Hacked - The Bad "Bet"

Episode Date: May 15, 2026

Or the very good bet, depending on how you think about it. In this episode, we start by discussing the strange, fast-moving world of prediction markets — platforms where you can bet real money on whether a head of state gets removed from power, whether a country gets invaded, whether the Fed raises rates — and where the prices themselves are supposed to be the point, a real-time probability signal generated by the crowd. Then we get to the case that just blew a very large hole in that theory: a Polymarket account called "Burdensome-Mix" that turned $33,000 into $409,881 in about a week, betting on the removal of Venezuelan President Nicolás Maduro — and the active-duty Green Beret who allegedly made those trades while participating in the raid that made them pay out. Learn more about your ad choices. Visit podcastchoices.com/adchoices

I've got a story for you here. Outlined in a court indictment and civil complaint, I'm going to lay it out for you. Hit me. Operation Absolute Resolve was the pre-dawn special operations raid on January 30 this year that seized Venezuelan president Nicholas Maduro. They grabbed him and his wife at a resident in Caracas under heavy fire, extracted him to the USS Ewojima, like 150 aircraft, Delta Force operators, helicopters, it's a whole big thing. They're on the ground for roughly two hours, seven service

members are wounded, Maduro's on a heliport in lower Manhattan two days later facing federal drug trafficking charges. This is not about those charges. It's barely about that at all, actually. What this is really about is a trade. Between December 27, 2025 and January 2nd, in 2026, which is to say like the 168 hours before the helicopters crossed into Venezuelan airspace, a user routed roughly $33,000 in a USD-pegged stable coin through a polymarket account under the handle burdensome mix and bought yes shares across four like very long-shot Venezuela-related event contracts. The market priced the odds of Maduro being removed at about seven cents on the dollar. We'll explain the mechanics of how all this works for anyone who is

unfamiliar, but basically right before Maduro gets got by all of these very fancy military operators, someone makes a really big, very specific bet that McDurro is about to get got by some very fancy military operators, like right away. And it happens. When that happens, that bet pays off huge. $409,000 U.S. dollars in profit, the bulk of it on this Maduro out by January 31st contract, which resolved yes, dollar a share at like seven o'clock in the morning on January 3rd, a few hours after the president announced the operation on Truth Social. The trouble with all of this has to do with who made that bet and their day job. Allegedly, sort of. Sort of. Master Sergeant Gannon Ken Van Dyke, 38 years old of Fayetteville, North Carolina.

The defendant on this indictment isn't a political prediction market hobbyist person. He is a green beret. Van Dyke has been an active duty soldier since September 2008, a master sergeant in the U.S. Army's elite special forces since 2023. He's on the same kind of unit that runs the kind of mission that just unfolded in Caracas. Specifically, on December 8th, the day prosecutors say he is in, in the process of planning absolute resolve. The day that started, he signed a Western Hemisphere Operations non-disclosure agreement

promising he would, quote, never divulge, publish, or reveal by writing words, conduct, or otherwise, any classified or sensitive information. 19 days later, this indictment alleges he starts buying those shares. Van Dyck wasn't allegedly just batting on the raid. It seems on the raid. At about 5 o'clock in the morning, January 3, 90 minutes after the assault element pulled Maduro out of his compound, a photograph gets taken of Van Dyke on what the indictment describes as the deck of a ship at sea at sunrise, wearing U.S. military fatigues carrying like the rifle standing with the other guys. That ship was the USS Iwo Jima. He uploaded that photo to a Google account from C, by which point the Maduro out position that he bet on at R.

already been resolved. This one's a dozy, Scott. I remember watching this as it happened, both the raid and the polymarket followed, and the arrest of master sergeant. Van Dyke, to me, this is the story where the rubber of our prediction markets insider trading hits the road of disclosing military secrets sure is illegal. Like, that's finally going to get resolved here. So we have the cases, a five-count federal indictment in the Southern District of New York.

a parallel civil suit going on. Van Dyke pleaded not guilty in an April 28th arraignment. He surrendered his passport and firearms. He's free on a 250K bond. I want to get into this story today because I think this is a fascinating, important story in the world of technology.

His defendant, Mark Garagos, outside the courthouse, said, quote, Mr. Van Dyke is an American hero, someone who is charged, unfortunately, was something that is not a crime. And therein lies the whole thing. A polymarket bet by a man who just allegedly finished doing the thing he bet would happen. Let's start this chatty chat episode with a very strange case of Master Sergeant Gannon Ken Van Dyke here on Hacked.

The scatting got called out in the last episode, so we're going to do it more. Did it? So we didn't like the scatting? I don't know if they didn't like it. They commented on it. And I felt seen. Ah. You felt seen.

I felt seen. I felt seen. We vary genre. We vary rhythm. We very rarely have melody. Today was a hard house beat next week. It'll be an acid jazz. After that, it'll be sort of a poca.

Poka, Poka. How you doing, Scott? We see you out there, Poka fans. I'm doing good, Jordan. How are you? Doing good. Keeping busy. I got the screaming cat here, temporarily not screaming.

I've got some weird tech stories in front of me. They are. The audience is used to your goblin. that's Jordan's cat's name by the way there's a little insider info my cat's name is Goblin and he earned it it's an accurate name we got some we got some interesting

stories for the folks yeah it's been a it's been an interesting couple it's been an interesting six months honestly it sure has there's a million things we could talk about but this is an interesting story to start with because it really hits a cross section of what's going on in the world of

you know prediction markets and the markets in general, politics, because you've now got political laws being drafted to prevent people from betting on these things. A bunch of the stuff going on in this. So I think this will be an interesting one, chat about it. It's an interesting technology story. It's, in a sense, a story of a person hacking something in the sense of using it as not intended. And in a much realer sense, that someone using a technology is exactly as designed.

So there's a lot of layers to this. and that whether you feel this was this technology being used as intended probably differs based on whether or not you work for polymarket. They would maybe disagree. Or Cali. Or Cali. Well, Cali comes up in this story and it's really interesting. Of course they do.

They do. They're where this didn't happen. Did you know, Jordan, that there are now hundreds of new prediction markets coming online. Yeah. That's really, really cool. Apparently, pretty much every online. online gambling and casino is opening up a prediction market as prediction markets are not seen as gambling. They can go into all the states that still outlaw gambling. It's interesting that it's the

gambling sites that are getting into prediction markets and not the market sites getting into prediction markets, though I'm sure they are too. And it gets to a real question. Is that even like media companies like are now launching their own prediction markets because they're way less regulated and way easier to open than, let's say, an actual online gambling site. Sure. So if you were, say, like, the score and, you know, you had allegedly had a large audience of sports fans that followed it, you could just whip up a prediction market to allow people to put skin in the game to interact with the outcome of sporting events, which is not

gambling on them. And there's such an interesting question. And we're going to dig into prediction markets. If you are part of this audience that doesn't really know what prediction markets are and you're sitting there thinking, please explain that at any point. Don't worry, we will. But there are these sort of like dual questions at play here of on one side, is it gambling? And as you swing away from yes on that, you get to the other side, which is like, well,

is it trading? Like, is it equities? Is it vulnerable? Is it, do things like insider trading laws apply to it? You know, you're not going to have an insider trading law for a roulette table. but you sure might for a stock. And where exactly on that spectrum does this technology sit? Because if it's just sort of perfectly balanced in the middle

and therefore ineligible for all legal protection, that's a really interesting place for us to be. Totally. Here, I'm going to, I think I'm going to, I'm going to jump in and take the start on this, if you don't mind. I'm going to say, I'm going to try and make a pro case for prediction markets because I think that they do have a really interesting, they have a really interesting place in the, ecosystem. So here we go. Here's my best chance is if you believe in open markets, which I do,

the wisdom of the group always exceeds the wisdom of one. So if you go to something like an old economics adage is like if you go to a country fair and there's a cow there and you have to guess the weight. If one person guesses the weight, the standard deviation of the air is massive. But as you get more and more guesses, the median of the guesses gets closer and closer and the standard deviations tighten up more and more. So the wisdom of the group exceeds the wisdom of one, which is what prediction markets are supposed to be. Hey, there's this potential outcome. Let's all essentially vote on it. But the difference is now the voting is done with money rather than just voting.

But in reality, what it is is querying the group for knowledge on a potential outcome, which does have validity and importance. Yes. Wisdom of the crowd, interesting. The utility of these things, as they are used to predict stuff, is like debatable, but seems like there's some. Yes. The question is whether it is a kind of gamble.

that reveals useful information and would therefore still be subject to laws governing gambling or is it a kind of commodities trading or trading at all that is therefore like subject to the laws governing that that also produces this useful information. I listened to an interview with Cal She's CEO once. Yeah. And they're from the finance world. Like they came from big investment banking. Makes sense.

