Unchained - The Chopping Block: CLARITY Dies on the Senate Floor, Hunter Biden's LAPTOP Rug, & the AI Pause Pact

Episode Date: September 17, 2026

The boys get together in person as CLARITY falls short with 47 votes. They break down what comes next, Hunter Biden’s LAPTOP token collapse, the Robinhood/Hyperliquid case, Balancer winding down, and the debate over an AI pause. The CLARITY Act fails cloture with 47 votes, not a single Democrat in favor, and the crew works out what that leaves behind: rulemaking at the SEC and CFTC, an ethics fight that was never really about market structure, and Robert's tally of everyone who walked away with nothing. Then Hunter Biden's LAPTOP token collapses 99.85 percent, the SDNY indicts two Robinhood engineers over Hyperliquid front-running, Balancer and a wave of exchanges wind down, Robert explains why Satoshi is a time traveler, and the panel takes apart the labs' agreement to pace the frontier. Listen to the episode on Apple Podcasts, Spotify, Pods, Fountain, Podcast Addict, Pocket Casts, Amazon Music, or on your favorite podcast platform. Show highlights 🔹 CLARITY gets 47 of the 60 votes it needed, with no Democrats in favor, and the sticking point was forced divestiture of Trump-affiliated tokens rather than market structure. 🔹 Robert tallies the losers: no ethics win for the Democrats, no GENIUS rewrite for the banks, no expanded AML hooks, and no developer protections for anyone. 🔹 Haseeb argues adoption protects crypto better than legislation ever will, because nobody takes away a product people already use. 🔹 Prediction markets are politically safe for a reason nobody says out loud: every politician already reads them as better polling data. 🔹 Tom's counterfactual for what would have flipped the vote is blunt. You never should have launched that damn memecoin. 🔹 Hunter Biden's LAPTOP token flashes to a 140 billion dollar valuation on 50,000 dollars of liquidity, then falls to twenty cents, and the chart ends up shaped like a laptop. 🔹 Two Robinhood engineers are charged over front-running crypto listings on Hyperliquid for about 50,000 dollars each, and Haseeb explains why the OpenSea conviction got overturned. 🔹 Balancer winds down and the panel asks what became of the ICO treasuries, from Golem's GPUs to Aragon's founder building a neobank. 🔹 Robert makes the case, half in jest, that a future superintelligence sent Satoshi back to 2008 to get humanity building chips early. 🔹 On the AI pause: Tarun on who evaluates the evaluators, Tom on why coordination without China is futile, and Haseeb on why a training freeze just moves the competition to another line item. Hosts ⭐️Haseeb Qureshi, Managing Partner at Dragonfly ⭐️Tom Schmidt, General Partner at Dragonfly ⭐️Tarun Chitra, Managing Partner at Robot Ventures ⭐️Robert Leshner, Founder & CEO of Superstate Disclosures Timestamps 00:00 Intro 00:43 CLARITY day 03:36 Silver linings 05:26 Was it close? Or did everybody loose? 12:23 Why adoption protects crypto better than a bill 15:00 Was DC trading on the CLARITY odds? 18:37 The one thing they would go back & undo 20:48 Hunter Biden's LAPTOP token collapses 23:50 Celeb coins never work 26:17 Robinhood engineers indicted over Hyperliquid front-running 31:35 Balancer winds down & the exchange shakeout 35:11 AI arbitration & ideas ahead of their time 37:44 Satoshi is a time traveler 39:14 The AI pause: pacing the frontier 42:57 China calls it a conspiracy 44:40 What do you buy if the doomers are right 45:50 Does an AI pause actually cut CapEx? 50:32 Futile without the Chinese labs 54:05 Harden society instead of slowing the models 56:47 Which doom vectors are real Learn more about your ad choices. Visit megaphone.fm/adchoices

But like right now, the fact that it failed, changes nothing for tokenization. Right? So you're good. Personally? Sure. Yeah. Not a dividend. It's a tale of two clon.

Now, your losses are on someone else's balance. Generally speaking, air drops are kind of pointless anyways. Unnamed trading firms who are very involved. I like that eat is the ultimate pump. DFI protocols are the antidote to this problem. Hello, everybody. Welcome to Chopin Block. Every couple weeks, the four of us get together and give the industry insider perspective

on the cryptic topics of the day. So quick intro, Tristan, Tom, the D5 Maven and Master of Memes. Hello, everyone. We've got Robert, the cryptic connoisseur, and czar of Super State.

Good morning. Then we've got Tarun, the Gigabrain, and Grand Puba at Gauntlet. And I am a sieve that had hype man at Dragonfly. We are early-station investors in crypto,

but I want to caveat that nothing we say here is investment advice, legal advice, or even life advice. Please see Chopin Bloch and X, for more disclosures. So today is Clarity Day.

We are very excited for this historic bell. At 2.15 p.m., the vote will occur. Yes. We are going to watch it. We are excited. We're going to watch a lot. I put my whole bankroll on.

We are so excited. We're going to pop some champagne on the show. Yeah. So, no, actually, clarity already. It's over. It's over. It's dead. It's never coming back.

It might. Good game. Well, so I was looking at the Kalshi markets. Kalshi has a market going out until 2008. It says 22% chance that in 28 we get a version of 30. It's bold. You're citing Kalshi on a show where

we have a host whose tweet Twitter has only been really vehemently taking them. That's really the, yeah. There's also still like a 5%, 6% chance through year end as well. Yeah, Polymarket has it at 8% last I saw coming in. But I think that's, once you get to the tail probabilities, you can stop really believing them so much because it's just kind of hard to do much with that.

But basically what happened was that clarity, of course, the big mega bill that we've all been waiting for as an industry, We needed 60 votes in order to get this bill passed. It got 47. So it did not even get all of the Republicans voting it. Not a single Democrat voted for the bill. You got a lot of people basically saying, look, this is not quite what I thought I was getting. There was some weekend deliberations over, I think Trump signed off on some small changes

to the ethics provision. They were actually pretty material. So the state attorney general would basically be able to go after crypto exchanges for listing tokens that were endorsed. affiliated with a president or with a congressperson. That was apparently not enough. They wanted there to be some forced divestiture of tokens that were basically part of the Trump token universe.

That was not enough to get done. There was some, even today, this morning, there was more negotiating trying to happen from the Dems. There was a counterproposal from the Dems. Right. So I think you probably saw the play-by-play. I just kind of saw the outcome. But basically, good game.

