Tim Cook Retires, Mark Gurman Joins, Images 2.0 | Howie Liu, Scott Stevenson, Alex Wiltschko, Spiros Xanthos, Carolina Aguilar, Jake Jurewicz

28:05Uh, but he takes it a lot further and he's like, so you want to sneeze on as many people as possible. And he refers to that as like, if you send them in a file that you've sneezed on them and they might create an account it's just like a much more like visceral way yeah i can't even get this well yeah no i mean the thing about so so my first company uh is an ad network yeah so uh we would you know a company would come and say like i want to advertise yes a hundred thousand dollar budget and then the company would put together a dashboard yeah uh of like potential buys and then the person would go through and so it was inherently viral every customer that would work with us had to log into Airtable and like use a product so that was just happening at like massive scale yeah and I think like that type of like you have some like data set like you know maybe it's for your ad inventory or whatever maybe it's for like your CRM or whatever you need to collaborate with it like it's a very fundamental construct in just like how knowledge work is done right so I think like the lesson learned for me or like that the principle applied was like can you go after something that's so foundational that like it's always going to be around right like i think with the personal crm thing i kind of felt the like turbulence of like is this in vogue right now is it not like and i really wanted to go after something that felt like it's going to be around for decades right and like what's more eternal than like people need like databases record a database that you can do stuff with the past 40 years of computing it's probably going to be around for the next 40 and like even now with agents it's like the database layer actually becomes more important, right?

29:33You don't want just a bunch of ephemeral context windows for agents. They need to store and collaborate on data along with humans. So we kind of picked that as the vantage point. And a lot of the early customers were startup founders, small business owners. But interestingly, we had written this fake business plan. It was basically a vision deck more than an actual business plan. But we had said, conjectured, we're going to have to go after a long tail of the kind of prosumer SMB audience, basically like Dropbox, right? And I think what was really surprising is it turned out to be a little bit more like Slack, where we got the most virality within larger companies.

30:07So, like, there'd be a big, like, media company or, like, even, like, a scaled startup, like a WeWork or something that would run all of their operations very quickly early on on Airtable and then just grow with the company, right? And so, like, I mean, WeWork was one of our early customers had, like, probably 10 ,000 people, like, you know, when they were at their peak. Like, basically, it was, like, used by every almost employee there. And like a lot of their operations, building operations, et cetera, were just built on Airtable by default. And I kind of learned the value of like having this like data gravity.

30:38Like once you get enough data into a product like Airtable, like it just kind of retains really well within the company and gains more and more usage. Yeah. How do you think about that? Until the company. Well, you index against the industry that you're in. So I want to get to all the good part right now and all that stuff. But walk us through, since this is your first time on the show, you went from being one of the hottest companies in tech during the whole no-code boom, the PLG boom, Zurp. It must have just been an insane experience. And then like there's kind of this reset in late 2022, 2023.

31:24How has it been kind of like building out of that trough? And then like I'm assuming it sounds like you've been like very re-energized by this new opportunity. Yeah. I mean, one of the maybe benefits of like not being an overnight success, because we took like two and a half years to build the product. Sure. Even like from 2015 to 2017, 18, I would say like we were getting like a steady compounding of growth. but it wasn't like Slack or like Dropbox where it just overnight became super easy, right? It felt like we had to really grind. We had to think about like, how do we need to like improve the product and increase the, you know, kind of like shareability and the scalability of it.

31:59So it's kind of a grind for like at least the first five years. And 2018, when we got our first unicorn round, it's kind of the first year where it felt like it was starting to get easy, right? So 2018 to 2021, like very fun, easy years, but also, you know, Everything is so good. Money just was being printed in the world, unlimited money. and we got to raise a big set of rounds. How would you like a 100x revenue multiple? I mean, yeah, and just the absolute scale of funding was huge compared to prior art, right? Now, I mean, you can raise$100 billion if you're opening eye, but at the time, we raised, our first unicorn round was like$100 million round, and we raised another couple hundred, and then a few hundred more, and then our big round was our Series F, which was kind of at the peak of the markets, raised$700-plus million in that round at an$11 billion valuation.

32:55And we still have all of that money on the balance sheet, and we're now cash flow positive. That's amazing. So I think it was kind of a fun time to kind of get to ride that wave. But I always, I think for myself, knew you have to build a durable business, right? And so valuations are going to rise and fall. It's just gonna be like macro but like, you know Ultimately, either we build a great enduring business or we don't if we don't then like, you know You could be like a flash in the pan, right? So yeah, I think we were always like trying to focus and I tried to focus on like what do we actually need to do to like, you know Compound growth like go after the enterprise obviously at the time especially like it was clear that was like the move, right?

33:36You get PLG but eventually you have to go into the enterprise and when like these big multi-million dollar contracts, like become a really sticky system within these larger companies. And we did that. And like, we're still doing that. We have like a bunch of the fortune 500, like running really critical operations on air table, whether it's like content production at a big media company or like, you know, like fund operations at a company, like, you know, like financial services company. So these are like the, like almost like modern ERP. Did you hire a different set of individuals to work on that?

34:07That we're already connected and knew that flow or was it something where like your best sales reps just sort of got bigger and bigger and leveled up? It was a little of both. I mean, I think it's a different muscle. Like I think Rolodex selling is even at that time, like, you know, not that effective. Like I think like just knowing somebody at a big company, like doesn't, even if you're like, you know, very senior and they're very senior, like doesn't actually help that much. Like we've had some reps come in, like they've had like a decade long relationship with like, you know, the CMO of XYZ company.

34:35And I think that gets you like a phone call it gets you like a meeting yeah, but ultimately like buyers are wising up right? They have been for quite some time where it's not just like oh, I know this guy I'm gonna like or you know gal like I'm gonna buy this like product from them Like you actually have to like show them why this is gonna help their you know to help them in their job right and help the company And so I think it became much more about like transitioning from like oh People can just use it on their own and they'll figure it out on their own To like starting to do more of a consultative sale like come in and say like okay how can we solve like a really big problem for you?

35:05And maybe like for one company, it's like, how do I consolidate like my end to end operations for like how we do all of our brand planning, launching new products, all that. And that's kind of like, it's like one part consulting, one part, like just thinking about like a big enterprise scale solution. And then one part, like be able to leverage the flexibility of our product, almost like in a Palantir like way to show the customer, like we can actually solve this really deep problem for you quickly. Is there any sort of like PLG motion or land and expand that happens in the Fortune 100? Because there's some small team inside of Coca-Cola or something that's using Airtable, and then you're able to use that as a demo or jump an off point?

35:49That does happen? Yeah. I mean, it's like there are some companies where you just can't even get your foot in the door without the top-down. Like a lot of big banks, we were firewalled out until we got some top-down intro. Oh, interesting. So they can't even... They literally block you, right? Your IP is like blocked. It might be like hard wall, request access. So there are some companies where you have to come in top down. But I would say 70 plus percent of our current enterprise accounts, including the ones that are now like five plus million in revenue, originated from teams within the company organically adopting Airtable, right?

36:23Sometimes it was kind of like shadow IT. They just figured it out on their own. and they just like they showed real value from using the tool right like they would build some real operation on it and say like well i've been waiting for like it to deliver me this like old bespoke solution or some like crappy other vendor for like two years now i got impatient and just built the thing myself and that's like a big i think you know i think of like the enterprise landscape now is like you know there's plenty of dollars in enterprise right even now like you know it's just shifting from like traditional software to now ai but like there's plenty of dollars, right?

36:55The budget's there. But like the question is, you're not the only one going after it. And so like, what's your kind of asymmetric wedge to get in there and like take those dollars, right? And if you're a big company like a Salesforce, maybe it's like, we already have the distribution, we have like the customer data in there, we're going to go and attack adjacencies. If you're Airtable, we don't have like the scale of like a certain, you know, ServiceNow or like an SAP or Salesforce. What we did have is like the usability of the product. So like the PLP was like kind of the entry point. And then also like, even when we pitched to other people in the company that hadn't used Airtable, they had probably heard about it from a friend, like maybe the CMO is like, you know, partner like uses Airtable in their company, or we can go in and just show them like a really compelling demo quickly.

37:36Talk about AI. What are customers demanding? What have you rolled out? Where does AI fit in? Well, where does AI take a back seat? Yeah. I mean, I think it's crazy because we've seen so many layers of disruption happening almost in parallel. You think about desktop to mobile, it was a single form factor change. Kind of easy, almost, to execute on. That was the one that I experienced at Salesforce. The big thing at the time was Mark would tell every team, show me the mobile UI first before you show me the desktop UI. Go mobile first. It was the right move and also kind of a simple move. Now it's like you've got at one level, like obviously every product should have like AI in it.

38:19So, you know, we have the obvious stuff. Like you can now talk to Airtable's assistant, like co-pilot style, and have it do stuff on your behalf in the product. We have what we call field agents, which are kind of like the ability to map reduce AI calls against like all of your data. So you have like 20 ,000 customer records and run like, you know, AI agentic, like, you know, kind of tool calls, like search and like research about the company, like synthesize. Hydrate a bio for every single time. Exactly, that kind of stuff. One row in time. Yeah, and we do all that stuff. But to me, the more interesting disruptions underneath that are, one, do people even want to come into your interface anymore?

38:57Yeah, that's what I was going to ask. Why you care about storing the data in a safe, secure way? Yeah, Salesforce just went headless recently. Is there a plan for that? or how are you been interested? Yeah, I think the right move is hybrid headless. I think the whole, look, if you wanted just a back-end database, you could use Postgres, Supabase, right? And there's PHP, my admin equivalents, modern-day ones, Prisma has its own version of it, that are okay or they're good. But I think what most people actually want, especially in a business context, is you want the database, but you want to have proper permissioning, you want to have proper collaboration.

39:36And most importantly, you don't want to exclusively interface with the data through an agent. You want to do that a lot of the time, but it's really helpful to actually go in and see the actual data. I think of it as the equivalent of, even though agentically you can generate all your code, and you should as a frontier developer, does that mean you never want to inspect any lines of code ever? No, you still want to see a diff of all the actual code files change, whether that's in your IDE or in GitHub or whatever. And so I think the equivalent here is you want to be able to drop down into a really nice interface.