Yeah. Like Jane Street, stuff like that. like hedge funds. And one of the things that they hated was that they couldn't hedge based on global events, election outcomes, you know, weather situations, any of those massive, you can't buy derivatives in the market currently that let you hedge against these outcomes. And that's where they kind of got the inspiration to create Kalshi apparently, was like, hey, wouldn't it be great if on election night you could diversify a portfolio by hedging

against the election outcome, which, like, as somebody who's interested in finance, makes a ton of sense. Yeah, if you're trying to financialize literally everything, a platform that lets you financialize literally everything would, of course, be compelling. Correct. Yeah. I understand the argument. I'm more, I'm actually going to insist we explain what these are. Yes.

So what is a prediction market? A prediction market is an online platform where you bet yes or no on real world outcomes. Elections, sporting events, whether the Fed interest rate decision, a geopolitical flashpoint like a president getting taken in the night, that kind of thing. You buy yes or no shares. At resolution, when the thing does or doesn't happen, the winning side pays out a dollar, like one U.S. dollar per share. the losing side pays nothing.

They had to pay to buy the share. Share prices float between zero and a dollar, which is sort of like the live probability that you were kind of alluding to you, right? Yeah. A seven cent share means the market collectively thinks there's about a 7% chance of the thing happening. Yeah, the easiest way to think about it is if, say there's $1,000 being bet on an outcome, let's say that outcome is like will Scott and Jordan record an episode today? And $930 is bet on yet.

and $70 is bet on no, then there's a 93% chance of a yes outcome. It's kind of the way that the market looks at it and the way that they set their odd lines. It's all supply demand. There's two big players. It's a polymarket. There are those sort of more...

I'm speaking generally here. Crypto-based runs on the blockchain offshore, uses stable coin kind of route. Worth a $9 billion, I think as of last year, worth more now. right. The other one is Kalshi, which is, it's U.S. licensed. It's regulated by the CFTC. It takes dollars. It requires full ID verification. I bring up this distinction because this distinction is very important to this story for sure. Yeah. Yeah. So Polymarkets international platform is

officially like it's off limits to Americans under a 22 settlement with regulators. Like that that whole thing is supposed to be blocked off by IP geolocation. As far as I understand, I'm not in the States, but that's my understanding of what's going on with Polly Market there. If you have a VPN, you have a VPN. This is a big industry. You've almost certainly heard stories about it, billions of dollars a week. Importantly, Kalshi is regulated by the CFTC, which is different from the SEC that governs

things like insider trading. fun little piece of trivia, though billions of dollars are moving through these markets, the CFTC has roughly one-eighth, the staff of the SEC, a not itself famously well-funded or staffed organization. So that seems important with all of this. Should we begin? Let's begin. With more lofty questions of gambling, or should we begin with the defendant and loop back

around to the lofty questions? I think we go story first. Story focus. We'll pull us in many directions, I'm sure. There you go. Okay, let's talk about Sergeant Gannon, Ken Van Dyke. 38 years old, Fayetteville, North Carolina. It's like a town next to Fort Bragg.

He's an active duty U.S. Army Special Forces. He's Green Beret since 2008. He was promoted to Master Sergeant in 2023. He's on the U.S. Army Special Operations Command. He's a senior non-commissioned officer. He's the kind of soldier that, like, other soldiers are taking their cues from. he'd signed a pile of non-disclosure agreements over the years, held a sensitive compartmented information clearance since 2018.

It signed the standard SCI non-disclosure agreement, promising to, quote, never divulge anything marked as, you know, special information. Top secret. Top secret. I have a list of, like, pieces of paper this guy is signed saying, like, top secret, totally understand. In December 8th, very, very relevantly here, he signed an NDA specific to Western Hemisphere operations,

like right before all this happened. I talked about that a little bit in the intro. An interesting detail from the indictment, because this is all PDFs. You can read through it. I've gone through a chunk of it. One month before the trades, Van Dyck uploaded a screenshot to his Google account of a Google AI search result, confirming that special operations files are, quote, not available to the public, and the quote, specific sources, methods, and full operational timelines are consistently withheld.

A sort of perfunctory Google search, it would seem as to what exactly did I sign a piece of paper saying I wouldn't disclose? I'm just doing a little background research into what I've agreed to. And then uploading the screen, that part I don't get, but hey, how you use the computer's up to you. The mission itself. Operation Absolute Resolve, January 3rd. The largest U.S. direct action rate in Latin America since Panama in 1989. Fun little trivia. 150 airplanes, 20 plus bases involved in the whole thing.

They really pointed a lot of energy at this. There was like a ground element of Delta forces going in like two, four o'clock in the morning kind of thing. Maduro and his wife, Celia Flores, gets seized under heavy fire and extracted to this ship, the USS Ewo Jima. And then Trump announces it at four o'clock in the morning on truth social. A very important moment in this very compressed timeline. 90 minutes later, this photo gets uploaded. I talked about this in the intro. It's like a sunrise photo taken of Van Dyke on the deck of a ship at sea.

He's kidded. He's got the fatigues and the right. There's the other dudes with him And the operator outfits He upload the whole gang's there The whole squads rolling deep He uploads it to his Google account And then shortly there after Maduro lands in downtown Manhattan

At this like heliport to face his charges So that all That all seemingly just happened The issue is to do with what happened Before that Which is the series of alleged traits You got 13 separate transactions between December 27th and January 2nd. These were importantly on Polly Market.

We'll talk about why they weren't on Kalshi in a minute. It basically was just a bunch of yes positions. We talked about the structure of these work, the yes, no bet. There's these four Venezuela-related contracts, mostly centered around the idea of Maduro out by January 31st. Then there's sort of like sub-genre bet. of like U.S. forces in Venezuela by January 31st. Trump invokes war powers against Venezuela by January 31st. Just a series of this is going to happen related bets or contracts.

If you knew, these would make great bets, great contracts. So this polymarket handle burdensome mix. I've got the wallet right here. I've got the wallet number. You can track it. Lord knows some journalists did. The account was created on December 26th, after he wired roughly 35,000 U.S. dollars from a personal bank account into a crypto exchange.

He uses a VPN whose exit node was set to geolocated in a foreign country. We actually know a lot about his tech pipeline from this entire minute. It's really, really interesting. It starts with a stake of $33,000. The bulk goes into that Maduro out contract. The market price, Maduro's removal is like a pretty big lawn shot. Most of the time he was buying, it was fluctuating, but it was under 13. sense. So pretty big upside. I always feel like this is one of those situations where maybe

you should have spread the money out a bit, made a few playbets on the side to bring a little bit less at less heat and attention to the fact that you just showed up on the platform, put a bunch of money in, and then put it all on red. Yeah, sure. And then red happens. Like if you, if you, maybe you put 20,000 on Maduro out and you put 5,000 on. GTA six comes out in 2026. Just some dumb shit. Oh, hurt.

Yeah. Just so you don't pop the flags as much than somebody that just comes in and is like, I'm only betting on one situation and I'm going all in on this outcome. Well, and seemingly, and we'll talk about this in a minute, that level of certainty is what flagged it off. It is not that he was the only person trading on this contract. He was actually a minuscule amount of the trading volume. It was, to your point, seemingly, the like, this one.

account just sort of burst into existence to make up $35,000 bet on a really specific thing happening. What's going on there? We'll get to that. But like that is very specifically, it seems why all of this popped off. Any technology company these days has some form of modeling on user behavior. Yes. And especially when it comes to a technology company that's in predictions, you know, has financial incentivization.

There's no way that they're not. auditing and scanning for people doing this exact situation. You show up, you've never had an account, you create an account, and a week later you put $100,000 on the situation. So there's that blitz you're describing, right? January 2nd, kind of starting at night, this like series of rapid fire trades. The share price is taking up a little bit during all this, but it's staying pretty low. Then, after that truth social post announcing all of this had happened at 421,

in the morning. The price rips. It goes from like 37s. It had already crept up to 37. It skyrockets to 95 cents in minutes. The certainty that this is going to happen skyrockets to 95% according to the logic of these platforms. Contract resolves at a dollar at 7 o'clock in the morning. And the final profit to these bets was $409,881. Roughly 404 of that was just on the Maduro out market alone. something fascinating about this because Van Dyck's situation does feel particularly agreed just given his role and to go back to what we were saying. Wall Street Journal reported that about $56 million in total trading volume across that Maduro out contract happened in this window of time. Van Dyck's Hall was just under was under like 1% of that. So once this starts, the volume explodes.

