History’s Largest Oil Disruption, Oil & AI, Sundar's New Pay Deal | Alex Epstein, Dr. Alex Wissner-Gross, Charles Lamanna, Julien Bek, Eoghan McCabe, Michelle Volz

You're watching TVPN.

Today is Monday, March 9th. We are live from the TVPN UltraDome, temple of technology, fortress of finance, ramp.com, timeismoney, SafeOdds, business, corporate cards, bill pay, accounting, and a whole lot more all in one place. We have a fantastic show. We were able to get Alex Epstein to come down to the UltraDome in person on short notice to get us up to speed on what's happening in the oil markets. Of course, oil is all— it's gushing all over the timeline, all over the Financial Times, all over The Wall Street Journal as part of the—

Alex had a feeling that oil would be important in the future when he wrote the book, Fossil Future.

Yes, which we have right here, which if you haven't given it a read, go pick a copy up and stay tuned for him joining us at 1:10 today. Let's pull up the Linear lineup and tell you who we have coming on the show today. Linear, of course, is the system for modern software development. 70% of enterprise workspaces on Linear are using agents now.

We have Dr. Alex Wissner-Gross over on Eon Systems. He went very viral over the weekend for connecting a biologically derived fruit fly connectome to a MujuCo physics simulated body. And we're going to be talking all about simulating fruit flies and what will come after that.

And we're also going over Microsoft Copilot co- Julian, which we're gonna call Coco. Coco, okay. Disney might like a word. They have IP in that space in particular, but we'll see.

I see they have it in the enterprise.

I don't think so. Julian's joining from Sequoia Capital. Owen McCabe from Intercom, our partner, is joining because they just raised a massive round. And Michelle Vohls is launching Pax VC. She left Andreessen Horowitz, I think, about a year ago and has been is now launching a new early-stage fund. Anyway, let's go through the timeline and see how people are processing the news of the Strait of Hormuz and the oil price spikes.

We— Fox News this morning. Yes, they were saying just basically their advice to the, to the captains was to man up and just send it.

Yes, that was it.

That seems effectively their advice. I'll read the—

Please stay safe.

I'll read the actual quote. A guy, Brian, said, "If you want to diminish the Iranian threat, if you want to make sure this ends with complete capitulation, show some guts and go through that strait." That seems very, very risky right now.

Stay safe out there if you are on a shipping vessel. Do whatever is responsible. There are some crazy twists happening. So apparently ships in the Gulf are declaring themselves as Chinese vessels to dodge attack. This is from the Financial Times. The Financial Times, at least 10 vessels have changed transponder messages in an apparent attempt to avoid becoming targets. A clutch of vessels trapped in the Gulf under enemy fire are adopting a tried and tested ruse to avoid attacks. They're changing flags, using transponders to declare themselves to be Chinese. There's always been a very odd, like, just like tug of war between how ships identify themselves, because often for tax reasons, they're bought in one country, operated by individuals from another company, but they fly a different flag to be able to go from one place to another. And it's all based on like the port systems. I don't fully understand it, but it is very interesting. At least 10 ships over the past week have altered their destination signal to read Chinese owner, all Chinese crew, or Chinese crew on board. About 1,000 ships are currently shut inside the Gulf and its immediate surroundings with a cumulative value of $25 billion.

And I don't know if you've seen some of the maps that show the strait. Where it looks like nothing is actually moving through. I think in actuality, a number of the ships are actually turning off their transponders. So you can't see the movement because they're basically moving a little bit, going through the strait, and then turning back on again.

Well, there are plenty of outlets that are covering the Iran conflict in detail. We, of course, will be focused on the business and technology impacts. I wrote about what oil prices mean for the AI buildout and data centers. It's just sort of interesting to dig into the deeper supply chain of artificial intelligence. But there are some posts that we should go through around the oil story. So crude oil is 5 standard deviations above its 50-day moving average. Statistically speaking, this occurs every 9,500 years. So the last time would have been about 6,000 years before Moses parted the Red Sea. Imagine what that did to shipping in the area.

Area.

Fanciful. What else is going on? The COBSE letter has been posting a lot about oil prices. US oil prices are now expected to rise above $100 a barrel this month. That already happened. This was posted on March 6th, with markets pricing a 66% chance that it happens. The last time oil prices were above $100 was July of 2022.

Let's pull up this clip from Landman.

I haven't seen Landman. Have you watched it? Is it good? Let's, uh, well, you want oil to live above $60 but below $90. And don't get me wrong, we're still printing money at $90, but gas gets up over $3.50 a gallon, it starts to pinch. It hits $100, every product in America has to readjust its price. $78 a barrel, that's about perfect. Yeah, it brings enough profit to keep exploring, but it don't sting as much at the pump. Unless of course you're in California. I mean, they tax the shit out of it out there. It could be $45 a barrel and it's still $4 a pump. I don't know how those son of a bitches do it out there. It's movie day. 2020, a barrel of oil was worthless and this place became a ghost town. And nobody's immune. Kids have to quit college. Trucks get sold or repoed.

We can pause it now. We got the gist of it.

Yeah.

Well, this is the largest supply— How old were you when you discovered that gas is really expensive in California, not just because there's a lot of demand for it out here?

I mean, I discovered it when I was filling up in Montana, and it was like $2 a gallon. And it's like $5 a gallon here. That was pretty wild. The first major gas price shock that I noticed was Hurricane Katrina in my life. I think I was in high school, because I was too young to—

I was talking about just the fact, realizing that so expensive because of the taxes that California puts on it.

Yeah, because it's nice to drive around here. Put the top down, the weather's good, so they're like, yeah, you're going to pay. You're going to pay more for the joys of driving an internal combustion engine car. Now, the price in California is aggressive. I don't know the structure of the prices, though. Is it percentage-based or fixed? Because that affects how much the price will move based on oil price shocks, because if the price per barrel doubles, but the tax is flat, you don't feel it as much here as you do other places. I don't know. Anyway—

I'm pulling it up, but you can run through this.

It's the largest supply shock by a factor of 4. The Hormuz blockade, which is current, 20 million barrels were lost in supply. The Iranian Revolution in 1978 was 5.6. The Yom Kippur War embargoes in '73 was 4.4. The Iraq-Kuwait War was 4.3 in 1990. The Iran-Iraq War in 1980 was 4.0. And the Ukraine-Russia invasion in 2022, which is the last time that oil spiked over $100 a barrel, was 1 to 3. So, an absolutely huge supply shock. And I'm sure it will have a lot of implications all over the economy. With triple-digit prices, here's what's going to happen now, says Policy Tensor. Markets will scream when they open tomorrow. VIX will surge to levels beyond what we saw in April. The sell-off will continue for some time as intermediaries shed risk. And the markets are red. They have been screaming today. The VIX futures curve already has inverted, bid up by dealers looking for insurance. This predicts a massive sell-off. The pressure on this captured White House now The pressure on this captured White House now being in earnest. That's kind of sort of oddly written. Anyone can tell them that if this persists for very long, it will destroy the Trump presidency and gut the GOP for a generation. The controlling factor here, as I have told you over and over again, is that the United States does not have the military means to reopen Hormuz. There's no military solution in sight. This means that not only does Iran have the strategic upper hand now, it means that Iran enjoys the unambiguous strategic advantage. All they need to do is keep the thing closed until he capitulates. I put my neck out far to call this in advance. And someone told him yesterday I was in the minority, perhaps even a minority of one. No longer. I was correct. Just calling a shot. The blob heads and scribes were incorrect in their assessment of the strategic situation, and now markets will price to reflect reality. And we briefly touched on Scott Sumner's blog post on Substack at the end of Thursday's show, maybe Friday's show. We didn't get a chance to read it. I actually read through it earlier today. It's pretty interesting. It basically makes the case, you know, it has a very controversial title, Freak Out. And the thought is that like he's doom posting. Maybe it's about AI, maybe it's about this particular conflict. He is more just reflecting on this dynamic between when the market freaks out, it acts as a moderating effect to policy. And so he gives a number of examples around like tariffs.

Or more specifically, the admin.

Yeah, yeah. Like the tariffs caused this massive circuit breaker, 5% sell-off in the market. And then that was internalized and very quickly adjusted. And there were a whole bunch of different carve-outs to sort of like create soft landings. And so he's actually— and what he's getting at is that after the fact, a lot of people say, look, you didn't need to freak out. Because the taco happened. Trump always chickens out. The actual proposed policy effect didn't go into effect. And his point is that, well, it's precisely because people freaked out that it didn't go into effect. So freaking out is good in Scott Sumner's mind, at least. Anyway, I love this.

Yeah, we have a solution.

We have a solution. If you're feeling the pain at the pump, pivot and get a horse. At oil at $110, the urban horse is the only option. Pulling up to the gas station on a horse is truly elite. I do want to know, what's the TCO on a horse with the food and the stables versus just keeping a— what is that, a Ford Taurus in your— I don't know, that's something else— in your garage? The horse really mogs at the gas station in particular, right? Because it's just making everyone feel so stupid. Yeah, we got to go back. One horsepower is all you need. Nick Carter says, seeing a lot of non-process trusting panikins on the TL. And then he followed up by saying, that's what I thought, because the oil prices spiked up and then they fell down. And we did not get $120 a barrel oil. We got exactly $100, which was still—

It's not like it's way, way, way too early to celebrate or anything.

Yeah, exactly. There's an interesting chart here, a really shocking chart from Oliver Grobe, Unreal Numbers, about Germany's nuclear power generation. So if you scroll down, look at this curve, this is truly the bell curve meme or something like that. JPMorgan estimates that had Germany not phased out nuclear power, the country would have generated 50% less electricity from fossil fuels and 84% less electricity from natural gas in 2024. Electricity prices in Germany would've been around 25% lower and the country would've imported half as much electricity. And just complete rise and fall of nuclear power generation in Germany. One of the craziest graphs, you don't see graphs like that very often in new technologies. Typically you see S-curves or you see exponentials. No one considers the models get better and then they get dumber. That's certainly the funniest outcome. Yeah. Very, very rough. Goldman Sachs sent a note to investors saying if oil prices increase by $10 and remain elevated for 3 months, US year-over-year headline CPI inflation would likely rise from 2.4% in January to 3% in May. Those are small numbers, but we're looking at an oil price increase of maybe $30, $40 over the baseline potentially higher. We don't know where oil is going to land. And so you have this weird tug of war right now with the Fed, where if oil prices go up, inflation goes up, the Fed has a mandate to curb inflation. That means higher interest rates at a time when the labor market is shedding jobs. You would expect a Fed rate cut, or a lot of people are optimistic about a rate cut. Trump certainly wants rate cuts. But if inflation is climbing, there's really no solution other than keeping rates higher or even raising them further. So, a real jam in terms of federal monetary policy, Fed monetary policy. Here's Art Cashin at the opening bell. This is a historical video. When was this? This was a long time ago. Let's play this clip.

This may be it. First, let me start out. Morituri te, salutamus esse.

And you know, that's the gladiator salute.

We who are about to die salute you. So it's going to be a tough morning. This may be it.

We are about to die.

Salute you.

Insane, insane aura for, you know, CNBC really doesn't get enough credit for being so innovative in terms of broadcasting and entertainment. Really some of the, some of the greatest clips.

I love one of the, one of the top comments. This is an old clip. Obviously, brother.

It's probably from the '90s. So, Zero Hedge says with oil at $111, total panic. And Memetic Sisyphus shares a clip that says half of the site for the last week has sounded like this. Let's see. Producers are telling me there is breaking news. The Asian financial markets have just opened to a huge sell-off. And we're gonna switch to that story right now. Good.

I'm glad I'm here.

Your thoughts, Tracy Jordan, on how this is gonna impact Wall Street?

Larry, I'm not an expert, but I do have a strong opinion. New York as we know it will no longer exist tomorrow.

Producers are telling me there is— Is that from 30 Rock? That's so good. That's so funny. Oil came way in from its overnight highs, says Joe Weisenthal. And the quote post is—

Chris Paul hits a huge 3 to cut down the lead to 42.

Absolutely, absolutely brutal. So let me tell you about Gusto, the unified platform for payroll benefits and HR built to evolve with modern small and medium-sized businesses. Stop making fun.

We're somewhat of a smodern business. Small and modern.

Small and modern. Smodern. Let me also tell you about CrowdStrike. Your business is AI, their business is securing it. CrowdStrike secures AI and stops preaches.

So we got to give you a shout out to George Kurtz for the Mercedes performance yesterday. Oh yeah. 1-2. 1-2.

Really good. George Kurtz.

Would you look at that? Takes a stake in the team. Suddenly they're at the top.

Yeah, he's on to something. Well, my essay this morning was titled, Why Is No One Talking About Oil? Of course, everyone is talking about oil. Oil discourse gushed onto the timeline this weekend. Crude prices spiked to nearly $120 a barrel as a broadening war in Iran threatens both transportation routes and production. The geopolitical and economic analyses are flowing. But what does this mean for AGI timelines? And a lot of people in the AI world are sort of tuning all of this out because they see recursive self-improvement, AGI, ASI, the buildout as more important. And I just wanted to sort of reality check the AI supply chain. To understand how does oil actually affect data center construction, AI production, token pricing? Like, is there any effect? My conclusion was that it's very moderate, but there are some interesting effects in the financial markets that are probably the bigger takeaway. But it's still interesting to hear about like, yes, oil actually is used in the production of AI, at least a little bit. So power has been at the forefront of the AI pushback. Like everyone's worried about these, local energy prices, these electricity prices increasing near the data centers. But, and it's become like a political issue, but pain at the pump might become a bigger story as gasoline prices spike. And that has been pain at the pump. Oh, it's been the most tangible sign of inflation and moves so quickly. You know, one jitter in the economy.

And it's a huge component, you know, people on the coast, people in tech, don't have a good sense for this, right? If gas— for a lot of people, gas could quadruple and they wouldn't really notice it.

No, no.

But if you actually look into the average American, how gasoline fits into their budget, it's a meaningful component of their monthly budget. So they feel it super intensely. And variable.

Yeah.

Cost.

There are so many different ways where a lot of Americans go on driving vacations that obviously is directly impacted by gas prices. And then also just psychologically, there's something about filling up at the gas tank where you see the number ticking up, and you're doing that on an every-week basis or so, that it's just so visceral. It's this thing that you have to stop and then go experience and watch the money flow out of your account in real time. It's very visual. It's very interactive.

Yeah, I remember I must have been probably 18 at this point, where I would just go, and for a long time, I'd just put it— you know, prepay for a certain amount.

I got $20.

I got $20.

Let's see how much I got.

And I felt like really, like I felt like the king of the castle.

Exactly.

I just put my card down and let it run up.

Yeah, yeah, yeah, yeah. How far it'd go. No, no, totally. So pain at the pump's been like a big, broad economic issue. It's so visceral. Like the actual phrase has a lot of power in politics and in like attack ads. Like your gas prices are going up under this person. That's why you got to vote for me. This is common. But in AI circles, the discussion's been much more focused on RSI now is the new acronym that everyone's focusing on, not AGI. AGI's here, we know artificial general intelligence. We passed the Turing test. The models can do things generally intelligently, but can they recursively self-improve? Are they RSI? Are we in RSI now? Is it coming? Is this going to be a fast takeoff? Is this going to be a slow takeoff? Well, Something's taken off. Dylan Patel said, being in SF is like being in Wuhan right before the pandemic. Something is happening, it's gonna hit everywhere, but so few people know it. So he's sort of echoing that something big is happening mentality. This idea of like recursive self-improvement, being AGI-pilled. A lot of smart people agree there's definitely something happening. There's still this question of like, what is sticky? What, where will, diffusion take longer, what will be sort of AGI resistant or ASI resistant for a variety of reasons, even if they're completely irrational and you could get a better thing from a computer, but people like the human version or whatever.

The irony is that George Hotz hitting the timeline to raise money makes me more bullish on acceleration.

Yeah.

Right?

Yeah. If not, he's obviously not historically been a huge fan of venture capital.

Well, he's not raising from traditional VCs, he said. He specifically said he wants to raise from like, basically like friends and family type of round. He does want $20 million. We have this in the timeline somewhere. Where is this post from George?

I'll pull it up if you want to keep.

Tiny Corp. He said that he's raising $20 million to buy a building. He's gonna buy a building. Here it is. So he said, this is from Tiny Corp, the makers of Tiny Grad and the Tiny Box. If Tiny Corp was raising $20 million at $200 million valuation, who'd be interested? Business model is basically this. Buy an $11.5 million building with 5 megawatts of power, link in our Discord. Wait for AMD to launch the RDNA 5 96-gigabyte cards mid-2024. 2027. Preorder 3,000 cards. Hopefully we can negotiate for $2,500 each. Build 500 $20,000 tiny boxes with 6 of the cards in, in each box. Run all the Chinese LLMs. Make $600,000 per month revenue selling tokens on OpenRouter. Market depth is there. This is 1% of OpenRouter. Uh, improvements to TinyGrad yield revenue improvements. Due to how power is priced in Oregon, it's only like $50,000 for the electric bill before the 4 megawatts, before they price for peak, not usage. We get like 3c kilowatt-hour power, 3 cents per kilowatt-hour. We can also make $100,000 per month leasing co-location space to Kama. Building and cards paid off in 3 years max, investment made back, low risk of being undercut, since we're using consumer GPUs and running the cheapest colo you can believe. If someone chill wants in, I'd do it. I'm not gonna fake hype, I'm not gonna hype fake tech, but demand for tokens is gonna skyrocket. Look at the open Claw install numbers. With crazy good optimizations, we could potentially get 3x more from the machines and we have electricity for 3x more machines, $5.4 revenue per month, then continue to scale from there, custom chips, et cetera. Et cetera. He's starting a NeoCloud. Are you starting a— yeah, he's going to be serving tokens. So that was incredibly bullish. Demand is definitely there. At the same time, he sort of took the other side of like, this is different. This is crazy. He says there's no AGI. There's no magic threshold. You guys see Auto Research change the random seed from 42 to 137 and OMG, it's AGI. It's over. Yet you critique the junior engineer for the same stuff. The cost of development is falling. Overpaid engineering struggles to compete. That's the story. And he says humanity has been recursively self-improving for centuries, so everything new is old.

I don't know. Obviously, I have a huge amount of respect for George, but at the same time, saying there is no artificial general intelligence when you're also arguing that engineers are struggling to compete with new technology that is replacing them? How is this not artificial intelligence competing with—

Yeah, so I think he is agreeing with that, but it's that there is an immense amount of desire for this binary moment. This is the singularity. This is AGI. This is ASI. The RSI is here. This thing is happening right now, and there's before and after, and everything has changed. And he just doesn't see it that way. I think he sees it much more like the internet, the mobile phone, like other technologies that have been rolled out.