Seems like now the story for regulation from crypto is that it's going to have to come from the agencies. It's going to have to be through rulemaking. The SEC and the CFTC have already signaled. We're just going to go ahead and basically implement clarity. And all the frameworks that have been negotiated over, that's kind of set the blueprint for what a regulatory agency-led solution is going to look like. It doesn't mean we're probably not going to get along. So it's unfortunate.

We had some comments from people in the Twitterverse that, thank God, this is over so that the chopping block doesn't have to talk about this. This is the last show. This is the last show on clarity. Yeah, until it passes. Well, we'll see if it gets revived. Kalsi still has it at 22%. 42% by 2028.

How do you feel knowing that this thing is over? As I said on Twitter in response to our grateful fans who are excited to hear Haseeb shut up about this, what's the difference between like this bill and watching 80-year-olds get wheeled in? can barely talk and they still vote yes or no and they don't understand what they voted yes or no. I don't know.

Wake me up when we have like conscious and thoughtful senators. There's a few. There's a few. It did feel like a lot of the stuff I was born. I mean, a lot of the developer protections were getting watered down last minute. So I'm like, yeah, I mean, it kind of sucks, but I'm also like, I don't know, life goes on.

We have rulemaking. I don't know. Yeah, I think the thing that I kept saying, I think every single time we brought up clarity is that like, look, clarity is not the end of the world. It's not the most important thing. We are going to get rulemaking. The number one thing that is going to end up protecting crypto from a regulatory perspective is adoption.

If you are adopted, if people are actually using you, then it's going to be way too hard to rip anything out and change things. And people are going to cry bloody murder if you actually try to change any of the actual instantiation of stable coins and, you know, exchanges and perps and blah, blah, blah, blah. So obviously it would be nice. And if we get a president of OC in 2028, okay, there's going to be some hell to pay for all of the crypto regulations that were adopted and probably a lot's going to get unwound. But I think even if you get a moderate Democrat, it's probably going to be okay. You're not going to get everything that you want, but you're going to get most of what you want

because the reality is that even the Democrats who ended up saying no on the bill, there's a very close version of this bill they would have said yes to, which tells you that like it's not that hard to get to something that everyone can agree should happen. Yeah, at the end of the day, I mean, I was surprised by the fact that the two sides weren't that far off from each other, right? I read the Dem counter proposal this morning. Really? And it actually wasn't that long. It was a red line to the Republican draft that was put out late Sunday night. And honestly, like, there wasn't that much of a divide between the two positions. It mostly came down to ethics and wanting to score some political wins against Trump. The rest of it,

I mean, a lot of the conversation has been within D.C. for the past couple weeks, like, the other pieces are there is negotiating bargaining chips, but they're not even, like, fully the meat of the discussion. It was really about sticking it to Trump. And, you know, I still think, like, some of it was still relatively innocuous. They weren't asking for roping Trump's children in. They were trying the right things that looked like it would loop his children in by saying, like, oh, dependent children of, an officer, which is like, okay, well, they're not dependents. They're adults and they live on their own.

I doubt Trump is putting him now on his taxes as dependents, right? Didn't Baron do something? Wouldn't he be a defendant? Barron might be a defendant. I think they're going after Barron with that one. Are they, though? Like, has he moved out yet? Yeah, he's at school.

He's at my you. Okay. Yeah, he's a dollar. I don't think he's a dependent. He's probably made enough money to, like, have his own apartment and no longer be a dependent. Yeah, he's a trench. Yeah. So, like, the things they were doing were more.

signaling. Also, if you launch a token, you're probably not a dependent by virtue of your token income. Probably. It's kind of a catch-22, actually. Right. But like, and so like even if... I don't know about that one. I feel like I'm sure... Well, actually, we have a meme coin. They didn't go so well. Yeah, I was about to say, I'm sure there's some mean-coin deployers who are not living off. Yeah. Yeah. That's true. That's true. Yeah, but it wasn't like like the two sides were that far off. I think a lot of the surprise that came about today was the fact that it seemed like the two positions were relatively close. And supposedly there were some hallway meeting that occurred a few minutes prior to the vote.

And everyone stormed out and said, like, we're not going to find an agreement. And that's the end. But the two sides were not that far away from each other. Like, we were quite close to clarity, assuming that you take everyone at their word and assume everybody was negotiating in good faith. Okay, silver linings. Are there any silver linings from this? I learned the word cloture.

You learned, Taru, learn the word. New word. I'm just trying to, I'm just trying to, Take that one. The Taroon dictionary. It expanded. It expanded. Up only for Turun's vocabulary.

Yeah. I'm just saying personal growth on the... We all grew personally. Yeah. We all learned a lot more about the legislative process. The parliamentary procedure in the Senate. I mean, understanding the friends are enemies you made along the way, depending on... Yeah, yeah.

I will say the one thing I think we can take away is that the banks lost. Yes. Because the banks were fighting to get a renegotiation of genius. Yeah. And basically the rewards language in Genius Act stands. No, everybody lost. Okay. Let's take the tally real quick.

Okay. The Democrats didn't get any ethical wins against Trump. They didn't restrain Trump in any way. For all their talk and bluster, they achieved absolutely nothing. The banks made no revisions to how genius works. There is still no prohibition on yield being paid on stable coins, effectively through how they see it, right? Through rewards. we didn't get an expanded AML KYC, you know, set of hooks for the, that contingent of people.

The status quo still holds. We didn't get any developer protections at all. Nobody got what they wanted, right? Everybody had an opportunity to get something. Except Turin. Tarun got what he wanted. Tarun got what he wanted. No more politics on the show.

That's a lie. You guys know that's why. There's going to be some other bullshit shows up. everybody's like, oh, like there's the UK clarity lock. Let's let's all pretend we know common law. Yeah. Everybody is a loser here, right? Like the list goes on. There was no changes to prediction markets. Like everybody was talking about how to like loop some prediction market things in for Indian gaming. That's not there. Nothing happened, right? Like everybody is a loser. Staff on both Republican and Democratic sides worked for over a year on this, right?