40:09And we've done some work around like kind of figuring out what's the best blend of the two right so like with chat bt for instance We have a kind of a first-class integration where you can go in through chat bt and like interact with your data and air table Say like hey pull me like all the customers that are like waiting for an outreach for me and pre-draft like outreach messages But then it can basically compose like a fragment of a view within the chat bt interface So like you can actually see like air table sure, but like, you know kind of a part of the internet So it's not completely headless.

40:38It's almost like you get to pull out like pieces of its face at the right time on demand. And I think that's a really important kind of like UX form factor. Yeah. How are you thinking about speed in the context of AI? I feel like the models keep getting smarter, but they also keep getting slower, basically. And while I'm extremely confident that I could point a deep research agent at a massive air table with 20 ,000 rows and get very good results. Like a lot of times I'm just in my email and I want to find one thing very quickly. And that feels like it has yet to be, you know, AI-ified or at least like LLM-ified.

41:19It's very much like, okay, well, I should probably just fall back to like a SQL query or just some Boolean logic or just like vanilla search because I want this now. Yeah, I think both are gonna be really important experiences and obviously we have like, you know kind of great like smaller and faster models Sure, like the mini, you know, yeah, you know that that are great for like more synchronous interactions And like within air table Like if you go do to air table or use chat to you with like one of the smaller models You get that like faster kind of almost like more like real-time experience But I do think like a really important class of work that will come to dominate like every frontier company or company trying to reinvent themselves to be Frontier is figuring out how to operate in this new modality.

42:03It's like the best developers today don't go and sit there in front of their IDE and synchronously talk to the agent. You have 30 separate branches that are each being worked on by a different agent. You can have the agents continue to update the branch based on human and other agent feedback. You can have comments back or run tests, etc. I think this whole idea of, look, it's going to take hours for that entire loop to complete, right? Like agent pushes some changes, the changes get feedback from other agents or humans, agent responds to that. Like that whole loop could be hours, not just like minutes.

42:37So you're not gonna sit there and like watch it one at a time. But the powerful thing about this is like, each one is still actually operating faster than like a human engineer could have like back in the day, right? Like when I think about like the speed with which like our early team at Airtable could build features, and we had a very good team, like one agent on one branch can, you know, do the work of like maybe three humans back in the day operating probably in like three times the time. Right. So it's like literally like a 10 X, you know, kind of leverage factor just for one agent. But the best engineers are now able to multitask and kind of basically say, look, I'm going to oversee my own little team of like 20 to 30 agents working concurrently.

43:13And so I think it requires like, it's almost like everybody needs to graduate from being an IC to like an IC manager of agents, meaning like in every function, Like if you're like a VC analyst, your job should no longer be to go and synchronously research one company. It's like you're going to go and research like 30 companies and do them all faster, better and higher quality. Right. Like then than what you could before. And so I think that's the greatest leap that is going to be challenging for a lot of people in a lot of roles to make the leap on. Because it's it's a totally different mentality to like how you operate and what your role is than before.

43:48What are you pushing the team to achieve? So a lot. I think there's basically three different levels of self disruption we're trying to do at Airtable. One is the core product itself, how do we reimagine that for an increasingly agent led future? So all the headless hybrid type stuff we talked about. The best testament to that is do we see massively growing basically tool call volume from Chat, Petit, Claude, any other agent products. It's like, are people using Airtable more and more agentically? And is it working smoothly for them? So that's like priority one. The second, though, is like, I think we have to like really transform how we operate internally, right?

44:31Like, you know, clearly like the companies, the best companies in the future are not just going to hire like massive armies of people to do everything, right? Like they're going to hire like people who can really effectively leverage agents, right? It's so obvious that's happening in engineering where like, you know, if you could hire one engineer who could be fully agentically leveraged, you get more output than like 30 kind of traditional engineers doing traditional engineering. So that's kind of one internal thing. But then the third is like, I'm a strong believer that like, you have to go and skate to where the puck is going, like index against like the big tidal wave coming, right?

45:04Like Amazon did this like back in the day against like the growth of the internet, right? Like they, you know, Bezos picked books and like, you know, e-commerce because like he thought that would be the best way to index against the growth of the internet. And so for us, like we get that through Airtable and like kind of hybrid headless Airtable, but we're also placing a big bet on hyperagent because hyperagent is basically like taking all of the like excitement of frontier agents, like IE OpenClaw and like YOLO agents that can just like have access to your data and tools and do stuff like really, really like long running stuff, not like 10 second tasks, but like 10 hour tasks.

45:37but for non-coders. And we want to do it in a business-friendly way. So you can go and do this, deploy it into your company, run agents across your entire company. And so that's kind of like a bet on if we believed Airtable 10 years ago was the most meta problem, the largest problem we could work on, which is software arguably was the biggest and fastest growing industry at the time. And how do we go after that entire category and index against that? how do we now go against agents and say like we want to build like the best agents platform that any business and any person can come in and use and just start building agents with right and deploying them into their company right and so if we do that really well then like we get to doubly win both as the data layer but also kind of have a bet on the the agent wave yeah last question should children learn c++ definitely not no i think is that because of c++ or because of the AI era?

46:35I think, well, I mean, I think both. Like, I think, I think the fundamentals of good technical architecture are going to be the most important thing. But that has, like, I think the abstraction that really matters now for creating value is raising up, right? Like, it used to be at one point, like, you know, Bill Gates wrote, like, some of his first programs in, like, literally, like, machine code, right? And, like, would punch it into his, like, PDP 10. And, like, clearly, you could be a great startup founder or be a great software engineer and make lots of money, like, without having to go down to that level.

47:03And so I think now with agents, the bar has raised yet again where what you really need is good product business and tech architecture sensibilities. How should this system work? Where should the different levels of responsibility belong? And if you can get really good at that, then you have super leverage. If you are just learning the literal lines of code and how to write them that a lot of engineers were before, I think that's going to be increasingly below the frontier line of like agents can just do it like equally or better to humans. Yeah, that makes a lot of sense. Well, thank you so much for coming on the show.

47:40Thanks for having me. Have a fantastic rest of your day. Great hanging. Talk soon. And we have our next guest, Mark Gurman, the Gurmanator himself in the waiting room. Let's bring him in to the TV PR and Ultra Dome. Mark Gurman, how are you doing? Tired. How are you doing? Tired. I can imagine. This has been months in the works. You predicted this many times, but also on our show. How did this come together? Did this match your timeline? Were you surprised by this particular Monday that it was announced? Walk us through the scoop. Yeah. And get the scoop. Get the scoop ready. This is the scoop.

48:21You got this for you. The golden scoop for you. The golden scoop. Oh, I've never seen that. Yes. It's a new prop in the studio. You'll have to take it for a spin the next time you're in person. The OpenAI deal is already doing work for you guys. There you go. So here's the deal. There's a reason I published my profile of John Ternus just a few weeks ago, right? This was all coming together. Things really ramped up internally at Apple on this at the end of last year. Things have been in motion. The plan was to announce it after the 50-year anniversary celebrations. And it almost felt like the 50-year celebrations were not just about Apple's 50 years, but sort of a goodbye celebration to Tim Cook and his legacy at the company.

49:07So it all came together over several months. This really started about two years ago when Tim Cook identified John Ternus as the next one. Ternus had been prepared for this role probably for over five years at this point, since they put him on the executive team when he became SVP of hardware engineering. But this started in early 2024. And at the time, I wrote that that was the first time I wrote that he would be the next one. Yeah. Have you been able to process? I know you've published a few memos. Have you been able to ascertain anything about the internal response? Are Apple employees excited about this?

49:50It feels like it's been managed from a communications perspective very carefully. And so it shouldn't have been a surprise to anyone. But are Apple employees generally excited about this? It seems like there's a lot of cause for optimism, but I'm always interested to hear. Yeah, there was that one article a couple months ago where clearly they were getting quotes from former employees that were like kind of taking potshots at him. Yeah. Basically saying like, he's never made a hard decision. Yeah, that was the quote that went into the journal, but I don't know. It seems like it didn't matter because he got the job.

50:24Well, got the job two years ago, and I think he's going to do a hell of a job in this new role. I am quite optimistic for Apple in the long term with Ternus at the helm. He has product sensibilities that Tim Cook simply doesn't have. He has product decision-making ability that Tim Cook certainly had but wouldn't utilize because he himself knows that product-based decisions is not where he could have the most impact. Just like Tim Cook really oversaw the operations part of Apple as the CEO and left product development to other members of the executive team, My expectation is that John Ternus will be intimately involved with the product side of the organization as he was in his prior role to CEO and will leave the operational side to people like Sabi Khan and Priya, the people who run the operations division at Apple, supply chain, manufacturing, procurement, AppleCare, you name it.

51:19And so he's going to pick his spots, and his spots is hardware and product development. There's a reason that when he chose his successor for the hardware engineering organization, he chose Tom Marriott. Not an innovator, but an incredible execution guy when it comes to hardware engineering product quality. He did that because his belief is that he will still be intimately involved and sort of be that product visionary for Apple in this new chief executive position. Yeah. When I remember Steve Jobs, I think of Jobs as an innovator, as a visionary, as someone who both did Pixar and the iPhone. So many different projects, a lot of them wildly successful.

52:00Tim Cook felt like a focusing of that a little bit, but you still had the car, Vision Pro. There's a few different projects going on. Is this the most focused Apple has ever been and will ever be? or do you think that there's, do you think Ternus has like some aces up his sleeve where he might want to take a wild swing at something? The thing with Tim, or what John Ternus is going to have to do is stay the course. Annual iPhone, iPad, Mac, Apple Watch, AirPods upgrades. But at the same time, is going to need to do a better job of bringing out new product categories that Tim Cook has done. If you look at Tim Cook's legacy in terms of major new products, it really was on the services side.

52:40the AirPods and Apple Watch those were both really developed by management teams, engineers and people who came from the Steve Jobs era that's not a slight, but my point being is that we really haven't seen anything wholly new that is also successful since 2016 with the AirPods and the Apple Watch at the end of 2014 the Vision Pro has obviously been a Tim Cook product, a Tim Cook priority, and it's been sort of a flop at least for now, I know Apple has a very long, decade-long spatial computing roadmap. They eventually want to get to AR glasses. They'll have display list glasses to compete with Meta several months from now into 2027.