Continuum.

Electricity. Yes, there is like a before and after, but you can only really define the period by maybe a decade, and you need a few decades to understand that moment. It's very hard to go back and— there is like the iPhone moment, and there is like the first launch of, I don't know, AOL. I don't even know what the iPhone moment of the internet was, just because it was sort of a slow rollout. Anyway. Live GPU clusters. Where's Jordy going? Oh, you're moving the goalposts. Okay. Jordy's moving the goalposts. Where are you moving the goalposts to?

Just over here.

Okay, but why? Is there a reason?

I mean, they have to be moved. George is moving the goalposts.

Oh, okay, okay, okay. He's moving the goalposts. So I said, you know, it's clear that the AI industry continues to grow and continues to need more and more power and compute as we've seen from, George Hotz's new project. That means large data center campuses. But if they're not in random office buildings that George is picking up on the cheap, they're probably going to be built with construction equipment. So what does this mean? They don't just drop from the heavens. They require building, which requires oil, but how much oil? And is oil a serious— is a serious oil shock enough to impact the AI buildout in a meaningful way? Spoiler alert, basically no. But live GPU clusters in the United States do not use much oil directly. Only 0.6% of US electricity in 2024 came from petroleum. We're much, much more dependent on natural gas, something like 42% of US electricity is natural gas. And so America ramped up natural gas production significantly over the past two decades. A lot of that was in reaction to the wars in the Middle East. Hey, we need to be less dependent on foreign oil. We need to be energy independent. And so you have the fracking boom, the natural gas boom, and that's where a lot of our fossil fuels come from today, I believe. Data centers only consume a single-digit percent of US electricity. So you're looking at 0.6% of a few percent is like the actual impact. So the short-term impact of high oil prices should be very limited on AI. Building out new capacity is much more—

That said, it's worth noting, I think it's something like Qatar produces something like 20% of the world's liquefied natural gas, right?

But not for us.

Yeah, not for us. So we're in a good spot, but this is gonna impact the data center buildouts or active data centers across the rest of the world if this continues.

Yeah, so when you're talking about building new capacity, building new data centers, oil is a little bit more involved. So diesel powers trucks, trains, boats, barges, generators, pumps, compressors, excavators, and tons more construction equipment. Petroleum is also broadly used for plastics, polyurethane, and solvents that all work their way into the data center supply chain. The biggest problem is delaying already tight schedules because of narrowly available components going out of stock. So the price of oil goes up. There's one marginal factory that can't produce one ingredient that goes into the rack and that slows things down. You have to wait a week while you find another supplier. That, that stuff can add up to just a little bit of a delay. This happened during COVID and the AI industry was already experiencing something similar with transformers. And so you don't want products getting stuck in transit or going out of stock. But the bigger problem and the one that people should be talking about, and I think you were debating with Dan Primack at Axios about this, is macroeconomics. So higher prices, higher oil prices lead to higher inflation. If the Fed has to raise rates to control inflation, capital formation for megaprojects gets a lot harder. So JLL has this estimate. The next 100 gigawatts of data center capacity could require about $870 billion of new debt financing. And so using this rough number, every extra 50 basis points of borrowing cost on $870 billion is $4.35 billion in annual interest expense. And so that's a, I don't know if that's a huge squeeze on cash flow, but it's certainly a squeeze on cash flow if you're a hyperscaler and you're going gigalong AI CapEx now and you're like, yeah, we're gonna do $200 billion this year. $500 billion next year or whatever across a couple companies. You start looking at interest expense bills that are already huge, but then they're going to go up by $5 billion or up by a full extra billion just with a 50 basis point increase, half a percent. So half a percent change in the Fed funds rate can echo through, even a quarter percent could probably be amplified into a multibillion-dollar cost line on your balance sheet or on your income statement as you're servicing that debt. Overall, higher oil prices do make the AI rollout harder, but not really for existing AI capacity. I wouldn't expect token pricing to change based on the price of oil, but there is this more, there's this bigger question of macroeconomic resilience during a time where our largest tech companies are digging through the couch cushions to find every penny to win the AI future. Middle Eastern investors pulling back. I agree with you. I do think that this is, a significant risk, 'cause if they pull back on big investments, makes these mega rounds more difficult, retail investors could fly to oil stocks or defense stocks if they think that this is like going to be a thing for months and months and months. It could all flip if there's a quick resolution though, which is what I'm hoping for. But if things continue, the AI industry walks an even tighter tightrope, so.

[Speaker] Yeah, I mean, the big question right now is the hundreds of billions of dollars of sovereign AI projects in the Middle East, right? I think a lot of those people are going to be like, do we want to send billions of dollars of GPUs over there?

And then also the money coming here is another thing where you might want to spend it elsewhere. Quickly, let me tell you about Turbo Puffer, serverless vector and full-text search, built from first principles and object storage, fast, 10x cheaper, and extremely scalable. And let me also tell you about Public, investing for those who take it seriously, stocks, options, bonds, crypto, treasuries, and more with great customer service. Whatever your investing strategy is.

We have to say happy birthday to Trey.

Happy birthday, Trey.

We love you.

We're very excited to have you. Thank you.

One of our strongest soldiers.

We appreciate you.

It's been an honor.

And have you been noticing that it's been hotter in Los Angeles?

I have.

Downtown Los Angeles is forecasted to approach 100 degrees Fahrenheit on Thursday and Friday in March.

Which is why—

not good.

We're going to do the weather. We have a new segment for you today. We're doing the weather on TVPN. We have our very own Ben.

We have Ben.

Hello guys, how are you?

We're doing great. Tell us about the weather.

What's happening?

Well, I want to start off by saying, as you can see, the weather today for the low, low of today is going to be 75 degrees Fahrenheit up there, high of 100 degrees Fahrenheit down there. But there's something I actually wanted to point out that I saw and I thought that was quite interesting. As you can see up here, there's a localized low pressure area up there and a localized high-pressure area right down there, if you can see that.

Does that mean rain?

The issue with that's not normally an issue and not a cause for concern, and it's not very common for this time of the year. However, today, later in the afternoon, these two areas are going to collide and they're going to hit each other.

No, really?

And what that's going to cause is a barometric pressure inversion. Okay, it might sound a little bit scary, but I guarantee there's no cause for concern. All that means is that hot air rushing in from the west is going to collide with that cold air rushing in from the east and it's going to cause a bunch of turbulence in the sky, moving all the airwaves around and oscillations in the sky. However, I want to add one more point. A byproduct of this effect is that all that humidity that dropped after that hot air moved to the bottom is going to raise up because of the water cycle, you know, evaporation and stuff. It's going to raise up into the sky, into those clouds, going to cause big clouds in the sky, and eventually all that water is going to fall down onto the ground. We're going to have big rain later in the afternoon.

Wait, it's actually going to rain in LA?

Yeah. It might sound crazy, but I just want it for all you guys at home. I definitely try to step outside with a jacket today, maybe a hoodie, just in case the rain comes. Don't try to be a big shot.

I'm fact-checking you right now, and the weather app says it's going to be sunny all week. Is this just complete fake news?

My team—

those are the numbers my team gave me.

This is complete fake news. No, no, no, this is the fakest news I've ever heard.

All that cold, the transatlantic current.

John, John, I'm— I don't trust I trust the weather.

Apple says it's not going to rain the entire month. There's zero chance of rain.

Did you look at the transatlantic current?

No.

The transatlantic current.

Precipitation, zero inches today, zero inches tomorrow, zero inches on Wednesday.

John, you're really going to trust an app? You're really going to trust an application that was probably vibe coded yesterday over Ben, who's doing the weather?

How did he get here? What happened here?

If you look right there, you can see a localized high-pressure area. That's a localized—

I think this is an overeager weatherman who's just looking for drama in the most boring weather market in America, which is Los Angeles.

I think this is the most important story in the world.

This is ridiculous.

You guys can look at the jet streams. They're coming in from the west.

I don't want to hear any more mumbo-jumbo about jet streams.

Get out of here.

You're done.

Great work, Ben.

Thank you, Ben. The fake news weather. Will return, maybe never, who knows. Let me tell you about Restream. One live stream, 30+ destinations. If you want to multistream, go to restream.com. And let me also tell you about Gemini 3.1 Pro. With a more capable baseline, it's great for super complex tasks like visualizing difficult concepts, synthesizing data into a single view, or bringing creative projects to life.

This is the future of the weather. You have a weatherman who gets into a live altercation with one of the other hosts. Who's just constantly fact-checking.

It's on site next time, Ben. It's on site. If you pull up with some fake news on TVPN's weather report, it's on site.

Ben, you crushed it.

Good job. So it is— it's not Adam Smith's birthday. It's the 25th anniversary of the publication— 250th. Sorry, 250th anniversary of the publication of The Wealth of Nations. Tyler, you have a copy?

I have two of our copies here. So yeah, usually I have one with me, but I didn't bring mine to work because we We have a few here.

Yeah, yeah, yeah, obviously. For those who don't know, give us the pitch for Wealth of Nations. Why should people read Wealth of Nations in 2026?

Yeah, I mean, this is kind of the OG economics book, right?

Was it? It was based— like, put it into AI terms. Is this like situational awareness type of—

I mean, this is like attention is all you need.

It was like the ChatGPT moment for capitalism.

Yes, yes. But I would like to, you know, I have a few quotes I would like to read.

Okay.

All right, let me open them.

Let's see. It's a big book. It's like 800 pages.

Yeah, it's very long.

And it's really, really dry. I mean, it reads like it was written 250 years ago.

Okay, so this is, I think this is kind of one of my favorite quotes.

Yeah.

Kind of the, you know, seminal. Every individual necessarily labors to render the annual revenue of the society as great as he can. He generally, indeed, neither intends to promote the public interest nor knows how much he is promoting it.

Yes.

He intends only his own gain.

Yes.

And he is in this, as in many other cases, led by an invisible hand to promote an end which was no part of his intention.

Yeah. Rational economic actors, homo economicus, thinking of the value for society is created by rational self-interest. It's sort of the spin on like greed is good. He was like an OG of that thesis.

The greed maxer.

Yeah, productive labor, efficient markets, free markets, don't hold gold, don't control trade. An important read in 2026, somewhat heretical. There are many people that disregard that book. Maybe they should not, maybe they should give it a crack at it.

Evergreen.

I think it's evergreen. I think it's evergreen. I mean, there was that debate over like Tyler Cowen over the weekend, and somebody was taking shots at Stubborn Attachments. For being like not advancing the discipline of economics. And my pushback to that was that I actually think it is an incredibly— it's just a fun read. It's like really well written and it's very entertaining and it's very thoughtful. But it's particularly interesting to hear someone sort of just rehash the tenets of liberal democratic capitalism in 2020, I think I think he read it in 2023, 2022, something like that. But like, there are a lot of voices out there who are anti-capitalist, anti-democratic, anti-liberal, and Tyler Cowen is just retreating to like a time-tested, time-honored economic philosophy that is deliberately not revolutionary. And I thought that it was a great book for that reason, and that's why I recommend stubborn attachments to so many people and so widely. But if you want to crack open the real hard stuff, pick up a copy of The Wealth of Nations. Why is one of those copies so much smaller? Is the font size just smaller?

Yeah, it's super small. It's also two columns.

Oh, two columns. Wow. Yeah. If you can get through that, you're truly elite in the global economy. What does the timeline have to say about the masterwork that is turning 250? It's now needed more than ever, says the Washington Post. On March 9th, 1776, 4 months before the American colonies broke with Britain over the issue of taxation, a little-known Scottish thinker published a long, dense book with an unpromising title— shots fired— An Inquiry into the Nature and Causes of the Wealth of Nations. 250 years later, Adam Smith is, by any objective measure, easily the most widely cited and widely quoted economist who ever lived. Astonishingly, his work still frames the central questions we face, not just about free markets, trade, and capitalism, but about the nature of human society itself, and even what it is to be human at all. Smith was born in 1723 in Kirkcaldy, Scotland, educated at the Universities of Glasgow and Oxford. He initially made his name not as an economist, but as as a moral philosopher. His first published book, The Theory of Moral Sentiments, offered a radical theory of how we form moral judgments. Radical because it derived from the creation of moral values, not from scripture or divine grace, but from human sympathy and mutual regard. The Wealth of Nations, as his second work, second major work came to be known, was an extension of that product. The book is not, as sometimes believed, a hymn to greed, a pain to market fundamentalism, or red-in-the-tooth claw capitalism. It was an attempt to understand how a commercial society could generate prosperity without collapsing into corruption. This 250th anniversary is not a moment for hagiography. It is an opportunity to recover a way of thinking that is directly relevant, indeed urgent, to the economic, social, and political challenges we face today. Begin with trade. In his own time, Smith's great target was mercantilism, or what he called the mercantile system, the idea that wealth consists of hoarded bullion and that trade is a zero-sum contest. Governments granted monopolies, imposed tariffs, and manipulated commerce in the purported pursuit of national strength. Producers were widely protected, consumers often ignored altogether. Against this, Smith argued—

One of these days we should have Tyler just do a speed reading of The Wealth of Nations live on the show.

Speedrun? Yeah, Wealth of Nations speedrun, any% glitchless. You need to do glitchless though. Glitchless. For sure, a glitchless run is the only way we can do it. No glitching. No AI summaries. I think that would take you probably 5 days. Probably 5 days. It's so dense.

Have you seen how fast you can rip a Rubik's Cube?

This edition is 480 pages around.

Oh yeah, I guess you don't have to read all the footnotes and stuff. You can kind of skip through it, but. It is a thick book, but the way it's written is just really not fun to read. It's like it reads like it's 250 years old. It's not like a page-turner. Well, we need to talk about the wealth, but first, vibe.co, where D2C brands, B2B startups, and AI companies advertise on streaming TV. Pick channels, target audiences, and measure sales just like on Meta. And let's also tell you about fin.ai, the number one AI agent for customer service. If you want AI to handle your customer support, go to fin.ai.

We need to talk about the wealth of Sundar Pichai. Pichai.

Sundar.

The new Pay deal worth up to $692 million.

Is this like 10 times what Tim Cook's making? No, it must be over, over time, right?

Cook's making around $70 million a year.

$70 million a year?

Combined.

So if he works for 10 years, he makes what Sundar makes in 3?

There's going to be a conversation. It is. I mean, but this is what we've been—

we've been advocating for this. So this is good. Yeah, no, no, we're in support.

But Google has increased Sundar's potential pay to $692 million over the next 3 years.

You know that Tim Cook dropped this in the Apple board members group chat as soon as it hit. He was just like, dude, this is a cool article. You should read this.

Drops it in. Just drops it in.

Pots? Yeah, exactly.

The bulk of his package comes in performance units with a target value of $126 million split evenly into two branches. The PSUs are valued by the parent company, blah, blah, blah. It could pay out as much as twice the target, a quarter billion, if it outperforms significantly, or nothing if it lags behind.

Gotta beat the S&P 100.

Um, Pichai will receive Waymo stock with a target— stock in Waymo with a target value of $130 million and $45 million in Wing Aviation, that's their drone delivery platform. Again, both can pay out up to 200% of the target.

They've got a lot of incentive pay. If he delivers and the company does well, he should be richly rewarded. Barron's was writing about the death of the Mag 7. Mag 7's over. Up and down Wall Street, the Mag 7 era ends, Barron's is calling it. They said, stick a fork in it, turn out the lights, hasta la vista. Say it any way you'd like, the simple truth is the Mag 7 trade is over. Finito. I love Barron's writing. Dead, the collective stock market outperformance of those 7 tech icons, Alphabet, Amazon.com, Apple, Meta Platforms, Nvidia, Microsoft, and Tesla is now a thing of the past. The group may still do okay and some of the individual stocks may even kill it. But the slam dunk, set it and forget it, run circles around the market era of the Mag Seven is gone with the wind. Recall that B of A Securities analyst Michael Harnett coined the term Mag Seven in 2023, pretty recently, referring to the 1970s John Sturges Western gunslinger flick an adaptation of Akira Kurosawa's 1954 film Seven Samurai. Have you seen Seven Samurai?

No.

Have you seen The Magnificent Seven?

The stocks?

No, the movie. The movie. The Western.

No.

It's about the Seven Samurai.

In what world?

Production team, has anyone seen either of those movies?

I think we watched them in school.

Okay. Okay. Which one?

You're saying you didn't watch them? You studied them?

Well, I haven't seen them in a while. Which ones have you seen, Tyler?

Seven Samurai.

Seven Samurai. It's a great one. It's a very good movie. Anyway, Barron's is bearish on the Mag7, but Dark Child—

Sounds like cope.

It does sound like cope. I think the Mag7 will do.

It's over for the largest and most profitable companies in history that stand to benefit some ways.

The actual argument was the growth-to-value narrative, the loss of cash flow as they increasingly invest in CapEx. The financials will look very different. There'll be margin compression, that type of thing. It's not an unreasonable take, but it is funny the way it's written. Anyway, Sundar Pichai took over as CEO in August of 2015. He's going on 11 years in the seat. Google's market cap has increased almost sevenfold from half a trillion to $3.6 trillion, briefly topping $4 trillion in January. Surge has made the Indian-born, 53-year-old former McKinsey consultant a billionaire. He joined in 2004 and made his name developing the Chrome browser and leading the Android division. He had been criticized for being too slow to adopt AI at the search giant, allowing OpenAI to release the first, first hit product, ChatGPT, in late 2022, but has since bounced back, releasing cutting-edge AI models and integrating the technology into its dominant search engine. Yes, he's done very, very well on that front. Pichai has also navigated a duo of antitrust cases brought against Google's search and app store businesses, avoiding the worst-case scenario of forced breakup. A third lawsuit is pending against the advertising network. Tyler, do you think that DeepMind counts as RSI capable at this point? Recursive self-improvement. So there was a take on the timeline where Basically someone was making the argument that xAI is cooked because while Anthropic is using Claude to build Claude and OpenAI is using Codex to build Codex. So both of those leading private AI labs are recursively self-improving. xAI was using Claude code to build Groq. And so they don't have the same flywheel of the tools that build the tools build the tools.

Yeah, because I think they lost access to Cloud Code. But then they still have access to Codex. So it's kind of like, do you count that as RSI? Because it's not really internal. But is Gemini— that's the question. I'm sure people at Gemini are using Gemini. I mean, there is a Gemini CLI. People don't talk about it as much. I mean, the model is very good, right?