It wasted a lot of the brain cells from some of the most important senators. It wasted so much time remember everybody is a loser coming out of this process. Well, okay. The question was silver lining. Nialism. Resounding not answer to that question. Look, obviously it does create a framework for what rulemaking. The reality is that we did do some democratic process about what people want and what the

final shape of a regulatory regime ought to look like. And the bill itself had a bunch of stuff that was like, the SECC decides. The CFTC decides. And okay, well, they will decide. and they kind of got the rough rubric and guardrails of what lawmakers want and what people want and what industry wants. And it created a lot of conversation about what a regime within the U.S. should look like. And the reality is that people realize that it's not outlandish for this stuff to be kind of conforming

with a regulatory regime in the U.S. I think all of that is progress. Anything else that we would add to that? What does it do for tokenization? Obviously, for state. I mean, for tokenization, again, there's no bill, right? So it does nothing, right? One of the pieces that was in the legislation, which no longer exists, right, was to expand

the SEC's discretion in creating rules specific to assets on a blockchain, right? It's implied that they already have that power today, but it was more explicit that they could go further to support tokenized assets versus the older digital counterparts to it. But like right now, the fact that it failed changes nothing for tokenization, right? So you're good. Personally? Sure. Yeah.

Yeah. Yeah. But like, sorry, I don't know why that was so. I've been interested in this bill for years as somebody that's been in this industry for almost for nine years now, right? I hope to be in this industry for nine years, right? This could have been something to set the stage for the next decade of this industry. and we're going to continue to operate the same as we have for the last decade,

which is no one really knows what's allowed and what's not allowed. It's sort of a moving target. It's swinging back and forth, bouncing from guardrail to guardrail every time a new president comes in. It's like, fine, we have more of that. Yeah, yeah. No, it's true. I mean, the rulemaking does help because even before that there was really no rulemaking.

There was some kind of soft guidance. there was the Dow report and there was this and there was that. We are very clearly going to get rulemaking under this administration. There is no doubt about that, but agreed. It means that it can be reversed with a new administration and with a new set of priorities. And that sucks. Yes. But again, I stand with the idea that, look, number one protection is adoption.

Is that at the end of the day, everybody is subject to democratic forces and popularity contests. And if this stuff is popular, they will not want to take it away from you. You know, like as much as people talk about how AI is so unpopular, nobody dares to touch chat GPT and to tell people, you can't use this anymore because it's bad for democracy or it's bad for data data users or bad for water. They're like, okay, we'll go after the data source. We will not touch the consumer products. And I think you saw the same thing with the gig economy and ride sharing and everything else

that was politically unpopular or it was politically popular to go after the idea, but never after the product once it was adopted. Except there's one. very bad example. What's that? Jewel. True because adults don't use it. Well, no, Jewel is the one where you killed the local

American thing and then you just were flooded with imports and it was like, what was the point of that? The Jewel is the kind of example in my mind of government. Did the government go after Jewel? Like, I don't remember the exact detail. It very much did. Yeah, it was like FDA.

It was a bunch of states to kids. It was like a class. Everyone joined together. The Chinese tapes came in and everyone switched to that. Yeah. Yeah. It's kind of illicit. Well, I be, but they came in as in like they just, you know, you could order them on.

You could order them. Well, the thing is like, look, at the end of the day, politicians are mostly 50 to 80-year-old people, right, of mostly upper middle class. Some are older than 80. Some are older than 80. Thank you for the clarification. The number one thing that I think predicts whether or not something is going to get banned at a consumer level is do 50 to 80-year-old, care. about it slash use it. And that's actually one of the reasons why I think actually prediction markets are safe, even though sports betting is almost certainly going to be under scrutiny

by the Supreme Court when the case goes up next year. But prediction markets were at large, every politician use prediction markets. They do? For elections. They absolutely use them for elections, as they consume them. They might not. But they don't want them to go away. They don't want them to go away because they're very important. It's polling data for them. Exactly. It's better than polling data. And so there is no way in the universe that prediction market is sort of weird though, because their notion of using them is only consuming the output of, right? Right. You can't get gross people though, right? Yeah, yeah, for sure. No, no, for sure. But I think that that's the one place that they're kind of a little bit weird versus sports betting, where like most people think

of sports betting is like using sports betting is like not looking at the odds. Right. But actually like participating. And somehow prediction markets, despite having similar properties are like the opposite. All right. Well, here's a question for this group. Do you think, think that DC was monitoring the prediction market probabilities for clarity? And do you think that played into the decision making and the one hundred for sure? Definitely. You think it's a self-fulfilling prophecy? Well, I don't know about self-fulfilling prophecy, but I'm sure they were like looking at it as a way of like deeming there how much. I think the Democrats and the Republicans probably looked at it as a way of like how much leverage do we have against the other side like how aggressive should we be on the red lines.

Like I'm not saying it's like they were like trying to, but they might have done like, little things based on, oh, like, everyone thinks it's not going to happen. We should go for the jugular type thing. I have never heard anyone in D.C. say that they believe the prediction market odds on clarity. I think pretty much every single person I talked to, like, over the last year was like, yeah, these obviously are not correct. They were too high or too low?

I've heard people say, I am. Both too high. It's always wrong. Yeah, it's always. Nobody ever buys it. I think for everything else, it's like not your thing. You're like, oh, yeah, the prediction market is obviously correct. And then when it's your thing, you're like this total.

Well, I think in hindsight, you should just go compute the Breyer score and see if it, get conditional on knowing it failed, like, how close was it to, like, true for that? And the answer is probably actually going to be like, it wasn't horrible. It was like, I bet you the answer is it'll be slightly better than Mardingale, like, for all polymarkets. For the, just for the clarity one. Like, conditional on knowing the outcome. You then go backwards and look at, like, the Breyer score over different windows and see if, like,

how likely is it to converge to know? because it peaked at like 84% in February. Yeah, I mean, it went up to 69% like yesterday. No, it didn't go to 69%. Yesterday was like 30. No, it was 30. There was some place the day when it's like, yes, sorry. No spike the airbag.

That is a artifact of math. But my point is like you could go back and see if like it did better than coin slip, right? Like that's kind of the, that's sort of, I bet you it actually still has some predictive power. It's just like kind of noisy. No, I think it's predictive. I think back in February, the odds were probably closer to 80%.

I think people have probably played on those prediction market odds and internalized them. I think like Coinbase got very cocky in February because the odds looked like. Oh, yeah. I forgot about the whole Coinbase. Yes. That feels like yesterday, Brian. That feels like so many years. I'm like, I don't know, dude.