There was just this one user that got in weirdly early. in specific that flagged the whole thing. That's all on Polly Market. I want to talk about KALSH for a second. Three days before opening the Polymarket account on December 24th, Van Dyke allegedly applied to open an account on KALCI. KALSI is the one that runs, it's called KYC, know your customer ID verification.

It is an American legal concept you see it in other countries as well. Basically for every new account, like you got to prove who you are. I upload your driver's license. And you've done it before. I have. Not on Calci, but I have to do. Van Dyck obviously couldn't pass this, given what he was seemingly attempting to do.

He pings Calci customer support on December 26th, 27th and 20th trying to fix it, gets nowhere. Polymarkets International site, which is, which is, which is unbound by any kind of, yes,

I'm not going to tell you how to use it, but you figure it out on your own. Quick Google search. get you there. Yeah. It asks for an email address. Oh, and a wallet to send the money to. Of course, an anonymized crypto wallet.

Crypto's fine. Crypto is fine. That account was open the same day he was rejected by Kalshi. Just a little quote, I liked in all of this. Former SEC Gensler-era policy lead Amanda Fisher said to a journalist, it's curious that the CFTC and DOJ didn't charge polymarket for not appropriately prohibiting U.S. betters from. accessing the platform.

Anywho, the cover-up. January 3rd, hours after this contract resolves, cashed out of the account, $437,000, $815,000 in this USDC, it's a U.S. dollar-backed, like pegged, stable coin, gets sent into a foreign cryptocurrency vault,

just like yield-bearing offshore, store your crypto product. This is all publicly visible. And within hours, some journalism starts. NPR's Bobby Allen and political researcher Tyson Brody had identified, wow, this one account seems really, really weird. This trading behavior is very, very odd. And they start posting about it, the story goes viral.

January 6th, three days after this trade first service and press reports, Van Dyke asked Polly Market, hey, can you go ahead and erase that whole account, please? Can you delete it? I lost track of my registration email. I would like you to vanish this account that is currently going viral. Please. Immediately.

He changes his crypto exchange like registration email to a brand new address. They created a couple weeks earlier. January 15th, he pulls $44,000 in Fiat dollars out of the vault, wired it to a personal bank account at a Texas headquartered U.S. financial institution. and therein lies the critical error. I thought the error was going to be way more cryptic than that. I thought the photo posted from the deck of the ship was going to be on the same VPN. I thought there was going to be some breadcrumb that led them back,

not just like, oh, I straight up wired it to my bank account. Yeah, he was anonymous on the polymarket, but not on the chain. Every trade was permanently kind of publicly inscribed so that when he went to cash out through a U.S. regulated exchange, like that requires the ID verification that wallet now ties back to his legal name. Polymarket says that it then flagged him internally and referred to the case to the DOJ. So you're starting to see them attempting to be like, we know we let your fancy man on the boat bet on our platform,

but like we found him. We're going to play ball. We want to keep our business. We're sorry this happened. They have a disclaimer at the bottom of their web page that says you shouldn't. Don't do crimes here. Don't do crimes here. Please don't do crimes here.

Yeah. They literally have that disclaimer. Just don't do crimes. Pretty much. That's going to be their Super Bowl ad next year. Yeah. And if you would like to pay us to make it, we'll make it really, really good. Yeah, yeah.

I'd love to make that ad. Just please don't make, don't do crimes on our platform. Don't do crimes on it. But you score it like a very earnest like financial like one of those feel good bank ads. We're just like, what are your dreams? What are you trying to create? With Polly Market, just don't do crimes. Not here.

Just not here. But like, do you want a boat? Don't, hey, don't do crimes. Anyway, let's talk about the charges. Five counts in the Southern District of New York. So this indictment was unsealed April 23rd. That's why we're talking about it now. And we can dig into the legalese of what, you know,

specifically what are these charges outline? It's basically like, hey, you can't do this. It's unlawful use of confidential government information. That's worth 10 years, theft of non-public government information. He's charged with both wire fraud and commodities fraud, which are really, really interesting charges given what this is all about. The total maximum exposure is like 60 years. He has to forfeit the accounts.

There's parallel civil action. Quote from the CFTC chairman, Michael Selig, the defendant, quote, took action that endangered U.S. national security and put the lives of American service members in harm's way. He was arraigned on April 28th in the Manhattan Federal Court. He pleaded not guilty. He was released on a 250K bond. His next hearing is June 8th.

We'll actually be in New York on June 8th. I don't think I'll be in town on time to go to the court, but I briefly considered it. Kind of wrap up with the defense here. And really, to me, what the big question here is. So I give you that quote from his lead attorney, Mark Garagos, which is Mr. Van Dyke is an American hero, someone who's charged unfortunately with something that is not a crime. And that defense theory is basically. like, hey, this is clearly super unflattering.

But it isn't covered by the federal statutes prosecutors are charging it under. That seems to be his argument. It's like, I'm not denying he did the bad thing. I'm denying the bad thing is illegal. This is from what I could find, like the first criminal insider trading case ever brought against someone to do with prediction markets,

which is really, really relevant. It's the first time the CFDC is invoking a rule that I'd heard of, but I didn't know it was called this, called the Eddie Murphy rule. Oh, nice. It's in reference to the film trading places in which two commodities traders manipulate the orange juice futures market using a stolen government crop report. So it's this idea of like you enriched yourself on a marketplace using this privileged government information. But that central legal question is like, does federal insider trading law actually reach this conduct?

Like you've got a factual story that is seemingly in agreement. It's like 13 timestamp trades, a wallet address, a VPN, a botched cover. cover up. It doesn't seem like they're fighting my bank account. A withdrawal. They're not fighting that. They're going to fight whether all of that stuff that it seems like happened is a federal crime under any of the statutes that the prosecutors are applying to fit it. Well, you'd think that the master sergeant would actually be in more trouble than just insider trading because technically by him putting the bet down, it's a disclosure of confidential information. 100%. So, so he would,

then be up on charges for, you know, a number of other federal crimes related to, you know, whether you're doing right by the, by your country or not, which would be much more severe than just, you know, white color financial crime, which, you know, we see and read about every day by people at all levels of power. Yes. On that point, you just brought up, like, there's a really weird quirk of these platforms, which is that you aren't. Van Dyke promised to keep secrets for the U.S. Army. He didn't make any promise about to the people that he was betting against.

And that is like a weirdly meaningful legal concept here. It's like the law has never been tested on whether that counts. Like he didn't, the U.S. government didn't lose any money. He just screwed over the people he won against, which is sort of like the function of these marketplaces. That's all separate from the question of like is a poly market bet a swap? which would make it eligible to insider trading. And then there's the wire fraud thing,

which is like, was anyone actually robbed? Like wire fraud, I think typically requires a victim that lost money or property. And like the government, who he signed the thing with didn't lose the dollar. People on the other side of that lost their money. But they were betting. Well, were they betting? And I just made the face to say, are they? Or were they?

But so that, but that's exactly it. That's exactly it. It's all of these like, oh, well, technically's. And it's like, the question is, can he sort of, of just like narrow path wriggle his way through those. I, it seems if, if, if this one case manages to narrow path wiggle the way through it, I would hope that the legal system closes that narrow path quickly. But the thing for me, again, just to go back to the bigger crime here, like say you are,

you know, authoritarian dictator somewhere in the world, would your due diligence of safety in protection, not then just include constantly monitoring prediction markets? Like, if we know, like, say you're Vladimir Putin, who's been hiding for two weeks because he thinks he's about to be assassinated, why does he think that? Maybe he saw a spike in prediction markets, which was indicative of insiders betting that they were going to kill him to make back-end money on it. So it is, it is, in fact, a way to communicate motivation and planning. behind a specific outcome to the market, which is a disclosure of private and confidential

information, which therefore would be treason. Sure. So if you really wanted to, if you really wanted to get into it, this person is probably going to be in a lot more trouble than just, you tried to make a quick buck. Yeah. If the government wins, like this is a precedent and therefore the government has a really big incentive to win. Like every prediction market insider training prosecution after this, if they win is going to run through United States versus Van Dyke or whatever it's called.