So I assume it is helpful. It would be interesting. I feel like we don't get as many little data points from on the Gemini and DeepMind team because it's a public company and they need to, you know, maybe break out financials at some point.

Yeah, I think like vague posting is probably, you know, more heavily looked at.

Exactly, exactly. Because you would, yeah, yeah, because it's totally possible that they're doing the same thing. I mean, Demis has been extremely, just very clear about the path this goes and he's very AGI-appealed. He understands the exponentials and whatnot. He certainly messaged it at a high level without actually saying, you know, definitively, yes, like we are at a point where we're not writing any code by hand, that type of phrasing that you see in the RSI crowd. Anyway, Pichai last got a stock award in December of 2022 worth $218 million, which was structured in the same way. His earnings are topped up by his personal security costs, which rose to $8.3 million in 2024. Earlier in the week, He sold 32,500 Class C shares at an average price of $303, worth roughly $10 million. The Bloomberg Billionaires Index estimates that he has sold about $650 million in stock since becoming CEO. They still own, along with his wife, he owns 1.67 million shares of Google, worth half a billion at the latest stock price. And Google's founders Sergey Brin and Larry Page still control the company through their ownership of of super-voting Class B stock, which gives them 56% of decision-making power. Interesting. I think you got to give Miami to—

Paper hands?

Oh.

He sold $650 million of stock.

No, no, no. He sold that, but he's— oh, and he only owns half a billion. So yeah, he sold more than that.

I'm just saying, hopefully, hopefully, hopefully he put it all into video. Paper hands for Shai.

Paper hands for Shai. But he's getting topped up. And he's diamond-handing. Diamond-handing.

I know. I'm just saying. I'm just saying. Looking at the stock chart since 2015, when he became CEO, he would have done pretty well just to not do anything and trust his own process.

Hey, maybe he has an amazing car collection or something. You know? There's a lot of things.

You got to rotate into metal, mate.

You can't drive a Google Class A share. You can't sleep on a bed of Class C shares, Jordy. You got to get liquid. No, you got to give Miami Tech its due because everyone's saying, oh, Miami Tech, they don't have any startups. There's no funding there. There's no— Miami Tech isn't panning out. Well, Google now counts as a Miami tech company because the Google founders bought houses there. You also got to count Meta as a Miami tech company because Mark Zuckerberg has a house there. So you put Meta and Google, that's like $6 trillion in market cap for tech companies.

All in Miami.

I count it. I count it. Anyway, moving on.

Owner of Four Loko is exploring a sale of the storied alcohol brand, sources say.

Wow. Let's dig into this. But first, let me tell you about Plaid. Plaid powers the apps you use to spend, save, borrow, and invest, securely connecting bank accounts to move money, fight fraud, and improve lending now with AI. And let me also tell you about Cognition. They're the makers of Devin the AI Software Engineer. Crush your backlog with your personal AI engineering team.

So the parent company of Four Loko, the canned alcoholic beverage that became a college campus sensation in the late 2000s before being reformulated under regulatory pressure, is exploring— I cannot believe they nerfed Four Loko.

It was too powerful. It was too powerful.

Like, you can just imagine the trajectory of the United States if it hadn't been nerfed. And then straight downward. Really?

Yes. It was so bad. It was so—

I thought people were having heart attacks.

Yes, it was terrible. Oh, really? So Four Loko was a 24-ounce can, so twice the size of a normal alcoholic beverage.

Trey, the birthday boy, says it was amazing.

It was amazing. OK. Now we have Buzz Balls. That's true. So Four Loko, the original formulation was twice as big as a normal can of Bud Light or Miller Lite or Coors Light, your normal beer, something that you would grab. So it was 2 drinks that way. And then instead of it being 3.2% alcohol or 4.5% alcohol, it was like 10% alcohol. It was the strength of wine almost. And so a single Four Loko was like 4 or 5 beers in 1 can. And then they also added like 200 milligrams of caffeine. So you would become incredibly intoxicated and inebriated, but then also incredibly stimulated from all the caffeine. And that spelled doom for many people. People would be very high-functioning but completely inebriated and discombobulated. And so they would get into all sorts of trouble, whether or not it was behind a motor vehicle or—

The discombobulator.

It was the discombobulator. It was also wildly illegal from an FDA perspective. You can't mix alcohol and caffeine in a single product. A rule that they chose not to follow, and they figured out how to make it. And then the FDA eventually gave them a warning letter, and they pulled the caffeine out, and that of course killed the viral sensation that was Four Loko. But apparently, the company continued to sell products that were compliant with the current FDA regulation.

Nerfed.

That's nerfed. The nerfed Four Loko continued to sell.

Somehow they're looking to and potentially will capture somewhere around $400 million on the sale.

[Speaker:TODD_HARRISON] The brand value. So people— it's hilarious that they're working with JP Morgan on this. That's wild. So Fusion Products is the owner of the brand. They're working with JP Morgan on the sale process. The potential sale underscores how ready-to-drink beverages have emerged as a growth category in an otherwise sluggish alcohol market. US beer and wine— US beer wine and spirits sales declined in 2025. We've talked to a number of guests about the declining alcohol sales as folks move over to more functional beverages.

Here's what the CMO had to say. For over a decade, our sales have been the leading flavored malted beverage— have been leading the flavored malted beverage market by embracing bold innovation, unconventional marketing, and a risk-taking attitude that delivers results year after year.

I guess they kept building. They also own—

The parent company also owns Mamitas, Pirate Water.

Pirate Water.

Basico Tequila. What is Pirate Water? And just Earthquake.

Earthquake. Earthquake seems like something you definitely want—

From the makers of Four Loko, Earthquake.

Pirate Water is a wild name for a brand too. I mean, it's a distinctive can. It looks remarkable. It jumps off the shelves. When the Four Loko nerf went into effect and they pulled them from the shelves because, Before they reformulated without the caffeine, they had to pull them from the shelves entirely. They were a hot commodity, and folks would go around stockpiling them before they were banned, and they were one of the most guarded resources in all of college campuses.

Wow. Earthquake is just a can that's just 10% alcohol.

Wow.

Do not recommend. Very risky.

Never quake.

Anyway, in 2010, the FDA sent warning letters to caffeinated alcohol beverage makers, including Four Loko. Saying that the caffeine in their products constituted an unsafe food additive. So the FDA has specific categories for regulation around what is an alcoholic beverage, what is a caffeinated beverage, and they never— the two shall mix. So they don't let you put these two things together. Of course, vodka Red Bull was the precursor to this where folks would order a Red Bull and then they would put and vodka in it, I think, but at a bar, which is made at the time of sale. And so that was legal, I believe. And this might be fake news, but I remember hearing a story that at one point Red Bull was sued because a gentleman had purchased a series of vodka Red Bulls, become incredibly inebriated, gotten behind the wheel of a car, gotten into some sort of accident, and wound up in front of a judge. And the argument was that he wanted Red Bull to be held liable for his actions while inebriated behind the motor vehicle. And his argument was that it was only because of the product, Red Bull, had been mixed with the vodka that he made the decision, that he was alert enough, stimulated enough from the caffeine to get behind the car wheel in the first place. This might be apocryphal, but the story goes that Red Bull made the argument that the caffeine actually made him more capable to drive, not less capable to drive, and actually increased the safety of the driver and the vehicle. And so they were held not responsible. But I don't know, it's just like an old wives' tale in CPG. I don't know if it's true. Anyway, let's move on to Phantom Cash. Fund your wallet without exchanges or middlemen and spend with the Phantom card. And let me also tell you about Labelbox. RL environments, voice robotics, evals, and expert human data. Labelbox is the data factory behind the world's leading AI teams. So we're all preparing for the singularity. What is this doing dishes? Just like, oh, if you're doing dishes, you're not worried about Dylan Patel's viral post saying that being in SF is like being in Wuhan right before the pandemic. Something is happening, it's gonna hit everywhere, but so few people know it. And that is the question of, are we in a software-only singularity where the dishwasher is life as normal? And every time I go to the COVID analogy, I start a clock because I remember tracking COVID in January. And I remember the viral post biology posted in February going viral. And then in March, the NBA shut down And in April, everyone was locked down. And so every time I see one of these, it's like COVID. I'm like, okay, let's see what it's like in 8 weeks. Like, 8 weeks is not that long. In 8 weeks, if I go outside and things look radically different and you talk to random people and it's wildly different, then we're like, okay, yes, that analogy was correct. But if it's more like in 8 weeks, we're like, oh yeah, like, you know, technology's getting more exciting and like tech companies are changing and like the market's moving, but like, The dishwasher is unaffected, the pizza place on the corner is unaffected. That's always how I, I just think people underrate how crazy a global plague was. It's literally like a once in 100, like the most insane thing ever. It's the craziest change to literally everything from the sports that you watch, you couldn't go to the movie theater, you couldn't go to dinner, everything was changed for like a full year, it was like complete upheaval in our society. And might happen, but 8 weeks, 8 weeks is a little fast for me.

And if you live in LA, probably still be able to go to the beach.

I think so.

So that's—

But maybe not, maybe everything's—

I did have a funny moment yesterday, I was telling you, but I'll share it again. I was talking with my car detailer who's become a dear friend, and he was telling me that the people in my area cancel their appointments at a much higher rate than the people in the rest of LA. And he was talking about running analytics on iMessage, and I was like, oh, how do you do that? And he's like, he's fully running OpenClaw. So he's now using OpenClaw to run his car detailing business. He's like, yeah, it's just always running in the background. To him, he was like, it's some combination of having like an EA or a CEO, like helping me manage the calendar, helping with comms, helping me make smarter decisions on routing and scheduling and all these things.

So when you book a detailing, meaning that this gentleman is coming to your house and cleaning your car, you send an iMessage to him, correct?

Yeah.

And that is what is so interesting about Open Claw, is that There were probably 1,000 companies that were like, we're vertical SaaS for auto detailing. You gotta do everything over email. And they just got eaten alive because people use iMessage, and iMessage has been this closed ecosystem that private companies could never access before. But it makes so much sense for him to say, hey, run through all my iMessages. And if I'm talking to my friends about the F1 race, discard that. But if I'm talking to a client about potentially booking a detailing, put that in a spreadsheet, organize that, and then run some analytics and tell me what the patterns are. And that's just something that should have been SaaS for a decade, but Apple has a walled garden that has now been smashed down by Open Claw. And I think it's really cool. And I think that that's what we're seeing the first use cases for Open Claw. Things that could not happen for business reasons, for logical reasons, sometimes for legal reasons, if you're talking about downloading illegal films, like the Napster analogy. But a lot of times, it's just Apple didn't want to do a deal to integrate iMessage into Salesforce.

The way to break down the walled garden was just to have a digital guy.

Basically, yeah, who just combs over all this. So is that a new type of software that should have existed that now is unlocked? Is it disruptive? It's not like he stopped using his previous CRM, probably didn't have anything. So now he has something.

Yeah, he's pretty tech-forward. He's tried every single SaaS tool that there is. But at the end of the day, having a digital guy that has access to all of your tools is just a much better experience.

And so I want to know what his bill is this month.

He said he spent a decent amount.

OK, I want to know what it is this month. And then I want to know what it is in like a year, because sometimes the maintenance of these systems can be really, really big. And then there's a question about, are we going to see per token deflation, or are we going to see prices— are prices being artificially held low because of competitive dynamics? So in a year, will he be spending more or less? Will he still be getting the value? These are all interesting questions, but I'm just excited that a digital product now exists that that could not or didn't exist before. And that feels incredibly positive and cool. So very exciting. Anyway, we have our first guest of the show in the Restream waiting room. Let me tell you about Console.com. Console builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. And without further ado, Dr. Alex Wisner-Gross. Welcome to the show. How are you doing?

What's happening?

Doing really well. Thank you, John. Thank you, Jordy.

Thank you so much for taking the time to come chat with us. Congratulations on all the virality. For those who haven't followed along so far, can you introduce yourself and the project broadly? And then I have a ton of questions, but I'll let you kick it off with an intro.

Absolutely. So quick background on myself. Originally from New York, undergrad MIT, studied physics, electrical engineering, computer science, math, PhD physics, Harvard.

Was there anything you didn't study? You studied everything.

Didn't study the humanities.

Okay.

Considered going to Cambridge to study chemistry and biology for a while, discarded it. Was told physics is a game for the young.

Okay.

So stayed in the US, did my PhD in physics focusing on AI and nanotech. Cool. I've started, invested in, advised, managed dozens of companies at this point. I've had a number of exits. And at this point, most of my time is focused on smoothing out the singularity as you guys were just discussing. So most recently, as you were gesturing, A company that I helped found, Eon Systems Public Benefit Corporation, just announced what we've been characterizing as the first multi-behavior upload, if you will, more on that in a second, of a fruit fly, which we think is a major step forward for the field. And quite frankly, in an era when, you know, I do another podcast, the Moonshots Podcast with my friend Peter, I talk all the time about tiling the Earth with compute. Right now, all of this AI infra buildout that we're doing to the tunes of trillions of dollars of CapEx, all of this is going to artificial minds. And E.ON is playing, I think, a very important role in doing some early pioneering experiments and developments to try to level the playing field of the future, not just artificial minds playing on all of this trillions of dollars of infra and data centers, but enabling emulations and, if you will, uploads of human minds, non-human animal minds, at the moment starting with fruit flies and aspirationally working towards mice and humans and getting the brains of a variety of human and non-human animals to operate in the cloud.

So walk us through the fruit fly experiment, why it was so exciting for you.

It's really interesting if you think back Remember back with the launch of ChatGPT, that was just a side experiment. That wasn't the main focus of OpenAI at the time. Some have characterized that as an unhobbling, if you will. GPT-3, the leadership of OpenAI thought at the time was sort of the main entrée, if you will. Similarly within Eon, I don't mind saying this was more of a side project that we thought it would be somewhat interesting to the world, But I think that the level of response and virality probably even took most of E.ON by surprise. So this is a video demonstration of an internal project to take a bunch of building blocks that were already sitting around for the past 1 to 2 years and put them together for the first time. So one of these was a paper that senior scientist at E.ON, Philip Xu, published in Nature in 2024, demonstrating that just from an emulation perspective that you could take the FlyWire, which was a well-funded project to capture the connectome of the fruit fly brain, that you could emulate certain circuits within the fruit fly brain. We've had that for a year and a half or so. Another project, NeuromechFly version 2, it's an excellent project to be able to formulate simulations in realistic mechanical environments of fruit flies. Another project from Erzdal et al. to formulate coordinated motor actuation of fruit flies. And then AI. So in summary, EON was really the first to put all these building blocks together that were sort of lying around. This is actually a tiny subset of what EON's working on. EON is working on scanning the connectomes of human and non-human animals at scale. And achieving whole brain emulation, which is the holy grail of the field. There have been many projects in the past that have attempted human whole brain emulation. EON is leveraging advances in AI, expansion microscopy, a variety of other techniques, but it was really just putting the pieces together. So we're quite honestly astonished by the level of interest in just this one demo.

I love it. I have a million questions. Let's start with, uh, uh, I/O. Like the inputs and outputs, like what are you simulating? When I pull up the Sims, that is somewhat of a simulation of a human, the, you know, a basic for loop over a bunch of random probabilities to say, "Are you hungry? If hungry, eat. Go over to the stove, cook food." That's obviously nowhere near the level of detail at which a human operates. Is there an abstraction layer here where The output is quantized to things that you would think a fly would do, like move a wing, or are you operating at the level of like choose to fly, or at what level of abstraction is the fly's output happening?

Yeah, this is what I was gesturing at a moment ago. So there's a paper by Özdal et al. that bundles motor actuations in terms of high-level representations.

Got it.

So a lot of people, sort of the people who got perhaps most excited in the neuroscience community about this video, some of them I think didn't bother to read Erzdal's excellent paper. So a lot of those, this is why I'm saying a lot of the pieces were sort of lying on the ground waiting for us to pick up and put together. So these like, these abstract motor movements were already available, just no one had ever prior to this bothered to wire them up.

That's cool. Then what is the, do you need a system prompt? Like, do you need like an initialization function? Or once you synthesize the connectome, synthesize the connections in the brain, things just start firing? Do you have to trigger something to start? Start the process of consciousness or whatever you call whatever the fly is doing?

No system prompt needed. On the other hand, in the early days of ChatGPT before instruction tuning, there was no system prompt either. Yeah. In the days of early language models, no one had thought of prompt engineering. So it's entirely possible that maybe in the future, some sort of prompt engineering equivalent for whole brain emulation will be essential, At the moment, there's no direct analog of that.

Help me compare some of the scales that we're talking about on complexity of a fly, complexity of a mouse, complexity of a human. I imagine that we're on some sort of exponential here, but how many orders of magnitude in complexity, size, number of neurons, something like that, whatever the metric is, How far away are we from humanity?

We're many orders of magnitude away. So the canonical estimate for the human brain is, depending on how you count brain cells, and there are multiple types of brain cells, call it order of magnitude 100 to 200 billion cells in the brain. We're many, many orders of magnitude, many orders of magnitude away from that. So this isn't going to happen immediately. We're not going to get human whole brain emulation immediately. On the other hand, I think it's important— I think back to the early days of driverless cars and autonomous vehicles, and what the world is missing right now is, I think, sort of a scale or a framework for starting to think about what that future looks like. We're at the earliest stages of that. Call it Level 1 or Level 2 uploading. A lot of people even get triggered by the term uploading. They prefer something else. And I think we need to start having these conversations in a critical, thoughtful way and start to define what is a framework for multiple levels of fidelity for uploading or emulation, or maybe we come up with a new term. But right now we're starting that discussion and we're trying to do it in a thoughtful, responsible way.

So you like the term uploading because you did not, you did not choose a random structure or an average structure of this particular fly's brain. You chose the exact structure of a particular fly's brain?

This is a very expensive structure of a very particular fly's brain. This structure was again captured through the Flywire project at great expense.

So that's like a clone, digital twin type of thing, right?

Aspirationally. I mean, again, we're at the earliest stages. I don't want to oversell. This is just a prototype. And quite frankly, this is a sliver of everything that Eon is doing. But the fact that Eon was able to put a few of these pieces together, build a prototype with relatively small effort, suggests that, you know, we talk in the AI community all the time about overhangs. Many people thought that the arrival of large language models was in some sense a technological overhang that we could have— one could imagine a thought experiment. Could we have had large language models 20 years ago if we simply knew what we should have been building? I think probably—

Could we have?