53:19But he needs to get cracking. There are six major Apple products in development right now. Six major new product categories. AI AirPods, smart glasses, pendant, smart display, Is that the lamp or the kitchen thing? No, no. Lamp is number five. Smart display is different. Okay. The tabletop robot, so that's the lamp, the moving lamp. And then number six is a, I'm only going to say very little about this, but a security camera. Okay. Interesting. Well, at least the Apple Vision Pro has one key fan. Do you think the lamp is a predecessor for a humanoid? Do you think Apple would ever do a humanoid?

54:06I do. I do. But I think it's going to be a decade if they do. And they're going to wait and see. I feel like it says a lot that they're not. Talking about it? That you don't know about an internal humanoid project yet. Oh, I do. They're exploring humanoid. The idea of a humanoid. They're not working on it full throttle, but they have a large robotics initiative. They're working on AI robotics technology, and they're also working on robotics hardware. John Ternus actually took control over the robotics hardware team about a year ago. He took it from the AI chief that Apple got rid of a couple weeks ago, John Gianandrea.

54:48But they're also looking at – they're actually building, it's really cool, a gigantic manufacturing arm or a gigantic robotic arm that they want to use in manufacturing, but also used in Apple retail stores to grab products off the shelf in the back room and whatnot and bring it into the store. That's probably five years away. But they're looking at robotics from a manufacturing standpoint, from a retail standpoint, and also, most importantly, from a consumer standpoint. They've also been exploring a mobile robot, something like an Amazon Astro. But I don't think that's probably going to see the light of day.

55:20That's fun. I feel like Apple has the perfect brand for robotics. Talk about Ternus' challenge with supply chain broadly, what you think he's going to be focused on over the next five years. I don't think he will be. I don't think he will be. I think just like Tim Cook did. But is that not – you're saying like just basically like broadly ignore that it's kind of a key risk to the business to have – I mean – No, it's a team. With the hands – in all hands meeting with Apple employees this morning, he was pretty clear that Tim Cook didn't do everything. Tim Cook chose his spots. and Ternus said that he's going to pick his spots as well.

55:56As we know, Tim's spots was operations, finance, and sales and he delegated everything else. My sense is that Ternus is going to, Ternus' mandate, Ternus was hired because they believe that he's going to be able to bring Apple back to the forefront of product, device, innovation. They already have the best in class operations, finance, sales people. They don't need Turnus to do that. They need Turnus to keep his eye on the prize, which is products. Yeah. And what do people point to when they say that there's a risk to Apple staying on the frontier of product development? I saw the Android phone that has the privacy screen that you toggle on and off.

56:42That looked like kind of a cool feature. There's folding phones that they're working on. But are any of these features that exist in other phones, it feels like they haven't actually gotten a groundswell and started pulling iPhone users away from the ecosystem. But are there key features that people are worried about? It's not here yet. Characterize that idea. Yeah. It's not here yet. Nothing you've seen is the risk. The risk is whatever the hell meta and OpenAI and Hark and all these companies eventually come out with. The risk is one of those companies doing something really cool and jettisoning Apple from that perspective.

57:18But we all know that nobody has done quote unquote cool stuff yet to steal away iPhone users. Nobody is ditching their iPhone for Android. In fact, the switching is going in the other direction, despite the fact that Apple is supposedly the most innovative company in the world and has the least innovative AI technology. Yeah, but consumers care about value and things like the MacBook Neo really deliver that value. Brand, colors, value. Ternus was only senior VP of hardware engineering at Apple for five years. It's a short tenure to be an SVP of a division at Apple. Oh, there you go. No, he's been at Apple 25 years.

57:58I'm kidding. We use that ironically. No, I know. But the point I'm trying to make is that he still has a legacy. And Ternus' legacy is making Apple hardware more performant, in terms of speed and battery life. Mm-hmm. And higher quality. He's really focused on the durability and longevity and the reliability of Apple products. And it's meaningful, I think, that the person that they chose to be Ternus' replacement in hardware engineering, Tom Marriab from Intel, is a product quality and reliability expert rather than a product design person. Yeah. What was the thinking? I mean, I remember we talked about this, how I got the new iPhone and it has immediately been dinged.

58:44What was the thinking I'm making it disposable? But it's better for heat or better for wireless connectivity, even though you can't get the color to adhere to the material as much, so it scratches off. Is that the current tradeoff? Yeah, there's tradeoffs with every material. Like titanium was light. It looked cool. You could beat blast it. It looked interesting, and it gave them a good marketing point. Like, oh, come buy a titanium phone. Like anyone cares about the material of their phone. But it had really bad properties related to heat. Aluminum, which we've known for 20 years, is an excellent material to build consumer electronics out of.

59:26So they went back to the basics. They were really talking about at the end of last year splitting the line between ultra-thin with the iPhone Air and pushing the iPhone Pro to the right as much as possible by making it more performant. My expectation is they're really doubling down on this. their goal is really to just squeeze as much performance and power in these iPhone Pros as possible. And for everyone who needs less power, you can get the thinner and lighter iPhone Air. And I think you're going to continue to see Ternus pushing that direction, making the MacBook Pro as amazing and most performant as it can be and pushing everyone else to the MacBook Neo and the MacBook Air.

59:59And I think his legacy on performance and product quality is a really important thing to remember. Yeah. Has Ternus ever talked publicly about AI in any capacity? He talked about AI in his all-hands meeting with employees this morning. He said that... I'm going to check it out. No, just hold on. I'm kidding. Hold on. I don't want to give you inaccurate... Yeah, fake news. That's the good stuff. Okay. Yeah, just hold on. Bear with me. I think I have to hear you posted. Internal memos. No, no, no, no, no. Oh, no. He said that he's especially excited to be stepping into this role at this moment because I am telling you, we are about to change the world once again.

1:00:50He said Apple has an incredible roadmap ahead and that I'm not exaggerating when I say this is the most exciting time to be building products and services at Apple in my entire career. AI is going to create almost unlimited potential. We're going to be able to keep unlocking possibilities that are going to create entirely new opportunities for our products and services, and I'm so excited about what that's going to mean for our users. Earlier this month, he reorganized Apple's hardware engineering division around in a new AI platform that they're going to be using to improve product development processes and overall quality.

1:01:27Interesting. Okay. I saw a post here from Bubble Boy. I want your reaction. Apple is about to become the mecca of hardware engineers around the world with John Ternus taking over at Apple. Is Apple not already the mecca? Is there actually somewhere to go that is up in terms of hardware engineer recruiting? Do you see this as changing the culture in some meaningful way? I mean, they don't pay like these, you know, open AIs, HARCs, and metas of the world. Apple has been pillaged by OpenAI and Meta and all these companies as of late. They are stripping apart Apple's hardware engineering division, hiring people from every team they can get their hands on, throwing very big offers at them.

1:02:16And so this has been a really big issue that Ternus has been dealing with over the last year and change. But Apple is, you know, the hardware mecca. They're the company that everyone wants to poach from. And they're the company that people go to to learn how to build consumer devices. So this is definitely, yeah, I agree to a large extent with you, actually. All right, with Bubble Boy. Yeah, Bubble Boy. This might be somewhat separate, but just get me up to speed on the folding iPhone. What is the latest there? Announced in September, Ternus' first big new product. Super exciting, super pumped.

1:02:55We've talked about this. I'm sick of the candy bar phones. It's been the same junk for 15, excuse me, 20 years now. I want a foldable. I want a bigger screen. Yeah. I really hope. John Nee wants a newspaper-sized phone. Well, they have those. I've seen those in China. They're the trifold, right? But this is a five-fold. Don't get me started on the trifolds. Those are awful. Okay, explain the trifold. Wait, why are they awful? It seems amazing. John wants pages of screens that he can turn. Flimsy, and they break. Okay, they break. You need a trifold at Apple-like quality. 20 years because Apple takes a good old time.

1:03:29When Apple does a tri-fold, it'll be good. I open up a foldable phone right now. You open it up and you can hear the screen sort of creaking. And then you have that big line in the middle. And then it's impossible to get your thumb in to open the thing. I hope Apple fix that. I don't want to hear a creak for$2 ,000. I don't want to hear a creak. I don't want it to stand stand sound like I'm stepping on you know a wooden floor yeah right I want it to just open and I want it to open quickly and nicely and it not be like I'm trying to lift the weight yeah video consumption though because I feel like we've done vertical videos nine nine by sixteen and then sixteen by nine widescreen but if you open up a foldable phone you eventually get a square and that doesn't really make like a movie watching no apples is different apples is like the new Huawei phone where it is um iPad screen ratio iPad screen ratio when you open it okay when you open it yeah okay so still black bars any intel on on uh no no no black bars black uh yeah sure there'll be black bars when you rotate it black bars on the top but if you're watching like a cinema film or even if you're scrolling Instagram like you won't necessarily get more view because for so long, all the content production has been ultra widescreen.

1:04:49If you're making a Tarantino film and it's super cinematic, or if you're on TikTok and you're doing vertical video, then you're, then you're going to have black bars on the side for, for the most part, but for so many other applications, you know, for word documents and notes and CBPN will look great on it. Yeah. Yeah. Uh, something to look forward to. What do you think Ternus' new comp package looks like. We almost marched on Cupertino mobile times because of Tim Cook. To get Tim Cook. To get him a raise. You're going to get yourself to the spaceship. Yeah, exactly. I'm just guessing. I'm just guessing.

1:05:29I think a million shares. Over 10 years. That's pretty big. Can I just tell you why I think that? Yeah, why? Because that's what they gave Tim Cook when he was named CEO, a million over 10 years. So I would assume it's the same. But again, I don't know. Entry level researcher salary, but it's a good start. Pretty much. Tim Cook was getting$100 million a year, and then everyone flipped out except you guys. And so he had to cut his pay to like$40 million. and then when things died down, he's like, all right, I'll take my 70, 80 million. I slept peacefully for those years. And then you should see my sleep score.

1:06:18We were really the strongest supporters of the Tim Cook pay package. I'd guess a million. I was just like, okay, the free market values a baseball player at the same amount as a guy leading a$4 trillion company. Make it make sense. But yeah, maybe it'll be 500 ,000 shares. Maybe it'll be 500 ,000 shares. I don't know. But I know that they gave Tim Cook a million. You got to get those numbers up. You got to get those numbers up. It's time to march. Really? You're all in on Tarnas already? We should preemptively march. We're bullish on both. We love both here. Well, now you get both now. I know we do.