Totally. And if they lose and the Congress decides to write a new statute trying to like make some part of this illegal, Van Dyke will still be remembered as like the case that exposed that huge gap. Either way, this is going to have a big impact on that. Well, and you want to talk about efficacy of the government to pass. of this type. Nancy Pelosi, who is retired, there's entire tabled acts called the Pelosi Act and things to stop Congress and politicians from maybe not insider trading, but you know, disclosure of confidential information to partners and husbands that insider trade. You know, there's a bunch of, this is all alleged. But, you know, Nancy Pelosi's net worth

went from hundreds of thousands of dollars to hundreds of millions of dollars during the course of her tenure in politics. And they have failed to stop or persecute anybody in that regard. Yeah, there's, I mean, Nancy Pelosi is a great face for it. There is a current sort of like golden era of insider trading and backhand dealing happening right now throughout, I wouldn't even just say the U.S. government, but certainly in the U.S. government. There's a lot of people that have figured out that if you have privileged information about the shit that's going to happen, you can make a buck. Totally.

This is just the first guy that did it by uploading a photo of himself in fatigues, mere moments after he launched an international event, the odds of which were so low. It drew attention to him. But like, this is happening from the stock market through polymarket. This is happening with people with privileged information. Well, just actually this year, Gildebrand and McCormick, two senators in the states, tabled an act called the prediction market act, especially banning anybody from Congress, president, executives, and branches, senators from being able to utilize these services at all.

Without knowing a ton about it, good idea. Exactly. Like, that's not an official take. I'm just learning about that specific thing. But I would say, broadly speaking, the more privileged the information you have, the less you should be allowed to act on it in a otherwise comparison. additive market. Totally. It seems pretty fair to me. Your mileage may vary. Super interesting story. Wanted to kick off the episode with it. Should we ad oasis? I think we should kick it over to the ad oasis. And when we're back, let's talk about birds. Let's talk about all kinds of birds.

Let's talk about birds. Do do do do do do do do. Do do. Think about the last time you heard a breach story on this show. It always starts the same way. Someone, somewhere, saw something too late, an alert buried, a signal missed, an SOC that just couldn't keep up. Arctic Wolf set out to solve that problem by rebuilding security operations from the ground up for a world where attackers are already using AI. They created the Aurora superintelligence platform, a fully agentic system powered by the swarm of experts. Instead of single-purpose bots or lucky-guess LLMs, this swarm is full of deterministic agents that handle whole entire workflows. Humans stay in the loop and on the loop to validate the critical decisions and keep everything trustworthy. And all of this is just off running on

their secure operations graph. A constantly updating intelligence engine fueled by more than nine trillion telemetry events every week and over a decade of real world incident response. The system reasons on real signals and real context not synthetic training data. And the result is the new Aurora agent SOC. It's the first SCC that is agent led by design. You get agent. that coordinate, agents that investigate, agents that respond at machine speed, and hundreds more that automate the repetitive work that normally buries human analysts. Arctic Wolf didn't try and bolt AI onto an old model. They rebuilt the model entirely.

What makes it even more effective is how it works with Arctic Wolf's concierge experience. The team brings customer-specific context directly into the platform so every AI-driven decision reflects your environment instead of generic assumptions. The automation frees your concierge security. team to focus on higher value strategy and proactive risk reductions while the agents handle the grind. If you want to see what trustworthy production ready AI and security operations actually looks like, go to arcticwolf.com slash hacked. Starting some new isn't just hard. It can be downright terrifying. Uh, you put a lot of work into a thing. You're not entirely sure it's going to work out.

You're taking a huge leap of faith. Uh, I've started a few things. Now I know I was right for believing and, you know, the idea of the product, despite all of those fears and hesitations. But boy, does it sure help when you have a partner like Shopify on your side? Shopify is the commerce platform behind millions of businesses around the world and 10% of all e-commerce in the U.S. from household names like, well, hacked podcasts, merch, to brands just getting started. You can get started with your own design studio with hundreds of ready-to-use templates. Shopify helps you build a beautiful online store that matches your brand style. Did I mention that that iconic purple shop pay button that's used by millions of businesses around the world?

I don't know why I wouldn't. I should. It's why Shopify has the best converting checkout on the planet. It also helps boost conversions, meaning less carts, sort of getting abandoned in the parking lot and more sales for you. It's time to turn those what ifs into sign up for your $1 per month trial at Shopify. com slash hacked. Go to Shopify.com slash hacked. One more time, that's Shopify.

dot com slash hacked i'm super excited to tell you guys about one of our new sponsors we got a new sponsor and they're called even realities and they make ai smart glasses which sounds crazy but they're actually real they showed up last week i got a friend that's had these they had the first version of these and he uses them for teleprompter when he does presentations i got these ones to do coding they have a full terminal integration into ai coding apps there's all also an entire ecosystem of people that have made custom plugins of which you can control AI assistance, sports scores, news, weather. There's a whole pile of uses for these things. Even G2 are productivity smart glasses designed to keep real-time support right in your view.

With teleprompting, conversation support, real-time translation, AI assistance, and more, they help you stay on top of the work and daily life. And unlike most smart glasses, they're designed to look and feel like premium I wear. There's no camera. They have a lightweight, 36 gram design that you can wear all day. The more context you give, the smarter they get. And they adapt to how you work and what you need. If you want to learn more about even G2s, go to even realities.com and see how everyday smart

classes keep helpful information in sight. So you can stay productive and hands-free throughout the day. And for our listeners, use promo code hacked at even realities.com to get 10% sent off the even ring one and or the even clip when you add them to your even G2 order. That's even realities.com promo code hat. Let's talk about birds. Do we want to talk about like AI birds? Do we want to talk about like old school birds?

This is a pallet cleanser. We just talked about a big techie crime story. I think either last episode or maybe the episode before, we talked very briefly about the technology underpinning crosswalks. We did. And people really seem to like that. People seem to really enjoy the weird tangent on not relevant but interesting tech stuff. The history of the crosswalk noises. And I went down a really interesting side road.

This isn't going anywhere. Just so you know. It had to do with birds. There was like birds flying around on the side of my window. And I was noticing there's a thing that birds do called a murmuration. You know this term? Nope. Yeah.

It's like, tell me about it. You're like, no. It's whenever you see a flock of birds and they seem almost like a single mass and it's kind of like moving through the air if you Google a bird murmuration,

it's like a big flock of birds flying in sort of density. And there was this question for a long time of like, man, how do those birds not just be smashing into each other all the time? Why aren't those birds constantly smashing into each other? That's such a tight tolerance for it to remain coherent. is an interesting question. The thing I want to talk about here is an artificial life program from 1986 developed by a guy named Craig Reynolds used to simulate the flocking behavior of birds.

He called it Boyd and Boyd's algorithm. In reference to a shortened term for this idea came up with of a bird-oid object. So like a bird-like object trying to predict the behavior of that, bird-oid got condensed down to Boyd, which also just sounds like the way the word bird would be said by a New Yorker. It's a Boyd. Boyd. Literally. This concept, this algorithm has been used in like film, video games, simulations,

science for years since. And it figured out that this like, the way they do that, this emergent behavior of birds is rooted in like a set of three rules. Basically, if every single bird observes these three rules, you get these big kind of murmurations of birds

and birds that never crash. Can they separate? Steer to avoid, like smashing into each other, do they align towards the average heading of where the whole group is going? And are they coherent, always moving towards the center mass of the whole thing? If you just have the first two, they'll not smash into each other and go in the same direction, but they'll eventually break apart. But if you add the third one in of cohesion, you get this like emergent behavior where birds can fly in these beautiful like single almost objects.

this was reproduced by this guy Craig Reynolds and he called it Boyd's algorithm and we've now just sort of known how birds work ever since and I thought that was cool and interesting and good I just want to tell you about that Scott no it's it's for a digression it is an interesting one I learned something that that is that is cool I've never really thought about it maybe that's a failing of my behalf now I grew growing up in the in the in the burles and watching flocks of birds take off all the time. I never ever sat back and said, wow, why aren't they smashing into each other? But now, next time I see one of these, I will be enthralled by it. Also, what an interesting field of study. Imagine your life's work.

Imagine you were the boy. Right. Like you had created. Kind of awesome. Yeah, kind of awesome. Like you cracked how birds worked for the first time. Like he did other stuff. And then his algorithms were like used in films and stuff and he worked on those but like yeah just like off the side of his desk cracked how birds i'm a bird simulation expert pretty cool pretty cool pretty interesting a real dead end narratively yeah yeah hey hey birds we're we're trying about all birds a i so oh yeah we did yeah there would have been a good transition there sure i wonder how they're doing in their their new ai service provider thing, life.

I wonder what happened to all the people that liked their shoes. For anyone that's unfamiliar, Allbirds the shoe company decided to do the strangest pivot I've ever seen and become an AI infrastructure company. As you do. Really proving as you do. If you simply couldn't have given less of a shit about your shoe business. Well, the shoe business had failed. So like they needed that.

