TBPN is a live tech talk show hosted by John Coogan and Jordi Hays, streaming weekdays from 11–2 PT on X and YouTube, with full episodes posted to Spotify immediately after airing.
Described by The New York Times as “Silicon Valley’s newest obsession,” TBPN has interviewed Mark Zuckerberg, Sam Altman, Mark Cuban, and Satya Nadella. Diet TBPN delivers the best moments from each episode in under 30 minutes.
You're watching TBPN.
Speaker 2:Today is Wednesday, 11/19/2026. We're back at TomorrowX day two.
Speaker 1:We got Lip Bu Lip Bu stage here.
Speaker 2:On the stage. John Gray coming on the show in just a couple minutes. He's gonna hop on very, very quickly. And then we got Travis Kalanick coming back. We also have Scott Wu, Jed Finn, Scott DeBoer.
Speaker 2:Fun show. Bunch of back to back interviews.
Speaker 1:A lot of goes.
Speaker 2:Day yesterday. Big day for us. 11 interviews, something like that. We really pushed it to the limit.
Speaker 1:No bad. Not as hard as Gavin Baker.
Speaker 2:Yeah.
Speaker 1:I think Gavin Baker is doing, like 18. He did 18 different either
Speaker 2:live interviews. 18 interviews on stage, plus our show, plus CNBC, plus other podcasts and stuff. Anyway, let me tell you about Ramp. Time is money. Save both.
Speaker 2:Easy use corporate cards, bill pay, accounting, and a whole lot more all in one place. And let me also tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches.
Speaker 2:We actually have some reporting about all the outages. What's behind the string of tech outages disrupting business and travel? There's more news. Everyone was like, this has to be a cyber attack. It feels like it's still downstream of AI, but much more related to patching bugs that were discovered in using AI tools.
Speaker 1:Here he
Speaker 2:So we'll dig into that more tomorrow. But we have John Gray from Blackstone
Speaker 1:What's going on?
Speaker 2:Here with us back on TBPN. Oh my god. And we have your song. This is your intro song.
Speaker 1:Right? My god.
Speaker 3:This is unbelievable. I feel like I'm on, ESPN Yes. With you guys.
Speaker 2:That's the goal. That's the front of a stadium and
Speaker 3:the whole thing.
Speaker 2:Well, we just heard you on stage. Thank you so much for making the time to come chat with us. How are you doing?
Speaker 3:I'm good. Yeah? I'm good. I'm I'm happy to be here. It's hard not to be excited when you're in Austin, Texas.
Speaker 2:Yeah. No. It's great. When you're when you're, not speaking and not, working, are you are you diving into history? Are you reading about the railroad boom?
Speaker 2:I I... I've listened to a few of your more recent chats, and, it feels like that's what you're drawing on more than .com. Well, I would say
Speaker 3:we're trying to find an analogy to what's happening today.
Speaker 2:Yeah.
Speaker 3:And .com was big, but this feels much larger. Yeah. And what's, I think, more similar to that late nineteenth century period was the huge capital intensity.
Speaker 2:Yeah.
Speaker 3:In .com, once the Internet was up, there wasn't much investment. Yeah. In this case, trillions of dollars are needed to build Yeah. And that changes things. And so There's
Speaker 1:no there's no obvious end in sight.
Speaker 3:Right now, we seem to have a very profound compute shortage and an energy shortage in our country. And so it has brought our firm in a massive way. Yeah. Because if you think about it... And it's nice here sort of metaphorically to be in Austin here
Speaker 4:in the center of
Speaker 3:the country because it really is the place where the tech world and the physical world come together.
Speaker 1:Yeah.
Speaker 3:Yeah. And that's why the conference being here makes so much sense because this AI revolution is a combination of these two things. Mhmm. It is a combination of this incredible technology, most of which is developed in Silicon Valley. But then it is the need to build out these physical things.
Speaker 3:Mhmm. And that requires enormous amounts of capital. And from those of us who are in New York, that's our job here to try to provide this...
Speaker 1:Well, I feel like in Texas, you get the know how too because it's like, you know, it's been funny to watch the, you know, sort of anti Elon copers over the last year being like, he can't just become a hyperscaler. Like, these these companies have been doing this for so long. Yeah. Yeah. Blah
Speaker 2:blah blah. Can he build a database?
Speaker 1:He's like, I'm gonna build them bigger, faster, cheaper.
Speaker 3:I totally agree on Texas. It's the know how and the can do. Yeah. It's both of those things that are here. And so what you see happening here is this coming together.
Speaker 3:And by the way, it's not just the AI infrastructure. It's defense. It's gonna be pharmaceuticals. Yeah. There's gonna be more manufacturing build, robotics, obviously.
Speaker 3:All of this is happening. As as investors, that's what's exciting is the scale of this. And you're right. It brings you back in history to when else did we change the world. Yeah.
Speaker 3:And that period in the nineteenth century Yeah. We went from horses and candles and living in wood houses to living in cities and having electricity. Yeah. I can't catch
Speaker 2:me on going back to horses and candles right now. It sounds pretty good. But I know, obviously, there were a lot of downsides to living like that. But what what is different this time? Because, the railroad boom, there were busts at various times.
Speaker 2:What's warning lights? What signals? What metrics are you following to to see how different this is?
Speaker 3:Well, I think back then, the issue was you were building railroads, in many cases, pretty speculative. Sure. And this is what happened in the telecom build out of the fiber in the nineties. You built supply and you hope demand came. Mhmm.
Speaker 3:And in this case, the demand is growing much faster. Mhmm. So if you look at our companies, their spending on models is up 20 fold in the last year
Speaker 2:Okay.
Speaker 3:Because of the productivity gains they're seeing, because of the efficiencies, the power, the type
Speaker 1:of wanna cut you off, but there's one other part of this analogy that I think is important, which is like railroads, you built them and they were only functional for one specific area of the country. And so you've... If you had excess capacity, you couldn't use that track somewhere else in the world.
Speaker 2:Yeah. No.
Speaker 1:That's good point. The the big difference with this analogy is like compute anywhere can be sold Yeah. You know, assuming that, you know, our our laws, you know
Speaker 2:If all of a sudden it's video models or open source wins or closed source wins or one lab wins, the compute will flow there. Or or physical AI.
Speaker 3:Totally. It's all gonna need it. But back then, what what you were doing was building the supply
Speaker 2:Yep.
Speaker 3:And hoping the demand would come. Yep. And in a number of cases, it didn't. It was also done with a lot of leverage as well. Today, the demand is running.
Speaker 3:Mhmm. And you're hoping to build supply to keep up.
Speaker 2:Yep.
Speaker 3:So it's a very different dynamic between these two things, and and the fact that most of this is done on long term contracts with the biggest companies.
Speaker 4:Yeah.
Speaker 3:So that again was done, you know, a bunch of bootstrap guys who are entrepreneurs who didn't have a lot of capital. Today, this is being led by massive hyperscalers
Speaker 5:Yes.
Speaker 3:With market caps that average $2,500,000,000,000 who have very little net debt Yeah. Who produce enormous amounts of cash flow. And so if you think about the data center build, you build this data center when you have a long term contract.
Speaker 4:Mhmm.
Speaker 3:It's not condos in Miami or Dubai during a boom. Yep. You build a bunch, and then there are too many.
Speaker 4:Yep.
Speaker 3:In this case, you have contracted demand. Yes. And that is very helpful. And I would also point out the supply is constrained today. Mhmm.
Speaker 3:It's hard to get chips. Yeah. It's hard to get power. Obviously, entitlements are getting harder. It's hard to aggregate capital.
Speaker 3:So the supply boom, which can get you into trouble with too much speculatively built, too much leverage, that's all being constrained. And that may turn out to be a better, healthier thing. It may mean, yes, some players who need...
Speaker 1:On on x, which a lot of the people here, are are avid enthusiasts of. There was speculation after the blue wave that that it was an inside job from the tech industry to to to slow down and prevent...
Speaker 2:Prevent overbuild.... Gloves
Speaker 1:to blow up, you know.
Speaker 2:It's like,
Speaker 1:what do you think about... So something I I've you know picked up on over the last year is like the high status thing in Silicon Valley is to be compute constrained.
Speaker 4:Mhmm.
Speaker 1:So if you don't have revenue to talk about, you talk about how like, oh I'm just so... We're just so compute constrained. Yeah. And it... And and people kinda eat it up and they use that...
Speaker 1:Venture capitalists will use that as a data point and say, oh, every company I talk to is like compute constrained. But I'm I'm sitting there being like, well, they're compute constrained because they wanna be able to train a model that might have demand. And they don't even know that yet. Yeah. So I don't really lean on that that much as a signal that everyone you talk to says they're compute constrained.
Speaker 1:You're talking to some of the biggest companies in the world that have a lot better real time demand Yeah. For their actual compute versus startups that are earlier stage and sort of speculating that they're gonna need a lot of compute.
Speaker 3:Yes. I I would say it's based. We do talk and invest in some of the smaller companies as well. But the bulk of these conversations are with the biggest companies, with the model companies. Mhmm.
Speaker 3:And they're they're constantly saying to us, do you have any compute Yeah. Today? There's some of the trading firms. The funniest thing is
Speaker 1:You should get little kit, like, like, I I know. Yes. You should get on your desk, like, a little... Like, some some some GPUs. I know
Speaker 2:Blackstone has a data center for something. Yeah. We're gonna throw some GPUs.
Speaker 3:The the the funny thing is that it would be worth much more if we ripped up the contracts and went to the spot market. I mean, if you look at what Elon has done. Yeah. Is by building... He's the one guy because of his balance sheet can build speculatively and then serve the market Yeah.
Speaker 3:At the spot where it's working and it's working for him. And as we see shakeouts potentially of smaller players, I do think that will mean better capitalized folks will will get this capacity. It probably means the hyperscalers and the biggest model companies will get more of it. But it is a global phenomenon, the shortage of compute. And it does feel like when we start to move to these personalized agents to more of the physical AI, I think this problem gets tougher Yeah.
Speaker 3:Because there's only so much energy and power. And, obviously, there are political dimensions. We talked a bunch on stage, did with Antonio, about the importance of us doing this in the right way. There's gotta be responsible building. We've gotta protect consumers, drive their electricity costs down, help them with property taxes, job creation, community give back.
Speaker 3:It can't be a handful of technology and financial firms who get the, you know, sort of they're the winners. It's gotta be broader than than that, and the economic gains here are sufficient that it can be. And so you're seeing a number of us say, we wanna do this
Speaker 2:Yeah.
Speaker 3:And and we're doing it in the right way, partnering with unions and local communities Sure. To get a lot of wins. Sure.
Speaker 2:Yeah. On the demand side, how are you thinking about just as a leader, actually underwriting huge token spends for your workforce? Because we're getting into a regime where you can fire off one prompt, make me an app for my company. It can... The bill can be $400,000.
Speaker 2:And that can provide millions of dollars of value, but you actually have to start thinking about the cash flows and all these different things. How have you been grappling with that idea of white collar team members spending inordinance amounts more
Speaker 3:than ever this think if it was on the excess use Yeah. Of of AI to the underuse. Mhmm. I still think we're here.
Speaker 4:Okay.
Speaker 3:I mean, yes, you always have a few super users.
Speaker 2:Yeah.