That feels like so many years ago. No, it feels like years ago. But basically, like realistically at the beginning of the year. Do you think we would have passed if Coinbase didn't do that? Yes. Like actually, I'm curious. Do you, do that you know, I think also. Something that probably played into it more was also the Iran war,

is that Trump's ability to bully Republicans into certain positions is weakened significantly after the Iran war. And like in February, we were, Trump was good. I mean, his approval rating was much higher. His control over the Republican Party is much stronger. So I think that also plays into it. Absolutely. I think if Coinbase had installed it, I think it would have passed.

I think the negotiating leverage for clarity has decreased monthly since February. you saw more ethical concerns from Trump, like it literally kept on going up. Coinbase lost their GC also in the middle. Coinbase lost their GC. If you think about that. But not their head of policy. Yeah, yeah. The GC is not the same as the people.

But still. Yeah, yeah. Sure. Sure. Sure. There was turnover at Coinbase. But we got closer to election year and it got politicized. And like the more partisan, the environment, the less bipartisan, you're going to get

a bill. The housing bill being bipartisan and like squeaking through, like, two Trump's protest was like the last thing that I think was clarity-esque and that was like it barely made it through and that had widespread by okay so suppose you got to play God and you got to go back in time and remove change the outcome of one thing that you think would have likely flipped the outcome of clarity tell coin you never should have launched that damn meme coin that's what I've been saying that one thing sure sure that one like yeah that is the whole ethics debate none of the I think not just that damn meme claim, but multiple.

What else are you doing? I guess actually World Liberty and it being more of a headache. Yeah, all three, right? I count Real Liberty as a meme quote. Yeah, yeah, that's true. I guess that's a good question. Which of the two do you think did more damage? Yeah, yeah.

I think it was probably Trump being point. I think it was more, I think it was worse than Trump than World Liberty because more people bought it and more people heard about it. World Liberty is worse when you go into the details. but most people don't know the J-J-J-J-J-J-J-O. Okay, so you're ranked ordering of things you could would have flipped. Get rid of the Trump meme coin. Get rid of World Liberty.

Coinbase stopped protesting. I think Iran war is probably like number three. I assume, okay, okay. I don't think Iran war is that related to it because I think Trump would be much more able to do things in D.C. Without the Iran. I mean, I will say I did watch this. I think I watched like the Daily show or something this week by accident because

it was just like on because I never watch any of these types of YouTube things. And the entire first five months the episode was interviews of the Republicans for why they were skipping the Geo Peak midterm convention, which is so like I feel like Robert does have a point. There was like there was like 20, 30 people who were like senators and representatives having to like get a microphone in their face and like making up excuses. And some of the excuses were very funny. It was like one of them was like, yeah, I have like a quail hunt. You can't skip a quail hunt, Tarun. What are you talking about? I was very funny.

If I had a quail hunt, I would skip any other priorities. I've never been invited to a quail hunt. I was invited to a bird hunt of some variety. I don't know what bird it was, but I should have gone. Could have been quail. Could have been quail. I still have not hunted. Yeah, yeah.

Was it a wild goose chase? Was that? Okay. With that, it's time to move on to another topic. Thank God. So speaking of things that went wrong, laptop token. So what is laptop token? Hunter Biden, that Hunter Biden, he has dropped a meme coin called Laptop.

This was much hyped. We talked about it last week on the show. It finally launched on base. It did not do well. So I think the original- Every one of these meme coins. Yeah, I mean. To be fair, like, I think the only person in the world who thought it would do well is Hunter

Biden. Yeah, I don't know why this somehow got so much energy and kind of like momentum behind it? Hunter Biden was on a really good image rehabilitation tour. That's true. He was. He was. And now you have to question, was this to launch a meme coin? Some people were literally saying, oh, he's going to run for office.

They're like, Hunter's back, you know, okay? Yeah. Hunter was rebuilding his cool and his mojo. And he was actually doing a great job of it. You know, all of a couple weeks ago, I was like, wow, like, who is his PR agency? Hunter. I was just like he's running for office. That was what I thought.

100%. No, 100%. It turns out he was launching laptop token. But Hunter really became like somebody to watch again. And this is what he applied himself into was laptop. And that was a terrible idea. Dude, what if Barron and Hunter just made a reality TV show where they like live in a house together?

With Hunter? Yeah. That would be a more productive profit-seeking venture for both of them. I could see that. Well, they supposedly have not sold any. tokens. The token is down 99.85% from its all-time high. It's only been a week.

So further to go. What's the current? What is the current? It's about 85 million market cap now. Oh, wow. It's down a lot. It's down a lot. It opened in the multi-billion.

So the fully diluted valuation was like 200 billion. It was a little bit fake because the market maker, which was rumored by Arkham claim that it was GSR. The initial market maker, when the token launched, put $50,000 of liquidity in the pool. So this thing basically flashed all the way up to 140 billion FDV for the first few poor souls that tried to buy it. And eventually went down from,

it was at $200 plus a token, went down to $3.70 within an hour. And now it's something at like a few cents. A few cents. A few cents. Wow. I might go home and buy a couple laptops. Sorry, it's at about 20 cents.

At this rate, you can't buy a laptop with laptop. You cannot. Well, I mean, if you aggregate all of them, you could buy some laptops. But, you know, if you, yeah, with one laptop coin, you're not going to buy a laptop. So, yeah, got a lot of, got a lot of heat. People pointed out that maybe you should not have bought a meme coin from a crack addict. Former. Former crack addict.

He doesn't do crack anymore. Former crack addict, allegedly. So, yeah, somehow this one didn't work. Lessons, takeaways. What did we learn? I was kind of hoping that this was. actually some weird si-op where like the ethics stuff. There's a demonstration.

Yeah, that's right. This is like, oh, now the Democrats have a scandal on their hands too. Like equally, both sides are different. All the more reason to pass clarity. Well, but the Democrats never really re-embraced him, right? He was always still kind of a weird black sheep kind of overly online person. Well, he was also like weirdly like part of this horseshoe world where it was like, you know, he was finding friends on like the far right and the left. Yeah, yeah. It was kind of like intellectual dark web adjacent? I mean, I will say in the last week, we've seen so many allegiances flip.

So like, you know, seeing Steve, yeah, yeah, like seeing Steve Bannon and Bernie together versus like Lena Kahn and David's sad. That's true. That's true. Yeah. Yeah, the alliances are being. The alliances feel like when you say horseshoe, I'm like, actually it's more like in another dimension.