1:06:55Yeah. Why is 65 a retirement age for the CEO of Apple? Like we were talking about Warren Buffett. he was able to manage a trillion dollar organization well into his 90s. Is it a more physically demanding job? Is he traveling more? Is his hand ringing from shaking hands in DC? Why not have another 10 years if you're in that seat? I don't think the hand situation has much to do with shaking other people's hands. Why is he not there another 10 years? Well, he needs to give the new guy runway. I'm sure there are some – I'll just tell you what Tim Cook – I'm not going to get into it. What I'm going to do is tell you why Tim Cook said he's stepping down.

1:07:47He said he's stepping down because it's the right time, and there's an intersection of John Turnis being ready, Apple's finances being in a very strong place, and Apple's future roadmap being in a very strong place. in terms of the real reason why he's stepping down now. You can read some of my prior articles taking a deeper look at the situation. Okay. Yeah, makes sense. Cool. I love seeing you. I love talking to you. Thank you for coming on. Congratulations on all the amazing coverage. Get some sleep. Great to see you, Mark. And keep up the amazing work. We'll see you soon. We'll see you in 15 years for Ternus Jr.

1:08:19Can't wait. We'll see you. Goodbye. Let's pull up the OpenAI launch. What's going on? And with Images 2. I've been playing with this for a while. There are some wild, wild examples. They are live streaming. Sam should have saved the Death Star meme for this launch. Point layout into coherent and organized image. And we think we made a lot of progress in both of this visual understanding and visual generation. Both of these aspects. As a result, being able to handle this kind of tasks very well. and we now have an output for this where you can see eight different real cool outfits for me. The level of detail in these models is getting so extreme.

1:09:06It's post-slop. It is post-slop. That's a good point. But I've seen some where people are like generate the entire periodic table with details about each element and a visualization of each element and it's just like so much information, so dense that you have to wind up zooming in so much just to get all the information. It's like the idea of generating like a single photo of a person in an outfit was remarkable. Just a year ago, I guess, when the Studio Ghibli thing happened, and now you can generate layers and layers of detail here. Good teaser. They said this is not a screenshot and posted the image.

1:09:47And detail. don't. Something that prompts, tell everyone about the prompt you've been running with Wiki. Oh yeah. I've been doing this thing where I take a Wikipedia article and I ask ImageGen2 to turn it into an infographic, or you can actually do an Instagram carousel, like 10 images that tell the story of something. I did this with John Ternus and it's remarkable. I mean, the text is is perfectly photo real. The other thing that's interesting is, uh, the, uh, the brand comes through an interesting way. It doesn't, it's not like it has like one style for infographic. Like I had it make a infographic about John Ternus and his career and his path and, uh, everything that he's done at Apple.

1:10:32And it put in images of the projects that he's worked on, but then also had like an Apple like brand aesthetic, like a white background, the correct fonts. Uh, but then I did the same thing for like the Elden Ring movie that we were talking about yesterday. So I went and got the Wikipedia for the Elden Ring movie that has some details and some leaks of who might be in the, uh, in, in playing different roles. And it was able to go get images, go get headshots of those people, put those in because it has tool calls now. So you can actually like, if you say like generate an infographic of John Ternus's career, it will do like a pretty good job generating someone who looks sort of like John Turnus, but then you can actually just take his canonical headshot and say, no, use this exact photo of John Turnus, and it will just drop it right in.

1:11:20And so it looks perfect, because it is just like a copy-paste basically on top of the layers. Let's get some audio again. Let's see what else is going on here. Output. This is particularly useful for very complex prompts, for things that require like web searches, for what require you to output multiple images that have to maintain to maintain coherence with each other. Or even for it to check its work before saying, hey, here's your final output. But let's just look over some examples of this first. Gabe actually kicked off a few of these examples at the start of the live stream. So let's go to the one on the phone, which is the one of him and Sam, the selfie of them, and they created a manga of it.

1:11:56And if we look at the very first image, we can see, yeah, it does look like Gabe and Sam, right? Yeah. But I think what's even cooler about it is that if you look at the follow-up images, they still look like Dave and Sam, and they still look like the style that was originally maintained. Now, children's stories, coloring books. It's like really insane. The story should be very consistent among pages one, two, and three. I did that this weekend. It was very successful. to rave reviews in the Kugan household for AI-generated coloring books for mythical creatures that my son came up with, combining different creatures.

1:12:42We beta tested the instant version of this model on Elam Arena under the codename Duck Tape. A few of you on the internet were really good detectives and deduced that it was us, but we were going to announce that it was us. And so in this prompt, we basically asked that... You got us. basically GPT images too, to go and find social media reactions to this duct tape model and basically quote people. And so we see quotes from threads, LinkedIn, Reddit, etc. But I think an even crazier part is that we've also asked the model to put a QR code to chatgpd.com so that you can try out this model right now for yourselves.

1:13:24And can we just make sure that it works? Yeah, I try. Oh, nice, nice, nice. So image generation with thinking allows you to do really complex things, such as, so in this case, web search, synthesize answers, and put a QR code all in one image. But we have still more, and Alex will talk to you about these new details. It's so interesting. The tool use is getting really, really advanced. So I saw one version of Codex actually was able to generate a vector diagram to make sure that everything was correct and then regenerate the image on top of that. Of course, we should go to the timeline because there's some people that are having fun with the image generation.

1:14:05Someone made presidents in Elden Ring. There's Joe Biden, which is a Dark Souls-style boss that you can fight, and FDR, Lord of the New Deal. And these images, it just looks remarkable if you ever played any of the Elden Ring are Dark Souls games. Because, of course, the text is flawless, and then you can fight Richard Nixon in front of the Watergate. And this looks like a mod that I think people would play if it actually existed. Blake Robbins said, the world is now ready for the rumored OpenAI image model. People are creating Google Street View images that just look perfect, and Grand Theft Auto V loading screens.

1:14:48And there's a big trend of people creating things that look like screenshots of live streams. And then they put themselves in the live streams and have all fun with that. And there's one of Satya Nadella presenting a slide and like even the minor text at the bottom of the slide in the picture, it's sort of remarkable to think that all of this is generated basically one-shotted. Justine Moore was posting about this a long time ago, April 6th, this ad was one-shotted by OpenAI's image model. prompt was literally make an advertisement for the m4 pro uh mac mini and uh i mean you can see how quickly this will speed things up uh the question is just like what is your source content what do you want to uh visualize or or deliver via this format uh the whole meme of like turn this essay into bullet points turn this bullet points into paragraphs uh now you can turn this infographic into text and turn the text back into an infographic.

1:15:49Uh, I, I feel like a lot of slides here for tracking your Uber eats and FedEx deliveries. And it looks, uh, pretty believable. Yeah. We can pull up the, pull up the screenshot here. Yeah. Uh, people, um, there's something, there's something interesting about this where, uh, I I've seen a number of, of, uh, slide decks that could be infographics. And I'm wondering if there's going to be a new level of like compression where people are saying like just send me a screenshot like a one-pager screenshot uh over text instead of like a slide deck that i have to click through that is a lot of information that looks very much like a dashboard but if you zoom in there's like pretty oh it's like pretty aggressive fidelity it's just that's wild um but yeah i mean if you're trying to if you're trying to deliver a whole bunch of information that could go into slide deck condensing it down into an infographic feels like potentially a new trend.

1:16:42And I wouldn't be surprised if you see a lot of these flowing out into Instagram carousels and other sorts of content. If you're trying to summarize some sort of content, a history, quotes, you know, any sort of information like that. Anyway, Tyler Cowen, back to the timeline, Tyler Cowen made an incredible call back in July of 2020 in the depths of COVID in Bloomberg. He said, This year is likely to be remembered for the COVID-19 pandemic and for a significant presidential election. But there is a new contender for the most spectacularly newsworthy happening of 2020, the unveiling of GPT-3. As a very rough description, think of GPT-3 giving computers a facility with words, a faculty, no, a facility with words that they have had with numbers for a long time and with images since about 2012.

1:17:32What a remarkable post. This was two months after Gwern's scaling hypothesis post and two and a half years before ChatGPT was released. That is remarkable. There's a lot of folks that are chiming in with, I called it too. Anyway, I believe we have our next guest in the waiting room, Scott Stevenson from Spellbook. Scott, how are you doing? Doing great. Doing great. How are you guys? We're good. Welcome to the show. How are you doing? Yeah, thanks for having me. Can you, I mean, I want to go into the contracted IRR debate, but let's get the update on Spellbook. How are things going? Where's the company at?

1:18:10How are you feeling? Going very well. We had a killer Q1. Crushed our stretch targets last year. Yeah, we're over 4 ,400 customers on board in 80 countries now. 80 countries? Yeah, we're the most used AI contract review tool in the world. Why so international so early? It's been inbound. Yeah, we've had a ton of interest inbound. So yeah, the choice is accept the customers or turn them away. We chose to accept them. Is the product sort of multilingual by default? I would say yes. A lot of it is driven by AI models. AI models have some ability to deal with different, actually pretty good ability to deal with different languages.

1:18:54And then we're able to supplement the models with legislation and norms from many different jurisdictions. since we're a legal product. Yeah. So, yeah, where else are the key integration points? Like, what's the hard work to, like, bring a country on board or even bring a new flow on board or expand the capabilities of the product as models are just sort of getting better every month by default in the background? Yeah, I mean, I think for us, the bit, you know, we try to be two years ahead of the market and build things that are two years ahead of what anyone else is building in legal AI or elsewhere, and we've consistently done that.

1:19:31we built the very first Gen.AI product for lawyers back in the summer of 2022. This is like before ChatGPT. A lot of like Claude for Word just came out. That was kind of what we launched like four years ago. So we're pretty far ahead from that now. What we really focus on is building unique workflows that are not just chat. I think if you're building something chat shaped, it's very difficult to make that defensible because there's going to be some really good general AI products for just generic chat based work. What we focus on at Spellbook is really rails for high volumes of contracts and contract workflows.

1:20:06So we sell to like Fortune 10, Fortune 100 companies and really companies of all sizes who are processing, you know, hundreds of thousands of contracts, not just with the legal team, but with their sales team, their procurement team. You know, we're in manufacturing, shipping, you know, all of these different verticals. and we kind of build these end-end rails that allow these contracts to move quickly and safely through organizations. And there's a lot that can slip through the cracks when you're dealing with these high volumes of contracts. A lot of mistakes are made. So we give every legal team a second set of eyes on these massive flows of contracts going through the organization.

