat them. But also, that's not what we're most excited about. What we're most excited about is that not only are we beating them today, but we're improving 1% every day. Yeah. And that is continual learning. That's the most exciting part about continual learning.
Very cool. Is part of this too, you know, you mentioned some of the application layer companies that are your guys' customers. Is part of this helping sort of reduce their dependency on the model providers and help change maybe the kind of leverage equation there.