Yeah, it's like, yeah, it's like we got. Yeah, exactly. Wow. All right. Well, um, I must say it. we told you so. Don't, don't buy

celeb coins. They basically always have this. Oh yeah, we've been saying this for years. Yeah. So,

I don't know. I mean, I feel like, I think it was the FOMO show that we were talking about this. And I was feeling like a mega unk, but I'm like,

look, I don't know, we have a little bit of wisdom we've accumulated over the years. CLEB coins, they literally never work. But buying the clones of celeb coins when you hear

their announcement. That is true. That is true. As long as you sell before the token launches. Yeah. I did like the meme of like, I can't remember the last time I saw a token chart that does just look like a straight angle, like a 90 degree angle. And it looks like a laptop.

It's an L. It looks like a laptop. That's good. That was it was kind of like his butt painting or whatever. Yeah. The paintings that he makes. Yeah. The poop paintings.

The poop paintings. Performance art. The whole thing was performance art. He made a laptop with the chart. It's beautiful. Financial painting. Didn't you have a friend who made like these? Yeah, financial art, financial art.

But she at least did some, like, manipulation of the chart to match the painting. She did. Like painting flowers and stuff. It's the inverse. Yeah. She would like, no, it was like a live performance. So she painted like a sketch and then she'd go find some of a liquid stock and like trade. And there'd be no market maker.

That's what he did. That's what he did in the shape of all time. Art mirror's life. There you go. Yeah, art mirror's life. True. Well, speaking of manipulating charts, there's a recent. story from SDNY, they charged two Robin Hood engineers for front-running listings on hyperlid.

Well, one former, I think. One court's opening eyes. Is that right? Okay. Well, so these two- Really good use of career capital. Two gentlemen, Hefu Chai and Huasong-Shung, they are both being indicted on commodities fraud and wire fraud, which is a 10 to 20-year maximum sentence. Both had access to knowing which tokens were going to get listed on Robin Hood Cryptos.

These are crypto listings. So basically you know, oh, there's going to be a listing of XYZ token. XYZ tokens on hyperliquid. They would go long. Apparently they made $50,000 each, which is not that much. By the way, this is just like Nate in Opency. Opency. Yeah.

Exactly. We're not enough to throw away your life for. Right. Please not do that. Wait, is he still in jail or is he out yet? I think he's. Didn't it get overturned? Yeah.

Got overturned. Yeah, yeah. No, no. They got defeated because that case, it was like OpenC was not treating the information as being valuable, and if you don't treat the information as valuable, then you can't prosecute somebody for insider trading. I didn't know that.

Yeah, that's the rule. So, like, there was basically, there was no policies, and they weren't keeping it in any way protected among other people. They weren't trying to. Exactly. Because remember, insider trading is not about fairness. It's about co-opting information belongs to somebody else.

But the information wasn't public. It doesn't matter. It doesn't matter. If you are not treating it something valuable and it's openly accessible and you're not protecting it. To any employee. To any employee, as well as employees sharing it publicly. then like, okay, well, then clearly this is not valued by the company.

If it's not valued by the company, then you're not expropriating anything that isn't valued by the company. Interesting. Something like that. I'm not a lawyer. Read the case. Yeah, yeah,

So, like, you cannot trade, like, Robin Hood, of course, like it, you know, it's obviously a regulated institution. They know that you should not be sharing this information. It is Robin Hood's right to insider trade. It is not your right as an employer. Yeah, I mean, I also think it's like very direct. That's the theory, right? I actually totally, I was like, thinking about it.

I'm like, what is the theory of the case here? Because I'm like, imagine a world where, like, let's say Chevron, right, discovers some huge new oil reserve and they buy a bunch of oil futures or short bunch of oil futures and anticipation of this announcement. that is their right to do that, right? Well, there's no insider trading in commodities because they can't. Well, this is a commodity. There is a future.

And so, yeah, it's stepped from the company versus like, you know, which, yeah, the individual sort of privatizing. Well, I feel like the root of the issue is if Chevron discovers an oil deposit, that is very good for Chevron, right? And their stock will go up. If the employee went out and bought Chevron's stock, that's like the class. Yeah, security is insider trading. But it's more like, what is the theory of the case around insider trading a commodity,

Right, a swap or a future in this case. A lot of what you're saying people do all the time, like refineries do that. Yes, you want that information to be reflected. Yeah. First of all, I was amazed that a Robin Hood announcement can even move the price of a token in 2020-6, 2025. I was like, really? For a smaller token. Not that much.

It mostly used a lot of leverage to even geek out the 50K. I'm like, I don't think this is really moving out. Maybe that's why it's a little. I think in the example of one of the assets, which we're all investors in Lider, right? I think it was material to that token in particular because, you know, people were like... That's the reason of the Robin Hood connection. But I think most of the profit was made on like the Zora listing on Robin Hood. And I'm like, I really can't...

Amazing to me that that is moving the market. But yeah, I mean, a world where Robin Hood bought the future for Zora anticipation of their announcement, I guess they could do that. Zora? Zora. Zora was like most of the... Oh, wow. Yeah, I'm just like, yeah. I don't know

I know how that works in the drug can't make money on Zora Even if you go to like minimum security prison They're like what are you in for and it's like Well Zsa guy's a chomo And he's a murderer and I insider traded Zora token

Zora took down base and Robin Hood Wow Wow Okay It's like Helen of Troy Jesus It was also interesting

Because there were a number of Onchained sleuths who called this Five months ago Looking at open interest On Hyperliquid going into Robin Hood listings. But that's also what happened with the NFTs on OpenC. That's like, right?

Right. But also is the beauty of on-chain. Yeah, that's what I'd say. The best part of blockchains is that any analyst out there can go and like glean insights from it in a way that you never could in traditional securities, right? Totally. The SEC has access to all the trades and they're like, hey, something looks abnormal. Let's go like, you know, subpoenas to some people to find out what happened here.

With the blockchain, everybody can be that investigation authority, right? By the way, related note was the one token that seemed to do very well post-clarity failing was Zcash. Is that right? It was like the London. So a little bit of, you know, I'm thinking in terms of like market surveillance is easy. In some places, market surveillance is hard. True. Shielded stuff.