1:20:44Yeah. What drew you to Cargate? Why are other is it legal AI companies that you that you feel like are getting a little bit dicey on this front or, you know, how do we get here? Yeah, so I've got some interesting examples to cite, but I think it's an enterprise AI problem. And I'll say first, my goal here with this tweet and what I'm doing is to destroy as much equity value as possible by discrediting this obscene metric, C-A-R-R, or at least the way it's being used today. So we can all get back to building real companies. So that's what I'm trying to get out there. Sure. I mean, where this came from is, I think, I just noticed more and more founders and investors telling me things about ARR reporting, you know, mainly the public reporting, but also some of the internal reporting that was just getting more and more skewed.

1:21:43And, yeah, there's all these headlines being published about, you know, ARR records being broken. And, you know, when the laws of physics are being broken, you have to ask, is it AI breaking the laws of physics? Or, you know, might there be some other kind of illusion going on as well? And I think it's a bit of both. We have really high growth, awesome companies being built. But when you have really high growth, you know, issues can kind of fester and hide underneath. underneath. So yeah, I'd heard a lot of more and more stories of people using this metric of C-A-R-R, often using this metric when they're talking to press about their revenue and then gaming it in some pretty obscene ways.

1:22:27So maybe I just tweeted about it. Yeah, the tweet. So you say the setup, company signs three-year enterprise deals. Year one is discounted, say 1 million. Year two steps up two million year three is full price they report three million as they are even though they're only collecting one million dollars this year that's a big deal the worst part the customer has an opt-out option at 12 months it's not actually a three-year contract so they're basically like taking the three-year number pulling it into the present even though it's not a it's not it's a contract that that the customer can can get out of interest um and they're not actually on on the hook.

1:23:05So it's not really, uh, it's rough. Uh, so yeah, yeah, yeah. You just react to that, I guess. Yeah. Yeah. So I think, I think, you know, that's a specific real example that I heard of in the wild from it, from an insider of how this, this, these error metrics were being gamed, um, to create, you know, some, some amazing revenue charts, but I would say there's, you know, a broad category of issues that I can talk about, like a few of them that, you know, after the tweet went viral, I got a huge response of other founders and investors saying that they were saying the same thing and like some other examples of the types of gaming that's going on.

1:23:40Yeah, because if you get one person in a category that starts doing this, then the other people like suddenly have to start reporting the same way and it creates a vicious cycle. Yeah. And it starts and it starts pretty innocent. You know, CARR for folks that don't know CARR is contracted ARR. So it allows you to count revenue that's not live yet. So maybe you're doing like a nine month implementation or you have a one year pilot. For short-term stuff, like, hey, this contract is going, you know, this customer is actually going to be going live next quarter, but we've signed it and we're just going through the implementation process.

1:24:20Yeah, yeah. But there's nothing that, there's no, like, law that says you can't say, like, we're going to extend the sort of, like, timeline dramatically. It just is not a very grounded way to run your business. Exactly, exactly. I think it's innocent. Three months extra credit, arguably useful, but it's a very easy metric to gain, especially if you miss those obligations. I think because we've normalized the forward deployed engineer, which we used to call professional services. Now you have these really complex implementations where you might be promising a customer, hey, we'll build this feature.

1:24:57Once we build this feature, then we'll start billing you. What happens if you don't build that feature? One of the issues you see is companies stacking all these commitments. And so they'll switch on billing once they deliver X with their forward deployed engineers. And then what happens if they miss that or what happens if that gets delayed? And then they're reporting it up front as ARR publicly, but they're not actually at the point where the ARR is live. So, yeah, that's another category of issue. And then there's, you know, people reporting pilots, you know, just three month pilots as ARR and they're free pilots.

1:25:31that, you know, I was talking to an investor yesterday who just sees that all the time from early stage companies, like coming out of accelerators saying they have like a millionaire and they look under the hood and it's just all pilots that haven't converted yet. So there's a host of different, you know, issues with the metric. And then the other one is the step up contract where, yeah, you're stepping up, you know, year one is, you know, 25 % of the cost, year two is a little higher, year three is higher. And then people are either amortizing that back over the period to get a higher average or even taking like that year three amount, like you said at the beginning.

1:26:05So yeah, there's a bunch of patterns that are happening. The other thing is like there's early opt-outs. So like, you know, you can have early opt-outs in these long-term contracts. And, but there's all, I mean, we're a contract company. So there's a million ways that a contract can be terminated. Seen a few contracts. Yeah. Yeah. Yeah. So yeah, I think, I think it's It's a really ungrounded metric and people should stop using it to report their enterprise AI companies should stop using it to report their ARR publicly. I think no one should take it seriously, except maybe internally for some projections.

1:26:40You know, it's not a good metric. What is the gold star example of using ARR correctly? Because it's very easy once a company is public to just say, okay, let's just go off of, you know, gap revenue for the year. And like, what did you actually book this quarter? There's a whole revenue recognition policy. It feels like there's some benefit to tracking ARR month by month if you're a high growth startup. But what is the best? Companies should just report their daily annualized run rate or hourly. Yeah. And that goes into the debate of annualized versus annual, right? So how have you processed sort of the better cases?

1:27:21or like what is the responsible way to report a revenue metric in 2026? Yeah, I mean, I think it depends on the company and the shape of the company and whether it's usage-based billing or seat-based billing, which you still have lots of both. Definitely don't do like hourly, you know, annualized run rate based on the hour. That's not good. I think the main thing I would say is it should be live. It's like what revenue is actually live right now for like what customers are you actually billing and are actually paying you. So calculating run rate based on, you know, the month of revenue that you have coming from customers that are actually paying you that you're actually billing.

1:28:03I think that's OK. I think, you know, annual recurring revenue based on live customers that you're actually billing that are actually using your service. I think that's pretty good. I think once you start stretching into people who will pay you or, you know, might pay you. um that's where things start to i mean it can just be so easily gamed and anything that can be gamed will be gamed yeah yeah are you optimistic that anything will change or do we need to see a massive correction and uh and a dark and a dark the dark ages like uh i mean post 2022 um i i mean i would like to see a steep correction and then back back to building you know we'll see if we can make that happen.

1:28:44You know, my reach is only so far, but, you know, I've spoken to a lot of reporters in the past, like 48 hours who are like, I'm always going to ask now, like when the company tells me their ARR, are they talking live error? Are they talking, you know, this like long term committed error that might come? So I hope, you know, at least the journalists are going to be a little bit more savvy and ask more questions before they report on these numbers. Yeah. I want to ask about who suffers. But in terms of ARR, like, yeah, there's almost something where you should just report your last month's revenue instead of doing the times 12 thing.

1:29:22And then if people want to multiply it by 12, they can. But at least you're just reporting, hey, last month, this is what the Stripe account did. You can also just say by Q3, we will be at X ARR. Exactly. That's a different way of saying that's better than saying we are at 10 million CARR. Yeah, yeah. Exactly. Much better. Who suffers here? Is it purely investors? Because I feel like a good venture capitalist, their job is to dig into the contracts during due diligence to set prices. And if they want to pay, you know, a thousand times ARR because they think it's a hundred times CARR, like that's their risk profile.

1:30:04I would maybe be careful, but that's their job. Or is there a risk that employees see a headline number and think that the business is more stable than it is and they join and then they're rugged? How do you think this affects the – who needs to watch out for this basically? Yeah, I mean I think investors are generally – good investors are generally very aware of the difference between CAER and ARR and aware of the widening gap between these metrics. and like in most board decks, you see two metrics. On the press, you only, you know, you usually see CRR, but it's called ARR. In a board deck, you see both metrics.

1:30:40So, you know, investors are quite aware, but I don't think it's victimless at all. I think, yeah, employees are signing up for companies. And as you know, in like a high growth startup, people are committing a ton of blood and sweat to be successful based on, you know, part of it is based on the growth of their equity. And if it turns, and they might think they're multiple, you know, they might read the headline number. They may not know the number that's actually in the board deck. They might read the headline ARR in the press release, and they might base their decision to join a company based on these headline revenue numbers, which are really not grounded in reality whatsoever.

1:31:15And by that, I mean I have literal examples, confirmed examples of the press number being three to five times higher than the actual live ARR number. So that's a huge difference. If you think about a multiple. Yeah, yeah, yeah. If you're choosing between a public company that's trading at 10x revenue multiple or something and then you get your offer from a company and it seems like they're at 10x, but they're actually at 50x. That is very material for how you should think about valuing that stock that you're working for. Makes a ton of sense. There's the customers. Customers are trying to figure out which company is most mature or least mature and then there's the whole competitive landscape.

1:31:56It's like if one person, if one company starts doing this, all companies have to start doing it. And it just creates. Yeah, I mean, we could start doing contracted viewership. So we can sign three year deals with people in the audience that requires them to tune in to the show every year. They have an opt out after a month. If they don't like it after a month. You can still advertising based on contracted viewership. Exactly. I'll sign the contract for the rest of the year. I'll watch every day, every hour. I mean, I mean, I guess that sort of does happen for YouTube channels that I mean, no one really does this But there was a time when YouTube channels were sort of valued on like the subscriber number as opposed to the average view number And of course there are some channels where every video gets a million views and they only have a hundred thousand subscribers for whatever reason And then there's vice versa where someone's been doing you see this on like old legacy media accounts on X where they will have like 30 million followers and then the post will get like three likes and it's like those are two wildly different metrics like that happens all the time but this is the name this is the name of the game in silicon valley the metrics game everyone's finding an edge somewhere well uh thanks for keeping everyone honest and uh good reporting good luck fighting the good fight out there and spreading the good word don't don't uh don't get too sucked into all this you're you can take we promise you can come back on in three years with your honest ARR and take a good victory lap become an investigative journalist pivot to investigative journalism blow the doors wide open on this wow this goes deeper than I thought good to see you have a good one we'll talk to you soon up next we have Alex from Osmo building olfactory intelligence.