They sold it. So it was worth some. Well, it had failed in the sense that it was essentially dead. So they needed to recombin. back to life as something else. So why not come back to life as an AI service company with tons of funding? Raise a bunch more money. You want to build some server farms rock on guys. Yeah, you're just going to be on the customer list for AMD and Nvidia waiting for chips like everybody else is. I don't know if you've been following the, let's talk. We talk economics,

finance corner for a second. Hey, lay on me. Finance corner. Man, man. Sound effect I haven't made. Hit me. Yeah, literally. I'm not sure if you've been keeping up. Yeah, literally. I'm not sure have you been keeping up Intel insane. Like for a company that was completely market uncompetitive like two years ago, had lost most of its, was losing in the progress of losing most of its server and personal computer chip market. Share had hit an all time low. You know, I think even as early as last summer,

we could have bought it for like $19 a share. And I don't know where it's at today, but it's probably around 140. I feel like the story about Intel has been just that they're bleeding their biggest customers. The famous one is Apple made the switch to silicon four or five years ago. I think there was a rumor they were going back to them for something. What do you attribute to the 19 to 140 switch route? It is a, so it's $127 as of recording.

483% increase in the past year, 5X. Pretty good, pretty good. Did Van Dyke make a bad? bet on it? Like what happened here? Well, pretty much. It's very similar in the sense that the United States realized that microchips were the future. When COVID happened and the chip pipelines dried and all of a sudden Ford was not rolling trucks off the production lines, you know, all of a sudden you realize that microchips that we buy from third-party countries, notably China, Taiwan.

This is rooted in the American investment. Yeah, yeah, yeah, totally. are completely required for every single piece of modern consumer product and industrial product, commercial product, government product. The government policy shifted to bringing fabrication of chips back to America. And Intel is the American chip company. So the government jumped in and it changed CEOs. And they've been killing it ever since. So when you say killing it in what way? Like killing it in the sense that there's money flooding into the stock or killing it in the sense that the company is performing better in a material way. I think both. They're producing new product that people actually want to use for stuff.

They're expanding their manufacturing capacity. The share price is way up. There's a bunch of things going on in that company. Investments and joint partnerships with other manufacturers. So I think it's all in all, I think it's doing. They've been doing very well in the market and in the industrial side of things. You know, I think with the government shifting policy, identifying it as a major risk and pushing forward, not just monetarily, but also policy side's been a big way for them.

Outside of the question of, you know, the American government investing in an American chip company, when you refer to them making products that people are excited, like, what is your sense of what Intel is up to on either the consumer, even the B2B side? because I have accumulated all of the plaque of the story that they're just sort of like, we ran out of ideas. Intel is no longer inside of anything. And we don't know what to do. And that's just sort of where the story is left for me.

I'm an Apple user, so I'm just not exposed to it. Where do you see them? Like, what kind of things are they doing that people are like, ooh, Intel, sick shit. They, I think they've been putting a lot of drive behind their GPU side. developing, they're not nearly a supported, of course. Smart. Yeah, good fair point. And by that means, like, they're no longer in the sense winning the market.

They don't have a big chunk of the market share, but they're at least structuring and positioning themselves to try and become a competitive player in a market that the supply is nowhere near the demand levels. Like, I don't know if you've tried to buy a consumer GPU recently, but like a premium from Nvidia GPU is, like, if you go out and try and buy a 50-90, you're paying like $5,000 for it, which is, you know, would have been the cost of the entire gaming PC three years ago. And now that's just for one part of it. Yeah.

Thus driving the cost of pretty much every piece of consumer electronics up, sort of down the supply chain. Like not obviously everything has an Nvidia graphics card in it, but somehow a Nintendo Switch two costs $50 more and no one can get Steam. boxes or machines or whatever. Like it's having all these weird knock on effects throughout the rest of the industry. Yeah. The, uh,

some of the other big things, like their new Zeon chips, like their server chips have really focused on performance per watts or bringing down the power consumption, bringing up the performance. Like stuff that A&D and video and Apple specifically have done very well. So they're just kind of, I would say they're retrofitting,

which is I think a good, a good, good way for them to go. Yeah. Catching up. catching up retrofay. And I say that not even smarmily. I'm like, yeah, you kind of need it to catch up and maybe you were undervalued.

Yeah. And can you? Because then if the answer is yes, then I kind of get it. That's really, really fast. They've always made, I'd say, performative chips. Like, their chipsets have always been pretty good. Yeah. Except for, like, they were just bottleneck by their fabrication cap capacities and, like,

how many nanometers they could get their chips down to. So even as early as, like, a few years ago, you know, they were still 12 nanometer chips when they're competing with AMD at like a six nanometer, nanometer, I think nanometer, nanometer, but I think we're even down to like four now. Yeah, we're down to almost two, I think. Yeah. So there's been a big pushes in just productivity and gains. And that's the big thing.

Like as you see the data center world kind of blowing up, you know, power, heat, massive, massive parts of, how do we get the most out of these things without it costing and environmentally taxing the most. Yeah, I think that I'd be really curious. I feel like I haven't heard this as much this argument, but I'm waiting for someone to start really making the longevity argument. Like, there's a lot of interesting stuff about the scale of the cost of the like data center buildout that we're living through right now, like just unfathomable amounts of money going into data center production. And you hold that against the cost of other infrastructure projects. highways, bridges, really physical, waterways, pipes, like the physical stuff.

And that's all well and good, if you believe in where the technology is going. But what's not great is that those things typically last 25 to 30 to 40 years. And a server farm is like, oh, that's a lot less. So I'm really fascinated to see which companies start coming out being like, let's just talk about these will last a decade. Like that as being the primary sales pitch for these things. You should use ours because it's more like a road. It's going to be here in 15 years or something. I think that that's a really interesting case that won't start feeling urgent until the wildly expensive server firms we're building now start aging out.

And the cost comes back. And we go, oh, my God, this is a subscription service. We have to keep building these things out. I would counter that by saying that if you believe that anything in the infrastructure world is not a subscription service, then you might have some rose-colored glasses on. Like roads have massive amounts of annual budgets to maintain them, improve them. Right, but not on the order of rebuilding the road from scratch every three years. It depends on what the climate's like.

But in the vast, no, those are different. Like those are really, I hear what you're saying, but they're pretty different. Yes, yes, yes. But here's what I'd say is that the data center itself, like the data center refers to typically a facility. the facility will not depreciate in seven years. Some of the hardware inside of it, but here's the reality is that in the IT infrastructure sense, servers generally have not had a longevity of that long anyway.

So every single system that you use, be it cloud or in a corporation, chances are they're constantly upgrading, changing and evolving their infrastructure stack rapidly, maybe every three years. So I think that Like Michael Burry, the big short guy, the doctor guy He came out and was all critical about the depreciation schedule for GPUs Yeah

And he said that that was one of the biggest flaws in a lot of these companies Is that people were depreciating them too slow That they needed to be on rapid depreciation schedules And it got countered pretty substantially by the market because the intention is, yes, there will constantly be innovation and growth in it, but old computing capacity still has usefulness. It still has utility. So essentially, like you're even seeing that now. So Elon built Colossus 1 for XAI and Tesla, built his first big data center. They're already built Data Center to Colossus 2. And they had a bunch of

excess capacity on Data Center 1 or Colossus 1, and he just last week signed a lease agreement with Anthropic. So Anthropics can utilize all that computing capacity. Even though they've moved their business forward onto newer stacks, all of this excess computing capacity was just sitting idle. So Anthropic was like, well, we actually need it. So we'll pay you for it. And yeah. So I think the, I think like any industry, like the, the, the, the, the, the, the massive AI data center industry is interesting and it's growing rapidly but it'll also given the amount of money moving in it it's going to innovate rapidly like the data center that was built three years ago is going to look a lot different than the data center that gets built five years from now

exactly but but it but it wouldn't improve without the level of investment and without the the market initiative so you're already seeing that with cooling systems with energy systems. You're seeing rapid improvements in all of the industries around the data center as they're all getting so much capital that they're pushing R&D forward to kind of progress it. So going away from diesel generators to modular, nuclear. Solar or whatever. Yeah. I'm fascinated to see that feels the like rapid growth and improvement of them over the next company. years, I can't see how that doesn't have some impact on the depreciation of the ones that came before it. And while there's always been that depreciation in server farms, there's an unprecedented

explosion in the manufacturing of these server farms. It's like that's just a, that's a weird economic situation to be in that we're all going to have to kind of just sort of watch how it gets managed, because some firms will manage it better. Some firms will be on the upgrade cycle in tighter little loops. Some will go too tight with it. Some will go too slack with the iteration. It's like, it's just going to be fascinating to watch how people manage these assets, knowing that they're tech and tech depreciates in a different way. Do Rhodes depreciate? Yes, but differently than a GPU. Let's let's let's let's not use Rhodes as the analogy. Let's use the electrical grid. Sure. So the electrical grid was put in, you know, in various stages over the

last hundred years. Grossly inefficient, loses a boatload of power to heat energy, you know, loss, you know. But here's the thing is that it's essentially largely a government piece of infrastructure paid for by the government. There hasn't been enough capital and market forces to push R&D forward on it. Like anytime that we transmit and move massive pieces of power across long distances, like you can lose 10%, which is a massive. Like if you were, If that was an economic and private entity, they would start optimizing to reduce that. And whether that's relocation of generation facilities, better forms of generation. It's the same incentive structure that we look at with things like alternatives.