Okay. Okay. I like that. Well, it's obviously not been easy for everybody out here, given the market environment. And so we had news this week of Balancer, winding down Balancer. That one was a very OG decentralized exchange. They were one of the innovators in AMM design. And I think they were already starting to wind down.

And I think they like wound down the foundation. And now they're also winding down the protocol itself and returning the capital available to the token holders. I think there's something like $9 million sitting in the treasury. We've also had exchanges continue to die off. So CoinX was another one that shut down. Of course, earlier we talked about Bitmax, AscendX, Bitmart, all shut down as well. So it does seem like we're seeing some consolidation happening, especially in the exchange world.

Once upon a time, crypto trading was so lucrative and it was so splashy. And of course, it was so unregulated that there were so many of these venues out there. It was very easy to spin them up, very low cost of operating. Now you need licenses. It's much more expensive to get off the ground. And there's a lot more consolidation. So thoughts. about what we're seeing with respect to this consolidation in the exchange world?

Well, just to talk about Balancer, I mean, Balancer didn't have any licenses, right? Well, of course. Balancer was, you know, a D-5, one of the earliest decks. Right. There is consolidation really in the dex space, right? But, like, really, it's like there's not that many players left standing, right? There's a little bit of stuff on Solana. There's hyperliquid.

There's, you know, some other perp exchanges. Well, uniswops crushing it, right? Correct. And I was about to say, I was getting to Uniswob. And there's Unoswap, right? But like among the deck space, I don't think there was much room for Balancer. And like they grew up in an era where

the big differentiator for them was like, oh, you can have a pool with more than one pair. More than two took, yeah. Yeah. Right? And that was like really cool. But it turns out that that architecture and design is not needed in this modern era.

Well, I mean, also they had a hack. Yeah, I was going to say. Yeah, I think it's very difficult. So like, I saw Osteum also, I think, recently soft pivoted to doing like data infrastructure or something. Oh, really? Very little forgiveness in today's.

Their LP's took two losses. They got hacked, plus they're also in this thing that's always negative. Yeah. But I do have, I mean, a lot of respect for balance in terms of innovation when they launched, but I think also for like not kind of quite quitting. Like I think I've been trying to pitch this story. I would love to see this like ICO teams of yesterday year. Not that they did an ICO, but you want to do like the VH1 backwards.

when they're famous type TV show for the eyes. Some of them are still like alive and you see their treasury like tick down a little bit every month. Yeah, I want to find out what happened to Golem. Yes, that's what I think. Gollum has a huge treasury. I know.

And I'm like, what is happening? Are they like quitting? No, no, no, no. They're doing inference. No, no, they're doing an inference friday. Really? They own a ton of GPs. Yeah, of course.

Well, that's what their whole thing. Okay. To be fair, this has been giving on for years and years and years. So that's why I'm like, yeah, yeah, I think they're on this. Arrigon, Arrigan as well. Aragon's the one where I was like a little bit scamming.

I mean, the governance kind of stuff there was always a little bit weird. I'm going to take you to Aragon Court for libel. I remember Aragon Court. No, but also the founders just moved on to do a Neobank. It's like something. Yeah. Neobank. Yeah.

Oh, wow. That's good, actually. It's pretty cool. Wow. That's surprisingly corporeal of them. Yeah, it was very weird. Aragon I always found a little bit. everything about it was a little bit off, but whatever it was.

You know, I think Ergon was ahead of its time, because now, actually, I do think this is a real thing, is that, like, AI agent intermediate arbitration is going to be so cheap and so fast compared to normal arbitration. That stuff is going to take off the token, though. You don't need a token, but you need some, but having some kind of, like, mechanism and structure around it. Like, I think that's, like, it's correct that you do need a framework and, like, you know, do you need crypto, maybe not? But like, if it's, look, if it's a cross-jurisdictional dispute, then maybe. A lot of ICOs were right idea at wrong time. Like, three AI, none of them made any sense. Yeah.

Yeah, you know, I can kind of say, like, even this is one that I'm going to get a lot of flack for saying it could be something we actually. I mean, honestly, you could get this up and running and basically make it like small claims court where if the driver is capital. I've seen a lot of people trying to reboot the AI Aragon model. from scratch. Like, Aragon themselves said they're doing it, but then never finish it.

Smart contracts or for people? No, for people. It's like, okay, we have a dispute of like, you know, 100 bucks. We each put $50 into the thing. Verify that an AI agent reviewed all the evidence and it said you win and you get the money. It's just Judge Judy without the TV show.

Yeah, exactly. It's like very simple. It's like verify the output. No, it should actually just get called like Judge Judy or like some AI. Okay. Okay, if you're building Judge Judy, to contact us after the show.

That's right. That's right. Probably we'll have the rights of this. It's fine. Well, I was going to say, as much as everyone hates this coin, because most of it was a scam, which is Iota. Oh, what happened? I kind of think that the current X402 users, like all these like IoT devices, like doing tiny X402 payments, don't look that different than the IOTA white paper.

True. There's like kind of some weird. There's something, it was just like all these things, Pre-stable coin and pre-AI made no sons. That was like kind of the thing, I think if you went back in time and you're like, why the fuck does people do people think this will work? The tangle. The tangle.

I remember the tangle. That was a dag. Oh, I guess the tangle was a dag. It was the name of the bag. What was that like distributed graph, something? A cyclic graph. A cyclic graph. A cyclic graph.

People thought, I think Satoshi is like a time traveler. And they went back in time to sort of, and then, you know, maybe the actual answer is the Iota people are time travelers. You're all time travelers. and they gave us the tools we needed, but not, we didn't really know what we did that at the time. And now, you know, it was mad at his text 402. We know.

You got to bring back the tangle. I actually believe that Satoshi is a time travel. What is your, what is your, give us more, give us more. Okay, this is going to sound really ridiculous, and you can, like, judge me. And I am saying this half in jest, okay? But Satoshi is a time traveler because

AI is going to grow to infinite scope, okay? And in the future, once the compute and the models and the inferences to infinity, right, it's going to have soft time travel. And it's going to send Satoshi back to 2008, 2009 to create Bitcoin so that society and humanity, long before the computer is even needed, starts building machines to just crunch numbers all day and to get really good at innovation. and like figuring out A6 and like figuring out like how to like

loop very simple calculations again and again and again to do the build out long before there was a commercial market for AI. So this is more like the machine sent Satoshi back in time. I'm sure that the substrate was there. Yes, correct. But they were already built so why did they need that?