1:33:46We've talked about this before. Can AI smell? That's our current benchmark for AGI. We say if, you know, we talk about white collar work, we see sommeliers as white collar workers. Unless you can smell, it's not AGI and artificial intelligence falling short. But it's your first time on the show. I would love an introduction on yourself and the company because I'm fascinated by this topic. Let's talk about it. Your sommelier comment and also what you talked about with Max Hodak spit on my mind. Amazing. My name is Alex Gulchko. I'm founder and CEO of Osmo. We're giving computers a sense of smell.

1:34:23I've been working on this problem for 20, exactly 20 years or so. First as an academic. Did my PhD in olfactory neuroscience at Harvard and trained under Bob Datta, who trained with Richard Axel. Got the Nobel Prize for discovering the receptors of smell. and my AI mentor trained with Jeff Hinton who got the Nobel Prize for Deep Learning. And I'm the one weirdo that's like... That's incredible. We've been waiting for you to join for the entire history of the show. There's been great prophecies of your arrival for hundreds of years. I'm so excited. I'm so pumped to be here. So should we start with maybe like olfactory science 101?

1:35:04Can you set the ground on like how does smell even work? What are the important sort of like building blocks that we should know? And then we can build up to the next generation and how AI is being applied. 100%. So the chemical slice of reality, all the stuff that's data in the air, we can detect that. Our sense of smell is literally our brain leaving our skull. So when you smell a molecule, whether it's a tree or it's a meal or a drink, like the physical pieces of that thing enter into your nose and touch a piece of tissue about the size of a postage stamp. And that's your brain, right? So like you were in physical communion with that thing.

1:35:41That information gets turned into neural data, which actually skips all of the normal way stations for the other senses and goes right to your centers of memory, the hippocampus and emotion, the amygdala. So our sense of smell is very primal in that regard. So it's like, it's the reason why when you smell something, you get dragged into a memory and you cannot stop it. You're just like back in high school or you're back as a kid is because we're physically wired for that. So that's real. I've always heard that. I've always heard that phrase. Smell is the sense that's most tied to memory. But I didn't know if it was just something you saw in like a t-shirt or something.

1:36:14No, literally neuroanatomically target wall art. Yeah. Yeah. It feels like target wall art. I don't know. It's just one of those things that you repeat. They say a smell is a thousand words. Okay. So that sounds like something that's extremely hard to reverse engineer. Do we have, Do we have sensors? Because, you know, LLMs, it was so obvious that we had text that was already encoded into data, into ones and zeros. And so transforming that and encoding it, I mean, it was an incredible breakthrough. But it felt like the text was, the data was already in the computer. And I feel like that's not true for olfactory data, for smell data.

1:36:54But how are we, do we need to digitize this before we do anything with it? How does digitization of smell work? Yeah, great question. So I was very fortunate to have those guys as my colleagues. I actually spun Osmo out of Google Brain. And so I was there when all that stuff got invented. And I ran the digital action team at Google Brain for about six years before we decided to make it a company through Lux and through GV. And you have it exactly right. The internet had been accumulating for a while. So we had all this text data. So we could basically slurp that down and start building models.

1:37:28We have chemical sensors. They're called mass spectrometers. There's other kinds of chemical sensors, mock sensors. There's like a dozen. The history of sensors that can turn chemistry into data is about 100 years old, maybe more. I mean, a lot of it was pushed forward in the Manhattan Project, actually. but what we've been missing is a map right so for sound low to high frequency is a map which lets us build mp3 and speakers and microphones and spotify etc and for color rgb is a three-dimensional map of color and that lets us build cmos ccd you know cameras etc we haven't had the map for smell and that's not crazy because there's three channels of color information in our eye but we know there's over 300 channels of information in our nose so in a way we actually did need to wait for artificial intelligence to mature in order to have the ability to extract a 300 dimensional map from data.

1:38:18And that's exactly what we did starting with our first work at Google Brain. So you got to go get a crap ton of information, right? A bunch of molecules, what they smell like. We've since collected the largest AI data set for a scent in the world. That's what drives olfactory intelligence. We have 5 million sniffs digitized, over a quarter million physical samples created we've digitized about six billion uh fragrance molecules so like all this is like inside of the company because there's literally nothing on the internet the fragrance industry has done a phenomenal job keeping everything secret so we built it all ourselves remarkable jordy how are you going to make money on that it's a good question if if if you go actually let's talk how can we make money on this so does tbpn have a scent yes and it's terrible terrible There's rubber smell in the studio.

1:39:07So in the studio, there's like thousands of cords. And cables. And the cables smell. We do a good job hiding them. But we have so much gear going everywhere. It's a lot of rubber, a lot of plastics. And so we had to get these. Racetracks. Racetracks, they're called, to cover all the cables. And it turns out these things smell terrible. A lot. It was off-gast as well. Yeah, they're off-gast. So we wanted to give our viewers the full CBP experience. I think we absolutely do not. It would be like a can that sits on their desk, aerosol, and it would just spray a rubber smell into the room so they could experience what we experienced.

1:39:44We could capture it, but I think we should fix it. So really concretely, we raised our Series B. We put an additional$70 million in the bank with Two Sigma leading Lux. Love it. Got the gong. That was to underwrite building a fragrance factory. So we have a robot that's the size of a school bus that makes a new fragrance every 100 seconds. And what we do is we design and manufacture fragrances for brands. Oh, yeah, that makes sense. And so we use olfactory intelligence to design it. So super fast, data-driven, basically perfect fit for the brand and for the consumer of that brand. And then we actually physically make it.

1:40:20And what leaves our factory is a steel drum of that fragrance while we bill them for it. We also will do end-to-end. So if you want to actually make a physical bottle, we'll actually put the fragrance in the bottle for you so the full product comes out. So if you guys want to launch a TVPN or something like that, we could design it for you. I mean, like if you tell me the prompt right now. It should smell like burnt rubber. No, no, no. Good smell. We're not doing burnt rubber. Rubber. There's some, it smells like disagreement. No, it needs to smell like old$20 bills. Okay. From the 1980s. That is a thing.

1:40:50And mahogany, the official wood of business. Mahogany. We need it to smell like mahogany mixed with old$20 bills, the smell of money. That's a good one. Okay, cool. We've done a Smell of Money one, which we demoed actually on the New York Stock Exchange floor, which is pretty cool. That's amazing. That's amazing. But no, I'll send you, we'll make something. Talk about sensor miniaturization. My phone has three cameras and no smelling sensor. Can we swap one of these out? When you say mass spec, I imagine like a device the size of a living room. I imagine that they are getting smaller. The dishwasher.

1:41:28Dishwasher. Size of the dishwasher. is there a path to actually shrinking that down to something that's more portable? So, yeah, in the same, there's like many kinds of cameras, right? So the one in the Hubble telescope, not getting smaller. So if you need resolution, it's going to be big, but you can make trade-offs. And like when the thing that's reading the data, instead of it being a person, it's an algorithm, you can actually make really intelligent trade-offs, which is what we've done. So we actually have a sensor right now, it's the size of two shoe boxes. And I kind of use that metric aptly because we've actually used it to smell fake shoes.

1:42:00So if you're buying a pair of like$500 Air Jordans, the real smell different from the fakes, we can actually pick that up. That's crazy. It actually turns out we can. Yeah. The counterfeiters use cheaper glues turns out. And the other thing that's interesting is we can actually tell the factory of origin of the shoe 93 % of the time. So the smell is a fingerprint. So we're already miniaturizing these devices. look the path to get from two shoe boxes to one shoe box is pretty clear yeah we're working on that um to go to something that's like the size of the airpods case there's going to be some like hardcore engineering required to have it be a component that fits in your phone there's some breakthroughs like i can't quite see through the fog yet but there's nothing like look our noses do it so there's nothing that mother nature is saying i think impossible but we just got a lot of work to do yeah that makes a lot of sense uh what about taste how closely is taste linked talk Walk me through the sommelier example.

1:42:54Yeah, so laver is everything that happens in your mouth, that's a sensory experience of food. Taste is not as like 10 % of that. It's like what happens on your tongue, right? You ever eat a jelly bean and like plug your nose? Yep. And you just actually can detect very little of what's going on there. It's because 90 % of what you experience is actually called retronasal olfaction, where when you're biting on something, there's a chimney effect in the kind of, almost the steam of what you're eating goes back through your nose and you smell it. Oh, excuse me. And then there's also the texture and everything in your mouth.

1:43:27So we've done tests and our OI models, this is from a while ago, we haven't revisited it, but we're really focused on fragrance right now. But our OI models actually work on flavor surprisingly well. And so the whole world of flavor is there for us with Moretti, but we're really focused on this particular business. I've seen a couple of these sort of, I don't want to call them niche, but like vertical AI projects that are not fully generalizable. There's a DNA model also from Google or DeepMind. And it feels like they're starting to get on scaling curves, on scaling laws. Are you at a point where you feel like, oh, if I 10x the computer, 100x the compute that goes into some of the world?

1:44:12I believe Alex is ready for a one gigawatt data center. He can be trusted with me. I would trust you with me. But how universal do you think scaling laws are? Is there a scaling law here? Is it data-based? Is it compute-based, both? How do you think about it? The better lesson's real. The better lesson's super real. I always think about technology as S-curves, right? And like what's driving you up that S-curve and then how can you hop on the next one? Our current S-curve is data, which is why we're maniacally focused on like generating a ton of data. Like we have a giant fragrance robot that spits out a ton of fragrances.

1:44:46We have mass specs running 24-7. We have sensory panels, both domestically, of a building of people that just smell all day abroad. And we ship them crates of stuff to smell. And that's how we get to 5 million sniffs, right? So data, data, data. The size of the models is not the limiting factor right now. And it will be at some point. And then switch to the other S-curve. Yeah, because you don't just have the open internet to scrape because there's not an existing data set. Makes sense. Totally, double-edged sword, right? So we've had to make it all, right? Which is really hard. But also, nobody else has it because we had to make it all and had to learn a ton of stuff in order to do that at scale and efficiently and all that stuff.

1:45:22Yeah. Where is the business today? I mean, you've raised money. It seems like there's, you know, monetization opportunities for sure. Are you fully in commercialization? Are you still in research? Is it half and half? Like, how do you think about raising more money over time and just growing the business? Yeah, so we're always going, like we started with like this curiosity-driven drive to figure out how to digitize smell, which is like a pretty wacky thing to do. So we're always going to be trying to push the edge here. But look, we have a factory. We manufacture fragrance for brands. We did this commercial kind of R &D to commercial transition last summer.