I thought we were very GPU-gated on that.

Well, the GPUs were being spent on video games for the first few decades. So we could have, in principle, like Markov models, we knew how to build Markov babblers. I was building Markov babblers in the '90s and early 2000s.

Yeah, it just feels like even if you try and port back Frontier models now to a, you know, 1080 Ti, which was a graphics card for gaming, you're gonna have a bad time.

Well, I mean, with the rise of SLMs and some of the amazing deflationary properties of algorithmic progress in SLMs, I think we will find ourselves in the world next 2 years where some of the earliest PCs could probably have hosted non-trivial conversations with LLMs. There's just that much algorithmic progress.

Extremely bullish for Apple. Anyway, not to put it in hardware terms.

What do you wanna do next? Like, where do you go from here? I know it's a side project.

Well, more flies or straight to mouse?

It's such a profoundly interesting question. So Eon is beginning fundraising. Eon wants to tackle both mice and men. And I think, The timescale will be determined by both technical capability increases and also fundraising abilities. So I think it's an interesting time. I like to say on my other pod that the singularity looks like all sci-fi tropes happening everywhere all at once. And one of these sci-fi tropes is most definitely mind uploading. And I'm doing my best vis-à-vis Eon Systems to push that part a little bit to the left so that the AIs don't have the Dyson swarm all to themselves.

Interesting. Architecturally, what does the neural network look like that you've built for the fly? Is it similar to a large language model like transformer-based architecture, or are you a beneficiary of attention?

Totally unlike the transformer. Totally unlike the transformer. Totally unlike the transformer. So transformer architecture, if you subtract off the encoder and decoder layers, looks— it comes in many variants, but the most vanilla variant looks like alternating linear and attention layers. So totally unlike the transformer architecture, at least if you look at it, it looks like more like a little bit more like a graph neural network. But really, that's such an AI way of framing it. It looks like a leaky integrate and fire. LIF model that we've had for decades from the neuroscience world. So it's just a graph of nodes that have leaky interneuron fire dynamics and they're firing at each other.

So what does that tell you about the nature of synthetic human intelligence? Like, do you think that there is a path to human-level AGI that does not involve the transformer? And instead falls along your path and your architecture?

Oh, sure.

Well, first of all, I would argue we've had AGI for at least 6 years now, 5 to 6 years at the very latest. Since summer of 2020 when OpenAI published their language models or few-shot learners paper. Yeah, GPT-3. I would argue that was an AGI.

Okay.

So I think there are many ways. My bet is there are many ways to achieve generality of intelligence. I published a paper a number of years ago arguing that not only is intelligent behavior a general process, I went further and argued that it is a general physical process, that you can even formulate intelligent behavior in pure thermodynamic terms. So my bet is intelligence is this very, very general effect and lots of ways to implement it.

Interesting. Do you, so do you, I mean, you've mentioned AGI 6 years ago. Singularity here and you're smoothing it. Do you have other sort of binary benchmarks that you're looking towards, like ASI, recursive self-improvement? Are any of these terms useful to you?

Oh, we're already there. I mean, we're already in the era of recursive self-improvement. All of the frontier labs are pretty public about it at this point. I'm looking past the singularity, so I spend most of my time on bets for what the post-singularity world looks like. So Aeon is one of those bets. Another bet, I have a company, Physical Superintelligence, that's trying to solve all of physics with AI and doing an amazing job. I just wrote a book with Peter called "Solve Everything," arguing that entire disciplines are going to get steamrolled by superintelligence and that what matters now is what disciplines we aim the superintelligence at.

Interesting.

What decisions do you make in your personal life, in the way that you live, that are based around your beliefs around technological progress?

There are a few that come to mind. So one is I'm hoping not to die. It would be a shame if I died sometime soon and get to miss the most exciting developments.

Same.

That's what gets the applause?

That's great. Yeah, that's a white pill.

White pill, very good. Bryan Johnson thanks me in advance. So, but I think also, I mean, there are so many other angles. Another is not trying to bet against the collective intelligence of the market. I've had a number of friends who thought they could outsmart all of the AI algo traders and day trade. And when they inevitably fail, they attribute it to bad luck or something other than just fundamentally betting against progress and betting against AI intelligence. So trying not to make the mistake of betting against collective intelligence of civilization and AI capabilities. I also, I'm trying to make bets assuming that the AI capabilities keep increasing to the point where they're ontologically shocking and trying not to duplicate effort of frontier labs. And I think that's perhaps something I don't see enough of in the venture community. So many new startups being formed now just aren't being ambitious enough. I have startups that newly formed or recently formed that are literally trying to grow new islands and new coastlines with AI. That would've been unthinkable years ago.

I have startups working— That's with physical AI? Or they're just dreaming them up?

It's with AI reaching into the physical world to steer ocean currents and grow new islands and new coastlines. That would not have been possible a few years ago. That's now starting to become possible. I have a startup that's working on solving interspecies communication, starting with dogs.

Love that.

That would not have been possible a few years ago. So I spend a lot of my time thinking about what the post-singularity state of the world looks like and how to smooth that out and bring it here sooner.

Yeah, what does the post-singularity world look like where you are able to upload yourself? I assume that feels like there's a copy of Alex in the cloud and then you still exist and not in the cloud. How do you think that that interfaces?

In my mind, if we wind up having the copy question come about, that's almost a failure mode. Speaking for myself, I don't want a copy of myself. Yeah, of course. Myself.

Yeah.

And Hans Moravec and others have written about this. What I would like to see in this space and what Eon is working toward, at least aspirationally, is this idea of a continuous transfer of consciousness so that it really is you. It's not a copy of you. It's not a low-fidelity facsimile of you. It should be a better expanded version of you that's still you. And whether that looks like moving from what Eon's doing right now with fruit flies to maybe replacing a single cell in a brain at a time, invasively or non-invasively, with a substrate-independent or substrate-migrated implementation. Some variant of that, I suspect, is where all of this will go so that you never have to worry, is it really myself or not? It will, by construction, be yourself. And I'm not sure of the precise timetable, but I think 5, 10 years from now, I think the world will see marked progress on the problem of not just whole brain emulation, but also transfer of human intelligence to new substrates.

That's very exciting.

This is very— New sci-fi corner.

I love it.

Every sci-fi trope everywhere, all at once.

Time travel?

That's a good framing.

Is time travel on the table?

We don't know yet. Okay. So it's an interesting question. We don't know whether the physics of our universe are compatible with of, I think, what most folks would construe as time travel. There are versions of temporal nonlocality that are consistent with the physics that we have right now, but not in a useful way that would be worth filming a sci-fi movie over.

What about faster-than-light travel?

We don't know yet. Same problem, in some sense, exactly the same problem as time travel. We don't know whether the physics of our universe will ultimately allow that.

'Cause it seems like if you don't get FTL, So like there's a, it's going to be a boring travel to go to another solar system.

I will say this. I think odds are pretty good that AI in the next few years will tell us whether the laws of physics of our universe are compatible with that or not and help us solve this. One of the reasons why I helped found Physical Superintelligence to discover any of such laws.

Very cool. Well, thank you so much for taking the time to come chat with us.

Yeah, great to meet you.

Congratulations on the progress.

Fascinating stuff.

Thank you. Luck, Wes.

We'll talk to you soon.

Cheers.

Have a good one. Let me tell you about Figma. No matter where your idea starts, Figma Make, Claude Code, Codex, or a sketch, the Figma canvas is where ideas connect and products take shape. Build in the right direction with Figma. And let me also tell you about Cisco. Critical infrastructure for the AI era unlocks seamless real-time experiences and new value with Cisco.. And without further ado, we have Charles Lamanna from Microsoft in the Restream waiting room. Now he's in the TV waiting room. Charles, how are you doing?

Good. Thanks for having me on, guys.

Thanks so much for coming on the show. Uh, why don't you give us a little bit of an introduction on yourself and then we can get into the news.

Sure thing. I'm on the product and engineering side at Microsoft. Uh, my team builds out things like agents, a bunch of the, the platform work. And our business applications and some low-code stuff.

So a bunch of odds and ends. And what, what's the latest and greatest in your world?

Yes, a lot of big news today. So today we announced Copilot Cowork, which really takes Copilot beyond chat so it can take actions and automate tasks and take delegation basically. We had some new agent capabilities in the Office apps and a bunch of agent management solutions.

Yeah, how are you thinking about the OpenClaw boom, people writing code locally versus Microsoft has Azure. Like, it's so easy to throw everything in the cloud. And I feel like one of the barriers that a lot of people have is like, do I need to go get a Mac mini? Can I just delegate to the cloud? Some of the apps don't just by default build an app in the cloud, or they don't have functionality. How are you feeling about that trade-off these these days.

Yeah, I mean, we're super excited by OpenClaw and all that means. I think it's like the resurgence of the PC. Like, what's the best thing you can give an agent? A computer. Yeah, let it run the apps, let it log in as you, let it message as you. I think like a Windows PC in the cloud is probably the main way people will ultimately use something like OpenClaw at work.

Yeah, how do you think about model selection? Obviously Microsoft has a, has time-honored partnership with OpenAI at this point, but Microsoft has their own AI lab and you can get pretty much any model on Azure. Do customers wanna pick models? How do you think about what's the best tool for the job in the products that you're building? But then also how much do you want to give to the user? 'Cause there's sometimes when I go into a chat app and I'm like, I just want auto, you should be figuring this out. I know the details of, pro versus thinking, but I don't want to think at all.

Yeah, so that's one of the things that we think is super important for Copilot and a lot of our products is the fact that it's multimodal. And the reason is because the best model for a given kind of task is not always the same, and it's not even the same from month to month or quarter to quarter. Like, if I look at Microsoft 365 Copilot, chat is like the main way people interact with it. Our auto-router is going to put you in OpenAI models because that's what we think works best with our work grounding in our work context. But then if you go to Cowork, the new announcement today, and you fire off a task and you want it to run for a long period of time in the background, use a bunch of tools, maybe generate some code, use like the terminal and stuff, that uses the Claude family of models. And I go to Copilot one place and get both. And then we're going to keep adding more models, um, models like from MAI and other places where it makes sense. But it's really about the experience the user wants, not so much picking exactly the version or the type.

Yeah. How have conversations been with business leaders in terms of deploying a product like Copilot Cowork? Because the ramp-up for software engineers was smooth and everyone was aware of it, tracking everything and going from, you know, better autocomplete to agentic coding was, was pretty intelligible at every step of the way for most software developers, but for someone who's lived their life in spreadsheets and email and maybe not really thought about building automations all that often, what does actual deployment of these tools look like?

Yeah, it's such a great question because I think there's like two pieces to it. The first is you have IT teams which are nervous with, if you imagine every single employee spinning up a whole bunch of agents that can create documents, send emails, run workflows. So they want to make sure they have the right governance and controls in place. That's why another big part of the announcement today was something called Agent 365, which makes it easy for IT teams to manage and deploy agents from Microsoft and other companies. So that's one part. And then you have the end users. How do you teach a typical Excel or PowerPoint or Outlook user this idea of agentic working? Even in coders, it's taken time and it's probably not fully diffused. So that's the thing that we think we're in a really unique position at Microsoft in terms of being able to do, because we can bring things like Cowork to users where they already spend their time inside of the Office apps, inside of Windows, and increasingly inside the Copilot app. And we can present it in a way where it's digestible to hundreds of millions of Office users, not just tens of millions of coders.

Is the majority of your business model still seat-based at this point? Because I imagine that there will be a transition at some point. How much of that transition's already happening?

Yeah, we're, we're still very seat-based on like the M365 side, but like the good news at Microsoft, we kind of know how to do consumption businesses because we got Azure. Yeah, we got a bunch of other very large capacity businesses. So the fact that we have user licensing and then consumption and pay-as-you-go licensing, licensing for capacity, all integrated with one commercial kind of vehicle with customers. We think we're in a pretty good position to make that transition over the next couple of years.

There was some, there was some chatter on X this weekend over how different labs are pricing tokens and plans potentially to take market share from each other. There's a little bit of a capital fight playing out.

How are you communicating, or more specifically, competing with the application layer.

Yes, yeah.

And so my question is, how are you talking to business leaders, CTOs, about how they should think about forecasting token budgets and spend and compute budgets if they're used to, okay, I have licenses and licenses maybe go up in price and my headcount goes up and I'm used to that, but all of a sudden it's like, well, everyone's paying $200 a month, but they might be $5,000 a month in a couple months, but also the models are getting cheaper and the inference is getting cheaper and the capabilities are going up and it's very hard to predict. What are you talking to, what are you telling to business leaders and CTOs?

Yeah, so the first thing is you have to embrace this technology otherwise you're going to be left behind right now because the productivity gains are just so significant. And if you go talk to a coder and you told them, hey, I'm going to take away like GitHub Copilot and the Agenta coding capabilities, they'd be like, I refuse to work in this environment. Like, it's just, it's just inhumane almost. So, so I think the, the same type of thing is going to happen for all information work, all office work. So 9 months from now, I think if you went to somebody and said, we're going to take away your agentic tools like Copilot, Cowork, they'd be like, no way, I'm not going to go back to the old way of working. So, so I think the first is there's a degree of inevitability because the benefit is so large And there's such strong pull from the end users. That said, everybody has budgets. So what you're going to see and what you're already starting to see is a bunch of controls where IT teams can start to manage exactly how many tokens or how much dollars each user can spend. And they're going to allocate that in a way based on where they think they're getting return and diminishing return. And it'll probably be like expense report approval. Hierarchy type thing. So you'll see all of that show up, but part of just doing work now is you need a token budget. That, that's the future of how we think about and how you budget for a team or a department.

Where do you have any exciting case studies on the non-software engineering agentic work side of the business? Because when, when agentic coding came to software development there was a shift in a lot of software developers' workflows where they went from writing a lot of code to reading a lot of code, reviewing PRs. They started operating at higher and higher levels of abstraction, but the decision what to build, how to prioritize things, how to build features, there's still conversations and software engineers are still doing work. What does that look like in the context of a finance department or operations department or sales group? Like where, when the rubber meets the road, like where are you seeing like, okay, this, this is ready. This is the next, this is the next sub area or category that will really, really benefit from the productivity gains from Agentic workflows.

Yeah.

I mean, a general thing and then maybe a more specific thing. What we're seeing is like, how do you measure productivity? Is itself a challenge? I mean, this is, this predates AI, right? Like lines of code, number of pull requests, number of bugs fixed. I don't think I've ever heard a good engineering leader say that's how I measure the success of my team, right? It shows up in other ways. Do you have a great product people love? Is your cost structure good? What's your install base look like? So, so I think we have to get better at capturing that final business outcome because that's what's going to get the CFO to open up the pocketbook. And continue to invest in more tokens and more AI. So if I go then take that and apply to kind of other disciplines, like what you don't want to say is, oh, our salespeople send a lot more emails or our finance professionals create 100 more spreadsheets. Like that's more like of a nightmare scenario for AI.

So 1 billion spreadsheets per company.

No, no.

I mean, exactly.

Yeah, we saw someone on the timeline say that their token budget went way up. But their revenue didn't go up. And so they're like, I don't know if it's, uh, if it's worth it. Like you got to show results.

Exactly. So that is what the question is going to be everywhere. But the good, the good news is, um, engineers historically have resisted clear measurement of impact because, you know, there's a little bit of art to it, but like sales professionals or finance professionals or customer service and support reps, they are so metric-driven. Like if I'm a sales rep, what's my quota? What's my attainment? What's my revenue per head? Every single sales department has all of those things instrumented. And we've seen with M365 Copilot, we kind of did a bunch of experiments and we published a white paper on it, like an almost 10% increase in revenue per seller just by giving them this, the family of M365 Copilot tools and things like our sales extension. Like, That's real dollars in the bank. Same thing for like back office finance processes. We've seen people use like our agents to do like accounts reconciliation. And we have like a great case study we did there where it was like 30% of all back office finance was automated out by this thing. So those are examples of real top line and bottom line benefit. And that's what everybody's going to be talking about the next couple of years. It's not going to be, you emails, documents, files, etc.

Yeah, yeah. I mean, you're in a high-level business role yourself. I find myself doing a lot of knowledge retrieval, deep research reports. But have you personally been using Agentic? I mean, this product specifically in your day-to-day? Like, even as something as simple as like a— Sam Altman told us that he has a self— he has some sort of like to-do list that does itself or something. I still don't understand exactly what that product is, but how are you using these tools these days? Yeah.

So the— for, for like Copilot Cowork, for example, I've loved that. Like my favorite magic moment has been around calendar management. I was telling the story to someone like I go look out 3 months in the future and try to see, you know, how crazy is my calendar and travel look like.

Interesting.

And I was able to decline 17 meetings for me.

Okay. That's magic for you right there.

Yes, exactly.

Nothing gives me more joy than declining a meeting. You sold me.

Yes.

And I'd say like, I went to Outlook and just saw all the meetings disappear, and it was like such a serene moment for me. And, and like, it did a deep analysis. Yeah, it went and said like, who's my manager? Who do I work with? Am I optional or required? What's the topic? And it recommended the 17, and I went through all 17, said, you know what, decline. And it was right. I mean, that's great. So like, that's one thing that I use it for for. Um, like another thing that I like right now is, I mean, I'm an engineering background, so I'm not going to go create the most beautiful content all the time. So I use it now to pull together information, design docs, meeting notes, and then go create either like architecture diagrams, presentations, or spreadsheets or docs. And I mean, I know it's just more doc generation, but it's very much more kind of like fire and forget.

So it works really well. Yeah. Yeah. I mean, even if it's just like the bones of a PowerPoint presentation with the actual data pre-populated that would just previously be a lot of copy-paste, go run this analysis, copy it over, make sure it's the same font. There's so much grunt work that can be automated with these tools. I'm super excited for that. Where do you see the design PowerPoint generation tool use? I mean, people have been talking about it's hard to create an RL environment with a verifiable reward around a great PDF, but it's easy to do that with coding. Do you have an internal dataset of the best PowerPoints ever created that you're RLing against, or where do you see that going?

Yeah, I mean, I think that is something that there's a little bit of taste in all of this because depending on the company, some companies like lots of infographics and lots of colors. Other companies like super fact-based, no images. So like, uh, you know, there's a little bit of an eye of the beholder, but absolutely we have a, a very robust harness and like gym environment where we're able to quickly evaluate models, prompts, tools, and our own like fine-tuning efforts to see if it's producing better content. So I think like this, like, uh, PowerPoints, Word documents, spreadsheets, it's gonna be like code.

Yeah.