Like, yes, we know that solar is not the best. But you know what? The only way that it gets better and it has gotten so much better in the last 20 years is by incentivizing the development of the industry. the same thing is going to happen in data centers. But to go back to the power grid that I was trying to talk about, you know, like most of the USA's power grid is 70, 80 years old. It's critical, critical infrastructure that has had very little investment. You know, it is the maintenance plan.

It is, you know, imagine trying to operate a data center from 70 years ago. Agreed. But that's my thing. I'm not mad at it. I like to see the market activity because all it's going to do is push whatever the windows and problems and externalities

caused in the industry are going to get cleaned up because the momentum's there, the capital's there, and the pursuit of optimization will make it so. That's fair. Pretty optimistic.

You know me. I'm the optimist on the show. I know you are. You're a market optimist. I totally. I have maybe a, I harbor, some skepticism when it comes to the like and the market will simply make it more efficient so they can be more profitable.

I'm like totally. Canadian internet's great. Like I can think of all these private market things rooted in physical infrastructure where they're like, why would we spend the money making that better? We've we've captured it. We make more money than God. No, no. But here's the flaw in your argument is that Canada has essentially a non-compete oligopoly when it comes.

comes to telecommunications. Nothing like that would ever emerge when it comes to AI. I simply cannot imagine something like that happening right now. In all honesty, I can't. Given the quality of open weight models and open source models, given the optimizations that are occurring in the public space, in the open source software, given the mess ups by companies like Anthropic, just fully disclosing harness code, and information sets, the, I think that there will always be competition. Like I can go out today and buy a computer like an Apple Mac Studio and essentially run my own anthropic super facility in my house on Mac hardware, which uses no energy, ultra fast memory

bandwidth, all the rest of it. You don't need like the, as the model optimization, so the same thing's happening in the models themselves where, you know, the first one that we made, we made this monster model. It's like, hey, we got this thing called GBT, and it's, we trained it on everything, and it knows everything, and it's kind of like bad at everything. And then people have been working on ways to optimally structure them. And then we've got a mixture of experts models. Then we've got, you know, that research field is evolving so quickly. And now it's being facilitated by the fact that the AI can help.

But your argument is the competition in those spaces is in some way different. And we see that where it's like, yeah, the oligopoly of AI. Sure, last week it was open AI. This week it's anthropic. Next week it'll be gem. Right. But market leader and oligopoly are compatible concepts. Like you can have 10 companies that have just like a whole thing.

But I take your point. I take your point. But then you've got Moonshot releases an open version of Kimmel. me K-26, which is in competition with like chat GPT 5.4. You've got, you know, Alibaba releasing Quinn models. You've got meta, releasing Lama models. You've got Gemini releasing Gemma models or Google. Like there's so many open weight models and so many people doing so many things that within the last couple of years, I've talked about how Apple and perplexity, there was discussion in the market about whether Apple was going to buy perplexity. And I originally didn't

Love it is a great idea, but now I've come full circle where I actually think it's a great idea, because perplexity is really shown that they know how to build the harness, the thing that runs the model, the thing that communicates with the technicality of the model, and gets the most out of it. Perplexity, that's pretty much been their business model since day one. They don't build their own models. And I think that that's kind of where the market's going. The models will change based on the situation, you know, whether it's running on your phone, your watch, your laptop, your organizational AI server, off-site cloud data center stuff. Plus it gives the ability for the organizations, like somebody like Apple who has all of

the surface with computers, phones. Siri, for God's sake, do something about Siri. It gives them all this surface that they can then tune their connections. So the way perplexity changes the harness to get the most out of the models in the back end, they can tune the interface to get the most out of the harness. So they can link the harness in the interface better because the harness is linked to the model. And I think that that's where we're going. I don't know. I don't know.

Like the more I look at the AI investment, I see the market value return on artificial intelligence. I see market penetration right now in the sense of who's used. using what solution. But I think as we go forward long term, most of those solutions won't be worth as much as people currently think they are. Does that make sense? To me, it reinforces the idea, though, that the availability of open source models as a competitive balancing force is less important in the future than it is currently just by way of the fact that like, if you're saying that the model is the interchangeable part and that we're relying on the existence, of open source models as a market balancing force.

I'm like, well, those two things seem at odds with one another. It seems like right now that's an insurance policy that is going to become increasingly less valuable. And I think that you're right that those models are going to be like the hot swappable thing. It's like you just sort of pick yours. It's like when I sign into a new iPhone or a new phone, I just like plug my email into it. I'm not worried about who my email provider is.

It's kind of just like a community issue. It's like, eh. It just gets funneled over to this thing over here. It all shows up in the same inbox. Whichever client I use, it all shows up in the same inbox. And to shift a little bit, I think that you are right that these things are going to become hotwired and commodities. And I think that has a really interesting effect, especially for Open AI who's getting into hardware. Like Apple, Apple Siri, we've talked about that a ton on this show.

We won't get into the consumer tech Apple side of things too much. But it is worth talking about that they just agreed to pay a $250. $50 million settlement to a class action lawsuit over how they marketed Apple intelligence before they knew what it was going to be. They didn't admit wrongdoing. This is just an interesting story for the last couple weeks. Basically, the plaintiffs alleged that Apple had overhyped a version of Siri that was super context-aware.

The thing that you saw in the commercial starring, I think Bella Ramsey, was this idea that because your calendar is on your iPhone, the iPhone, Siri would be able to check your calendar looking into the past, figure out who was someone you'd met with at a certain, event and conjure that answer in real time. They postponed to those personalized series features, citing engineering challenges. And as a result, this lawsuit happened saying, hey, a lot of people bought iPhones because you were running ads saying that they'd be able to do that. And now Apple has basically agreed to this $250 million settlement up to, I think it was up to

$95 per device. And that is literally from an era where they thought they were going to be developing probably their own models to your point, deploying them on the device. And now we're looking at like, you can just plug any one of these models into your device for any individual task. You can have five of them wired in. So like so much has changed over the last two years of where we think the value in this market is. And it's all kind of encapsulated there. It's interesting. And that's just it. Like, for me, like it is a true, a lot of my belief is grounded in the sense that there still are going to be open weight models accessible and that people will still provide them.

If that goes away at some point, I think you start to see not-for-profits spinning up, training their own models from releasing them. I think that there will always be competition in that space because the value is too high. I still don't understand how Apple hasn't done anything about Siri. Even with their minuscule Apple intelligence models running on our phones, they still could have made Siri not so terrible. Any day now. I don't know why. The fact that they're taking so long to do anything worthwhile tells, well, no, it

shouldn't tell me. I'm hoping is an indication that they're actually going to do it right. And they're not just going to rush stuff out like they did with that commercial that got them sued. And their settlement is going to be pennies on the dollar compared to what the enterprise value change was for their share. So I don't think they really care. That's an interesting question of did they make more money on the stock pop after running

that ad. Then they paid on that. I guarantee they did. Yeah, guaranteed. But yeah, so I think, you know, the more and more and more that I play in this space and get to know it and tune it up and really dig into it. I think we're going to a world of specialization, just as we did with laborers. You know, we had generalists.