Speaker 3:But today, we're really trying hard, let's say, Blackstone Yeah. To feed this and make it work in all our processes. So we're trying to encourage use.
Speaker 2:Yeah.
Speaker 3:Yeah. And the returns are so significant. Yeah. Right? I mean, at our portfolio companies I was talking about, you can process a rental application 90% faster.
Speaker 3:Ancestry.com is 30 times faster processing your family history when it comes in. And so at a firm like ours, we're trying to capture all this information. In our hedge fund business, how do you take all the earnings reports that are out there, all the calls, everything hedge funds are saying, you can boil this down now in real time and make better investment decisions. So I would say at this point, we're more encouraging people to use the technology because the early gains are substantial. I do think there is some later state where we look around and say, hey.
Speaker 3:Are we driving, you know, a Ferrari to pick up the groceries? Do we need to optimize which models are there?
Speaker 2:I like the idea of driving Yeah. Groceries. That sounds fantastic. You know, there is...
Speaker 1:I think another comp that's worth using here is like marketing organizations. Organizations. They They they allocate money to marketing and then a lot of it ends up being completely wasted and ineffective. And that doesn't mean you should stop marketing. And in fact, you should still be aiming to spend more money on marketing as long as your overall efficiency Blended cost.
Speaker 1:The sort of blended efficiency. And your
Speaker 3:learning machine.
Speaker 1:People people have these stories of like, yeah, one person spent way too much money. But it's like, okay, did you learn that that's not a good use of tokens and are you reallocating them? And I think like the fact that marketing teams have been comfortable with like, yeah, we spent $50,000 on this thing, it didn't work. Or or in some cases, spent a million dollars on this camp... Campaign and it it...
Speaker 1:We're not gonna do it again. Mhmm. But you learn and then you you optimize and and you
Speaker 3:And that's what we're encouraging the companies. And there are just a lot of things that are pretty straightforward. I mean, creation. Companies can create their own ads now
Speaker 2:Yeah.
Speaker 3:At a fraction of the cost. Yeah. And so you're really empowered when you wanna do this sort of stuff. And to me, it's exciting. We've just gotta do a better job.
Speaker 3:We've gotta push our teams to do better. We created a company with Anthropic called Ode, which is a deployment company. Their biggest challenge is serving everyone who's calling them for business Sure. Having enough people. So I still think when when I think about diffusion Mhmm.
Speaker 3:I think about us very early days. Yeah. There's a lot of room to run here. And... But I do think at some point, yes, more cost sensitivity will come into play, but I don't think that's the primary focus now.
Speaker 3:Mhmm.
Speaker 2:How are you thinking about AI AI and how it intersects with biology and biotech? Yeah. There could be a really broad boom where we're seeing lots of different companies take drugs through trials, more IPOs. There could be, you know, the the the the big biotech companies just gets bigger. The labs could do things.
Speaker 2:How are you
Speaker 3:thinking about that? Well, the interesting thing is on the front on the front end of innovation Mhmm. This is gonna lead to many more potential therapies. Yeah. Because you're gonna be able to do sort of synthetic chemistry and biology in a way you never could.
Speaker 2:Yeah.
Speaker 3:And that is hugely exciting. And when we think about the future of cancer care or neurological diseases, pulmonary, I I think it's gonna be radically different. I think this is the most promising thing Yeah. That is out there. I think the limiting factor you have to remember is you've gotta go through the physical world when you get to trials.
Speaker 3:Yep. And that's when you go from this huge funnel
Speaker 2:Yeah. Of ideas to code on AWS.
Speaker 3:Yes. So It's just not We have a big business that invests in phase three trials. We've announced a number of major deals in the last few weeks with J and J and Pfizer and so forth. Yep. Those trials cost a billion...
Speaker 3:Multiple billion dollars. So... And they involve human beings Yeah. And processes and the FDA, and then you've gotta do it globally. So I think the exciting part here is all the new therapies that will come.
Speaker 3:Mhmm. I think the hard part remains that tight part of the funnel where it's like, hey. You gotta go through the real world.
Speaker 2:Yeah.
Speaker 1:And You're... I'm hearing that capital might be a constraint, and that tells me it's your problem. You gotta figure this
Speaker 2:out. Problem.
Speaker 3:So we like... When there's capital as a constraint and we we have expertise, knowledge in a space, and we can bring capital, we see that as an opportunity. Thank you for
Speaker 1:your service.
Speaker 3:Yeah. Well, but I would say in this case, what we love about it is you can be
Speaker 2:a real force for good. Yeah.
Speaker 3:Yeah. Anything we can do to accelerate human health Yes. Is the best. Yeah. Of course.
Speaker 3:And as we talk about this AI thing and all the concerns and negativities, I do think the positivity. I think Antonio said it... What was it? He wanted to go from PDoom to PBOOM or something was his line. I think life sciences
Speaker 6:is the tip of the spear.
Speaker 2:For sure.
Speaker 3:There's a bunch of other things.
Speaker 2:Grounded. People, they, you know, they like the chatbots, so they don't. But, like, if if if you if you cure your disease, if you get rid of your allergy or even just get rid of the cold... The common cold, people will be happy.
Speaker 3:Yes. For sure. Make a huge difference. And by the way, entrepreneurship, the ability to start companies, you're seeing that in company formation today. Yeah.
Speaker 3:Education getting democratized. You don't have to live in Cambridge, UK or Cambridge, Massachusetts. They have the greatest professor in the world. Yeah. It's amazing.
Speaker 3:There's this blue collar job boom. There's a lot of good things. There will be challenges. Yeah. And as a society, we've gotta make sure job transitions, skills training, we've gotta be sensitive to this.
Speaker 3:Yeah. But I do think there's a lot of good things that will come from what's happening. So that's why I'm excited to be here.
Speaker 2:Thank you so much for coming on the show, John.
Speaker 1:Great to
Speaker 2:catch to
Speaker 1:see you. You guys are great.
Speaker 2:Thank
Speaker 3:you. My sports moments.
Speaker 2:I'm so excited. It's close
Speaker 4:to the stage.
Speaker 3:You're a star.
Speaker 2:Thank you so much for coming on the we have our next guest joining in just a second. Let me tell you about Console. Console builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. And I'll also tell you about public. Invest across all asset classes, automate anything with AI agents, run trading strategies, risk management
Speaker 1:Buck
Speaker 2:movements, and monitor markets with AI agents. We're investing on public.
Speaker 1:Scott, president and CTO, of Micron with us.
Speaker 2:Welcome to TBPN. How are you doing? Good to see you. Thank you so much for taking the time. Would you mind, kicking it off with a little bit of introduction since it's the first time on the show?
Speaker 6:Sure. Absolutely.
Speaker 2:And we're gonna adjust this right here. There we go.
Speaker 4:There we go.
Speaker 2:Production likes the way we do that.
Speaker 6:Yeah. So I'm president of Micron Yeah. And focused on technology and products and our long term road map
Speaker 2:Yeah.
Speaker 6:And focus with our customers on delivering great products.
Speaker 2:Are you are you... What goes into actually, making sure that you're on Moore's Law, that it's getting better every year? I've always been fascinated by the idea that simultaneously, computers get more dense, more efficient, more effective every year like clockwork when you zoom out. But when you zoom in, it's gotta be crazy on a day to day basis and a month to month basis. Battles.
Speaker 2:How how is it that that we have a smooth line? Is that just Definitely definitely working together?
Speaker 6:A pretty crazy
Speaker 4:Yeah.
Speaker 6:Situation, but... And different through time. Yeah. You think really big transitions through memory over the past twenty years in terms of, growing bits per wafer and and improving the cost structure. And then, you know, the past few years, especially with AI, more focused on performance and customization Yes.
Speaker 6:Less bit growth... Raw bit growth in terms of technology, but more and more performance coming into the memory products and more and more co design with our customers.
Speaker 2:That makes sense. And then now I imagine that there's so much demand that it's not just enough to make each wafer better, to make each chip better. You also need to just make way more of them. What has been the plan to date? What does the future look like for your manufacturing footprint?
Speaker 6:So definitely given the situation, the tightness of memory across really multiple years now. We're majorly focused in expansion across the globe, certainly especially in The US with more than $250,000,000,000 of investment coming. A massive facility underway in in Boise, Idaho right now. Congratulations.
Speaker 2:I saw the news. I was very excited.
Speaker 6:Yeah. No. We're really excited about that. And and a big bigger... Even bigger facilities underway in in Clay, New York.
Speaker 2:Okay.
Speaker 6:And expansion in Virginia also. So just a a number of of the biggest projects we've ever done, some of the biggest industrial projects The US has ever seen, and and bringing more memory to to the market.
Speaker 2:Three American cities. Is that just because you're purely patriotic, or is the is The United States uniquely good at building huge manufacturing centers now? This seems... It seems risky. It seems contrarian to not go to the cheapest location where the labor's cheapest.
Speaker 2:Yeah. I'm I'm a bald eagle, and I applaud you for building in America, and I love it. But why? Why why is it... Why why is it
Speaker 4:a good strategy?
Speaker 6:We're... It's a good strategy for a number of reasons. But also I'll add, we are... We have two expansion projects in in a number of other countries as well across across Asia. And it's just a lot of focus uniquely in The US now because of...
Speaker 6:Primarily because of the customer base
Speaker 3:in
Speaker 6:the And certainly the aspects of having more US supply to match up with those customers is really important in today's world. And so it's really balancing our output across the globe. But definitely the opportunities in The US, our RDs in a number of countries, but our base is in The US. And you know, more and more customer partnership becoming important in how we build our products.
Speaker 2:Yeah.
Speaker 6:So it's great to have manufacturing Yeah. In The US close to those customers.
Speaker 2:When you build a manufacturing hub, what are the key ingredients to making that go smoothly? Because we're seeing data center pushback, and we were always saying that it was odd that for a moment they were getting rebranded as AI factories because people don't like shoe factories in their neighborhoods usually. Is everywhere in this country. No one wants a smokestack. And...
Speaker 2:But I imagine that you've put years and years of effort into making the the best, most efficient manufacturing centers you can. What's the secret to success?
Speaker 6:We have. And and, you know, there's a few secrets to that success. Our our headquarters course is in Boise, Idaho. That's where the company was founded. Yeah.
Speaker 6:And, you know, we've had a a significant positive impact of the community in in Boise, and we're very conscious of interaction with the, you know, the local environment with the cities that we're in, certainly in Boise, but also in New York. And making sure that we partner with those cities upfront with the community colleges, with the local schools, with the whole ecosystem. In Boise, we recently made significant donations to improve roads and housing. So I think we've been very focused throughout the the process, but really throughout history and how we interact at every city that we're in. Yeah.
Speaker 6:And have it in... You know, a positive environment. I think that's one of the big things about New York where, you know, there wasn't a lot of industry where we went into there, just the aspect of bringing the thousands of jobs into that economy and making sure that, we hear from the community and understand how we're gonna impact them in a positive way.
Speaker 2:And those are more permanent jobs I imagine because you're manufacturing an ongoing business. We have not
Speaker 6:just Micron either. We bring a substantial influence of all the partner companies that support the infrastructure And fab
Speaker 1:facility Not for some to mention like retail in the surrounding restaurants, you know, all all these other
Speaker 2:For sure.