And you, you're starting to see it. Already, but it's going to be like code for sure.

Yeah. I mean, like a lot of these things are like XML under the hood anyway and can be puppeteered like code anyway. Well, congratulations on the launch, Jordy.

I hope you don't mind. I've been calling Copilot Co-Or Coco.

Coco.

So hopefully it doesn't stick unless you like it.

Yeah, it's a very cute name.

But congrats. Congrats on the launch and great to meet you.

Thank you so much for taking the time to come on the show. Cheers, Charles. We'll talk to you soon. Let me tell you about 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. And let me also tell you about Graphite, code review for the age of AI. Graphite helps teams on GitHub ship higher quality software faster. There's an interesting data point from Apollo featured by Restructuring here. AI is making starting a business easier and the numbers confirm it. Founder. Did you see this? Weekly business formation is exploding higher.

So previously— This matches what the Collison brothers said when they were on the show. Yeah. They're seeing not just more business formation, but businesses growing faster.

It could be that AI is actually making businesses harder to start, but Stripe Atlas Atlas is making it so much easier that we're seeing an upward trend and it's fighting back against that. Now, of course, these are— Jeff Weinstein is just holding up the world. These are multiplicative, very clearly. But Stripe Atlas does make it very easy to start a business. US business applications jumped insanely after the COVID pandemic. In 2019, we were seeing about 60,000, 70,000 new business applications. I think a 4-week moving average. We've been over 120K for years now, and AI seems to be driving that trend upwards by about 20% over the last 2 years, which is great news. More small businesses, they are extremely edifying, extremely valuable. I highly recommend.

And we got another white pill.

Give it to me.

The data center buildout has allowed Loudoun County to lower property taxes by 38%. Since 2010, translating to average savings of about $3,400 per year for homeowners. This is great.

Love it. What's happening here? Data center square footage in the millions has gone from 5 million square feet to 40 million square feet. That's a huge jump.

Yes.

This is like Northern Virginia. This is like AWS East 1. Okay.

Yeah. So, so homeowners are paying They went from 1.2% of whatever their house is. So if they have a million-dollar house, they will be paying $12,000 a year. And now they're at 0.8% property tax rate because they're taxing those data centers. This is the easiest solution to every problem. If the companies are profitable, tax them. If the— if we already tax every business, we tax them by default. If somebody makes a lot of money, on an AI stock, tax it as capital gains. You get the money. The money shall flow. And so good news for the folks over in Loudoun County.

Jacob says America once named things plainly. Place, trade, surname. You had business, International Business Machines, long-distance telephone. Studebaker. Studebaker and Bethlehem Steel.

That International Business Machines logo is— goes unbelievably hard. Is this real or is this like a reimagining of what the IBM logo could have been? Because I love this logo. What a very cool thing.

That is apparently the original IBM logo.

IBM logo.

Wow.

We don't know how to make logos like that in this country anymore. Seriously. I mean, that's what kicked all this off. Somebody was upset about the granola rebranding.

America's innovation slowdown is primarily due to the amount of Lord of the Rings that founders have to flip through before identifying a name they can raise off of.

No, you can just start a company and name it International Business Machines. Whatever you want to do, just name the company exactly that. Technology and Business Programming Network. It's that easy. In other news, 2 Chainz.

2 Chainz revealed he was an early investor in SpaceX.

This is huge.

He says, it just felt like I got in like early Bitcoin.

There are some crazy early SpaceX investors. I don't know what Elon was doing with that early roadshow. We know some of these folks, very cool community.

Said, get me 2 Chainz.

Get me 2 Chainz. 2 Chainz got in at some point. And Gabriel over at OpenAI says 2 Chainz is about to be 3 Chainz. Well, let me tell you about Okta. Okta helps you assign every AI agent a trusted identity so you get the power of AI without the risk. Secure every agent, secure any agent. Let me also tell you about the New York Stock Exchange. Hopefully we'll be seeing SpaceX there soon. Want to change the world? Raise capital at the New York Stock Exchange. Can you imagine 2 Chainz and Elon ringing the bell at the New York Stock Exchange? I can. I hope it happens. Well, we have our next guest in the Restream Raider room, Julian Beck. From Sequoia Capital. He's a partner there. Welcome to the show, Julian. Good to meet you.

Hey guys, how you doing? Nice to see you. We also hope we can ring the bell very soon.

Holy, holy logo mug. Yeah, holy logo.

You know a few of those companies?

I do. We're partnered with Cisco. Thank you for backing them early, making TVPN possible through your investment 30 years ago.

Cisco paved the way.

Since this is the first time on the show, why don't you give us just a brief backstory on yourself, how you wound up at Sequoia, what you focus on, what your day-to-day is like.

Sure. So first, thank you both for having me. It's an honor to be there. I know you had a lot of my partners as well, so we're very excited to be on the show. My name is Julien. I'm a partner on the team. I'm based in London, but I spend a lot of time going back and forth to spend time with the rest of the team in California, and I started as a founder a very, very long time ago in high school. I built companies in hardware before, uh, and so that's why I've been investing mostly in software, um, for most of my career.

Yeah, wow, you, you, you were like, yeah, I gotta get out.

Are you thinking about going back to hardware? Is software dead? What are you thinking?

Yeah, well, I mean, two things can be true at once. I think hardware is really interesting. I'm spending a lot of time thing about physical AI. But at the same time, I've been investing for the last 10 years in application software. So I want to remain true to that. And this is kind of the things I've been spending time on recently.

Yeah. Take me through your current thesis on how software is changing in the age of AI, in the age of maybe some sort of takeoff. AGI is here, according to Sequoia Capital. Amongst many other people. The labs are recursively self-improving, and it seems like, you know, Sam Altman made the claim that, like, if you want to build a startup, you shouldn't build a company that is in the direct path. While more and more products seem to be in the path of AI and AGI, how are you thinking about the changing dynamic with software as we know it?

Yeah, that's a really important question, and I think We've been listening to Sam a lot as shareholders. Yeah. And recently just talking to our founders, a lot of them are wondering if they're just an iteration away from the models replacing what they're doing. Yeah. And it puts us in the position of thinking, what is coming next? What is going to be defensible? Yeah. And for us, we've been making that prediction last week that the next trillion-dollar company will be a software business that masquerades as a services firm. And we shared that thesis with everyone online. The reality is if you sell tools today, you're really in the line of sight for the models and you're effectively competing with the next generation that they're going to launch. Whereas if you sell the work, you're actually benefiting from what the models are doing and all the billions of dollars that are going towards AI. So part of that prediction was saying for every dollar that's spent on software, $6 are spent on services.

Sure.

Until now, we could only really go after the $1 because we were building tools. Now that we can actually deliver outcomes, companies should think about capturing those $6 instead.

So the, I mean, the obvious, like, first step is just get to outcome-based pricing, move away from a seat-based model, probably. Would you agree with that as sort of like the most immediate change in business structure that needs to happen? Or if you're starting a new company, maybe don't start with seat-based pricing on day one? Or is the, like, seat-based pricing is dead, maybe that's like an overblown characterization? How would you— how do you think about just like the business model moving forward?

Yeah, we think both can work. They just serve different purposes. Okay. We have companies in our portfolio that started as these copilots that are helping the workers, and then we call them autopilots, the one who, you know, do everything end to end. Many of them are transitioning from copilot to autopilot to do what you're describing. A company that comes to mind is a business called Sierra that was started by Brett Taylor, that basically helps companies with customer support, which is pretty direct bullseye use case of AI. And in this example, they're actually charging the customers per outcome, right? Per resolved ticket. And so if you spend $20 per ticket resolved with humans, they charge you, say, $5 instead, right? And that's driving ROI for the company. It's driving an outcome that's very measurable. And so that's the first more obvious company that's really working doing that. Frankly, until recently, outside of software engineering, we haven't seen a lot of this fully owned end-to-end outcome driving. But the models are really progressing quickly. And for everyone who's been vibe coding the last couple of months, I think we've all been impressed that we can code a full website just end to end without any skills, frankly. And basically our prediction is that this is going to happen to other categories. And we've been mapping out these categories and helping founders come up with frameworks to prioritize where to attack them and how to do it.

Yeah. Sorry to interrupt. One thing I'd be curious to get your take on. So if a company is just selling enterprise software that helps people accomplish a task, that feels very threatened by, you know, a foundation model company that maybe makes an agent that can also do said task and then you don't even need the software. So if a company transitions to just selling, selling the work, selling the service like you're saying, you would agree that they're still going to be under, under threat from the labs who will also presumably be selling intelligence or agents that can go and do the work. But you're saying it will maybe be more closer to a fair fight. Is that the right way to think of it? Because I still think, let's say if somebody is like spinning up a, instead of making a legal AI tool, they make an AI-native law firm and they're selling the work. I could still go to a ChatGPT, a Claude, a Gemini, and a future version of the model. I can just say like, hey, I need to do this thing., can you help me do it? So there's still going to be that competition, but you're saying that's a much better sort of competitive environment to be in than just selling like the software that enables somebody to do the task. Is that roughly right or is there something I'm missing?

Yeah, I'll just add one framework that I think is helpful for people to clarify what that means. Basically everyone is conflating the model's intelligence with what humans are really good at, which is judgment. And I think if you strip those two concepts, that you really understand what the future will look like. And so right now, the models are becoming really good at everything intelligence-related. So think of tasks that are rules-based, where just logic gets you to the goal, right? Whereas the judgment is everything that humans really spike at, which is more are harder to define, right? It's instinct. People call it taste, call it experience. And those things are things we think the humans will remain really good at for a very long time until intelligence is able to absorb these skills into it. But it's just really hard, right? Because we give the models a lot of credit for their intelligence, but they're just trained on set amounts of data. And so we actually think that this hybrid motion of coupling the intelligence from the model with the judgment from the humans is going to help you build the next generation of companies, which we call these autopilots.

Right.

So to your example, in the law firm, you will automate a lot of the intelligence work that's very much codified, which is what the model will do. If you put the employee at the center, someone who's really experienced, who can talk to a jury, who can market pause at the right time. Those are all the things that you learn only from real-world experience that a top lawyer will be able to experiment, but that the models will omit today.

Does that make sense?

Yeah.

I still— maybe I'm too AGI-pilled, but I just assume that a sufficiently advanced agent would be able to capture and accumulate the ability to—

Drink 2 AGI pills. Just talk to Dr. Alex Wissner-Gross for 30 minutes, and he's like, I'm uploading my brain right now.

No, it's not that. But it's just the idea of, you put a smart person into an organization where they don't have any experience, and they start doing things, and they start getting feedback from other humans in the organization. Over time, they accumulate enough knowledge and domain expertise that they're able to excel at something that even using the customer service example, you've always been able to outsource customer service, and you can go work with a firm that is excellent at customer service. The alternative has always been to effectively hire a really great leader for that part of the organization, have them build out a team and a structure and an approach and execute something that would be a similar result. There's like a cost trade-off, but—

It does feel like overall there's like maybe some pressure on the B2B2B model or like the B2B2C model. If instead of, to use your example of customer support, There was a time when software companies would sell phone tree software to, not to companies directly to do customer support, but to customer support outsourcing agencies that would then deploy that. And now you're seeing more companies go directly to the firm that makes the customer support experience. End-to-end, go directly to the company that needs customer support as opposed to having this like middle level. And it almost makes me feel like we're maybe going to rotate even a little bit further into D2C. When I think about law, there's a lot of stuff where the software would go to the lawyer, the law firm goes to the person, and if the person can go straight to the law firm or the legal model, you wind up with a little bit more B2C see activity a little bit more company directly to the service provider or the technology provider. But again, this feels a little bit dependent on specific market dynamics. I'm interested to know your reaction to that, but also what does hold in your mind in the near-term AGI scenario? Because there's been a lot of debate over how will marketplaces perform. Reform, or are network effects still super sticky in a world of AGI, or what are the powers, the 7 powers that force us to stick around?

Yeah, I guess part of it, to kind of close out my thoughts, like I agree with the entire thesis in like the relative short term, like the 1 to 2 years, but then beyond that, there's still some uncertainty around. There always is, yeah. Like I don't think it, solves the sort of discount that even traditional SaaS companies are getting today.

Sure, sure, sure.

Yeah, Julien? On that point, I think you're right, and basically what we realize is that no one really knows, right? I've been in many partner meetings recently where everyone just is astonished at how quickly things are moving. I've been doing that for 10 years, and it's just by far the fastest rate of change I've noticed. But some of my partners, you know, Alfred Lin, Pat Grady, they've been doing that for a lot longer. And they also acknowledge that everything is moving so quickly. And so what we're trying to figure out is also not what is going to change, what are the things that will remain true no matter what, right? Customers will always want things faster, better, cheaper, right? And so that's kind of the thesis that we're describing. They're just categories that require more intelligence than judgment. And so we think that in the immediate, the ones that require more intelligence will be actually really interesting for these autopilots because you can have, you know, instead of having 10 humans in one AI, you have one human and 10 AIs, right? And just that ratio is just going to shift as the models get better and are able to absorb what we used to call judgment from the humans that are is turning to intelligence. And just to put like a practical framework to that, you mentioned outsourcing. It's actually the best wedge right now is just to focus on outsourcing if you want to capture this services spend, because there's 3 reasons. The first one is that outsourcing has already been agreed to a third party. So it's very easy to sell, right? Because it's already a vendor that's external to the company. The second is that there's an existing budget that's assigned to it, which you can peg yourself against. And the third is that it's actually very straightforward to swap a supplier. It's not so straightforward to reorganize your company, right? We all talk about headcount reduction, you know, in terms of ROI, but actually putting that in practice takes time, whereas swapping a vendor is very quick. So I actually think that And that's the prediction, is that the companies that will succeed very well with this autopilot model will start with where there's a lot of labor spend that's being outsourced today. We mentioned customer support. We'll mention also things like accounting, tax, insurance, and so on and so forth.

Is there a world where in the next 5 years or something like venture capitalists truly shift away from software and go back into rare metal mining or something crazy like that. I just keep coming back to Daniel Gross, now at Meta, venture capitalist. I think one of his first AGI trades, the first bet was like, in the world of AGI, is copper mispriced? And it was. It doubled in price over 2 years. And I'm just reflecting on this idea that like, if the job of venture capitalist is to back software companies, that's one thing. But if it's just to allocate capital in a way that can have exponential upside and single investment can return a fund—

Well, the thing that I think you said, would it ever make sense to stop investing in software? I would hate that because we obviously love enterprise software. But it has been funny, a lot of people have pushing like, you know, the models are so good, we just need better harnesses. Yeah. And to me, this is giving me flashbacks to early GPT days where people were like, ChatGPT is really powerful. You just need the right harness, right? All these like wrappers.

Yeah.

Yeah. That had explosive growth very momentarily. Yeah. Until—

and so I do— most AGI-pilled person ever over here. I love it. I love it. But what do you think about the more physical aspects of the job, like rare metal mining, for example?

Does that seem like something people will be working on in a couple of years?

Yeah, I mean, everyone's talking about the fact that energy will be the biggest bottleneck to train these models. So it's very natural that you would go there. Personally, that's not really where my skill set is. I really hope you're wrong in that we'll find other areas in between.

I think the real path will be somewhere in the middle.

Hopefully, right? My way of doing things has always been to focus on talent. And so right now I'm just always asking the smartest people I know, who do you know who's the smartest person? What are they working on? And you're right, more and more people are going more into the physical world. But right now that probably means things like robotics and things like that. People are making a lot of money trading commodities, but that's a different world.

Yeah. Who knows? Maybe that's the job of the venture capitalist in 2030, just trading copper all day long.

Exactly.

You mentioned using AI to build a website despite maybe skills that have languished since your time in software or hardware. But how are you using AI tools these days? Are you still doing mostly knowledge retrieval, market research for your job in venture capital, or have you actually built tools that have accelerated certain things? Have you automated any of your workflows? I'm just kind of curious how people are using AI these days.

Yeah, well, I don't know. I hope my competitors are not listening.

So they are. So you can say I haven't touched it. Actually, it's useless. Yeah, exactly.

You should uninstall organic capital.

Pen and paper actually is underrated.

Exactly.

Back to the basics. Pen and paper. And don't try cold code. It's really, really direct. Now, jokes aside, it's been really a transformation, and particularly the first quarter of this year. I mean, I use it for all kinds. The work is basically sourcing, picking, winning, company building and harvesting, right? Then I really have workflows that fit all these boxes. So it goes from things as at the end of the week, it will just scan my entire calendar, look at all the meetings I took, look at the granola notes, which is a transcript, do an analysis and help me try and find the key questions and the due diligence. And it'll help me just prioritize, right? Because this job is just so much about prioritizing. It's a power law, right? There's only 3 meetings in the year that will really, that could change the course of your career. And so making sure that you're attuned in those meetings is really important. So that's really what I would say Claude has helped me do right now. And I'm sure OpenAI will release something next month that will be amazing and so on. You just have to experiment. And I think that's also what we've seen with things like Open Claw being so top of mind is like this culture of experimentation is not just in software engineering now, but it's transpired across the entire knowledge workforce. And venture capitalists are not an exception.

I love it. Well, thank you so much for taking the time to come chat with us.

Great to meet you.

Have a great rest of your day. Thanks, guys.

We'll talk soon.

Goodbye. Great. Let me tell you about ElevenLabs. Build intelligent real-time conversational agents. Reimagine human technology interaction with ElevenLabs. You know, if I was a venture capitalist, open Claw every morning at 5 AM. Text me the name and market cap of the company that I passed on 3 years ago. I need inspiration to grind harder. That's the secret to success, as is Lambda. Lambda is the superintelligence cloud. Building AI supercomputers for training and inference that scale from one GPU to hundreds of thousands. And without further ado, we have Owen McCabe, the CEO of Qualcomm, a huge supporter of TVPN as well. Welcome to the show.

How are you doing?

Great to see you. Great to see you both also.

It's been too long, man. You have— you just have a voice for radio, live streams, podcasts, all of the above. It's incredibly calming.

Yes, I love it.

Okay, what would you like me to say? I can record a meditation for us. We could kind of start up the hour that way.

Do you have time?

Do you have time to meditate these days?

It's kind of the only thing that keeps me sane.

That's good. Well, you have some new funding for a new project. Break us down. What happened? What's the plan? How much did you raise?

Take us through it. Yeah, so we raised $250 million.

Let me hit the gong.

Congratulations.

Did that gong get bigger since I was here?

It did get bigger.

It did get bigger. It would be cool if it got bigger every single season, right?

Fast takeoff. Just wait, there's a fast takeoff.