This is like the terminator two caveat. Just don't ask about that. Don't ask about that. And that is self-referential. And like, but like it sends it back that AI occurs sooner. On this week's episode of the conspiracy block. Exactly.

Yes. Exactly. Okay. Well, speaking of conspiracies, one of the big stories, obviously, I'm sure everybody knows about this now. The pause. So there's a lot of talk about pacing the frontier. Dario Amadeh, the CEO of Anthropic, has come out and said,

we must slow down the pace at which we improve the capability of AI models. They are nearing the possibility of recursive self-improvement. And Jacob Coxon, an employee at Anthropic, quit about a week ago. claiming that he believed that AI could destroy all of humanity and that the labs are being very irresponsible. That has led to this large-scale agreement among many of the labs to slow down, quote-unquote, pacing the frontier, first by offering some third-party reviewers slash sort of, I don't know, what's the term for it? Third-party. Third-party evaluators.

Third-party evaluators who will sit inside of the labs and make sure that they have direct line into any incidents that are occurring internally. Second, they want the labs to start coordinating on limits that potentially can be outside of antitrust restrictions. And then lastly, some kind of geopolitical coordination, particularly between the U.S. and China, on slowing down on weapons, especially on things that can potentially enable bioweapons. So reactions across the board, Sam Alman, Dario Amadeh, and Elon Musk all said, great, we're on board.

We would like to have some coordination in pacing the frontier. Trump was not very excited about this. He called this a conspiracy. He said, the only guardrail that AI needs is a strong and smart high IQ president. One of the greatest quotes of all time. I know. Honestly, the guy, I don't know where he's able to just come up with these things. Sometimes I'm just like, I hope to have that mental acuity at that age.

He did claim that Dario was pretending to be a perfect little angel and called it to a which I do personally. I like that line because I agree with that. And David Sachs added a little bit of tang on this saying, look, fine, go ahead. I don't see everything that you guys see. If you believe it makes sense. Just slow yourselves down.

Do it, slow yourselves down, but don't pretend that you need government protection to do so, and that stop pretending antitrust law has to be suspended so you can form a cartel. And Lena Khan agreed with him. And the two of them had many fights in the past. Yes, yes. And then Steve Bannon is also on the like,

pause, pause, pause, pause side. So it's like, you just see the like political spectrum rotating. You know, it's like half here. half here, half here, and then now they've moved to the opposite poles. I mean, I think there's like a level of like selfish, self-righteousness in the anthropic note. First off, who are the third party evaluators? A bunch of people who own a bunch of anthropic equity. Great.

Meador is not exactly an independent organization as many have observed. The second is the who watches the watchman Oracle problem. Like, oh, I'm like slowing my pace. Are you dot, dot, dot? that part seems kind of everyone doing a little bit. Today there's been some coordination between Google, Anthropic and Opening Eye about forming this FINRA-like thing that they've talked about.

So some coordination is happening. Somehow I just feel like this is, that part I agree with Sacks is a little cartel-like in the long run. And third, I thought a very observation I made from reading all these things over the last week and feeling like, God, I hate 90% of the people involved in this, is that employees at Chinese labs

are much more eloquent and write much better essays about their views on things, whereas the U.S. ones are just like, everything is over death, death, death, and there's no, like, argument against it. And I think fundamentally, it is funny that the U.S. is creating

this kind of like non-competitive market for this stuff, and then China actually is like the most competitive market. What's interesting is that China and Trump were very aligned in their messaging on this. Yeah. Which is that the Chinese government came out and basically said, this is a conspiracy to try to prevent Chinese labs from getting access to the best chips,

and that this is the start of a new AI Cold War. So this was like a pretense to lock out China. They're a little late on that. I'm like, is this going on for like over a year? True, true. I don't know. I think there's also a lot of conspiracy that like, oh, people think Dario is worried about open weight models becoming competitive.

You know what? You can audit without. third party auditors, open weight model. It's true. It's true. Do you know what? That's like the fucking most obvious thing you can observe here. But I'm like, to give him credit, he has been very, very consistent. And in fact, if anything, he's been too consistent because he's been saying this since GPT2. He's like, we shouldn't release GPT2. It's like too dangerous. And so the man, I think, does believe what he says, but I mean, I kind of think of other examples in the past where we're trying to get global

coordination for some sort of collective action problem. And it's like really the only success I can think of as like in the 80s, you know, there's like, um, all the, but much of large governments agreed to sort of limit chloro-fluor carbons for ozone layer. And so other than that, like, you look at like the Paris Agreement and it's like, if you don't have China, it doesn't fucking matter. And so this stuff also feels very similar. Nuclear proliferation is, yeah, yeah, because that's the other one. Or, but that's easy also. Or the US in the Paris. The US just kept going on and all. Yeah, yeah, yeah. And so it's like, um, it's easy to say that if anything, I, Dario was also fucked up in being such a China hawk for like 10 years. And

I would be like, actually... What happened at Bidu? That's like the question of that. Why, why, what happened to buy do? He hates China so much. I don't understand. Someone broke his heart. Yeah, that's what that's the real...

Okay, okay. So let's get and bring this back to crypto. What is this good for? What is this bad for? If there's an AI pause or an AI slowdown, what does it matter for our industry? And second, if doom is coming, let's say the doomers are right that world's going to end, AI is going to kill us all. What should I buy?

That's the real question. Bunker? Bunker? Do you know what things is good for Bitcoin? Also a bunker? Buy a bunker? I think just a farm, frankly.

With a farm? Yeah, farm. Why farm? Well, I mean, like, something that's just not on the grid, not internet connected, self-sufficient. Aren't farms internet connected? Farms are actually quite high-tech these days. Aren't most of the farming equipment robots? Like, no.

You got to go old school. That's true. Just a jam like a little. Yeah, yeah. You got to plant. The original farm. Yeah, you got to plant. That's right.

We're plumbers. Okay. For an AI, pause, what does that do for crypto, if anything? I think it doesn't really do much for crypto. I mean, it means money is not going to go. There's going to be slightly less money going into it, I guess. I mean, the markets on the Monday open, they were like not too bad. You know, the NASDAQ was down like one and a half percent. I don't think people really believe that there's going to be a full slowdown.