Then we got into fields of specialization. You could become a generalist in a field, or maybe you were a hyper-specialist working in an esoteric place inside of a field. And I think we're going to start seeing that with models. Like if you are a law firm, you will have a model that has been hyper-specialized in case law, precedents, you know, etc., etc. And there will be a custom harness built around that model that takes the most out of it as possible. And I think the company that's showing me that the best at doing that right now is perplexity. you really like the heart you really like the front end basically for all this shit that they're making the front end's different the mid layer the mid layer is what i'd say the harnesses like here you got

the model of the base the harness in the middle and then you've got the ux perplexity's ux leaves some things to be desired but i think apple's an amazing ux company and if you gave them an amazing midlayer company like perplexity to go through every functional component and interaction in an operating system be it a phone, a watch, an iPad, a laptop, a server, and you allowed perplexity to wire in, you know, functional AI at those moments. I think that Apple would leapfrog. It's interesting to imagine with them, knowing that this is kind of maybe their new route. This idea of like, you want Apple notes to rewrite your note, you tell us what you want to do that. It's like, oh, okay, I see where you're going with this.

it's going to be interesting to know what the like five years after that's going to be is it that they buy one of these companies and start really promoting it weaving it in and sort of promoting that internal option do they Sherlock the entire AI industry at some point and just go like well now that we are the surface for all of this when everyone is just using it

through their iPhone and switch something over and go and just sort of like decimate all these things it's like it's a it's a really interesting deal to make because obviously you want your model in front of all of the iPhone users. That's the, of course, huge surface area. Massive.

He's definitely giving a lot of control to one company. In other Apple News, just what we're rambling about it, Tim Cook obviously stepped down. I retired, took his billion dollars and exited stage left. Good for him. He added, I think somebody, I heard a joke that he'd added an entire zero to the back end the enterprise value of Apple since he took over. So I think he did a great job as a managerial person. I don't think, as somebody that fell in love with the company Apple based on rapid innovation, definitely didn't see that under the Tim Cook era. So I was happy with their appointment to their

new CEO, John Ternis. He's a product, product person. Which makes sense with what we're talking about if Apple's saying, like, we're just going to go back to being the hardware front end, company. You put your fuel in our engine. That's fine. Totally. We'll build the UX. You take care of the rest. Your AI plays to own the hardware. Like honestly, super smart. Be like, we're just going to own the hardware. Yeah. To speak nothing of the Mac mini of it all, we're just going to own the surface that you engage with these things on. It's like honestly pretty smart. Stay out of the suit. Well, I can get it. Open AI spent billions of dollars acquiring John Ives for the same thing. about that. Apparently it's a phone. Everyone is saying now it's the rumor mill is that it's a phone.

Yeah. Interesting. Best of luck. The thing for me, best a look. The thing for me is that like a phone is now no longer a phone. We use the word phone colloquially to discuss the mobile computer that we carry. The amount of people that actually push the phone button on the computer, on their mobile computer is infrequent these days. I don't get as many phone calls as I used to. But the extensibility of having a mobile computer that comes around with you everywhere that has fully AI power. I'd be interested to see what they do.

Very curious to see what they do. I bet it's Johnny Ives, so the hardware will be nice. I'm curious. It's like I'm not a hardware guy and I'm not, I'm not a mid-layer guy to borrow your sort of framing. But I think about design and front-end the kind of experience of using a designed object pretty regularly.

And I'm really curious how much they commit to the chat layer. Like I think you could make a version of a phone that is like pretty unpleasant to use if it over-relied on the text and natural language interface. Like full stop. I think it's like you could, you could really tune this wrong and be like, yeah, just ask the phone to book that for you. Be like, that is 27 syllables when I could previously do it in four taps. You haven't made it.

I'm curious. That's one of my pet peeves with chat AI right now is I ask you a simple thing and I get back a A novel. That's a novel. Like I don't want them. And I'm just like, I'm like, give me, give me three bullet points. Like, give me your recommendation pros and cons. That's it.

Like, I don't want you to discuss the world with me. And it just be succinct for the love of God. I heard someone liking it to like the brittleness. It was like, we escaped. I can't. remember who said this, but it was like, we escaped the command line in terms of how we use computers because it was like, it was just very brittle. You had to type it. You know the exact syntax. I'm borrowing that framing of brittleness. And then we moved to the mouse keyboard and

eventually multi-touch, but this idea of just touch it, touch the thing that you want it to happen and it will that removed the burtleness. Large language models don't have the original brittleness of the command line, but they introduce a new set of brittleness. Like, it will respond, but it might just make shit up that isn't relevant to what you're trying to do, and it might write you a novel when you need two words. So it's like you have, the future seems like a really thoughtful blend of these two. And if you're the chapaw company, I'm curious where you fall on that. Well, that's the, that's the mid layer.

That's the harness. We know that the model, and it's also the model tuning, obviously. Like, they've spent a lot of time tuning the model to be very descriptive in their responses to make sure that they discuss every case that they made an assumption on. Just also for liability's sake, where you're just like, you ask me this. And, you know, it's impacted by these 75 things. So let me go through all of them in great detail. And you're like, okay, yeah, great. Just tell me where to go eat brunch.

Exactly. Like, yeah. Yeah. I'm really, if it turns out to be a phone, I'm really interested to see where that goes. Like, how good a camera does Open AI make? Like, just all these weird questions about, like, phone engineering and, like, the post-processing pipeline of images and, like, all these weird little things that have just nothing to do with their core business? So it's like, do they just buy nothing? Like, do they just buy a mid-sized Android phone maker and, like, repurpose it?

Be like, you work for Jonathan I. They don't even need to. The reality is, like, you can go on Alibaba and buy bulk Android Snapchat and phones and have them branded. Yeah, exactly. It's not, like, you don't need the hardware engineers anymore. Yeah, the problem's been solved, so they don't need to focus on that. The thing for me is, is Apple's being successful because of, and in part by the ecosystem. You know, I have an Apple computer.

I have multiple Apple computers. So I bought an Apple phone. And because I have an Apple phone, I bought AirPods. You know, they might not be perfect, but damn, are they easy to use? Better than the competition. Yeah. Better in the competition. If I want to really listen to music, I'll go sit in like my listening room or I'll use a set of high-end headphones.

But like for daily use and phone calls and walking around, I'll listen to a podcast while I get groceries. It's, they're amazing. So the ecosystem becomes a thing. So to just show up with the Trump phone, the Open AI version of the Trump phone. And here's the other thing is given the fact that code is now a commodity, a building of code, everything comes down to the product idea and product design because the AIs are rapidly prototyping things at speeds unseen by companies in the past. So even Apple released a service update to a piece of their software this week, and they left their Claude instruction file in it accidentally.

So like Apple's using Claude code internally. Yeah. And nobody's surprised by it because it's barely a gaff. Output efficacy. Yeah. Of course. Yeah. The Jensen Wong quote of like if I hire a half a million dollar engineer,

I want them to spend a quarter million dollar in AI tokens is coming true. Like the more productive my human capital becomes. the better, the better investment in the person. So, so the, yeah, to just build a phone and crank it out, even if they come up with the world's greatest way to interface with the AIs, Google would probably be able to replicate it inside of 28 days. So it's like, what kind of market moat are you going to have to keep your business sustain? I would bet their argument is going to be, like, the vision of all these things is just like,

well, if we have unlimited personal context about you, because you input everything into something and you receive all your messages through something, and maybe it's just sort of like got a camera on it and you're every time you take a photo, like the more context these things have, I think the argument goes, the more sort of powerful and oh my gosh, they become. And so if you don't own that physical ID layer, if you don't own that object and that object's manufacturer can put any kind of a privacy wall between you and that information, you can't get that that thing, that precious troncha context thing, that pile of context about the individual user. So who has all of our context, though?

Well, right now Apple and Google to an extent. Apple and Google too. Increasingly less, but for enterprise work, a thousand percent they do. Like that's what they own. Sachin Adela is a smart person. And it's like, okay. I lost the microphone, like the Windows phone race. So I've lost the personal context, but I can get the business context.

And so I think the question is like, is chat GPT sticky enough to get people to want to feed their context into it? And face you just made same face I'd make too. Not compared to like, no, not on its own. Unless Johnny Av comes out with some hardware that's just shit hot and everyone's got to have it. I'm not sure that like, but chat GPT knows all my stuff. I'm like, I'm not sure that's a value ad friend. Yes. I don't know.

that to me to loop back to the earlier conversation about model portability is that to me is the fault in all this actually after one of the last episodes where I asked I was like oh man I wish I could invest in data bricks somebody messaged and was like oh you can actually buy these venture funds

but they're trading it like 18 times value of the VC value so like if they bought a 100 billion dollars of equity in a trillion dollar raise for open AI, they want $1.8 trillion from the market to buy that same equity. And I was like, there's no chance. Like there's people buying into open AI like it's going to be the next coming. And I just can't see it.