There's a fast takeoff in gong size over here.

It's growing exponentially. Except this was actually, I think, the largest we could find in the US. So we gotta go.

Yeah, no, there's always bigger. I kind of have gong envy right now. I wish I had a gong.

We'll send one over.

Let's talk about that later. Well, we'll talk.

Anyway.

We'll talk. So $250 million, what's the plan?

Yeah, yeah, so I mean, we've been building this thing called Fin for 3 years now. Service was just this obvious place where AI was going to, you know, blow up. And all of a sudden everyone else started to chase it. We were about a year ahead of everyone else. And we started doing, you know, customer queries and then more complex problems. Originally it was answering questions for customers and it's fantastic. You know, the typical response rate with a human was 1 to 2 days. You'd have to deal with the beautiful imperfections of humans. Unfortunately, these bots are just far more perfect than humans and they answer in seconds. And of course they're cheaper too. So we started to say better, faster, cheaper. These things are really better, faster, cheaper. So that was a tremendous success. Finn is the biggest service agent in our category, nearly $100 million ARR, 8,000 paying customers. You know, some of the biggest logos in the game. Thank you, thank you. From Anthropic to, I don't know, Snowflake, the cool guys like Polymarket and Vant and everyone else are on Finn. So wonderful. But where does it go next? You know, clearly it's just so obvious as each of these categories develop that there's just more to explore. There's none of these categories are in any way fully baked. And that's one of the interesting things about kind of all of these AI spaces that we're still kind of pulling the thread and seeing what's going to show up next. And the next big play for Fin is not just answering service queries, but really getting deep into every stage in the customer lifecycle. And so we call that the customer agent, and this money will primarily bring the customer agent to market.

Okay. So when I want great customer service for a company that I'm running, I want, yeah, you know, search the knowledge base, answer the obvious questions, handle the refunds, track. That's part of the customer journey. Are there other touchpoints that are unlocked by AI along that journey?

Absolutely and completely, yes. From day zero on the website when the customer has very basic questions, when they get more interesting, when the business wants to start to qualify and understand who that customer is, when the customer actually then wants their hand held when they sign up or they want to actually engage and learn and meet salespeople, they want to book a demo or they want to talk to a virtual salesperson. I mean, that's just the very top of the funnel alone, but to get into the application. If it's B2B, there's all the onboarding, all the detailed help required to get someone successful. If it's e-commerce, it's actually finding products and getting it into the cart and bringing them back for more shopping. So there is really the, literally the full customer lifecycle of work to be done by agents, and customer service was just the little sliver. I want to know what's happening off camera though.

Oh, we just have our next guest walking into the studio. Got it, got it, got it.

So I'm just—

Hopefully they're less interested than me.

Yes.

I want to talk about some of the best customer service agents that I've hired have sort of wound up— Real people. Real people have wound up in sales roles almost. Like, they have a fish on the line, they're reeling them in, saying, "You don't want to cancel. We actually— this product does exactly what you want." Right. And the fact that it was delivered one "Hey, you're one day late. We're happy to refund you, but it seems like the underlying issue is that you actually want this product instead of this product. Why don't we get you over here?" That's this different skill set that comes out in the best customer service agents. How are you thinking about actually driving incremental sales through customer service agents?

Yeah, I think one of the interesting things is that a lot of the work in these functions isn't entirely separate. We just needed to draw lines between them when we were running human organizations, because you can't possibly have one person who's outstanding at service, sales, marketing, success, etc.

That's the thing that's going to go away with AI. Yeah, and the SaaS market maps too complicated.

Totally.

A pain in the ass. But unfortunately, that pain in the ass is coming with AI, because I just don't believe you're going to have these separate sales agents and marketing agents. You're not going to have these things fighting with each other, having different contexts and information, different memory in different parts of the customer lifecycle, different messengers, not going to happen. There's going to be vendors that will provide one seamless solution across the customer lifecycle. And that's a big thing we're passionate about. And one of the very fascinating things we've seen, we've been building this for quite a number of months now and we're going to launch the first part of it next month. One of the fascinating things we've seen is that When there's an underlying relationship with the agent, and then there's actual connection in the moment, and then there's clear intent, the context is primed for selling and encouraging the user to do something next. If you're on a website and you get a super annoying pop-up that says, "Hey John," or "Geordie," in fact, they're not even going to use your name. They're going to say, Do you want to learn more about our whatever? And you're like, get out of here. It's like not the right time or place. Whereas if you're actively discussing, it can ask you questions back and it's so much more effective. And so we've seen just these things, the service agent, when it starts to do sales, for example, and vice versa, they're far more effective. And so the real magic from AI across the board, and of course this agent world, is going to be when the agents can do things that humans can't.

Yeah, one thing that stands out is like right now you have this spikiness in dialogue as a customer of a business, right? You might have a question coming in and it like creates a moment and then it goes away. And then you get maybe transition to someone else and you're talking to a real human. You talk with them a little bit and you get passed to someone else. And like there's something about creating, I think, the customer experience of having this continuous conversation conversation with the organization that you're a customer of that I think will be super powerful because it just means that, you know, a split second away the entire time. I was talking with some guys that own a number of like big car dealerships recently and I was talking to them about AI and they weren't fully sold yet 'cause they use like ChatGPT, but they haven't had anything spread across their whole organization., and I was explaining to them how there's so many different moments where like a customer lands on their website and they wanna know if a car is available, but somebody gets back to them too slowly, and like they, they, the light was like finally going off for them. Do you think that there's a number, how many industries out there have been kind of resistant to even using software from Silicon Valley to date that are going to kind of take a second look at all the different tools and things like Fin now that they maybe have had a cool experience with AI in their personal life and they're basically saying they're willing to kind of look because maybe the historical SaaS platforms, they tried them, but it wasn't quite magical yet.

Yeah, there'll be so many spaces, so many verticals that some smart entrepreneurs will benefit from bringing sexy technology to that aren't immediately obvious to those of us in Silicon Valley. You know, we tend to sell to ourselves. It's really insular. So all the names I just mentioned there, like Anthropic and Snowflake, they're all the cool brands in our world. However, there's like car dealerships that are, you know, I don't know, maybe it's not as big as Anthropic, but they're pretty damn big.

Depending on the car dealership, I would put car dealership above Anthropic. Topic and Snowflake personally, if it's like a Ferrari dealership, Lamborghini dealership, like just on the cool side, we gotta give it to you.

On the aura factor, sure. For sure. Maybe not on the economic factor. But there are so many industries that just don't have our attention. And so there will be, you know, there'll be companies like ours that sell into many verticals, but really, you know, we'll focus on all of the verticals that are adjacent to ours. And just tens of billions of dollars of opportunity just in that alone. But there'll be people who will say, we're going to make agents for dentists, car dealerships, or certain government bodies, maybe even for tax bodies. There'll be people who will do this because so much of the effort is not just in building the technology, it's actually creating the trust, as you say, Jordi, with the actual people who will have to deploy this technology and make it work for them with their legacy systems. There's a lot of very unsexy work that will be rich and rewarding for some people who bring it to these places.

Talk about the decision to raise debt instead of equity.

Not something you see. Yeah, I mean, yeah, I know. I just realized I had this kind of blind spot for this as an option for capital. And there's just so much equity opportunities out there. We raised a bunch last year for is secondary. But when you're in our dynamic where you're accelerating as quickly as we are, like every year we're doubling our rate of growth pretty much, there's no way we're going to be able to achieve the valuation that we're pretty confident we'll get next year. And so, any equity raise is going to just be massively dilutive. We calculated that raising this in debt is going to cost us about a tenth the kind of price to shareholders that it would cost in equity. So this is, you know, the incumbents have disadvantages. They tend to move a little slower. They have older businesses they need to wrangle with, but they have access to phenomenal amounts of debt. Like $250 million is a relatively small amount of debt for the size of our business. And I think that you're going to start to see a lot of these older SaaS companies as they pivot to AI and they actually find new growth, start to lean on debt as an opportunity to invest rather than super dilutive equity. And honestly, I just think it takes a little bit of maturity to take that route.

There's so much pressure from employees.

Yeah, it's not just maturity, it's confidence.

Yeah, to know your business. Well, I like to think so. But the reason I say maturity is that there is so much temptation to go and attain that latest hottest valuation. Your employees want it, your previous investors want it, media wants it. But if you're able to tell them we don't need it, actually, it's way cheaper for us to get a lot more capital with far less effort, then you can actually thread the needle in a way that I think is going to work much better for big companies like ours.

The gong over here rings just as loud for debt as it does for equity, so don't let them—

Yeah, I didn't actually— You guys are hiring a ton. You said You're bringing on 650 new people this year?

Yes. Yeah. So that's gross. I think net it'll be a little less than that. It's a phenomenal amount of hiring. I fear it. I celebrate it with great caution. There are just countless stories of the hubris of CEOs who boasted about how many people they've hired and it's all ended in tears. So it's actually not a point of pride, frankly. I think small is beautiful, but we are growing incredibly fast and we need all the help we can get.

And you've been at this long enough to know the impact of what, you know, 100 people, 600 people, what impact that will have on your organization. So I have the full faith in you. Thank you so much for taking the time to come chat with us. Great to see you. Congratulations.

Good to see you both. We'll take care and keep doing your thing.

All the best.

Have more amazing stuff.

Goodbye.

Let me tell you about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy apps, servers, databases, and more while Railway automatically takes care of scaling, monitoring, and security. And let me also tell you about Sentry. Sentry shows developers what's broken and helps them fix it fast. That's why 150,000 organizations use it to keep their apps working. And without further ado, we have Alex Epstein. He's the author of Fossil Future. Live in the TVPN Ultradome. Welcome to the show. Good to see you. Yeah. Thank you so much for coming down on short notice.

What are your texts been like the last 48 hours? Is it over? Because you're— because Jon and I were thinking, who's our oil guy? We had one.

We had one.

Yeah.

And I'm local too. Yeah. If I'm in town, it's great. I always like coming in person versus—

Thank you so much.

So much.

The other stuff.

Yeah.

Appreciate it.

So yeah, lots. I mean, there's so much in energy, but I'm happy to talk about this.

You lead the conversation however you think it should go. What are— what's going on?

What—

how should we even be thinking about oil prices right now? The impact of the war in Iran? How do we— how should we be processing this news?

Okay, so we did this. So there's a million things to comment on. So I'm going to try to segment what I comment on. We did get involved militarily in Iran, which I would just say in my non-area of expertise, in general, that is a good idea. I think it's a better idea to have Congress involved if you can get them involved. And I think it's a good idea to have a lot of expertise on international oil markets being brought to bear. Sure. When you do this. And I think this administration does a lot of good stuff on energy, but I think it's pretty clear at this point that that there was not maximum expertise brought in on this.

And so the standard issue when you're talking— And like indicators there would be like the Strategic Petroleum Reserve hadn't been refilled.

I mean, we can talk about that. So let's make sure to get back to that one. But the main thing is just, you know, rough numbers, you know, Strait of Hormuz, which Iran has control over, is 20% of the world's oil production is flowing through that. Every day. And so whether you're looking backward or looking forward, there is no replacement whatsoever for opening that strait and keeping it open. And one of the things we've heard from different people is, oh, we have a lot of options on the table. Like, there's a lot of things that we can do. There are definite things that you can do, but none of them is $20 million, right? So it's just, I mean, that's, that's more than U.S. oil production just flowing through that one place. So yeah, we can talk about the other things, but I think the first thing people need to recognize is If you do not get that thing open, then you have dramatically higher oil prices. And then the other thing to get is oil prices really matter.

On that street for Muse, dumb question, can't you just go around?

I mean, you can try to some extent, but there's a reason why 20 million a day are being routed through.

Because it actually comes out of there as well, right?

Yeah, there's just, I mean, all of these things are very, very optimized. One of the things about just understanding the energy industry is like things are just very optimized in terms of where the infrastructure is, You're doing all of these things. So it's— there are other— and we'll talk about this. I mean, there are certain pipelines you can pipe more oil through. There are alternate routes, but we're talking about in the millions a day, maybe. Most of our other options are sort of in the 1 to 2 million barrels a day. And just so people know, barrel is 42 gallons. So you're talking about almost 100 million gallons a day for this stuff. But yeah, so you don't have a lot of great options. And I think that's very important for— I'm going to give some options. But the number one option is keep that thing open. And there are basically two ways to do that. One is you just win the war, if you want to call it that, a lot more quickly. Like if you get some form of surrender and you have friendly people controlling it, then guess what? It can open. That's one thing. So I'm not an expert on how to do that, but schematically, doing that is very effective.

The other thing, which is a little bit— Game theory for Iran is they actually benefit from global markets being in turmoil because it gives them some leverage over any type of negotiation.

I mean, this is the biggest damage that they can do. I mean, clearly we've seen they're not going to do— they're not going to send missiles to Israel and get rid of Israel, let alone do anything to the US. But they have this incredible control of the one of the centers of the world economy, and they have a bunch of stuff going on there. So let's talk about— so option 1 is you get a surrender and they don't use all of their options. But then let's talk about their options. So they have mines in this thing, so they can blow up mines. They have missiles. Maybe the worst thing they have— and this is a change in recent years, and I— you've had Palmer on here and Ethan on here— like, the drone thing is just a total game changer because you can have a small vehicle you can launch one of these, you know, Shaheds out of the back of a truck. They can go really far, and that's a lot harder to deal with than a fixed installation.

Yeah, they have something like 500 miles of range on a typical—

Yeah, depending on—

Yeah, right. So a truck drives up to a beach, drone launches, as soon as it finds that cargo tanker that has a bunch of oil on it, it hits it.

Yeah, and then there's the— you're worried about that.

Sure.

As you're going through. So you have to think about Those are the kinds of threats you face. So how do you deal with that? I mean, basically you need the US to lead some sort of convoy where people have sufficient security and economic assurances that'll go through. So what's involved in that? Well, one thing is, if possible, you want to get allies involved. And it's even possible you get unconventional allies like China, who has— well, they depend on this. Now they have a bunch of— unlike us, you want to go back to SPR, they've been filling up their reserves. So they have more reserves relative to their imports ports than we do relative to ours. If you think about ours, we have about 400 in the SPR, and we can release about 4 million a day. So we don't have as much, but still, they really care about this, and a protracted thing is bad. So, but for sure, Japan, South Korea, India. So one thing is just at a high level, if you're doing this convoy, can we get these countries that are aligned economically, maybe China, but definitely these others, can we get them involved? Because then it's an addition. They have military presence there. That's an additional threat to Iran if they attack one of their ships. So, but that's a macro thing is just, can you do this alone or can you do it with allies? I think ideally you would do it with allies. Then there's a question of what you do about these various things. And I think the biggest, I mean, the other macro thing you can do is just in some way credibly threaten Iran and say, hey, if you, and I can't give the details of how to do it, but if you, If you attack us, if you attack in the Strait of Hormuz, it's going to be very, very bad for you. So just at a high level, can you make that threat so they understand this is going to be very, very— now, you know, wiped out a couple of layers of leadership. So they might be taking these things seriously. After that, I think the biggest thing from what I've heard from military experts is just these drones. Like, what do you do about these drones? How do you give any kinds of assurances? And as far as I can tell, we don't have any one perfect solution. But you can do a certain amount of stuff stuff from the air. You can do a certain amount of stuff from the ground, although it's very, very expensive. And then most cost-effective where you can do it is you can take out certain stockpiles and facilities with the proviso that these things are decentralized. And of course, I have no specific knowledge about where they are.

But in theory, if there's the suicide drone, the Shahed, that's— there's a factory and you take out that factory, you have reduced that.

Right. So that's going to be it. And, you know, I know you guys had on Ethan Thornton, who I'm friends with, You know, he talks about this a lot, is just this cost asymmetry issue, is you just want to be doing things, even if you're way wealthier, you do not want to be spending $3 million to take out $30,000.

Which might be what's happening with a Patriot missile battery.

This, this, this. Yeah, that's the kind of thing that happens. So you need, you need a combination of those kinds of things. And then the other thing. So that's all the security stuff. So you get the allies involved, you take these actions specifically against Drones. I think the— what was I going to say about the other thing you can do?

I'm just making sure.

There's still just— well, like the other big thing. Oh, the insurance.

Okay, insurance. Yeah, the insurance. So, and this has been floated by the president and it's a reasonable idea, but my understanding is the insurance vehicle we have, the development corporation, does not have enough funding. So you might need to go to Congress, but basically you want to be able to say, hey, we're going to lead this as the United States. We're going to lead it with allies. If Iran does anything, it's big trouble for them, and we can counter their specific attacks. And if something goes wrong, you're insured. You're trying to create that confidence to get enough people going through, and then you get a certain amount of stability, and then you start to get closer to your 20 million barrels a day. Either that, or you just defeat them really quickly, or both.

So, so the other factor is just the loss of production, disruption to production. In plants shutting down, plants being damaged. How does that factor in? Do refineries get brought back? And I'm talking about an allied country.

You mean just the trend over time or what?

Yeah, generally. I mean, just seeing the videos coming out, you've seen places, I think, like Qatar, Kuwait, places like— yeah, basically just saying, yeah, we're going to pause production because we don't even have a place to store this, and/or they're suffering damages, and it doesn't make sense to keep refineries online.

But this is all just downstream of, do you have the route open? Yeah, got it. Like, if you have the route open, then you're good. It solves everything. If you don't— the way to think of the other things, so we do have some other interesting options, we have to think of them as these are temporary stopgaps. Like, if anyone says— I mean, you hear some crazy things. People are like, oh, Venezuela. We talked about this last time. Venezuela is less than a million. It's just, I mean, it's beyond crazy. It's just impossible. It's not like Venezuela has the ability to just increase their oil production by any significant amount in the near future. I mean, we're talking about companies considering going in harm's way to get some incremental boost.

So Venezuela is essentially useless. So best-case scenario in the next year, Venezuela goes from under a million to 2 million?

No, I mean, 2 million would be I mean, if we made it our entire goal in life, I don't know, but why would you for that?

Venezuela was at like 3 million, like, 10 years ago or something like that.

Yeah, some years ago, yeah, yeah.

Right, but there's so many— But there's a lot there.

It's not gonna happen. It's just so, yeah, so Venezuela, I would not even, that's just in the category of, it's not even a bad idea, it's just not an idea. There's nothing, there's nothing there. No, no, that's right. So the bad ideas, by the way, are, The worst idea imaginable is ban oil exports from the United States, which I have heard floating in conversation.

Seems somewhat reasonable.