Speaker 6:Yeah. Thousands thousands of jobs in Yeah. In all these locations is you know, it's a big positive impact. So we're really proud of it.
Speaker 1:Yeah. Talk about when do you have to tell customers no? Not on the like not on the demand side, but like they want you to do something and you're working against you know, multi year lead times and you you just have to say like, sorry, no. Because like, usually in a perfect world, you're like, yeah, we could do that. But in in this game, it's like you're you're constantly iterating.
Speaker 1:You're on your own road map. They have their road map. Everybody's trying to predict like what capability will be needed in two years, three years, four years. Yeah.
Speaker 6:You know, it's a it's a great question. I think one aspect of that is we've massively increased our R and D spending over the last few years and to support more of the yes answer on that to our big customers and be able to understand what their roadmaps really involve. And very much so, we're in the middle of discussions with those customers in ways we never used to be in terms of their products for three or four years from now and making sure that our products and roadmap line up with them, where we need to build in features or co design products with them. So we've... A big part of the answer is we've expanded our capacity to be able to say yes more often to the the big trends.
Speaker 6:Yeah. And... But absolutely, we have... We also have an evolving road map where they don't always know where they're going And and at times, we can't support every customer. And, you know, we have to just respectfully be, you know, part of the process and make sure that that we align things and they understand when we don't have the capacity to do certain things.
Speaker 2:Yeah. I mean, Micron's been called the the bottleneck. I I I hope that's a badge of honor and not a not not not a critique. No.
Speaker 1:It means you're important.
Speaker 2:It means you're important. But it's, it's kind of a twofold bottleneck. Right? Because it's supply. People want more of the great chips that you already make.
Speaker 2:And then the individual models can be bound by a particular architecture. Yep. Talk about the evolution of the AGI in Micron because different manufacturers have had, oh oh, well, there's a crypto boom. Maybe we should do a little more, but may... Maybe not.
Speaker 2:I I don't wanna be the end of the bullwhip. Right? Internally, culturally, what has it been watching this boom? And then how how confident have you... Or how have you built confidence around, okay.
Speaker 2:This is the end of history in terms of architecture. We can build against this road map for years.
Speaker 6:Yeah. I think that's a really interesting question. And, you know, our our company, not just in terms of the products we're making right now, but we're very focused on AI utilization internally. And it's dramatically changing how we design products, how fast we can design products. And at the same time, even though we're becoming more and more efficient, we're also expanding our company in terms of people because it is a very labor intensive thing.
Speaker 6:It's kind of a combination of those two things. And AI for our company, of course, because of how important and how centric memory is to AI, it's really a key piece of everything about how you know, tokenomics work is is memory capability going forward. The company has raised up in terms of understanding they're part of a a bigger and a different kind of Sure. Part of the ecosystem now than ever before. And so there's a lot of excitement there and great adoption of AI.
Speaker 6:And it's also exciting to be really connected with our customers in ways that we never used to be. When memory was a commodity even five years ago, but certainly through history, memory had a very different role. Today, with with memory centric computing, it's it's just such an exciting time to be really part of how how AI takes place. And we're just at the beginning of AI. Yeah.
Speaker 6:So it's...
Speaker 2:Talk a little bit more about how Micron uses AI internally. I imagine two years ago, there were meeting note takers and email summaries like everyone had. But today, are you seeing the latest models actually be useful in design in these really hard math challenges? Is that generalizing? Because we don't see a lot of, like, memory design benchmarks.
Speaker 2:We do see the Millennium Prize problems. And you can think, okay. Maybe that would, generalize. But is it? Are you feeling acceleration?
Speaker 6:Yeah. It's it's pervasive across across all aspects of our product development right now. That's actually the place that it's implemented the most. Of course, like every like every big company, we're we're utilizing it for workflows across Yeah. Segments absolutely.
Speaker 6:But really, the engineering is where we've taken the lead on actually impacting the timeline. And in targeting to get designs done in half the time that we got done before. And able to do twice as many designs with the same number of people in that period of time. So really changing things. We do, of course, coding, but we we're doing more and more of design verification and physical AI in terms of building it into how we simulate process, how we simulate thermals, how we simulate the physics of our chips.
Speaker 2:You know? And do you have a house view on whether you think the big models will just learn to do this and be... Get rated it out of the box, or you will need to fine tune with proprietary data either in partnership or using open source or doing something else. Do you have a view on how spiky the intelligence is relative to your business?
Speaker 6:So there's definitely aspects of our business that have to be internal, and we have to bring the models in and optimize them around And especially process technology, the bringing those forward can't really be done without bringing them internally. There's other aspects like certainly the design verification pieces that are very common across the industry that are just impactful right off the bat. And then we can partner with, you know, a variety of of AI companies to to move that forward. So I think it's a mixture of both in terms of your question.
Speaker 2:Yeah. Depends on the... You gotta use the right tool for the
Speaker 6:We do.
Speaker 2:That makes sense. Yep. Thank you so much for coming on
Speaker 6:the show. Great talking to you.
Speaker 4:Have a
Speaker 1:great rest.
Speaker 2:Thank you, Russ. Every day at TomorrowX. And we will bring in our next guest, but not before I tell you about Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in store, on mobile, on social, on marketplaces. And now with AI agents
Speaker 1:We got Renee. Let me also tell you about
Speaker 2:MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Don't just build AI. Own the data platform that powers it.
Speaker 2:We have Jed Finn from Morgan Stanley. Welcome to the show.
Speaker 5:What's going on,
Speaker 4:How are
Speaker 2:you doing, Jed?
Speaker 7:Great to have you.
Speaker 2:Great to
Speaker 1:have How
Speaker 2:has your TomorrowX been so far? You enjoying?
Speaker 5:Excellent. It's amazing to hear.
Speaker 2:And Morgan Stanley is a sponsor.
Speaker 5:I saw
Speaker 2:them on the ticker. Correct?
Speaker 5:We absolutely are. It's nice to hear a broad range of perspectives.
Speaker 2:Yeah. It's great. Why don't you introduce yourself first time on the show? Tell everyone about your role.
Speaker 5:Yeah. It's great to be here. Thank you for having me. I run the wealth management business at Morgan Stanley, is the largest full service wealth management firm in the world. We have about 8 Thank trillion of assets.
Speaker 1:Really wanna know how many assets that are rep... Are represented in this this room because I I I might be... You might hit... It's chunky. It's chunky in this house.
Speaker 1:I'm gonna guess closer to a to a to a trillion than...
Speaker 5:You know, don't know which of the speakers are still here.
Speaker 1:Yeah. There's a lot
Speaker 4:of prospects here as well.
Speaker 5:But yeah, 8,500,000,000,000 of assets. We have 21,000,000 clients and we have our adviser channel, our etrade channel, and our workplace channel. Yeah. It's what we focus on.
Speaker 2:Yeah. Oh, yeah. You you acquired Solium.
Speaker 5:Right? Solium in 2019.
Speaker 2:That's right. Yeah. For employee stock options and
Speaker 5:So there's a public equity admin component to it. And then there's a private cap table management component. So we're the cap table manager now for nine of the top 10 unicorns Yeah. By market value for 75% of the top 100 and for 50% of all, almost a thousand unicorns.
Speaker 2:Yeah. And and that it goes for like pre IPO access as well.
Speaker 5:Exactly. Well, it's it's a glide path. Sure. When it's two people in a garage Yep. You just need someone to track who owns what.
Speaker 5:We can do that as you... I didn't
Speaker 1:I did not realize you you had gotten that that that much market share like very seemingly like very quietly because I mean that the the the I don't know the the all the cap table like upstarts and even Carta, they do a ton of... They do they do quite a bit of marketing Mhmm. Around this category and and seems like you guys have.
Speaker 5:Carta is a great partner of ours, actually. And when those companies outgrow the private side, Carta refers them to Morgan Stanley to go private to public seamlessly. Yeah.
Speaker 2:Yeah. How is the firm thinking about the IPO tomorrow? Anthropic is going out, maybe pricing later today. How are the... How's the client base reacting?
Speaker 2:What can you tell us about Morgan Stanley's view on Anthropic?
Speaker 5:There's a lot of demand. Yeah. There's a lot of demand. I mean, it... It's tough.
Speaker 5:Right? Because the TAM for intelligence is kind of all the dollars. Yep. And, you know, if you think about it just from an upside downside perspective, it's really tough to predict what the upside could be, but it's really large. Yeah.
Speaker 5:And the downside, course, is what you just put it in. So there's a a lot of excitement from our client base
Speaker 2:That's great.
Speaker 6:Both on
Speaker 5:the retail side and on the institutional side.
Speaker 2:That's great. That's great. Is this... Are are you taking any lessons from SpaceX to hear? Tell me about what was going on at the firm during the SpaceX IPO.
Speaker 2:That was previously unprecedented. It feels like we're in another moment like that.
Speaker 5:Yeah. Well, just like anything else, when you do it a couple of times, you start to build this operating rhythm. And so we know... You know, SpaceX was great. It was a success.
Speaker 5:They're a really important part of the firm. Yeah. But there are things that we learned on the technology side, on the testing side, on the allocation side. And so we've just further strengthened those processes and worked with all of our market partners in particular to make sure that, you know, the rails work, and, you we're feeling pretty confident and optimistic.
Speaker 2:How are those rails looking different today in the age of AI versus a few years ago?
Speaker 5:Well, the private to public isn't a huge difference. I think what's different is the narrative, the market opportunity, how people wanna play the different market opportunities that evolve. Obviously, there's a wide range of views on is it a bubble, no bubble, what's gonna happen. Yeah. And so I think there's just another level of discussion that needs to happen as people think about committing capital.
Speaker 2:Yeah. And then in terms of the clients, how do... How are they responding to the adoption of AI tools at Morgan Stanley in wealth management? Because I can imagine people are like, hey. Look.
Speaker 2:As long as the spreadsheet of my net worth is correct, I don't care if you use Python or an LLM. It just needs to be accurate. But they probably still wanna have a conversation with a real person. How is that evolving?
Speaker 5:Yeah. Look, that's our core thesis. Described it very well. Where we're rolling tools and capabilities out in phases. Right?
Speaker 5:So we're starting with our financial advisors who now can interact with their book, their clients' portfolios in a much more natural language using a number of agents to build portfolios, change portfolios, you know, tax loss harvest, etcetera. We're also building the capability for ultimately our clients' agents to have that curated experience with our system, but we need to build the rails around that. So there's been a lot of talk about, are you letting agents into the system? Yes. Or no?
Speaker 5:And for us, we are ultimately going to, but we need to make sure that experience is curated. But clients are gonna increasingly demand it and we obviously respond to client demand.
Speaker 2:Yeah. There's also an interesting thing where... I mean, I'm I'm sure with many of your clients, they interact with a banker over email or text message. And agents, if they can send an email, they can interact with Morgan Stanley one way or another. And so that that that's sort of inevitable.
Speaker 5:It's not one or the other. Yeah. Let me give you an example. We were early with a tool that we built called debrief, which was essentially listening to a client adviser discussion. Sure.
Speaker 5:At the end of the discussion, what it would do is summarize the discussion Yep. Send an email to the client Yep. And then put on the notes in the internal
Speaker 4:That's great.
Speaker 5:CRM tool. The next iteration of that is an agent is listening and actually taking the action. Sure. So if you say, I wanna change a beneficiary
Speaker 1:Sure.