Like they just don't have a moat. The moat is that they make good models. But guess what? If you go look at the model arena billboards and stuff, lots of people can make a good foundational model. It's coming down to who, like even the competitive differentiation now you're seeing is like Claude Co-work, Claude Codex. That's all harness. That's all mid-layer. There is some model tuning going on for sure, but perplexity's mid-layer where perplexity is sitting and building and growing and developing skill sets is now what the frontier model providers, the foundational model

providers are now competing in. It's like, hey, look how good our tool is that uses our model. Sure. So the models have become less and less important. You know, we're almost sealing out on the models, but getting really good at building the mid layer that communicates with them. I'm, I'm interested in this because I think this is going to be where we see the fork of like, what is the application of this technology for people versus software developers? Like that Totally. That gulf of like, it's so hyper useful for this group of people. Yeah.

That are making the tools for the other people. But it is untested how useful it is for these people. And there's all these cool theories and stuff that you'll be able to do this. You will do this. Like totally. Not happening today. So how do we get to those tools that make that midlayer really powerful for the average person is like, whoever answers that question is like, oh, you win.

Like you made that midlayer. valuable to people and now you've turned everything else into just like a water line or a commodity for your layer that provides value to the actual end user and not just software developers. I think that's an important distinction. Totally. But I will say that the focus on software development has facilitated the speed at which the entire transition is happening. Of course. You start with what it's useful at. Totally. Rational thinking. Like law, wild, what it's going to do to the law. It can read legal precedence and build a, build a case. So anyway, I still want to focus on, I was going to make a point there, I don't want to

digress too much. But the thing that the biggest shock to me, like when we talk about the AI bubble, I think there is a bubble in the anthropic world. I think that there is a bubble in the open AI value. I don't think that AI as a societal change, you know, disruptor is a bubble. I think there will be massive changes from it. The big thing that has confused me is how. Apple Siri is a case study in the failings of a massive company that has an unlimited amount of money

could have built the world's greatest personal assistant utility. Could have been on the forefront of this. And we're still seeing this. Apple, like the people that have the surfaces to make the disruption, Microsoft, Apple, aren't doing it. Even Google Gemini, the way that they've integrated Gemini into the office suite, Google Suite, is really bad. It's like the, like if you have an unlimited amount of ungodly resources and power and money, and you have the software developers that are now enabled by AI agentic engineering,

you're getting 10x the output out of every single one of your software engineers. How have you not built better products to utilize your surfaces to make this transition? That's the thing that really blows me away is the companies that could win it all are losing in my eyes. Like they're behind. Like why is a little tiny company like Anthropic pushing out co-work and OpenAI pushing out Codex, which is their coding tool that they're now rebranding is like, it can also help you do regular stuff. Same thing cowork is.

It's just the open AI harness for Claude code, but repurpose to like business things. operational efficiency stuff. So why is Microsoft's solution co-pilot chat? Like you guys have trillions of dollars. You had all the opportunity to nail this. Yeah, some of the best

software engineering minds in the market. And like co-pilot chat is your solution? But then so too should we all be using Windows phones. And it's that great thing of like, you can have all the money in the world, but if you don't have the product or the idea, it doesn't matter.

A buddy of mine is a, like a, he is a developer, but he works in product now at a tech company. And he was outlining how the relationship had always been, I'm on the product side. Comes from an engineering background, but he works on product now. And they had a series of different engineers, like, say three or four below them. And it was the idea that being like engineering takes so much longer that we would come up with concept, like product concept working with marketing. sales and design, and then we would allocate it to the engineers, and they would have to take time to do it. That's why we had multiple of them. It took time for them to execute and to make these features that the product wants. And that in the last two years, that relationship is

completely flipped, where they're executing on the idea is faster than the product pipeline can come up with them. And I'm imagining that relationship multiplied over the scale of a company like Microsoft or Apple or any of these things, where it's like, you suddenly don't have nearly enough resources allocated to the what and you're way over invested in the execution and the how. Like one of those people can now do what the other three were doing. And you have this asynchronous that has emerged. And the design problems that would lead you to not releasing co-pilot in the shape that

it is currently in aren't being tackled because you've misallocated all of your resources. And I think that that's a really interesting thing for all these companies to have to figure out is like what is our relationship between engineering and design? in World War Engineering costs what it costs now. Say you gave me a job tomorrow, VP of AI integration at Apple. First thing I would do is make a bunch of tiny teams. Designer-led product manager-led. Like, U-X first, and then a couple of engineers,

and I would send 50 teams away to compete and be like, bring me back. Like, here's one small task. everybody bring me back on it yeah skunk works bring me back an independent solution to each of these which one did we like

we're going to case them to everybody everybody's going to get to vote trial them pick the one that works maybe you get a hybridization then like we really like this from there and we like that from there great let's put that together

and see if it works and do that over and over and over until you figure out how to integrate it because the reality is is like a tiny little team of highly competent engineers matched with a competent project manager, design focused UX person, could bang out a product in,

like I can bang out a product in a month as a person. Imagine what you could do if I was being paid to do that, not just doing it while I watched tennis tournaments. Yeah, I think that that bigger investment in the design side of things, it's like you are talking about the design process. It's like ideation prototyping and testing. Like that little TikTok tick, it's like reorient around that because it feels like you're a little bit letting the like, you're letting something else lead the process. And it doesn't seem to be working. Like so maybe go back to that fundamental.

The harness is the thing. Like you don't need to have a chat interface. The harness can have the chat for you. You just need to give it a simple instruction. And if you've got all the surface area and UX surface area that's provided by the operating systems, You can integrate tiny little instructions that go into a background harness that pull context from the apps that it has access to, your calendar, your messaging, your emails. It would be simple. It wouldn't be simple, but it wouldn't, it's not challenging.

It's not like we need to, like, solve a new problem. And it just, it shocks me every day that these massive companies that are so well funded and so well capitalized are, yeah, and they're not progressing. They're not rapidly moving. Like, the thing that's giving Anthropic and Claude and OpenAI and Codex momentum in the market and valuations in this Goldie Locke scenario is the fact that they're agile. They're moving quickly. They're trying things out. And they're showing successes rapidly where a lot of the bigger companies aren't. Like, when was Google's last model?

Like, I know they released like 3.1 or something. But they really haven't changed any of the interfacing. there's nothing really that's changed about how you interact with AI. They just have a new model that's a bit smarter. And it's like, that's nothing anymore. Like, I would actually prefer an older model that's a bit dumber with a really good harness than the best model with like just a chat interface. Yeah, I wonder if Google at this point has started going like,

we never really owned enterprise of a lot of this stuff. We owned a consumer chunk of the. the market. It's always done really well for us. It's very compatible. Like, low cost, high volume is great if you have an ad business bolted on to the side of it.

Because I'm like, I actually don't care about your bill being that big. I care about your usage being high because there's a trillion of you. Like, that worked really, really well for them. And maybe they're like, you can go develop software wherever you want.

We're going to control the chat interface on the iPhone and the email and all of the internet. You know what I mean? like they're going to start segmenting based on what's useful and eventually someone's going to really own software development someone's really going to own the chat interface on a phone someone's going to really own each of the major market segments which are going to become calcified and clear over time and we'll kind of like just sort of end up not quite back where we started there might be a new entrant or two and an old entry that's gone but like it happened it's happened every time since probably going to happen again I would I would make a bet on Polymark

it. Do it. And I apparently could. Me and Sergeant Van Dyke. Is there any, me and Sergeant Van Dyke, just to bring a full circle. Just to bring it full circle. Anything else I want to close on? No, I think that was a big enough rant. I think we solved it.

I think it was a good one. I think some people are here for the rants. Sergeant Van Dyke, Boydick. Boyt. But we didn't. Not a big deal. Did you want to talk about birds?

Did you want to talk about? words. Well, I already talked about birds. What did you want to talk about? Just if it flies as flies. I thought I already just mentioned that before. Famously. Famously. No, there's nothing else that I really need to chat about. This is just wild, wild times out there. It's a very interesting time. Yeah, there's going to be, maybe we should do talk some cyber crimes about AI, not just about AI because there's some pretty interesting stuff going on. Yeah. But maybe we'll say, I think we got some of those. I think for a future up, It's going to be a busy summer, so we might be coming at you with a hotline hacked or two while Scott and I are gallivanting around to various places.

But save that for the next one. You make it sound so nice. Like we're going to be gallivanting and not just sitting in meetings. A meeting is a gallivant for the mind. My oh, my. My, my. All right. Thanks for listening.

As always, we appreciate you tuning in. We'll catch you in the next one. Take care.