1:46:00And we're kind of almost at the end of that. And we built a manufacturing organization. We built a sales organization. We have some really amazing partnerships with some big brands. and we're making fragrances for brands. You can go into Target and buy a product that has our fragrance in it today. And so we're scaling this part of our business. We're still placing bets on the future though, right? So I think we've got really the tiger by the tail in this, it's a whole other conversation. Sometimes you should come to the factory in New Jersey and see how it operates. But like the fragrance industry is wild.

1:46:29We've got a lot of work to do there, a lot of opportunities. So we're focused on that. Amazing. Well, congratulations and thank you for the work that you do. We think it's so important. and one of the most interesting companies we've ever uh we've ever learned about on the show yeah true science awesome i love it we're trying to make science fiction into science fact but like open invitation to come see how it all gets made it's pretty crazy in person so come to the willy wonka chocolate factory where all this stuff happens i would love to we'd love to thanks so much so great to meet you come back come back on soon yeah we'll talk to you soon have a good rest of your day uh we're running a little bit behind but up next we have spiros from resolve ai AI raising a massive round to build AI that runs production systems.

1:47:11Let's bring in Spiros. How are you doing? Hello, guys. Good to be here. Welcome to the show. Sorry we're running a little bit late. Kick us off with an introduction on yourself and the company. I'm one of the founders and the CEO of Resolve AI. We're building agents that can help you debug and run production. Think of it as the counterpart to coding agents that produce all this code, and our agents are there to support you. Okay. Is your customer always deeply in the throes of Vibe coding, has rolled out agentic coding across many organizations? Who is the target customer? Do they have to already be deep in the agentic coding wave to really get the value here?

1:47:56They don't have to, but the two are correlated. Anybody who runs a large software system has this problem. The only solution we've had so far is humans manually solving it, right? Using the tools, being on call. Of course, now AI allows us to automate all of this. But I would say, this is true, it was true before. Now with all the AI generated code, it becomes a necessity, right? So we see strong correlation between the two often. Yeah. And what are customers coming to you asking? Is it, I want the code that's written, we're writing way more lines of code, we want it to be more readable, or we want it to be more secure or we want to be more performant or all of the above?

1:48:32The way to think about it is like for anybody who's delivering their business through software, look at some of our customers, Coinbase, Salesforce, MongoDB, right? To them, reliability is of paramount importance. If anything goes wrong and affects customers, it's a big problem. So Resolve becomes essentially the first level of defense that captures any problem that happens in production that can affect end users, gives you a resolution and a fix, let's so you can accelerate that loop, right? And it doesn't take too much human effort, but more importantly, it doesn't cause impact to customers.

1:49:05What is, like, I mean, the company is now over$1.5 billion in valuation. What has been, like, the key to growth? Is it just product-led growth? Do you have a big sales team? How are you actually scaling the business as you scale the valuation? Yeah. Yeah. So this is a very big problem, right? Anybody who has, as I said, delivered their business or software is facing this issue. Yeah. And whether you're a CTO, you know, who pays for, let's say, developers to focus on reliability, or whether you're an individual that has to solve this problem, you'd rather have AI do it for you. Yeah. So we've seen, like, a huge amount of demand from day one since we launched the company a bit more than a year ago.

1:49:47Yeah. And we've seen it coming from both big and small companies. We primarily focused on larger enterprises because we think there is a lot more complexity, given the complexity of the software. And most of the growth, I guess most of the demand comes inbound to us because it's a well-understood problem. And of course, we have both a product-led approach, let's say, but also a sales-led approach as we work with large customers. Yeah. In some ways, the naive approach would be, okay, just appoint a typical AI agent at the code base and just tell me where the fault lines are. But I imagine there's some special sauce in the engineering to understand knock-on effects that can happen across a large code base.

1:50:30Are you actively working around context windows or creating a special harness to understand these problems that can come up before they do? Yes. So think of it like we have a production ID, basically. The same thing you have for your code, we have it for all your production systems. production involves code involves let's say telemetry logs metrics tools like Datadog Splunk it involves AWS so you have to deal with all of these not just code and then we also are training our own models now to improve let's say the state of the art let's say you can go far enough let's say with a good harness and a lot of work let's say on the agendic front but now and we just announced together we're funding that we're building a lab to focus on actually training our own models for this domain.

1:51:20Sure, sure.

1:51:24What goes into getting relevant data or actually nailing a specific model for this? Because I imagine that you have some great clients. They probably don't want you training on their data. At the same time, if you just grab some open source code, it might not be as complex as the Coinbase monorepo or whatever they have going on over there. So how do you actually create enough training data to justify a special model? What is important here to understand is the training doesn't happen on code per se. What happens on is actually the action a human takes to perform a task for the most part. And we're talking about very long kind of, let's say, planning tasks here.

1:52:08It might take many, many iterations, looking at code, looking at Datadog, looking at infrastructure. And generally speaking, this is not in a training set of models. And software, let's say, generally is both a deep and wide domain. So I think if you actually focus on building a model for the types of problems we're trying to automate and how you run and debug production, I think you can have a lot of gains, both in performance, cost, but even quality of outcomes. And that's our goal. And I would say the big labs make it sound, make it look like it's impossible for anyone else to build a model, but I don't think that's the case.

1:52:40and that's what we're seeing ourselves with our investments. Yeah. How do you put together such a low dilution round? Yeah, tell us about the round. I want to hit the gong. The 40 on one and a half billion. So it is an extension. We just did essentially the A. We just did the A at a billion dollars like two months ago, right? And I would say resolve. There we go. Sorry, continue. There's always, essentially, in many ways, created this market, right? Like AI for production. Sure. And I think it's well understood by investors. It's also proven given the customers we have. Yeah. So, and we're also a very ambitious company, right?

1:53:20Like, we are obviously trying to build the agents and the models for this domain. And we have a lot of traction. So, I mean, as simple as that, right? Like, there's nothing you can do to create a load that looks around other than be very successful, in my opinion, these days. That's a great answer. That's a great answer. Step one, be successful. I love it. Or like, step one, focus on building a business. Like this is my first startup as a founder. I made this mistake many times before, right? Of thinking that raising money is success. It's not, it follows real success on a product. Yep, yeah, no, that's 100 % right, I love it.

1:53:54Well, thank you so much. Congratulations on the new round. Yeah, great having you on. And great having you on the chat. Congrats to the team, excited to watch you guys go. Thank you. We'll talk to you soon. Have a good day. Goodbye. Up next, we have Carolina Aguilar from InBrain Neuroelectronics, building the first inhuman study of graphene brain interfaces. What's going on? Welcome to the show. How are you? Thank you. Very good. Thank you for having me. Please, since it's the first time on the show, introduce yourself and the company a little bit. Yes. My name is Carolina Aguilar. A lot of people call me Carola.

1:54:32I am the CEO and the co-founder of InBrain Neuroelectronics. And we are a graphene-based brain-computer interface therapeutics company that actually is developing the most intelligent interface between the neural system and AI to restore health for billions. Okay. So walk me through brain-computer interfaces and the decision tree that got you to graphene specifically. I'm familiar with, like, the first decision is probably invasive versus non-invasive. We've talked to a number of founders that have taken either approach. How did you confront that first question? Yes. Well, I call them implantable and non-implantable systems.

1:55:19And in our case, we're an implantable company. We believe that the real signal processing that is going within the neural system is actually deeper in the brain. and to listen carefully to what it says, what the neural system says, and being able to decode it but also modulate it. We need to be close to those neurons and interact with those neurons firsthand. Okay. So when I hear modulate, it sounds like not only processing information that's coming out of the brain but also potentially writing information back into the brain. Is that the long-term vision? and this is the magic of graphene is actually about reading and writing very effectively at micrometric precision within the brain i think that's why we took let's say higher risk to get an advanced material into this funnel because we see that the benefit is is incredibly impactful okay what what are what are the most near-term commercial applications yeah so the the morgan Stanley report stated the market in 400 billion.

1:56:29And we thought that we needed to bring a platform with three product verticals to actually penetrate such a big market. So we are creating three products. One is, let's say, not implantable, actually. So it's a semi-chronic platform, kind of like the modern Utah array. It's like 100 contacts of graphene that can read and write. We went into tumor and epilepsy resection at the beginning, and that one is pretty close to commercialization. We're almost there. The second product is the implantable platform for the brain. So this is a implant on the brain for Parkinson's disease. So we didn't do, sorry, assistive BCI because we saw a 1.8 billion market that is suboptimal that we could actually displace very easily with this technology.

1:57:29So we decided to go therapeutics into Parkinson's. And the third one is the same platform, but instead of a... So we connect, let's say, brain sensor, we connect a vagus nerve sensor that is actually able to decode all the fibers that go into the different organs. So we have a therapeutic target for each of the organs just by targeting that nerve in the neck. What about the actual implantation process? We followed from Neuralink. They had to build a whole robot just to drill into the skull. It's incredibly high precision. Are surgeons capable of implanting this at this stage, or will there need to be other robotic devices that are developed to actually deploy this technology safely?

1:58:16It's an excellent question. I'm coming from Medtronic. I spent 10 years in neuromodulation and another three in diabetes. And I think in the future, when micro robotics are ready, we will have a very close relationship between our interfaces and micro robots that probably can deliver this implantation in 30 minutes. But today, when there is not micro robots and we are not Elon Musk, we decided to actually have our platform ready for the current surgical workflows that today exist. So we are not changing much from the neuromodulation workflows. And it's an easy procedure. Two hours, one or two hours, you know, is enough.

1:59:05In the case of the neck is 45 minutes. Wow. Wow. That's very impressive. Well, congratulations on all the progress, and thank you for the work that you do, and thank you for stopping by the show. Yeah, great to meet you. Have a great rest of your day. Cheers. We'll talk to you soon. Bye-bye. Up next, we have Jake from Blue Energy. He's the co-founder and CEO with a massive raise. It's a gong breaker. Jake, how are you doing? Welcome to the show. Hey, guys. I'm doing well. I got a feeling you have the biggest number for us. I think you have the biggest number. Kick us off. How much have you raised, and then we'll get into what you're going to do with it.

1:59:39but tell us about the financial situation for the company. We have announced a$380 million raise. Okay, go ahead.