Speaker 5:Or I wanna open up a trust account Yeah. It will actually do that all in the background Yeah. Now. And the adviser just hits agree. Agree.
Speaker 5:There's still a human in the loop
Speaker 2:Yep.
Speaker 5:For now. Yeah. And as regulations evolve, we'll see how that ultimately shakes out. But it just making the next step just seamless. So everything becomes easier and simpler for our advisors and our CSAs so that they can do the thing that matters most, which is spend time with clients.
Speaker 2:Yeah. Yeah. That makes a
Speaker 1:lot sense. How do you think the the private markets in the, let's say, small business category, 1 to 50 bill... 50,000,000,000, let's say. How do you think, activity there needs to evolve? I imagine given your position at the firm and and how many clients you guys have, you've seen some, like, wild structures and and attempts at getting exposure.
Speaker 1:And and I'm sure you guys have clients come in that that tell you, hey, I'm doing this deal and maybe sometimes you
Speaker 2:You bought what?
Speaker 1:You found this company at two if you could snap your fingers, what would you what would you... What changes would you make in the terms of the overall market structure?
Speaker 5:Sure. So the first thing that I think is clear is that the difference between private and public ought not to be a binary distinction. Mhmm. Right? What it ought to be is levels of disclosure and a gradation of that and then levels of investor sophistication.
Speaker 5:Mhmm. Right? So that seems like a structural thing that ought to change. I can't predict when that's gonna happen. But, you know, if you wanna share nothing with your investors, then it's just gonna be the biggest institutions in the world that can invest.
Speaker 5:Sure. If you wanna go all the way to Sarbanes Oxley massive disclosures, then retail investors Mhmm. Then there's probably some categories in the middle about level of information and sophistication. That's one thing. The second thing is a lot of the innovation that's happened in the marketplace, particularly on the private side, has been done not necessarily in concert with the companies themselves.
Speaker 5:They've been done to the companies. So there was a much published, you know, wrapping in SPV of shares that did or did not exist. And some companies got upset. There was a well published spat over the summer between a tokenizer and a tokenizee who didn't want that to happen. Yeah.
Speaker 5:At the end of the day for us, we're an issuer centric firm. Sure. We are there for the company and we are building the capabilities to make it easier for those companies to be able to access capital markets on their terms. Sure. That's a very different mindset from what a lot of the, you know, market
Speaker 1:Hostile approach of like, I don't care that you don't want us to do this. We're doing it because we have demand. It's just like
Speaker 2:He's on the side of Matt Graham. Yeah. Yeah. We love to see it.
Speaker 5:But but the other thing is, you know, the value proposition to the investor and the employee is always clear. What is the value proposition to the company? I mean, I think you have to be very clear about why it's good for you to broaden the distribution, and they have to support it because it's their company, and they're building the equity over time.
Speaker 2:It makes a ton of sense. Jed, thank you so much for coming on the show.
Speaker 1:Thanks for having See you later.
Speaker 2:Take care, guys. See you later.
Speaker 1:Have a great event.
Speaker 2:Thank you so much for coming on TBPN. Let me tell you about Figma. Agents, meet the canvas. Your AI agents can now create and modify your Figma files with design system context. Let me also tell you about the New York Stock Exchange.
Speaker 2:Wanna change the world? Raise capital at the New York Stock Exchange. We have Scott Wu coming on TBPN. Welcome to the show. Scott, come on down.
Speaker 2:Come on down to TomorrowX. How are doing, Scott? Good to see you. Here he is. That headset on.
Speaker 2:What's up, guys?
Speaker 1:How's
Speaker 2:going? Co founder, CEO, Cognition. How you doing?
Speaker 1:Needs no introduction at this point.
Speaker 4:Yeah.
Speaker 7:No. I'm doing good, dude. Thank thank you for having me. It's, I've been here for for like a day or so in Austin. It's cool to see the energy.
Speaker 7:A
Speaker 2:nice What? Well... Crowd. Well, take me through the latest, announcements from Cognition. Right?
Speaker 2:Because I've heard a bunch of things, and I I need to reset because I think some of them might not have been launched yet. So I want you to set the table for me so I don't accidentally leak some things. Good. Good. Lot of exciting stuff, but what's actually out?
Speaker 7:Good call. So yesterday, we just released Devon bots.
Speaker 2:Okay. Good. Yeah. Okay. It's live.
Speaker 2:So then heard about that.
Speaker 7:Yeah. So we can talk about that now. Great. But, yeah, we had a nice launch on Twitter. Okay.
Speaker 7:Yeah. Yeah. I've seen all that. I think in one line, basically, it's these persistent agents that can actually own outcomes instead of tasks. I think it's basically taking Devin one step closer.
Speaker 7:I mean, the whole vision
Speaker 4:Yeah. Of
Speaker 7:course. Two, three years ago was, you know, the AI software engineer, the virtual teammate, so on. Yeah. And I think it's taking it one step closer to that where it has memory. It can take on a full project.
Speaker 7:It can really, really delegate and run things on mass. Yeah.
Speaker 4:It can
Speaker 7:own a goal. So let's separate
Speaker 2:the technical from the functional. What what... At what level of abstraction are we prompting at this point? Are we just saying, like, build the website, make me money, like, make me happy?
Speaker 7:The EBITDA machine. Right?
Speaker 2:You know,
Speaker 7:it'd be nice if we had some more revenue. Yeah.
Speaker 2:Because because we're we're we're going up the abstraction layers, I think, and we're closer to the top than we've ever been. With engineering
Speaker 1:tasks specifically, it's like, I wanna build this application, and then we're gonna get to a point where you guys just give them a price and Yeah. You
Speaker 2:hit Yeah. Confirm.
Speaker 7:I... No. So I think that's right, I think it is already going like much closer. I mean, a a good example of this, so obviously, Devin has a mobile app. Yep.
Speaker 7:We've had an iOS app for some time. Yeah. We have an Android app now. Amazing. There was not a single human who worked on that.
Speaker 7:Wow. And so there is just one Devon bot whose job is basically like, look. Look at the iOS app. Whenever there's new changes that come in Mhmm. Build the whole thing, test it out, make sure you make it mirror the behavior of the iOS app.
Speaker 7:It just works. Okay. It's it's kind of
Speaker 2:So the prompt is basically like Android app, please. That's right.
Speaker 7:Okay. Yeah. Yeah. Similarly, it's like, you know, we run a lot of obviously like inference kernels and things like that. We have this always on dev and that's just like, okay.
Speaker 7:Just go optimize. You know, it's it's actually a very nice use case, I would say, for bots. Yeah. Makes you feel the RSI at least like a little bit because it's like you have a very clear... You know exactly what the token output is supposed to be.
Speaker 7:You can run all the benchmarks in terms of like what your GPU throughput is and so on. And it's just like as you're finding optimizations for your specific, you know Yeah. Set of prompts, you can go and do that Yeah. And and improve your
Speaker 2:So then under the hood, how much routing is going on? Like, can can... Try and just maybe scope it in, like, orders of magnitude. Like, can can Devon Bot actually call a model that is extremely expensive and then also simultaneously delegate to a model that is a 100 times cheaper? Is that how wide we're talking?
Speaker 7:Or That's right.
Speaker 2:Yeah. Order of magnitude?
Speaker 7:It's probably... A 100 is maybe a little bit... But it's in the probably like 10 to 30 x range Okay.
Speaker 1:In terms
Speaker 7:of like the biggest to the smallest. Yeah. And so that routing I think has been... Obviously routing has has been a big thing probably for the last six months or so I think for for all of us of of like really seeing that. I think actually a lot of the the big diff too is just like going through all the little product kinks that make it a really really good experience.
Speaker 7:I mean, think recently a lot of these guys like Instinct and Muse Yeah. And and Grokbot and so on have done that on the the kind of personal assistant side of like solving the product experience of what does it mean to have an agent that has access to your stuff and actually makes decisions and always, know... And I I would say one of the earliest actually of those was Poke Yeah. Which is the interaction team. And that's part of the reason that we were so exciting excited about working with that team.
Speaker 7:I've known Marvin for years and years. But but I think incorporating a lot of that, but obviously from the perspective of a teammate at work. And so what does it mean to actually have it be, you know, every dev and bot, for example, is a separate instance that you can talk to in Slack. Yeah. It's other people on your team can be working with that simultaneously.
Speaker 7:It can go and proactively look at channels and message and all all of the things that you'd you'd want it to go and do.
Speaker 1:How are you guys working with customers on cyber right now? I think obviously we've seen outages driven by attacks. We've seen outages Yeah. Driven by people saying we need to take our we need to take our systems down Yeah. Fix them before, you know, a bad actor gets involved.
Speaker 1:I'm sure you guys are are really focused on this, but but
Speaker 7:For sure. Yeah.
Speaker 3:Talk about the moment.
Speaker 7:We've been doing a lot more lately. We brought on actually Alex Stamos who is the the CSO at Facebook
Speaker 2:That's right.
Speaker 7:As as our CSO. He's been fantastic. I think at a high level, the you know, people talk about the agent swarms and agent forces. I think security is one of the most clear cut applications where it's just very clearly the case that, you know, doing a code scan once a month or something is just not going to cut it. Like, if if if AI agents are attacking you twenty four seven, then you need to have the agent defense force to actually go and, protect things as well.
Speaker 7:We've been... We we we really... We did a big security release actually, about a week and a half ago, and... But we back to have lot
Speaker 2:of Is that driving net new customers to Cognition or is that pure cross sell?
Speaker 7:It's it's both for
Speaker 1:sure. Okay.
Speaker 7:Yeah. And I mean, I I think especially with... As you said, I mean, all these incidents that have been been happening over the last couple weeks, I think every major enterprise in the world, as as they should Yeah. Has been thinking a lot more about this and how do we... You know, it's like one one of the common things that that we hear is like, we did a code scan.
Speaker 7:Mhmm. We had found like 2,800 warnings and vulnerabilities and like, now what do we do?
Speaker 1:You know?
Speaker 7:And so so like figuring out...
Speaker 1:Company that that responded to like finding a bunch of vulnerabilities by hiring like 50,000 temporary...
Speaker 2:I think there was Bank of America or
Speaker 1:something. Yeah. It was like Bank of Bank of America like did had early mythos access in the end. But the response was like we need 50,000 new people.
Speaker 2:So bang, boss.
Speaker 1:Scaling up the contract for us. Yeah.
Speaker 2:I I
Speaker 1:might have the numbers off by
Speaker 6:a lot.
Speaker 7:Unfortunately, there's not enough software engineers in the entirety
Speaker 2:of the Yeah. Because you're going up against what we're gonna have to go do. You're going up against infinite minds of hackers.
Speaker 1:Yeah. I was only off by 50 x. It
Speaker 2:was 1,000.
Speaker 1:Thousand contractors. We're still at the time. Yeah. It's like a thousand.
Speaker 2:A thousand. Adding a thousand people. How big is Cognition now?
Speaker 7:Cognition today, our entire EPD org is actually about 75 people still. Wow. The whole company is around 500.
Speaker 2:Okay. Wow. So haven't smashed the butt... The higher thousand people just
Speaker 7:We have, but it was... It's a Devin button for us.
Speaker 4:For sure.