I want the oil. Of course, because we have oil, so we don't want to sell. So, I mean, so many different things. But one obvious one— not obvious one, but it's obvious if you know how these things work— is the refiners in the US do not match up well to the oil produced in the United States. The refineries in the United States are based on heavier crude, like crude is rated in terms of weight. And sourness. OK. And so our refineries are mostly for heavier crude. It's actually one reason why Venezuela has some appeal. Canada has a lot of appeal because they have heavier crude.

Yeah.

And we're going to talk about— we're good at refining it. We're going to produce a lighter crude.

Yes. Because the shale revolution was just this dramatic shift. Yeah. Very quickly in oil production. So our refineries are not primarily equipped. That's why when we ended the crude export oil ban, I forget, it was 2014 or something like that. It was just this huge unlock. So it allows the global market—

allows us to produce for a global market. You can't look at American oil production as a monolith. It's not fungible. We import heavy crude and export light crude.

Is that roughly correct? That's roughly correct.

Yeah.

So that's one thing. It's just insane to do that. But in general, think about it. What is the value of higher prices? Is it stimulates production? We're going to talk about all these different ways where we want to unlock oil. So you want to tell American oil producers Hey, we're going to actually totally screw you over. It's actually going to be worse for you maybe than it was before because you don't even have a market. So you're going to strand all this oil. So that's terrible. There's also this financial manipulation idea which has been floated, which I think is terrible. Hey, let's manipulate the financial markets basically in a way that you're selling futures short on a certain time period and then you're buying them. What you're trying to do is lower the near-term price and raise the long-term— or Treasury.

Yeah.

Right, so you're— A very American way to try to solve a real-world problem.

A lot of people like that.

The only solutions involve unlocking oil in one way or another. You do not want to screw up the markets so that people are less inclined to unlock oil, nor do you want to screw up the markets to destroy information. So again, the main unlock is reopen that strait. There's no way around that. But then we can talk about other things. So one category is what you call spare or emergency. Emergency capacity. Okay. And this is—

there's some uncertainty here, but it— Quickly, this is separate from the strategic reserve?

Yeah. So, well, emergency capacity is the strategic reserve. Oh, it is. So spare means you could be— you could pretty easily be producing more per day, but you're not because you're not happy with the price. And this is mainly Saudi Arabia is considered Exhibit A in terms of amount of this. Now, there's debate among experts about how much they have. Yeah. But some people estimate they have 1, 2, 3 million barrels a day of spare capacity they could ramp up on some kind of—

And do they know how much capacity they have and they're just not sharing it?

Or does no one know? I mean, that's always been the story.

They know.

That's been the story of the Gulf, right? They're not going to— if they just flooded the market with every single barrel that they could possibly produce, they would be getting a worse, much worse price.

Yeah, and there's this cartel arrangement that does— I mean, in a sense, everyone has this on different timescales. You think about US shale producers. If prices were sustained at $100, guess what? There's a whole bunch of shale deposits that they could produce at $75 a barrel and make a fortune on that they're not going to do 4 weeks ago when it was around $60 a barrel. But in terms of the thing about the Saudi oil is it's produced at a much lower cost. So this is them saying, hey, we have some of this on the table. And then the question is, you need to be able to produce it. And then you need to be able to transport it. So you have estimates around maybe you could get 1 to 2 million barrels a day. Yeah. So that's one of them. In terms of emergency, if you look around the world, you don't have a lot of immediate spare capacity besides that. So we have what's called emergency capacity or Strategic Petroleum Reserve. You know, it's, it's— Jordy mentioned that we didn't fill up our Strategic Petroleum Reserve. I think this is unequivocally a mistake. I mean, if you look at the whole thing of the Strategic Petroleum Reserve is you want it for when you need it. And the best possible time to fill it up is when prices are low.

And is that physically a place with a bunch of barrels that we—

there's a bunch of different parts of oil. Well, there's multiple different places, but, but you can think of it.

We basically cache the physical oil in America or somewhere we own it or can access it. Yes. And it's sitting there going unused until we're—

yeah.

Now there's issues of Biden. Biden definitely misused it. So they They used it to basically have lower gasoline prices during midterm elections, even though, parenthetically, their entire policy goal was get rid of fossil fuels, which means you want the price to go up so people can't afford it. It was a very cynical kind of move. They did it in a way that degraded the facilities. So let's say we're at about 4 million barrels a day we can get out of that thing. We have 400 million in there. We should have over 700 million. We could have easily filled it up. I mean, it was a perfect time when you're talking about $50 barrel oil, $60 a barrel oil.

[Speaker:TREVOR_BURRUS] Yeah. And can I do the basic arithmetic of 4 million a day, 400 million in the reserve, 100 days of oil? But that's not actually— if we're talking about relieving price pressure, you could trickle it out over a year and have one quarter of the effect.

[Speaker:JOHN_DEL VECCHIO] But just think about $100 million a day, 100 million barrels a day is roughly the global market. [Speaker:TREVOR_BURRUS] OK, OK.

4 million is just, that's its maximum throughput capacity.

Got it. It's kind of an analogy to batteries.

Sure.

We're talking about batteries. So there's a certain amount, like with a battery, if you hear there's a 1 gigawatt battery installation, that means that's the, it usually means there's 1 gigawatt for 4 hours.

Sure.

But if you wanted to do half a gigawatt, then you could do 8 hours, right? It's the same deal with this. But keep in mind, this is not 100 days of US oil demand. And by the way, for the refinery reasons, we can't just supply it all with our SPR. It's just the maximum output is— but 4 million barrels a day is something we're talking— that's one-fifth of what's flowing through the Strait of Hormuz. So yeah, it's, it's a thing. But, but yeah, you can think of, yeah, maybe we'd be willing to do 1 to 2 million barrels a day from ours. Now, interestingly, the International Energy Agency— we are part of the International Energy Agency reserve program. So we have about 400, but they have about 1.5 $400 million. They have $1.4 billion. So we have allies around the world that can also release this oil. So maybe we could get 1 to 2 million from them. Now, at least last I checked the news, they hadn't agreed to do this because they don't think it's an emergency. But that's what it's there for. So notice the pattern is—

At what point would they think it's an emergency?

I don't know. I mean, and we can put pressure on them. I think some of these determinations have to be made in conjunction with the military determinations. Decisions. You can't make them in isolation. Because if you think about what's the military objective, how close are we to that? Or how close are we at least to opening this up via high-security and financially backed convoys? Then you can think about the timetable here. If you think about, well, the Strait of Hormuz is going to be closed for 5 years, let's just get used to being poor. Blackbilled. No, I mean, this is— people don't get oil. Like, oil is— there's a reason why I focus a lot of my life on oil. There is nothing more— there's no material in the world of energy that is more valuable than oil, because oil is just so unique in terms of it has this very high energy density and portability. And there's nothing like it.

I mean, nuclear has high energy density.

What about plutonium? Yeah, yeah. It doesn't have the same portability.

5 years, we're gonna be rich. We're gonna have energy too cheap to meter in 5 years. I talk to a lot of nuclear founders.

Well, look, I love nuclear as much as anyone. They don't even have a portability solution though for the near future, unless you're talking about aircraft carrier or an icebreaker or something like that. So I'm just saying, the world runs on mobility, and oil is ideal for mobility.

Portability in the sense of like using nuclear to actually power trade and commerce and moving ships around. Yeah, yeah. I mean, even if you can power the grid, you will be poor because you will not be able to trade. Yeah, trade is really important.

Trade is really important. Trade is— and also, it's within the country too. We don't have the nuclear trucks to move things around. Sure, sure. Obviously, you want super cheap electricity and you want super cheap transport fuel. And we want to see if we can electrify some of this stuff. If you can do it cost-effectively.

Yeah, there are some people working on electric bulldozers, for example.

It's very hard. And by the way, guess what? You need a lot of oil to even get that whole supply chain started because you need the Yep. So anyway, bottom line, yes, I meant that thing about we can all be poor very seriously. Very seriously. So we have, if you look at the spare capacity, I think that's all of them. So we have the rest of the IEA, we have ours, and then to my knowledge, we have Saudi/UAE, them using spare capacity and then piping it where they can. So maybe combined, we're talking about 6 million a day if you use all of those. I mean, somebody could imagine $10 million. And all of these have time, particularly the emergency reserve things, these all— like, at least the Saudi one is you're sort of— you have a deposit, a big deposit here. It's just you have a small reserve. So you can't do this forever. And of course, there's a risk of if you deplete the SPR more, what if you really, really need it in the future? And then there's a bunch of nearer-term stuff that you can do, and some of which I like just because it could be done. Anyway. So number one, to piss a whole bunch of people off, is let's stop using this Jones Act. Do you know what this Jones Act thing is? I've heard of it, but explain. So the Jones Act is a set of restrictions that say that you can only transport things among ports in the United States if you have— everything is basically American. So it's owned by an American, it's run— I don't know, all the—

it is all—

yeah, it's basically American ship, American crew.

Yes. So it leads to all these crazy, crazy things, like we end up importing things from— like instead of importing it from one part of the country to another, you end up importing it from some really faraway place. It leads to that kind of thing. But in this case in particular, this— as Californians, we should all recognize this. We don't have oil being piped into California. So we want very efficient maritime transportation, and if you suspend the Jones Act, then you can get more efficient transportation, which means We can at least have lower prices here, and we can have less of a price shock. But in general, I want people to get used to having no Jones Act, because I want there to be no Jones Act. So that's a good thing to do.

My favorite idea here is— And I imagine there's a bunch of powerful lobbies that will fight and kill for the Jones Act, because there's an entire industry built around—

Yeah, it's the idea that we can only have a good shipbuilding industry. I don't know what people think is happening with our shipping, Like, our shipbuilding industry's not going as well as other people's. It didn't exactly work. Yeah, I mean, it worked to protect certain people's—

I'm just saying, if you look at American shipbuilding, it certainly has allowed us to be competitive on a global stage.

The shipbuilders would say we're early, but I take your point.

Early, but you've seen— but I'm just saying, like, how many decades of declines do you need? I think we're early.

It's still early. It's like the smiling curve, like churn. It goes down and then goes back up.

So, that's the bull case. That's the, all right, well, it's definitely not gonna help us near term. So, here's the most interesting thing, which is a pet issue of mine, because we're, like this administration, I should say, does a lot of good stuff on energy. I say that I, you know, I'm in a fortunate position where I get a lot of people ask for advice and stuff like that. And I feel like one of my jobs is to tell people things they're not doing. A lot of things they'll just do right on their own. But Canada is not one of those things right now. We are just absolutely sleeping on Canada as an opportunity. So Canada has no people and infinite resources and is really friendly.

Yeah.

So they have these oil sands, you know what these things are? No, explain. So they're just like these deposits of oil. It's like nature basically committed an oil spill is the way to think of it. So there's all these oil sands. For years they were not very useful.

They weren't viable because it's not just you drill drill down, black gold comes up. It's like buried in the sands, need to be recovered.

There are different kinds of things. You can mine it out directly, or you can do what's called in situ, which is you heat it up underground. Oh, and it liquefies. In any case, we— they have just this unbelievable oil seep. They have way more oil deposits than we do, way more oil reserves, but they have on the order of a third of the production, and in part because they have absolutely terrible policies, which are their fault., but in part we have had terrible policies, including getting rid of the Keystone XL pipeline and this kind of thing. So, but if you just think of Canada, it's just, we should be so, they have water. I mean, they get no people, 1/10th the population, infinite raw materials. They got everything. They got timber, they got natural gas, they got oil. Like, and we're talking about Venezuela, they're a friendly country, right? They got smart people there. They haven't been driven out of the country 'cause they're afraid of getting killed by Chavez or Maduro or whatever. So this is an opportunity to say, hey, Canada, let's work, let's have a task force. You guys are at a low in terms of, we don't have pipelines, but we can do rail. At least we can do rail, maybe trucks, but at least rail. Rail transports from Canada are really low right now. So let's have an initiative where maybe you bring a couple hundred thousand, depending on your expectations, maybe we can add a couple hundred thousand barrels a day. So that's one that excites me, 'cause I wanna be, Like the US-Canada superpower thing is this huge opportunity that we're sleeping on because of the idea of, oh, we should only be doing it in quote, America. But this is just, by the way, unbelievable. I think they have the best uranium in the world.

Yeah.

It's maple syrups up there.

Yeah.

I don't have a Canadian personality, but I love that place. No, strategically that makes a ton of sense.

And they're very friendly. Yes. They're very friendly.

Maybe we should be friendlier. Yes. You know, we should, and we have trade agreements coming up. So there's that. The other thing is we have some, you know, we can increase capacity here, as I mentioned, depending on prices going up. So there's that. But that might be a couple hundred thousand barrels a day. But so high level, open the Strait of Hormuz. That's the only thing that really works. Then you can talk about Saudi, Mideast spare capacity, our emergency capacity, and IEA emergency capacity. And then below that, you have Canada.

Jones Act and US production. That's great. That's a really, really thorough list. Thank you.

I appreciate that. Anything else, Jordy? What are you going to be paying attention to most closely?

Do you watch the price chart?

Is that important to you? Well, it's important. I mean, I'm— it's weird. I work with— I run like a kind of a network of energy executives, but I'm focused on the policy side of things. Sure. So I'm often laughably out of touch. I mean, I know in general within $10 usually what it is. But as a policy, I don't invest in energy.

No, because you're looking at longer-term, 5-year—

But also, like, I'm deliberately avoiding exposure to the commercial part of it.

Of course, of course.

You want to be independent. Yeah, I want to be independent. That makes sense. Yeah, I'm not— but at the moment, yes, I am looking at the prices. But I don't do predictions, but I'm just saying, I think there's a service right now in just laying out the options for people, because even in the government, I mean, I mean, not everyone has a lot of expertise in oil. And this is the kind of thing where, yeah, it's kind of why I think you texted me this morning. I mean, sometimes you need to know about oil today, and sometimes you don't. But when you need to know, you really need to know.

Yeah. Well, it seems like whether you're in the business community or just humanitarian, everyone should be rooting for a swift and peaceful resolution.

Yeah.

Do you have any insight? This might be totally out of your wheelhouse, but insight into how people running businesses that are most impacted by fluctuations in oil prices, like, are they— were they— have they been actively hedging over the last couple weeks? Are they doing— like, how do you run your business when your input costs can go up dramatically, potentially at a much higher rate than you can adjust your own prices?

I mean, just in general, these guys are sophisticated, I think, within the bounds of doing it. I don't think you can hedge indefinitely for this big a price kind of differential, like somebody is going to say. I mean, but it's also— you think about there's on the consumer side, but on the production side is even more dramatic. If you just look at the margins of these guys, I mean, the oil industry is just fascinating. I mean, I just remember anecdotally when I came out with my first major book, The Moral Case for Fossil Fuels, in 2014. Even I could see speaking engagements correlated to these prices, because you had a big price crash around them.

And then you think, wow, like—

We don't need Alex. Yeah, it might be nice to keep the troops motivated, but, you know, they'd probably rather be employed. But it's a hard business, and one of these things is you just— the people who survive have a very strong constitution, and they can manage the risk very, very well. I was talking with a bunch of these guys, and actually, some people we know in common, and like, some of the better ones, some of the two of the young superstars in the industry, they were just saying, like, we love the chaos, because we can just— no, we can handle it, like, when the prices go down. Because that's— you think about what they've been dealing with over the last year. They've increased production, prices have gone down. And you look at this, this administration is very focused, I would say too much focused on having prices go low. This is one of the variables here is, although the administration, I don't think was fully, was not fully prepared for this situation. In general, they are maniacal about prices going low. Like Trump in his mind has a very strong focus on $50 a barrel, which I think he feels like that's always the perfect price for oil. And the people in the oil industry say, wait a second, like we've had inflation of everything else. Why does our one product have to stay at $50? And that's part of the appeal of Venezuela, rightly or wrongly, 'cause it's not gonna make a big difference soon, but it's like, how do we keep oil prices low? But that's hard to be in that industry as a consumer, as a producer rather, and it's hard as a consumer. So they have financial instruments, but the main thing is you just, You need to just have the flexibility in your business, in your business model. And it's, it's oil is again, it's the most valuable material in the world of energy, which is why everyone uses it, but is also why it's so inelastic. And you have these price fluctuations. So you just need to come to terms with that or somebody needs to figure out something better. But it's, it's really hard. And of course, policy-wise, we can have better policy. To make it more stable, but it's still that thing where it's so valuable that guess what? You get a little extra demand, a little less supply, price goes up a lot. It's different than other things which you can substitute for much more rapidly.

You're out of a job when we build the Dyson Sphere.

Oh man, I'm working on this AI thought leader thing we talked about last time.

So I got plenty to do.

We'll cross that bridge when we come to it. I got plenty to do with my time.

Thank you. As long as we need energy, you'll be doing just fine.

All right, guys, thank you so much for coming on the show. Let me 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. And let me also tell you about AppLovin. Profitable advertising made easy with Axon.ai. Get access to over 1 billion daily active users and grow your business today. And without further ado, we have our next in-person guest live in the Ultradome. Michelle Vohls, how are you doing? Good to see you. Thank you so much for coming down to the TVPN Ultradome. For those who aren't familiar with your venture capital career and your marathon running career, introduce yourself.

Yeah, so, uh, I am Michelle. I was previously at Andreessen Horowitz. I just launched my own fund, Pax, which is a $50 million early-stage venture fund focused on foundational categories. Categories.

So, I was on the American Dynamism team, Andreessen Horowitz, similar themes to that, just earlier stage.

Biggest lessons from— Can we hit the gong first?

Oh, yes.

We gotta hit the gong. I think Michelle should do it herself.

You wanna hurt anything? Okay, here you go.

Hit the gong. You're here in person.

Go for it.

The mallet's already sort of broken.

You won't break it.

Smash that.

Good.

It's hard.

He's in kind of tennis form. That was good. That was good. Anyway, let's start with like lessons from Andresen Horowitz, like what did you learn? What are your biggest takeaways? What are— what's the playbook that you think you will be continuing? And then we can go into what you think might be different.

Yeah, I mean, I think in the categories that I look at, it is common for founders to need to raise a lot of capital. And the ability to be able to fundraise for that capital ahead of maybe what are traditional milestones in software investing is very important. You need to be able to be a magnet for capital, for talent, and sell a vision, because what you're doing is very hard with fewer obvious proof points, maybe, at the early days.