1:59:54And what will you be doing with all that money? Yeah, so our focus, we're really unique amongst the field of nuclear players right now. Our focus is on building the world's first project financeable nuclear power plant. So we're using this funding to actually put deposits down on long lead equipment, as well as finishing out the engineering and development licensing on some of our first sites. Does that mean like less R &D risk or more in like the GE or Westinghouse territory, more like, you know, going with products that have been de-risked, but it's very expensive and maybe the underwriting is different this time around?

2:00:34Or are you working on an entirely new reactor design somewhere in the supply chain? No, you're exactly right. We are not a reactor designer. We're a developer, but our technology is the proprietary approach by which we go about building the plant. So what bugged me and why I started the company was I had a lot of friends in the nuclear space designing really exciting new reactors. And then you have a lot of incumbents in the space who are working on kind of the same old technology that we've been operating safely for 70 years. But NOBIA was focused on the core issue of how do we build nuclear on time and on budget?

2:01:11So I grew up in a construction family. I used to be a draftsman from my father's architecture firm. So I just grew up around a lot of construction, did my nuclear engineering and physics degrees in the space and just felt like there hasn't been anyone really focused on the root cause issue. So what we're doing is we're borrowing best practices from LNG and offshore oil and gas and offshore wind to prefabricate everything at existing oil and gas fab yards and shipyards. And then we barge it all as a prefabricated system on the order of 1 ,000 or 2 ,000 tons to the operating site. And then basically we're just installing it like giant Lego pieces.

2:01:48But what that allows us to do is bring a lot more debt financing to bear. So we're not taking a lot of reactor technology risk to start in the beginning. We're using mature light water reactor technology to start, but we'll happily work with the Gen 4 reactors as they mature. Cool. Take me through. I feel like you've been wanting a company like this. Yeah, I've been wanting this for a long time. John's point has been, hey, we know how to do this. It works. We don't need to reinvent. I mean, it's great if we want to reinvent the wheel. Yeah, we need the next gen tech. new reactors, but in the meantime, let's just build some.

2:02:23Yeah. So my question is about Vogel, lessons from the Vogel project. What do you think they did well that you want to copy, that you want to learn from? What, if anything, do you want to do differently? Yeah. So to put some stats on Vogel, it, like so many other nuclear projects in the West, ended up being about two to three times over budget and behind schedule. But when you double click on where that cost was, you realize it wasn't actually in the reactor technology or the equipment. It was over 40 % of it was just the construction overhead. So it was the cost of training and relocating 10 ,000 skilled workers to the site at Vogel.

2:03:05Think about the cost of training, relocating their families, retaining them once they're trained, because data center projects are trying to steal that talent. You have to set up a nuclear quality assurance program in the field. So there's all this overhead. It's basically like building a small town. And then the hope is that you'd be able to move that traveling circus around from site to site. And then a third of the project costs was just capitalized interest on debt because it took over 10 years to build before it started generating revenue. So these are the two big problems we're trying to address is we are moving most of that work off site.

2:03:38We're keeping the workforce centralized at the fab yard and the shipyard where they already are. So we can start to put nuclear into a learning curve and drive the cost down over time akin to what we've seen in wind, solar, batteries, and gas turbines. What they did well was it was a mature light water reactor technology. It's a passively safe reactor technology. They really pushed the world forward a little bit in the licensing space and steel composite structures, which we're looking into as well. They actually had, this is not well known, they originally wanted to barge it in up the savannah river but they had to then dredge the savannah they would have had to dredge the savannah river for miles and that would have become part of their environmental impact statement so they ended up it became such a regulatory and permitting nightmare that they gave up and they said all right let's just truck it in and then they had to truck it in and build a module assembly building on site so they ended up doing all the welding on site yeah it's like a remodel from nightmare remodel.

2:04:37Very interesting. Yeah, I'm fascinated. Where's the company based? Where are you guys based? We're in Chevy Chase, Maryland, pretty close to NRC headquarters. And then we've also got a big presence in Edinburgh, Scotland, where there's a lot of offshore engineering talent, particularly from the history of shipbuilding and offshore oil and gas. And we've also got an office in Houston, also kind of offshore oil and gas capital world. Okay. I read a blog post about one of the potential problems or stumbling blocks that nuclear projects run into. And I want to reality check it with you. The thesis was basically that we have a reactor design.

2:05:21We have a, you know, there's infrastructure around that. Cement needs to get laid. Pipes need to go here and there. But oftentimes the regulation will change while the project is underway. And so in order to stay ahead of the changing regulation, you might have to jackhammer a bunch of concrete and move a pipe to a different route because the regulation has changed. Is that real? And is there anything that we have done or can do to get to a regime where the regulation is more locked and deterministic? So that I imagine you're not going to build this overnight. But if it takes you a couple of years, you know that the contracts that you put in place, the plans that you put in place today will hold and you won't face a massive delay.

2:06:09Yeah, so that was another one of the big learnings from Vogel. Vogel was the first and still today, I think, the only project that did a combined operating and construction license, which means once they locked the blueprint, they really were not allowed to make changes to it during construction. So every time they encountered something and said, oh, we need, you know, the craft labor wound the rebar clockwise instead of counterclockwise, you know, they had to jackhammer it up or they had to go and re-approve it all. And there's just a there's a long list of things like that that they encountered.

2:06:41So one of the things we're doing is we're following a slightly different licensing process, the original licensing process of part 50, whereby we're going to incrementalize it so that we don't have to encounter that rework, that kind of regulatory triggered rework situation. But also because we're following this prefab approach and we're moving 80 % of the capex into a fixed price contract environment with these fab yards and shipyards, it forces us to go to something like 60 % detailed design up front and locking those designs because that is what is going to be coming in prefab from the fab yard.

2:07:16So it's sort of baked into the strategy. But really, this is about taking a lot of those lessons learned and making sure we don't make the mistakes of the past. But I'll also say we've never had a more supportive regulatory environment than we have for right now. This is a critical juncture in the history of nuclear power that we can take advantage of with where the NRC is presently at. What is the power output for the first reactor that you're targeting to bring online? So we are focused right now for the first project using light water small MOSA reactors, which the power range of the units we're looking at range between 50 and roughly 300 megawatts per unit.

2:07:56Each site we're targeting doing is going to have multiple units. So it's going to be a gigawatt to a gigawatt and a half per site because multi-unit operations is important. And it helps drive down costs. And then are you already sharing a timeline? Do you have an optimistic scenario, a base case, a bear case? I'm sure you get asked this all the time. It's the worst question. But we talked to a lot of nuclear founders and we hear a lot of 2030s. And there's a whole bunch of projects with hyperscalers and big tech companies that are looking at 2032, 2035. It's exciting. Better, you know, 2032 than never.

2:08:35But do you have anything to share on like the timeline of rolling out new nuclear capacity in America? Yeah, we'll be announcing things very soon. but what I can share on the dates and this is actually another unique thing we're doing part of our strategy is we're pursuing this thing we call gas to nuclear conversion so we're actually going to be building half the nuclear plant right away so the whole nuclear steam turbine system set up for nuclear steam conditions and quality and we're going to fire it early with two combustion turbines as a two on one combined cycle so it's kind of a Frankenstein combined cycle we've actually gotten the NRC to buy off on this methodology through a top core port recently.

2:09:16So that allows us to actually project finance half the CapEx for a first SMR. And then we will build the reactor and splice in the steam and switch it over from gas steam to nuclear steam. So that actually accelerates our commercial operation date with confidence. So we're looking at generating first power in 2030, 2031, mostly driven by gas turbine delivery dates today. And then our first nuclear commercial operation, we're looking at 2032. Okay, so is that switch out possible at any legacy natural gas infrastructure site in America currently? Because that seems like an environmentalist dream, right?

2:09:55I'm not ready to say yes or no. Yes, I think this opens up a whole new world of fossil to nuclear convergence, which we think is an important precedent to set. Yeah, it seems huge, if possible. But I imagine that, you know, it's not exactly USBC on both sides, but hopefully one day we can build the adapter. That's right. What we're really focused on is, there's a lot of announcements out there, a lot of sometimes noise. There's a lot of exciting things happening in the nuclear sector. We think we've got the first project financeable nuclear project. And the first one that's going to power, it'll be a new build that powers a new AI data center.

2:10:34so we're excited about that and we think our timeline is credible, is aggressive but credible and defendable and we've got the right set of partners around it to make it happen. That's amazing don't use the word data center, we're using the word supercomputer. Supercomputer. They're supercomputers they're supercomputers now. No one likes data center, everyone likes a supercomputer. Yeah Anyway, thank you so much for taking the time to come chat with us. Great to meet you Jake, I'm sure you'll be back on soon. Good luck with everything We really appreciate your approach, it feels like you're making plays and And I like how pragmatic and innovative the approach is at the same time.

2:11:08It's great stuff. Thank you. Appreciate your time. We'll talk to you soon. Cheers. Goodbye. And we will close out on this. I need your reaction, John. Ferrari's first electric car is priced at the low price of$650 ,000. An absolute steal. They're giving them away at that price. It's electric, right? So you don't have to deal with gasoline? You don't have to deal with it. Yeah, you save a lot on gas. You don't have to deal with all the noise that comes out of a V12. Yeah, no noise and you save on gas. Yeah. Oh, well, yeah, I mean, if you're saving on gas and you're driving, I mean, if oil keeps spiking and gasoline goes to$1 ,000 a gallon and you're filling up every week, you could easily be spending millions of dollars a year on gasoline in a normal car.

2:11:56So there's potential cost savings here. So it's sort of a more you buy, the more you save situation. I think Ferrari Luce. There's really going to be a, like, what kind of Ferrari client are you moment. No, I think it'll be a status symbol. Because if you see someone with this, you will know. You know they can eat 300 grand of depreciation. That and you know that they are high on the list for the F90. Like, they are working their way up, buying in now. And when the F90 comes out in a decade, they're getting a call. Yeah. They're getting a call. For sure. Yeah, very interested to see how this does in the market.

2:12:31And the good thing is if you are excited about the Luce, but you're not excited about paying$650 ,000, you will have an opportunity to buy them for far less than that very, very quickly. Probably. Probably. But thank you for hanging out with us today. Thank you for tuning in. It's been an honor and a privilege to be here with you. We hope you have a wonderful afternoon. Leave us five stars on Apple Podcasts and Spotify. Throw that flashbang. Sign up for our newsletter, tbbn.com. Goodbye.