Speaker 7:So we just... You just Yeah. You just run, like, a thousand Yeah. Yeah. And then that's kinda
Speaker 2:So doing a lot of inference. What about training? Are you doing post training? What's your view on spiky intelligence, the value of training? What what what are you doing these days?
Speaker 7:Yeah. For sure. So so recently we put out Suite two and then we put out Suite two Fast. Yeah. And basically what we've just seen is I think across the board as we were saying it, you know, the the kind of intelligence saturation on a lot of tasks means that there are so many models now that are good enough to get the job done.
Speaker 7:Mhmm. And now what you care about is not just the zero one binary of can it solve this logic problem
Speaker 2:Yeah.
Speaker 7:Which used to be the bottleneck for your agents. It's just literally that. But but now it's more like, okay, like, how how does it respond to you and does it do things in the style of how you want them to and of course, you know, speed and cost and so on. With us, we've just found that, you know, doing doing mid training and post training, especially with a lot of the unique data that we have in Devon Yep. Has allowed us to put out put out really good models for that.
Speaker 7:And so, you know, our our our most recent one, we took the the the recent Inkling model from Yeah. Linky and did a ton of post training on it on very specifically these coding tasks that we literally, you know, see our users going and using in Devon. Yeah. And we we were able to see scores pushed like a ton.
Speaker 2:Yeah. Yeah. So you're doing you're doing post training. You're also doing mid training. Yeah.
Speaker 2:If I follow that line, feels like pre train is next. Just help me understand because you're you're extremely well capitalized. Yeah. We're we're several years out from the big GPT four runs, and those, if I gave you the budget that OpenAI had when they did that, I feel like you could do... Is is there something where you don't want to for a different...
Speaker 2:For a particular reason? Like Yeah. Why not just do the whole stack?
Speaker 7:Yes. If if we could build GPT four. I I think you're probably right that we could build GPT four, but, you know, one of the beautiful thing... You You don't need to.
Speaker 2:Okay. So why bother? Because Sorry. It's already there.
Speaker 7:Think the pre training, I I mean, it seems clear that these will get bigger and bigger. Sure. Will go from billions to tens of billions to hundreds
Speaker 1:of billions. You don't
Speaker 2:wanna be on
Speaker 7:that track. It's it's just very different, I would say, from from what our kinda, like, core competency is. Sure. And and I think
Speaker 2:Yeah. We we Yeah. You have the edge of the data, and you have the value that you
Speaker 6:can get.
Speaker 7:Data, knowing exactly use... What use cases we want our customers, you know, our our customers actually want from Devin, what behaviors or what metrics, you know, we're trying to optimize. For example, even like as folks know, it's like really hard to have good benchmarks, to have good evals and so on. Right? I mean, a lot of what we do and spend our time on is figuring out, okay, what exactly does it mean for Devin working on a task to be like a good job or not a good job?
Speaker 7:You know, because it's not as simple as we're saying as just getting the PR Yeah. Done in a way that's functional. It's like, okay, well, was it scoped correctly or did it do things, you know, did it use the various utilities you already have instead of like writing a bunch of net new stuff that it didn't need Yeah. Did it go? And so, you know, solving for things like that, I...
Speaker 7:It really lies in the in the realm of kind of mid training and post training. So that's why it's made sense.
Speaker 1:Changing the steps to... I'm sorry. I have a question. How do you how do you try to get a read on... Like what's what's your philosophy on hiring executives?
Speaker 1:You guys have have signed so many people with just incredible track records and pedigree. But how are you evaluating them that the best work of their career is is still ahead because everyone's at different times hired somebody where Yeah. They had this like crazy run but it was something about that company at that moment and they can't replicate it. And so Yeah. I imagine like you know, reading people in in in poker helps you, you know, roll roll into your executive hiring process.
Speaker 7:I appreciate that. No. I I I I think it's what would I say? So so we... Yeah.
Speaker 7:We brought on a number of execs recently. I mean, Chris Degman joined as our CRO. He's been he's been amazing. Alex Stamos, obviously. Paul Gray wall as our chief legal officer.
Speaker 7:Mhmm. You know, brought brought in Brandon who ran all of marketing at Ripley and took it from from basically zero to like a billion plus of of revenue run rate at Ripling there. You know, a couple of thoughts. I I mean, I I think obviously there's there's no substitute for folks who have really really worked with them for years and years and seen how they click and I think there is to your point. I mean, the...
Speaker 7:There's a there's a difference between the kind of leader who follows a playbook and that playbook just happened to be the right one for that company at that time versus it's just very clearly the case that they they figured out all the right first principles things and they knew what was needed for x company, you know, Coinbase or Snowflake or whatever Yeah. To go and succeed. The other thing I would call out which it sounds kind of dumb but you know, is is like when you hire from from from companies, obviously, it's not just a function. For example, everyone who's hired great engineers thought like, okay, what are the companies that have really great products Mhmm. That are just like really tasteful or have like really tough like, you know, well architected things.
Speaker 7:I was gonna say the names but I realized that maybe they would be upset at
Speaker 1:me for saying that. But but
Speaker 7:you know, it's like hire from the from the companies that are known for that Yeah. Really great. And and the same is true for for everything, you know. It's like, if you're gonna hire a sales leader, like, what are the companies that really scale Dominated sales. Yeah.
Speaker 7:And dominated Yeah. You know, in in the previous era. Yeah. And and now it's kind of the same except it's all like five times faster than that even. And, you know, Snowflake, obviously, it's like Chris took that from zero to like 3,000,000,000 plus of revenue run rate.
Speaker 7:It was one of the best enterprise sales companies ever probably. You know, similarly, like, I think, you know, Art, for example, who who joined and and runs all of our, you know, partnerships came from Brex where he was a chief business officer. I think Brex did an amazing, amazing job of doing all the, you know, the partnerships and and and really landing that well and so on. Right? And so basically figuring out not just what are the the the the great companies and the great kind of like startup experience that folks could have had, but what are the companies that are actually known for the particular skill
Speaker 2:For sure.
Speaker 1:That you're trying to
Speaker 2:hire for. For sure.
Speaker 4:Yeah.
Speaker 2:Yeah. You think there will ever be a black market for hard math problems that have never hit the Internet? Because everything that's hit the Internet will be eaten by the models. Yeah. But if I if I had in my possession, hypothetically For for the pleasure
Speaker 7:I see.
Speaker 2:Of persuading math questions. That has not been touched by an LLM. Are you in the market? If I had a math problem that was, that was going to be very hard but tractable to a human, but it's never been seen in the training data. No one's ever thrown a model at it.
Speaker 7:Unfortunate... I'm actually pretty far on the other end. I'm I'm so... Okay. I'm I'm an amateur mathematician, not a real mathematician.
Speaker 7:But I would say the the the full opposite, which is I think the LLM and and all the AI models over time will actually be much better than we are at coming up with
Speaker 1:Oh, interesting. Creative math problems.
Speaker 7:Like, I wanna show up and do the
Speaker 2:Oh, okay.
Speaker 7:Programming competition or whatever where the LLM came up with, like, the coolest problems.
Speaker 2:Yeah. Yeah.
Speaker 7:Yeah. It's like, wow.
Speaker 2:Yeah. So maybe the next step is designing the next the IMO Yeah. Question. Exactly. Like,
Speaker 7:coming up with the questions.
Speaker 2:Yeah. But but creating them in a way that certain teams can be better at one and it's super competitive.
Speaker 1:Come up with a problem that would stump me.
Speaker 2:How how have you processed the the millennium prize challenges falling and specifically generalization?
Speaker 7:Yeah. Yeah. My my my two cents on it is you know, I had this tweet a while back that just as soon as they they beat us on these math competitions, honestly, the the rest was obvious. Mhmm. Which I think is is is still really the case.
Speaker 7:I mean, people talk about, oh, is AI plateauing?
Speaker 2:Yeah. Is all the
Speaker 7:NVIDIA stuff? Is it whatever blah blah blah? Is it like is it time for a correction? And I I always, you know, the first thought I think is like, guys, we just solved Yeah. Like that.
Speaker 2:Yeah. Go back to the actual progress.
Speaker 7:Always. The question is is really like, we can debate all the time, know, people love talking about the stack and where people capture value and whatever. Yeah. But like, I think it seems very clear that the total net impact of AI, you know, is can we solve drug discovery? Yeah.
Speaker 7:Can we build all these like really, you know, can we come up with novel technologies? Can we go improve all the software in the world? If we are literally solving like the hardest math problems known to man that people have spent like probably tens or hundreds of millions of hours trying to go do and this they can go and do in a couple of days. Like, I think we will.
Speaker 2:Yeah. We're simply walking down the path.
Speaker 1:Yeah. No. I think... And I think we're gonna look at this period as like a golden era where everyone had access to exceptional intelligence, but the models didn't have like creative agency. Yeah.
Speaker 1:And it was an opportunity for humans to just like...
Speaker 7:You're saying we'll be in this golden era until the models themselves have the creative agency? Yes. Maybe.
Speaker 1:What I'm saying. But but but it's like right now it's like
Speaker 2:advantage. Take advantage.
Speaker 3:Take advantage. Yeah.
Speaker 2:It's a fun time for sure. How how do you think about... By
Speaker 1:the way, you do you do a lot of media appearances and you know talks and stuff. Can you believe how many how many of these Gavin and Antonio are doing? He he did... Apparently he did 18 talks yesterday.
Speaker 2:Wow. No. No. It crossed a few days. Oh, two days.
Speaker 7:I'm doing my fireside with Gavin as well.
Speaker 2:Amazing. Yeah. He's...
Speaker 4:Yeah. It's
Speaker 1:like that
Speaker 7:that'll be my... You know, that... For for me that's the the whole kind of event. But for Gavin, It's just what it is.
Speaker 2:Yeah. Yeah. It's crazy. What about... How how...
Speaker 2:Do you feel like you're developing a muscle for being able to predict the sequence of events? Like, earlier this year, I was I was reflecting on the fact that the hugging face sci fi agent swarm hacking thing happened before I saw the first person book a haircut with ChatGPT. Yeah. And it's like, nowhere in the sci fi story is, like, the agent swarm coming before the haircut. Yeah.
Speaker 2:And like people are just now like booking restaurant reservations, but we've solved I... We've solved the IMO before that. It it feels like an odd sequence of events, but do you have a better intuition for the way things diffuse?
Speaker 7:You're very generous and even... I have to say for sure I don't. But but but no, it's it's interesting because I agree. It's a lot of these things are non linear. I mean, there's a there's a Parthi term jagged intelligence Yeah.
Speaker 7:Which I've always really liked. Yeah. Great. It's basically just like, you know, think about these these things as levels of difficulty, but those are the levels for a human,
Speaker 2:you know.
Speaker 7:Yeah. For like an AI, it's like this thing is really hard and this thing is like actually really easy, which apparently includes solving all of mathematics apparently. Yeah. But but but but... Yeah.
Speaker 7:I I I think sequencing is there for sure. I I would say...
Speaker 2:Anyone can do the dishes. Very few people can solve a Millennium Challenge problem. And yet, we we have got to do the dishes. Yeah.
Speaker 7:I I think the... You know, seeing agents really take off for example, I think Mhmm. Encoding agents in some ways were kind of like you got a bit of a precursor in terms of like what things were going to get there with agents. Yeah. And naturally what did happen in coding agents, you know, in 2025 and to 2026 was basically like, okay, if you have the clean kind of like verifiable tasks like that was gonna work first and big repetitive things that you were going and doing over and over again.