And I think identifying that in founders early is something that Andreessen Horowitz is very good at. Yeah. In American dynamism companies, hard tech companies, it does feel like there's high CapEx requirements, high human capital requirements. You're not going to automate everyone on your supply chain on day one. And so there is this, like, how you talk about growth, how you talk about scaling and success and finding product market fit, it almost feels back to like the, OK, we're scaling DAU, and let's focus on that metric instead of profit in software. What does it look like in hard tech? Is that the, okay, we have a program of record or we have these SBIRs coming in or we have this long contract that is very solid, but we're not going to draw down on it for a few years, so we have a capital need. But how do you think founders in hard tech should be talking about what progress looks like at top line or whatever top line means?

Yeah, I mean, I think that's the hardest part. Like in software, there's such a structured set of metrics that everybody has been trained around, like growth rate, like weekly active users, revenue. It is so different company by company in these sectors, and that's why the storytelling is important, because if you're talking to later-stage investors, they might be trying to make a pattern recognition off of a pattern that doesn't quite exist yet. And so in defense companies, like, yes, there are government contracts you should be trying to get, you should be working towards programs of record, but there are like different intermediate proof points before that. I usually explain it to founders as like the 3 things you're trying to make progress on are people, product, and then like revenue or traction. And so it's like, are you hiring the best people and the people that are experts in the different categories that you need? So maybe you need a mechanical engineering hire, maybe you need a government sales hire, maybe you need software as well. Are you Are you making progress on people? Are you making progress on product that's tangible? Can you show that there's developments? Maybe that's you've built something and it's gotten to a test milestone. Maybe it's in the hands of users. Are you making product progress? And then are you getting some sort of signs from customers that they are willing to pay for this? Are you getting contracts? Are you getting LOIs? Are you getting something that shows that there's a commercial use case for this? I think sometimes in like hard tech, there's a million names, hard tech, deep tech, frontier tech, American dynamism. I think sometimes people skew more towards thinking it's just deep tech, which is like, I would describe— Science projects. Exactly, like science risk, like nuclear fusion is probably like the most extreme example. That's hard, like it's like, of course there's a customer if you can do this, but like, we'll see what the timeline is, hopefully soon. But I think in a lot of the categories, categories I look at, there are commercial proof points you can make. There's usually a latent demand or an immediate demand. It's like, how do you show that you can get those customers and they can signal that they will buy from you?

Yeah. There was this viral Citrini article that said software is dead, sell everything. How are LPs thinking about hard tech as a category with that backdrop? Of like the SaaS-pocalypse. I imagine it makes things easier for you, but has it actually rippled through the LP community to the degree that your investors are more excited than ever to have a hedge against chaos in automatable software, all of that?

Yes. I will say this with love to all of my LPs. I think founders are typically like leading indicators. Certain categories. Like, they, they are at the cutting edge. Like, they are seeing around corners. Often, like, great founders will see something before VCs do. Like, VCs are lagging indicators. Like, once something is hypey, it's like usually too late to find a good early stage investment. LPs are lagging lagging indicators. That once something is very mainstream and there starts to be some signs of returns, then they're like, we should pay attention to this category. And so I would say I think that was more surprising, partly because I live in my own echo chamber and constantly hear about all of these great companies building in defense tech, hard tech, and industrial tech. LPs weren't as familiar as I expected. So there was a level of education. I think there's still some wariness of, there haven't been a ton of exits yet. I mean, I think that might change if SpaceX IPOs. It'll be one of the biggest. IPOs ever in history. But I would say like the appetite for it grew over time as I was fundraising.

On SpaceX, yeah, I mean, great point. It could be like $1.75 trillion is the number that's being floated. It's a staggering amount. It's gonna return like not just like one venture capital fund. It's gonna return like 20 venture capital funds. Like everyone who invests invested at early stage is gonna see their fund return from that. But all the LPs had to go on a 20-year journey with like a million continuation vehicles. How are you thinking about timeline messaging to LPs? Because 10 to 12 years is what I've heard recently is kind of standard, but I've talked to some VCs that say, yeah, like most of my LPs want their money back in like 6 to 7 years. And yet in hard tech, it feels like it might be longer. Are you communicating a specific a particular timeline, or is that even a relevant conversation?

Yeah, I think every fund roughly pitches 10 to 12 years. That's sort of the expectation. I think it's actually an interesting catch-22, because if you had ownership in SpaceX, your LPs actually were like, hold on to that. Yeah, totally.

Keep that.

I think it's going to keep going up. But maybe you would have an option to take some off the table and return some money money to LPs if that's what they wanted. But often, if it's like a really, really good deal, LPs are like, oh, I think it's going to keep going up. But I think it's like case by case. I actually just was talking to LPs about this. And they were like, sometimes it makes sense, and sometimes like hold on. And I think if you build a lot of trust with your LPs, like they're usually along the ride with you.

What GOs are you spending the most time right now? Obviously, it feels like we have so— so many founders that are just like in our kind of neck of the woods here, but are you spending a bunch of time in Texas, Bay Area?

Where do you expect to deploy most of your dollars? Yeah, I think like there's 5 geos. It's San Francisco, LA, Austin, New York, and DC. I'm actually seeing like a significant amount of companies in DC. Austin, I would say, though, is the biggest surprise for me. I know people have been trying to make Austin Austin happen for a while, and I was skeptical. And now I think it's sort of been an explosion of great companies.

It was kind of the same thing with LA. There was this long LA movement a while ago. Snapchat was sort of like the first big public tech company, or one of them, to come out of LA. And then the El Segundo thing just happened completely organically without any of those people that were cheering for LA. And it just sort of popped up, and it seems like Austin's going through a similar thing. What's unique about the companies that are building in DC? Are they thinking more demand generation first, more work on Capitol Hill first, and then maybe they'll set up a satellite manufacturing office or work with like a contract manufacturer to deliver their capability?

Yeah, I think like you're seeing founders who are working at companies like Palantir, for example, that have a big office in DC. They've maybe relocated to DC. DC, they're close to their customers, and then they spin out, start a company. Usually these companies will end up opening an office in New York as well. It's just a deeper pool for engineers. But there's a surprising amount of really strong engineering talent in DC that is like pretty fixed, like they wanna stay in DC, they have a life, they have a house, they have a family. So it, it can be a very good place to start a company.

What's the view on shipbuilding? Jordy says It's impossible to build a ship in this country. You're saying it can't be done.

Is it possible? We need to figure it out. Otherwise, we're gonna be behind. I think, I mean, like, the optimistic take would be like, we can make it possible. We should make it possible. The pessimistic take is if we don't make any changes into the regulatory environment or the cost of building these things, it might not be, and that would be a shame.

Yeah, that makes sense. Strategy with How much do you like to fund? Yeah. Are these like $2 million lead, pre-seed type checks?

Yeah. What's the deal? Yeah, on average, $1 to $2 million checks, like early stage, try to be in like the pre-seed, seed, occasionally a little bit later rounds. Like to partner really closely with founders. And I think the special thing about the early stage, which is just my favorite stage of company, is you get to be very aligned. It's like a significant check for me if I put, you know, $1 to $2 million into your company. And I am very motivated to help you be successful to get to the next round. And then for that next round, I can be much more strategic in helping you get there because I can't lead your next round. Whereas I think sometimes at the multi-stage funds, it can be a little bit tricky to give advice for future fundraisers because you might be able to participate participate. And so you have to be more delicate about it at a big fund.

And as a small fund, you just get to be totally aligned with the founder. One last question about sort of the category targeting. American Dynamism made a lot of sense as a theme around defense tech and hard tech. And you put a manufacturing company in there. One of the interesting subcategories of AD that always popped out to me was something around education, like reeducating workforce education, also just like education technology broadly. Do you think that that's within your purview, or is that something that you're sort of carving out and focusing more on the deep tech, hard tech, physical products?

Yeah, I would say I like to look at things that sit at the infrastructure layer of society, which could include education and housing, sometimes even healthcare. Often things at the intersection of hardware and software or tech and government or just like doing something fundamentally important for like the broad majority of Americans. I think American dynamism is this beautiful phrase that can be an umbrella phrase as well. I sort of embrace that as well with PAX, but at the early stage, like what you're really betting on is people. Like there is so little to understand except like the team and a dream. Like are you building something something in a category that's very exciting, or you're taking the big swing, do you sort of have a right to win in that category and some sort of unfair advantage? I like founders that have, like, worked in the industries that they are building in before.

And do I believe you can be that magnet for capital and talent and everything? Yeah, where are the big talent pools these days? There's the SpaceX mafia, the Palantir mafia. It feels like there's an Anduril mafia now.

Where else is that?

Probably gonna end up a BasePower mafia.

BasePower maybe in Austin.

In Austin eventually. Yeah, I think Base. I think Saronic is starting to see a few things. Basically any company that's reaching escape velocity, or like escape velocity and product market fit, has grown really quickly, has hired exceptional people. There are some people that really love that 0 to 1 stage, and the company gets big. I think we're seeing it at Anduril. We saw it at Palantir. We continue to see it at Palantir. Where people learn how to grow quickly and then want to apply that to their own company.

Amazing.

Well, thank you so much for coming by. Congratulations. Thank you.

I'm sure we'll have a lot of your companies on. Let me tell you about Vanta, automate compliance and security. Vanta is the leading AI trust management platform. And let's go over to some hard tech that's happening with Stargate because—

By the way, John, Apparently Trump said something to the effect of, I think the Iran war is very complete, pretty much. Oh, okay. Well, so we may be on the verge of a new era.

Yeah, crude is back down to like pre—

like Friday, pre-weekend. What does that mean? $80, $90? $83. Okay. $83. Okay. Well, maybe he was watching. Maybe he was watching Alex Epstein and Alex said, hey, We got to wrap this up. We got to stop. The price is too, too darn high, says Alex Epstein. Oracle and OpenAI allegedly dropped their Stargate expansion lease. Meta swooped in the same day. Ben Pulledine says that's not weakness, that's a hot real estate market. And then he goes on to comment that the real story is OpenAI passed on 6 additional buildings because power won't be ready for a year. The power is, is, is the gating factor here, not the chips. Interestingly, we are in the power bottleneck, the energy bottleneck. And by then, Rubin replaces Blackwell. So why are you going to build a Blackwell data center when you could build a Rubin data center? So why expand with last-gen chips? The 8-building, 400,000 GPU base build continues. This is smart capital allocation, he says. And so this is based on reporting from The Information about Oracle and OpenAI sort of adjusting the plan for Stargate. I am particularly interested in digging into the Stargate of China. They have a project that's the Chinese Stargate funded to the tune of $36 billion. Not, not up at half a trillion, but still a very big number. It's a big number and it's growing. And I've been, I've been digging into it a little bit, trying to learn as much as I can.

We'll do a deep dive probably with someone at Semi Analysis who knows a lot more about it. ChatGPT is back at the top. No way.

The App Store leaderboard.

You know, we do have to—

Adam GPT says GPT 5.4. It's for gods. You know what that meme is from? Big Chungus.

It's a good one.

Anyway.

Yeah. So I'm kind of surprised by this for the one reason being, I expected Anthropic's been spending a ton of money on marketing.

They started ramping it up. Yeah, yeah, yeah. I see that all over the place.

Around the time of the Super Bowl. I expected them to basically be willing to spend whatever it took to stay on top.

Yeah, but I mean, it's capital war, so both sides are going to spend a lot a lot of money. Yeah.

And, you know, uh, there was— what's up? Wasn't the rise in the Claude, like, ranking just because of the, the DOW news?

And now that's, like, not on the full page. So, so, yeah, that's what I was getting. I was like, when the Super Bowl ad dropped and we had some harsh words for the, the nature of the characterization around ads, our favorite thing in Claude, led us to launching Claude with ads. Uh, but our question was, will this Super Bowl ad move the needle for the consumer app. Like a Super Bowl ad is a massive campaign. Millions, tens of millions of people will see this. It went viral. Like it's, will it work on people? Will people see this ad that's like sort of deep in the weeds around how ads work in LLMs? Like, will this be enough for people to go and hit the download button? There's a hurdle there.

Will they do it?

And they did it a little bit. I think Claude went from like, like 26th to 18th or It started climbing then, and then you're right, the Department of War narrative really, really shot up. But as with any story in tech, it has to go back and forth and back and forth. And the memo that came out on last Wednesday probably sort of reset the narrative there and changed how people were thinking about that. And also these charts are based on momentum, and so it's very, very hard to state a lot.

Yeah, Claude had gotten to something like a million downloads a day according to Mike Krieger. Oh yeah. And at the same time, ChatGPT last month I think did like 55 million downloads.

So it's still not outpacing. Sure, sure. Paramount+ is number 4 in the free apps, although how can it be free when you have to subscribe?

Sort of.

Yeah, that's the whole thing.

But bullish for the Ellison team. Chris says UMF's ruined agency so bad that I'm I'm going to have to start using, saying gumption.

Gumption is really good. I like gumption. We hire for gumption.

AI gumption. AI agents.

We only hire for gumption here. What's the noun version of gumption?

I don't know. Friend of the show, Amjad, said not having coding experience is becoming an advantage. And Shuria says not knowing how to cook is becoming an advantage.

CEO of McDonald's. I love it. Amjad has some great takes though. This is just an easy riff. It's all in good fun. What else is going on? Lots of debate over the singularity pricing, as Will DePugh put it, because Barely AI, which I believe is Trung Phan, says, "Cursors internal analysis shows how hard Anthropic is subsidizing Claude code." Last year, a $200 monthly subscription could use $2,000 in compute. Now the same $200 a monthly plan can consume $5,000 in compute. It does feel like there's a capital war. I'm very interested to dig into the history of capital wars in winner-take-all markets, but also in oligopoly markets.

We will see— You should write the book on capital wars.

Capital wars are fast. I lived through it during the Uber and Lyft era.

There was a flippening for a little bit where Lyft was at the top of the charts.

The friendly rideshare company. It was very friendly. The cars would pull up with a pink mustache on them. And it had a very, very fun, whimsical branding.

And you would sit in the front seat of the car because you were a friend driver. I can see Claude having a pink mustache.

This is a real thing. What are you supposed to do? So when Lyft launched, people see Lyft and Uber as commodities at the time, but Lyft was revolutionary, and people were super blackballed on Uber, because Uber at the time was only black cars. So they didn't allow anyone to just hop on the network. It was not a three-sided marketplace or whatever, two-sided marketplace. You had to have a livery to TCV number, I forget what it's called, but like you had to be registered as a limo driver and you had an Escalade and normally someone would call you and say, "Hey, there's a wedding and we want you to come shuttle people around at this wedding," and they would pay you for that. And Uber went to those folks and they said, "Hey, you are registered to legally drive this car as a professional driver. We want to book you through our app." And so the black black car became Uber's main product, their only product. And then Lyft figured out through some interesting strategy how to allow people to just show up with a Prius or a Toyota Camry or Corolla or whatever. And so drivers on launch day, I remember like the launch week in San Francisco, would just show up in just some random car. But Lyft gave them all pink mustaches, huge pink mustaches. The team can pull up a picture of the huge pink mustaches on the front of the Lyft cars that were very whimsical. And it was remarkably cheap because Uber had been somewhat subsidizing the black cars. They hadn't been taking a huge margin. They were losing money. But the Lyft cars were way cheaper because someone was just showing up with like their Toyota Corolla or whatever with their pink mustache on it. And so you would hop in the front seat because this driver was your friend. There you go. The pink mustache on top of the— what is that, a CRV or something? And so you would hop in the front seat. They would ask you to pound the driver. Give him the knuckles. Give him knucks?

Give him knucks.

This is a thing. No way. I'm almost sure. So you'd be like, what's up? And then you would get in the front seat, and they would drive you. Of course, if you had a group, you'd be in the back seat.

But then—

Everyone had to give knucks, or just— I think the knucks might have been optional, but encouraged, recommended. 'Cause it was like, hey, we're buddies. And we're fighting the war on the Borg.

We're going up against the Borg.

But you were encouraged to get in the front seat. Front seat. Front seat in the Lyft. In the Lyft, when you're taking a Lyft, your friendly Lyft driver shows up, you sit in the front seat, you give him nugs. This is the thing. Pink mustache. Chungus. That's crazy.

Big Chungus sends his regards. All right.

You got to hire a car, but you have to be absolute boys with the driver. You do. But truth be told, a lot of the early Lyft drivers, were like super cool tech-forward San Francisco people. And so like you'd have a lot to talk about because like people were like, it wasn't professionalized at all yet. It was very much like two-sided marketplace. You'd make a lot of money as a driver. Like there was no margin compression yet. It was like free money everywhere for everyone. And it was a capital war and Lyft raised a ton of money and came into the market with a disruptively cheap product and wound up ended up delivering a really fun experience that was very, very cheap. And then Uber sort of fast followed and launched Uber— what's it called? The normal one, not black. UberX. UberX. Yes, UberX. And UberX was bring your own car. And then the rest of the game played out. And it became a capital fight. And for a while, like, the meta was, oh, you need a ride. You open Lyft and you check the price, and then you open Uber and you check the price. And people would check both prices constantly. And then over time, the liquidity and power loss stuff came into it. John, do you remember Uber Helicopter? Yes, that was a partnership with Blade. I think it was a partnership with Blade, which was the Uber for helicopters. And it was a thing, but for a few weeks or something. Did you ever do it?

No.

No. Because Uber's had a whole bunch bunch of different stunting products like Uber yachts or Uber— they did Uber ice cream where you could get ice cream delivered. That was like the precursor to Uber Eats. They did an ice cream delivery truck where every Uber driver checked in with the field office, got a cooler in the back with a bunch of dry ice and a bunch of ice cream, and you could press a button and for like $2, they would bring you an ice They even did, I think, Uber Kittens. And you could ask Uber to bring— Uber Dogs or something. You could like— Cats and dogs or something. I remember that. Yeah, it was very funny. Anyway, lots of stunts, lots of ways to get attention, go viral. And that was sort of the launch strategy. They would do that for whenever they launched new markets. They would come in and they'd be like, we need to go viral on the local news. We need to let the town know that we've arrived. And so lots of back and forth in the capital wars of 2010. 2023, 2024. It was fun times. I lived, I fought through them.

What else is going on?

I think that's a good place to call it. Well, thank you so much for listening. We will see you tomorrow at 11 AM sharp. Leave us 5 stars on Apple Podcasts and Spotify.

Sign up for our newsletter at tbpn.com and have a great rest of your week. It has been an honor and a privilege. It has. And I hope by the time the show goes live tomorrow, the war is over.

Me too. Me too.

That would be—

I would love that.

That'd be quite nice.

Thank you.

Thank you.

Goodbye.

Cheers.