Speaker 7:And then once you started to get to a point of like getting the consistency in the models where you could branch out and and do all sorts of actual like one offs. Obviously, that's when it made sense for some of these kind of personal agents to go and take off as well, right, like booking the haircut Yeah. Or or or something like that. I think... Yeah.
Speaker 7:It's it's it's an interesting... I I would guess that the... There's a lot more actually to come I think in the next shift of like really, I mean, as as we're saying, you know, with with with dev bots and all this, but but like moving more and more towards outcomes. Yeah. And I think the...
Speaker 7:One one way to put it is I think there's still for sure just like a massive diffusion bottleneck. For all the other things that we talk about in AI, it's like there's definitely still a big piece of like teaching folks how to go and use it. Totally. Whereas if you can actually go and deliver the outcomes, to say the obvious, I mean it's just a lot easier to go use that versus okay, here's this whole new tool that you have to learn all the details and you're gonna spend data. It's it's like, no.
Speaker 7:Like, I can actually just work with a teammate that's like, okay, here's this project that I wanna get done or here's this thing, for example, I keep, you know, buying me a new pair of pants or something that I just want to happen.
Speaker 2:Yeah.
Speaker 7:And the agent is good enough that it can just own that. It actually frees me from having to think about all the context
Speaker 1:of Have about... So you I can very easily imagine like persistent agents in the context of cyber security Yeah. Because it's this constant like defensive mechanisms Yeah. Scanning, scanning new code, things like that. Have you thought about...
Speaker 1:And and I can I can imagine like the the sort of like task, you know, like buying outcomes? That makes sense. But what are some other ways that you could see persistent agents showing up in in like general software engineering work. Yeah.
Speaker 7:Yeah. Yeah. It's it's very much this kind of like self driving software future is is maybe the way to put it. And I think the... You know, there are...
Speaker 7:If if we're gonna super generalize, it's like there are software tasks that happen because somebody came up with something that they wanted to go do net new and they go and do those. And then there's a set of software tasks that come up because there's just that from... Okay. Customer reported a bug. Alright.
Speaker 7:We have to go fix the bug. Or like, there's a new version that came up. We gotta go upgrade to the new version. All of the latter tasks kinda should be self driving and that's kind of what we're seeing. And so for example, like a lot of the early, you know, the beta users of of for like Fireworks or Reducto or Deckingon and so on.
Speaker 7:And and this is the kind of thing that they're going and doing. It's basically like, you know, for example, you have to go monitor your entire system for for performance. Right? And up until recently, it would still be like there was a human that's looking at the dashboards and pulling out things and saying, oh, wait. Why is this particular?
Speaker 7:Why are customers telling us that this particular flow is slower than it used to be? Okay. It seems like there was a regression here from last week. Maybe this commit, you know, did this change or something, so we have to go fix that and let's go look at the the code and and see what went wrong. Now it's like you can actually just have a persistent Devon bot that just goes and and owns that.
Speaker 7:Right? It's just like, okay, go through all the customer channels
Speaker 2:Mhmm.
Speaker 7:Listen in on the feedback. If you see consistent things where people are reporting that there's some like new regression that came up, just go and spin up a sub agent that goes and runs that and then investigates. And if you find it, then go just go fix that bug. You know? Yeah.
Speaker 2:How are you thinking about Devin for bio? Bio Devin. I I mean, I imagine everyone is predicting, like, a huge boom in bio. Yeah. You know, opinions vary about the timeline and how fast that takeoff will be, but, you know, every clinical trial has probably used Microsoft Excel.
Speaker 2:They will probably all use AI in one way or another Yeah. Regardless of how one shot able cancer cures are
Speaker 1:in their I near want my drugs, farm to table, handmade.
Speaker 4:Handmade? No AI at all.
Speaker 1:You're going
Speaker 2:to the witch doctor. Yeah. I want a mortar and pestle Yeah. Grinding it up. Yeah.
Speaker 2:But, I mean, I imagine that you're already in contact with many biotech firms. But how do you see that product going? What are you doing to get yourself set up for if a boom happens there, you're ready, you're in the right place?
Speaker 7:Yeah. Yeah. First of all, for pure bio, you know, drug discovery, I'm... You know, very different kinda field and I expect there's gonna be a lot of really great companies and really great products Sure. Doing that.
Speaker 7:Obviously, I think for us, no, we've seen some of these signs. I mean, you know, Brian Johnson actually was a... Was an early beta user of Oh, as way. Well. And he used it to go build basically all the tools and analysis that he does on his measurements and make sure like things are, you know, up to par.
Speaker 7:Are there any things that like I should be noticing or concluding from this? He obviously has his own entire kind of like Yeah. Panel of tools that he goes and looks at and basically used a Devon bot to go and make all of those systems, but customize them the way that he wanted. Yep. I think there'll be a lot more of this in terms of just like, you know, I think the...
Speaker 7:We're super happy to to... I've been super excited to go and work with folks who who kinda like wanna work on particular applications. But in general, like this form of like, I'll call it like personalized software. Think we're gonna see a lot
Speaker 5:more of.
Speaker 7:And and it's basically just like there are so many problems that you wish you could solve or you would wanna have, you know, a tool that goes and does. Like that's all software. The reality is just like it doesn't make sense at all to go pay a human team of 30 engineers, you know, to go work for a month to go build this app for you. But like now actually, it is just like Totally. You know, having having an agent that goes and owns that can be your own personal
Speaker 1:How are you thinking about your own software and vendor supply chain in this cyber era? Are you wanting to take more things in house just because you can own it end to end and and potentially have more confidence? Obviously, building something yourself can create its own set of problems and and issues but what's your thought process there?
Speaker 7:Building stuff yourself has its own set of issues to your point. I mean, I I think the... I think being very thoughtful about like what kind of data you're sharing with vendors. You know, it's like it's not just here are all the vendors that I have. It's like, okay, which ones are getting customer data?
Speaker 7:Those have to be absolutely absolutely patched down. Sure. Which ones are getting... Know, there's there's just like different levels of obviously trust and diligence that you wanna go and do depending on on what you're doing. And and I think there's like a lot of actually like a lot more eighty twenty that folks can be doing even on just thinking about, okay, like this this this particular kind of like thing that you're working on or like, does it have to have, you know, customer sensitive data?
Speaker 7:Like, are there ways that you can go and anonymize or whatever you need to go do that that makes all that safer? I would say I I I don't, you know, despite us being, you know, building Devon and all the things, like I I don't think that is the the the case that we're at the kind of full like SaaS pocalypse build all software in house thing. And a lot of that is actually because yeah, mean to your point, it's it's like there are a lot of details involved in going and managing software still. And I think the the reality is like making it 10 x easier to go and build software is amazing. It gives you a lot of leverage to go do a lot of the things that you wanna go do.
Speaker 7:It doesn't actually change the math of like what is your core competency that you really care about versus like what are all the other things. You know, I mean like something like Salesforce with which people have probably... I don't know. You could... Somebody can do the back of the envelope calculation of how many hours of software engineering we'll put into that.
Speaker 7:Maybe it's like, I don't know, ten million hours or something. Yeah. You probably don't wanna spend one million hours going and doing it. Like you could, but but, you know, the the purpose of software, I would say, is to kind of like to do the replication that like frees you from a lot of those things. Yeah.
Speaker 7:And and so I think there's like still obviously like a lot of that that that should happen even as this gets more and more efficient.
Speaker 2:How much time are you spending on naming? It feels like that's like kind of the hardest part of an AI company is like you come out with... Hardest part, John. No. No.
Speaker 2:No. Sometimes it is... It... It's it's sometimes, like, vague because, you know, Devon was a bot, but then you need to reintroduce the concept of long running Devon. So Devon bot, you know, and and and...
Speaker 2:But every company's gone through this with, like, the different model name. It's been the ultra fast and the old... And the max and ultra and all these things. It's all very confusing.
Speaker 7:Yeah. Yeah. Yeah. And it's also as as you mentioned, it's like all of it's kind of unpredictable too. It's like I think the the joke with like the the the open air anthropic model, each one was like, it means it's logically consistent at the time.
Speaker 7:Yeah. They didn't realize, okay, there was gonna be a new reasoning model. Yeah. Exactly. Wanna have like a
Speaker 2:small model. Yeah.
Speaker 7:No. We we thought of like a bunch of different names for Devon bots and so there were
Speaker 2:Yeah.
Speaker 7:Super devins and there's auto devins.
Speaker 2:Yeah. We
Speaker 7:were saying, oh, should we use Otter? Because Devon is an Otter, you
Speaker 2:know Okay. Yeah.
Speaker 7:Like, odds.
Speaker 2:Yeah. Anyway Okay.
Speaker 7:We had... In the end, chose to obviously, like,
Speaker 2:keep it
Speaker 7:pretty pretty simple, but but but... Yeah. Yeah.
Speaker 2:Sometimes you can ship... People say ship your org chart. Sometimes you can ship your architecture.
Speaker 4:You can say, okay.
Speaker 2:We have you have a reasoning thing, the user might not care. Anyway, thank you so much for coming on the show. Scott, always a great catch up. Great. Thank you.
Speaker 2:We have TK coming on. Let me tell you about Cisco. Critical infrastructure for the AI era, unlock seamless real time experiences, a new value with Cisco. And let me also tell you about Codex. Codex is a powerful workspace for getting work done with AI agents.
Speaker 2:Whether you're writing code, analyzing data, creating content, or automating business workflows, Codex helps you move projects forward from start to finish. And TK is about ready to come on TBPN here at TomorrowX on Wednesday, 11/18/2026. While we're bringing him in, let me tell Christmas. Let me tell you about Christmas.
Speaker 1:Here he is.
Speaker 2:How many days? Thirty seven days till Christmas. Welcome, mister OTK. How you doing?
Speaker 4:Is that
Speaker 1:boy on?
Speaker 2:Please
Speaker 1:throw him on.
Speaker 2:Here we go. Yeah. There you
Speaker 4:go. Switched out my little Alienware ones. Was nice. Always walk around with.
Speaker 7:Okay.
Speaker 4:You guys... Alienware Pro.
Speaker 1:Oh. Those are... It's kind of like a little bit of a throwback. Were you a gamer growing up?
Speaker 4:Guys, it's legit.
Speaker 1:It's legit.
Speaker 2:Doesn't Dell own Alienware now?
Speaker 4:I don't know. Might. I think he bought it. I saw him and he's like, dude, Alienware.
Speaker 2:Nice. No. No. No. He owns it.
Speaker 2:He owns it. That was... When I was growing up, that... Getting an Alienware PC was like the Ferrari of gaming. It was like you had to get Alienware.
Speaker 4:So I'll be anywhere. It'll be super noisy. It could be here. Yeah. Could be at a Beyonce concert because you never know.
Speaker 2:You never You gotta take a call.
Speaker 4:Rocket. Throw it off. You're locked in.
Speaker 2:You're locked
Speaker 4:in. You don't hear anything. Doing deals. Other people think you're just locked in. Yeah.
Speaker 4:It's...
Speaker 1:Like on a wake like if you're at wakes wakeboarding or water skiing.
Speaker 2:It was waterproof?
Speaker 4:It seems like it might be. It's pretty interesting.
Speaker 2:But it
Speaker 4:might fly off. Yeah. This is this is some... That's a really interesting...
Speaker 1:Yeah. You need that. I I I can see you water skiing in pure silence.
Speaker 2:Woah. I kinda
Speaker 4:feel like I'm at like NFL Combine or Super Bowl or something like this. I'm very excited about everything going on.
Speaker 2:Yeah.
Speaker 4:That cowboy hat's also legitimate.
Speaker 1:Yeah. Do you the function of a cowboy hat?
Speaker 4:Yeah. It's about about the sun. Yeah. It's so good. So like the way you know so if you're on Lake Austin all the time.
Speaker 4:Yeah. Yeah. You know this. It's not even like... Yeah.
Speaker 4:Of course. Yeah. I put a cowboy hat on when I'm cruising around captaining the boat with my friends. Yeah. Cause like I'll I'll just get roasted otherwise.
Speaker 2:Yeah. Course.
Speaker 6:It's a
Speaker 4:real thing.
Speaker 1:It's a
Speaker 4:real thing. Works.
Speaker 1:It's a real thing. This last couple weeks had to have been like PTSD for you?
Speaker 4:No. Mhmm. Not PTSD. Just like there's wisdom Mhmm. That was gathered over time.
Speaker 4:Mhmm. I mean I went through it
Speaker 2:Yeah.
Speaker 4:For sure. And there's a friend who was going through it. And for me, it's so clear what it is. I was probably one of the early people to be like, alright guys, here's the 12 principles of how to fend off an op. Yeah.
Speaker 4:But that was like Monday of last week.
Speaker 1:That was that was like...
Speaker 4:That was Monday of last week.
Speaker 2:How did you identify that it was an op? Like it triggers that for you?
Speaker 7:You know what I mean?
Speaker 1:It's just like...
Speaker 2:For a lot of people, it's just like, oh, there's a bunch of news articles. Something's happening. No one's behind this. It's just
Speaker 4:just me. It's like, you know, it's a sixth sense, you know. Yeah. I don't know. It's like the...
Speaker 4:It's like when you see those murder mystery things or those true crime things and there's like this woman who basically knows where the crime happened and all the details but she's not connected anyway. She's a psychic. Yeah. That was... That's my psychic.
Speaker 1:You're a psychic for corporate ops.
Speaker 4:I'm I'm a psychic for these types of things because I've seen a lot of them. Mhmm. Mara, for for me it was six months Yeah. Of ops. Yeah.
Speaker 4:One a week. Yep. Steady as she goes. Sunday night. Every Sunday night, it would hit.
Speaker 4:You're, like, sitting there. Literally, like,
Speaker 2:hot work.
Speaker 4:Sunday Sunday afternoon, you're like this. You're like, oh, what's gonna happen today? Wow. Six months. Yeah.
Speaker 4:Beatdown. Beatdown.
Speaker 1:Right. Well, the craziest thing with this timeline is we... Like from from our point of view, we... And then October tenth of last year, we were talking about Mega Speed. We were talking about Alice Wong.
Speaker 1:We were talking about Shell corporations. Like Yeah. Whole situation was was like interesting to say But the but then, yeah, you have this thirteen month lag before anything really happens. It's an odd. Yeah.
Speaker 1:Obviously.
Speaker 4:Yeah. It's like she's a spy, but wait, we don't really know what the definition of a spy is nor she's probably not a spy.
Speaker 7:Mhmm.
Speaker 4:Yeah. There's like these export control problems, like actually they didn't real... What do they do illegal? Well, they didn't. It was following the law, but the spirit was somehow...
Speaker 4:They're like, but but then they introduced... He he introduced Alice to Alice Wang to to Alibaba, which of course is the first thing a Chinese spy needs is somebody else to introduce them to Alibaba. That's like, of course, oh, and by the way, that introduction was happened before there were any export controls in the first place. Come on. Yeah.
Speaker 1:Yeah.
Speaker 2:Yeah. So how do you fight back against an op? I mean, you you you put out a post with some detailed information.
Speaker 4:Was I I should just get the printout. Maybe I should just pull it.
Speaker 2:I'll pull it up.
Speaker 4:I could I could tell you little pearls of wisdom on on each of them. Hold on a second, guys.
Speaker 2:Let's dig through it.
Speaker 4:Is it your pinned tweet?
Speaker 1:Lightning round. Op defense lightning round.
Speaker 4:Let's find it. Hold on. I'm pulling
Speaker 2:it right now. Jensen back.
Speaker 4:I did... I've done a lot
Speaker 2:of tweets this week. A little time on my hands this morning to do some writing. Send along to someone who you know who may benefit. Practical principles.
Speaker 4:This is literally... By the way, this is super interesting because A
Speaker 2:modern day struggle. I have it right here if
Speaker 1:you wanna read it.
Speaker 4:Alright. It's like I need to read my own stuff. It's so weird.
Speaker 2:Number one, own what you did but fiercely resist accommodating the mob. Resist owning half truths or actions slash behaviors you didn't So
Speaker 4:what happens is these ops are... They're... It's about mob mentality. Mhmm. So what happens is the large mob of people basically is brainwashed into thinking you're a bad person and did incredibly bad things.
Speaker 4:Now there's usually some thread of truth Sure. But it's a fabrication in sort of the nature of what occurred etcetera. Right?
Speaker 1:And the truth here was the affairs. For for instance. Yep.
Speaker 4:Yep. Then you're like, okay. Well, the problem can happen if you're a human and you're like empathetic generally. You're very susceptible to just taking on what the mob thinks. That truth no longer is what is true or objective, it's only what the mob thinks.
Speaker 4:Mhmm. And you've gotta avoid that at all costs. And you need to own the part, your part in it, but you need to fiercely resist the psychological need to... Or psychological vulnerability of taking on what other people think is true. Mhmm.
Speaker 4:Because if you don't resist that, you're done.
Speaker 2:Mhmm. Two. The op is not about your past actions. It's about your enemies and this event as their opportunity.
Speaker 4:So this is a little bit about...
Speaker 1:And that's and that's what why my alarm bells were going off is because we... It wouldn't have taken much opposition research to figure all this out Mhmm. Thirteen months ago. And in fact, I do believe that a bunch of people knew exactly what happened.
Speaker 4:There there's like that one short seller who put it out there Yeah. May or April
Speaker 2:or Yeah. Why didn't that get traction but this did?
Speaker 4:And like the heat may have been part of the op too. He he may not even know it. But like somebody said, hey, check this out. Okay. Yeah.
Speaker 4:You know? Yeah. That's how this stuff happens. Yeah. Interesting.
Speaker 4:So the... So on number two Yeah. Look, again, if you do number one and you own your truth Yeah. You own what happened and what didn't happen, don't own what didn't happen, You need to realize you need to clearly identify your enemies Mhmm. And move forward.
Speaker 2:But do you need to go public with an attack... Counter attack on your enemies? Or should you sort of turtle shell and keep your mind right?
Speaker 4:Look, guys, I did the turtle shell thing.
Speaker 2:I don't Yeah.
Speaker 4:I wouldn't recommend it.
Speaker 2:Okay.
Speaker 4:I think you get... Look, the bottom line is you need to get a war room up Yep. And like strategically handle it. Some things you go, you know Yeah. We don't need to do this.
Speaker 2:Yeah.
Speaker 4:Some things you go lean forward Totally. And it's a strategic response. Mhmm. And not just one at a time, which is where I was. Sure.
Speaker 2:But... You're playing defense
Speaker 4:every Monday morning.... Reaction. No Sunday night. And you get the you get the text from your comms person. It's like, hey, we need to talk.
Speaker 4:Yeah. And you're like, you literally bend over. You're like, oh. Oh, no. The pain begins and it's almost like a physical response because it's a Pavlovian physical response because the pain you went through the four, the week before that Yeah.
Speaker 4:10 before.
Speaker 1:They're just training you. That's Sunday night pain time.
Speaker 2:Pain It's
Speaker 4:it's really... It's crazy.
Speaker 1:But... Yeah. And I do remember
Speaker 4:saying that's why it's crazy I don't have PTSD.
Speaker 2:Yeah. Yeah.
Speaker 4:Yeah. But if today
Speaker 2:Yeah.
Speaker 4:Anybody texts me and says, hey dude, give me a call. You see, I'll respond and go, I'll
Speaker 2:I'll respond and go, dude.
Speaker 4:Can you put this on a scale of one to 10? Like, I'm I'm throwing balls right now.
Speaker 1:I do think that there was some... I do think there was clearly some... You know, I don't know. I don't know Jensen. Heard him talk.
Speaker 1:Mhmm. Obviously, bunch of hours. But I do think there was some kind of denial going on because you hired the crisis comms firm last year. And you had every opportunity to lead from the front and potentially like take a leave of absence, which still would have sent the stock down. Mhmm.
Speaker 1:But you wouldn't be in the situation where you resigned. Yeah. And then now you have to get other people to kind of rally and say, hey, we we we actually want this guy back.
Speaker 4:Yeah. Now look, I think this gets into the personal side, but like Jensen is a very, very good person. He's so good that that, you know, when there are personal failings, it's like devastating. Mhmm. And I think it just...
Speaker 4:It probably... And again, I... This will be a little bit of speculation on that, but like it's... It just took a little time for... To to work through that, I think.
Speaker 1:Mhmm. Yeah.
Speaker 4:And and again, there's no shirking of responsibility. Mhmm. But if you if you get affected to such a degree that you can't then go to, you know, point two and point three here. Point three is no matter how bad it gets, nothing is as bad as it seems. I'm happy to talk about that as But like, you can't get to the other points if you don't first go, I accept my truth.
Speaker 4:I take responsibility for it. I did things that... Not me. I'm
Speaker 2:just saying Jeopardized.
Speaker 4:You know? Yeah. Then you go and you say, what what is the best version of me today? Yeah. And it's not...
Speaker 4:By the way, it's not letting NVIDIA go to the suits. Sure. I promise you that's not the best version of Jensen Yeah. Today. Okay.
Speaker 4:I promise. Yeah.
Speaker 2:We we we Yeah.
Speaker 1:Debated this. Going into the earnings call yesterday, I mean, the the team obviously did the absolute best they could given all the chaos. Sure. But it was like... It just immediately was like, it wasn't...
Speaker 1:It was a different company Yeah. Like, away.
Speaker 2:Yeah. Take me deeper on the pitfalls of letting a company go to the suits. Because I can imagine, you know, a Tesla without Elon. It just collapses. It doesn't work.
Speaker 2:But then I go to an Apple post jobs where, yes, you could say the innovation slowed down. There were product misses, but financially, that company continued to perform arguably because of all the work Steve Jobs did. Yeah. But is there no chance for NVIDIA under suits?
Speaker 4:Here's here's the thing. Yeah. First, the Apple transition was a planned one. Yep. K?
Speaker 4:And it was an optimistic one. Totally. What I mean is the day after Steve passed Yeah. They weren't talking shit about him. That's a good point.
Speaker 4:They weren't like Yeah. We need to rid the ranks of all people that are for Steve. Totally. Didn't happen at all. They celebrated
Speaker 2:Yeah.
Speaker 4:Him and what he built.
Speaker 2:They also had Apple University where they were training everyone in the Steve Jobs mindset in the
Speaker 4:in the methodology. Yeah. Obviously, you end up in a different Continue culture. Yeah. Okay?
Speaker 4:What can happen when a new person comes in and it's done by a... There's a board element to it like there is here, is that the board of course wants to look smart Mhmm. And not stupid. Mhmm. That certain decisions were made that they approved or didn't approve, whatever it is.
Speaker 4:Right? So okay, they brought the new person in. They need to look smart. So they need they need the new guy. They need that decision to look good.
Speaker 4:That new guy, his number one KPI, you guys wanna guess? Keep my job.
Speaker 2:Yeah. Keep his job. Like, no chaos.
Speaker 4:Keep my job.
Speaker 2:Like Yeah.
Speaker 4:The idea of like when the new guy ultimately finally get a CEO, that person, his or her job will be keep my... The number one KPI is keep my job, which means that is a natural resistance to Jensen coming back. Yeah. And so the... You know, ideally you can get an executive that comes in Mhmm.
Speaker 4:That is also empathetic and open to Jensen coming back because if that person is not really gnarly, then you gotta go... Then you go annual meeting sort of proxy votes for for board members, that whole thing.
Speaker 2:Yeah.
Speaker 4:Which is still a thing. I think that that's gonna happen in June. There's some chance you could do it earlier. Okay.
Speaker 2:Yeah. Yeah. So are you optimistic at all about Jensen coming back? Do you think it's 5%, 10%,
Speaker 4:50%? What does Poly say
Speaker 2:right now? I think it's been at, like, 1%.
Speaker 4:Come on. Shut up.
Speaker 2:It's gone to 8% when there was some rumor, and then the board shut it down with their letter. And so they've been trading letters back and forth. I'll have Jordy actually pull it up.
Speaker 4:But Poly's obviously got got it it wrong. Wrong.
Speaker 2:So you're a buyer?
Speaker 4:Yeah. Not for the price. Dude, I'm a buyer. I'm a buyer. Yeah.
Speaker 4:Look, I I... It's hard to handicap sort of like on a percentage basis.
Speaker 2:Yeah. Because it's just... Is
Speaker 4:that the most important thing that could happen happened yesterday, which is Jensen saying, I'm down to be back. Yeah. And by the way, that means that means he's not he's not ashamed of trying to come back. Yeah. He's like, this is the right thing.
Speaker 4:Yep. He's done all those steps. There's like a bunch of them here.
Speaker 2:Yeah.
Speaker 4:And then it's like...
Speaker 1:Prediction marks, it's have him at 11%. Okay. 11%. 11%. I'm still a buyer.
Speaker 1:Yeah.
Speaker 2:There's a chance.
Speaker 4:I'm still a buyer.
Speaker 2:Okay. Yeah.
Speaker 4:So so so yeah. So him leaning forward then means all people Mhmm. Which is a majority of the people that work in NVIDIA Mhmm. Majority of the shareholders
Speaker 5:Yeah.
Speaker 4:And I would probably put out there, majority of humanity Yeah. Is super pumped to have Jensen back. Yeah. And so those people now have something to hang their hat on because the guy they want back has said let's freaking go. Yeah.
Speaker 4:And he was like, lowercase LFG? But we're gonna get uppercase. It's it's gonna happen.
Speaker 1:We had a funny moment last week, of course, in LA. We were talking to an actor. They said
Speaker 4:That's like so everybody. Yeah. So like your waitress.
Speaker 1:Yeah yeah yeah. He's a real real actor. Professional. Gets paid to do the job. And we we asked him like what do you...
Speaker 1:He didn't know who we were. So we were just kind of messing with him. We were like, do you think about this whole AI thing? And he said, well, just think it's so so beautiful that human love and lust took down AI. It ended AI.
Speaker 1:Referencing Mhmm.
Speaker 7:You know, NVIDIA. Storytelling.
Speaker 2:This is the end of AI. There won't be any more AI models. There won't be any image models. No more video models. Like, it's all over.
Speaker 1:The techno revolution will continue.
Speaker 2:I I I think you're not gonna wanna tune into the IPO. Look. Think here's
Speaker 4:the thing. Right? There there's there's the op thing and my support of founders generally. So there's an op thing. There's my support of founders generally.
Speaker 4:There's Jensen being a good guy who's... Okay. Personal failings, but a good dude who deserves support. And oh yeah, there's one more thing which is the future of humanity is on the line. Mhmm.
Speaker 4:Do we want an overlord? Mhmm. Or do you want to have... Do you wanna use, consume, utilize, benefit from AI on your terms?
Speaker 2:Yeah. Yeah. Gavin was talking about this.
Speaker 4:That's what this is about. Yeah. I don't know. I I don't want an overlord. Yeah.
Speaker 4:You know, I'm just putting it out there.
Speaker 1:Yeah. What if they gracefully love you?
Speaker 4:What is it... What is... What was it? The movie Her? Yeah.
Speaker 4:Okay. So, you know. Oh, dude. Okay. Nice.
Speaker 4:That was the essay. Okay. Did Dario's essay. Yeah. Okay.
Speaker 4:But on Her Yeah. It was so wonderful and romantic and you know you know what happens at the end. Yeah. Like the AI just gets bored of you.
Speaker 2:Yeah. At least.
Speaker 4:You're like peace out. You know?
Speaker 2:That'd be a weird outcome. I I haven't thought about that. No one's modeling that. Everyone has Pee Doom. No one has Pee b Boom.
Speaker 2:Boom is different than p. The AI just says Okay.
Speaker 1:But I
Speaker 7:just take
Speaker 4:the doom
Speaker 2:and you
Speaker 4:flip it.
Speaker 1:Look. I I like the I like the the effort. Yeah. I'm a little concerned that people just associate artificial super intelligence, intelligence explosion with like literal Armageddon and explosions. Well,
Speaker 4:no. That's because you have Dario out there talking about it all the time.
Speaker 2:Yeah. Yeah.
Speaker 4:Which is crazy. Like, dude, watch out. It's gonna kill everybody. This is a large percentage shot of killing humanity.
Speaker 1:Yeah.
Speaker 4:Oh, and then the next day is like, dude, we're getting a bio lab up
Speaker 2:and running.
Speaker 4:It's gonna be awesome.
Speaker 2:Is a whole mixed messaging sometimes. Like, what's going on now? Yeah. One thing that's absent from here, I think, is get all the employees to sign a petition. Is that helpful?
Speaker 2:Is that a sideshow? What do you think about that? It's been surprising. I saw... I've seen NVIDIA employees public...
Speaker 2:Publishing on their public Instagrams, bring him back. So clearly ground swell.
Speaker 4:Start, but it it look... Piticians are nice. Yeah. Walk out better.
Speaker 2:Walk out better?
Speaker 4:Oh, yeah. Oh, yeah. Like... If they all walk out.
Speaker 2:Yeah.
Speaker 4:Okay? If the topic is...
Speaker 1:Came here, filled the stadium. Jensen walks out, c o two cannons.
Speaker 4:I'm back. If they walk out Yeah. Like I'll be at HQ with a wafer encrusted megaphone Mhmm. With founders, COs Mhmm. Their armies.
Speaker 4:Mhmm. Then the LFG goes from lowercase to uppercase. Okay. By the way, I'm saying all this. There's gonna be an op against me starting tomorrow, Derek.
Speaker 4:They're gonna be like, we gotta take this guy out.
Speaker 2:Hey. Hey. You're you're just speaking speaking your mind. Yeah. I think it's fine.
Speaker 2:I mean, you're the one you're the one to, yeah, I'm kind on this. Everyone's everyone's happy to hear from you this time on this. It's very helpful. Yeah. So thank you.
Speaker 1:What's the update on that? Let's move Talk about you.
Speaker 2:Yeah. Last question.
Speaker 4:Mining awesome.
Speaker 2:Yeah.
Speaker 4:I think we're probably, you know, a huge number multiple bigger
Speaker 2:Mhmm.
Speaker 4:Over the next twelve months than we are today on mining. Because it's just going through that ultra hyperscale or hyper growth. The whole thing I like to say, the vertical, it's so vertical that the curve turns to the left. So we're... So that's happening on mining.
Speaker 4:We're really excited about that. That's a lot of fun.
Speaker 2:That's great.
Speaker 4:Food, we got food robots. Our food robot factory. Yeah. The factory that makes the food robots is going up in Pittsburgh q one. Yeah.
Speaker 4:So then we're just gonna be printing a ton of these machines that then help restaurateurs make a better living, put food on their own table. Super exciting.
Speaker 1:Do you
Speaker 4:What about couriers? What about automated couriers that are moving the food?
Speaker 1:How does Pittsburgh feel about like what what have you learned from the the sort of failings of data center developers to like get people excited about infrastructure being built in their in their backyard or town? Is that...
Speaker 4:Look, I think that's also... I mean, I don't wanna call everything enough, but that's that's like a big red herring. When when everybody got really nervous about the the data centers, look, I understand that it affects near term plans to like grow, train models Sure. Do inference. It's a big deal.
Speaker 4:But like, you're not gonna have a situation where every single state and every town and every state is like, I don't want this.
Speaker 1:Sure.
Speaker 4:Somebody's gonna be down for the 100,000,000,000, $200,000,000,000 in infrastructure spend. Yeah. Somebody's down.
Speaker 2:Yeah.
Speaker 4:Yeah. So it's going to go where it needs to go and there'll be certain states that are super great. But obviously, it slows us in the near term. But it's like it's a bit of a red herring. To be honest, I mean, we had know that.
Speaker 4:These things don't... You can easily have closed loop water systems. Yep. When you do... No.
Speaker 1:But I mean... But I even in your new factory in Pittsburgh, what is that? What is... Are you going in there and like, do you want the mayor to be like pumped that you're that you're coming in? Or is it more abstract than that?
Speaker 1:Like there's an existing... We
Speaker 4:should we should do that. We should do that....
Speaker 2:Wrote the playbook on this with Uber. You go into a city, get a bunch of people excited, and then they're your advocates. And I've never seen an AI company do this. And I and I know that in a certain town with that is like AI
Speaker 1:AI companies don't meta. Don't need
Speaker 2:a data center in your backyard
Speaker 1:to get the benefits of AI. So it's different. Uber, you need an Uber there to to be able to call a car.
Speaker 4:On the food side, we've got a pretty cool event that's happening sort of like q '1 ish, like maybe end of, you know, in that neck of the woods. We call it robot day. Fantastic. It's gonna be good.
Speaker 2:I'm very excited to cover It's gonna be That's gonna
Speaker 1:be amazing. Good.
Speaker 2:Yeah. So Well, thanks
Speaker 4:so much for coming on
Speaker 2:the show.
Speaker 7:Good to
Speaker 4:see you guys.
Speaker 2:We'll close out.
Speaker 1:See you.
Speaker 4:Yeah. We'll close
Speaker 2:out with TBPN. And Spotify. Sign up for newsletter tbpn.com. And we will see you tomorrow back in LA, back in the Ultra Dome, 11AM. GT six is watching Texas.
Speaker 1:We better be recording this on your boat. Let's go
Speaker 2:on the boat. Go on the boat. Let's do it.
Speaker 4:I mean, if it's the next couple months, we'll be in wetsuits.
Speaker 2:Oh, okay. We're in. We're We're in.
Speaker 1:In. We're We're We're