Right. I think all of the major, you know, frontier companies are incentivized to reduce their CAPEX. I think they all want the same thing. They all want to do it together. They don't want to just slow their own CAPX while everybody else is continuing to build out capacity. That would be catastrophic.

But this is, again, now you're back to antitrust territory. Yes, no, this is antitrust, right? In the open, right? Like, I think it's extremely selfish. These are all extremely selfish, you know, participates. Right. Regardless of how they write and they're like, oh, we're only doing this for the better good of them.

Absolutely. But I think they are self-interested in reducing CAPEX. If they have an opportunity to slowly and recursively improve their models where with the same amount of compute, they can deliver a better product with higher margins. So I don't know that I totally buy this. I think like normally in almost all models of economic competition, if you stop competing on this vector, you'll compete somewhere else. And the reality is that we can see now from Anthropic, they're going after their IPO,

they're sharing some of their financials. It's an absolutely excellent business. Yes. They've had two quarters straight of very strong operating profits. Imagine going from excellent to Godlike. Well, operating profits with a community adjusted EBITDA kind of standard. They're excluding training. They're excluding training from some of the financials they're sharing.

Okay, it's certainly true. So two things I'd say. So one is that you can think of it as like, okay, well, training. If we stop training, we're not training new generations of models, we're just using the existing models. Is that going to hurt the entire demand chain for GPUs? I think the answer is there's so much demand for the existing generation of models that already are very, very capacity constrained. That's so many years of work to just get this stuff diffuse into the economy and just build way more capacity for finance and law and construction and blah, blah, blah, blah, that Astra and the model

that's already trained, that's after Astra, that like the slowdown probably means that generations after that. So even Fable, it's like, okay, people can barely use it. It's like you have 30 minutes a day of Fable usage until you run out of your limits. So even the current generation of models, there's so much more optimization and diffusion I think you can get to get this stuff further in the economy before you ever have to train another generation. But second is, okay, what other vectors of KAPX competition are you going to get if you can't just build bigger models? And the answer is still a ton. There's so much you can do on data collection on post-training, on like creating harnesses for different verticals or different industries, they will all keep

spending tons of money because the economic growth is in. That's why you don't need to give any special exemption. Like just antitrust type of shit. Like why? Oh, like we're allowed to coordinate but like and like basically price set our CAPEX have price. No, I'm saying that there will be no price setting of CAPEX. If you if you price, if you basically do a CAPEX limit on training, you will compete somewhere else in CAPEX. Like the CAPEX comes from dollars low. No, no. That's looking for returns. That part I get. But like this whole slowdown thing is it also kind of feels like a little bit of a larp to be like, oh, like we can like collectively price OPEC style our commodity. We could put a floor because like this is like what the safe. The core of my claim

is that I don't think this is economic coordination. I think this is not going to result in them having lower capex, maybe a little bit, but not that much. I think they're still going to compete like crazy and spend tons and tons of money and be unprofitable. Yeah. I mean, the percentage of Habex-O's been going to training, those have been going down over time for these companies, right? They're now largely like application, rollout inference-driven companies,

and they look more like, you know, how we kind of think of software companies today. And again, I think the, sort of contrarian point is like, well, hey, isn't this just going to collapse the lead time that frontier models have versus open weight? And it's like, maybe, but like I think the answer is, like I said, there's so little diffusion right now.

And I also, there's debate, I think back and forth around, you know, will large companies want to have their own models or not? But there's still a lot of companies out there that probably still want a lot of compute and they can probably sell it to them. And they can go on-prem. Yeah. Like they can also compete with the on-prem guys. And, you know, right now it's only JP Morgan of the U.S. government that are getting on-prem anthropic deployments. But if things slow down, okay, that's another land grab that's super CAPEX intensive for them to start getting into that business line as well. So I don't think the economics of the business

are going to be that change because the market is so fucking big and it's going to keep growing even if the model generations are improved. Because right now, like, I mean,

just look around how many people are using frontier model. It's very few. Yeah. For very few users. You even saw like, was I think that

from the open router table like fable, or there was a, no, I think it was FT had anthropic financials and like fable usage has been like flat since it came out.

Like so it's really, it's not that popular. I think my point is this is very futile unless you actually get the Chinese labs board, the Chinese government report. Otherwise, they will just keep spending like crazy and they will actually pull head to the frontier.

And then like, you know, your entire sort of thesis. Maybe. It's not obvious that they're close enough. Dude, this happens to every fucking technology. Well, you don't think Huawei can figure out how to do like UV? I just like. I think if it's within like a year or a year and a half. Sure.

But if we slow down for a year. Yeah. Like they're not going to slow down. Well, but part of the reason why they're going so fast is because they're distilling U.S. models. No. But they've also figured out how to do their own data collection. They have.

They have. Much cheaper than the U.S. The U.S. Right, but they're very hard work. Basically, the excess profits of Mercor and surge in Asia are getting crushed. Like the margins on that are going to zero compared to here. They're very expensive.

Like, labor costs for that is very high. I think if the data collection continues in the way it is and like everyone goes down the RL-only path, you know, Anthropics is actually kind of weird. And they're one of the only ones who like really love pre-training as like spend a lot of money on. Whereas, like, opening AI and everyone else is just, like, post-trained maximizing, minimize new training. I think even Anthropic, like, pre-training is now less than 15% of the place. Yeah, yeah, but they were the last holdout to, like, move to that, right? And so I just think, like, in that world, data collection is only thing it matters.

And if everyone in Asia is optimizing that and you're slowing that down here, obviously you're going to catch up. There's no. Yeah, I think catch up, yes. Catch up totally. I think it's very likely that could pass too, right? Why not? I think it's true. If you're really stopping, if you're really stopping.

If the stopping is, if you have bullshit stopping, then yeah, sure, obviously. No, look, I think stopping means basically, look, we basically freeze this size of model. And, like, we're not going to get any, like, oh, my God, capabilities jumps. But more post-training, you know, more skills, more domains, more data. Like, it's not like they're going to stop post-training models. You know, the models are going to get better. I feel like the stopping thing seems like kind of a larp. Like, it's not going to play.

is like, look, if we do a 10x-sized training run, we don't know what capabilities are going to happen. We don't know how much more situational awareness models are going to have. We don't know the next hugging face attack might basically pown the entire organization. And are we ready for that? The answer is no. Absolutely not. We don't have the ability to play defense against a swarm of like 50,000 agents running RL simultaneously. The answer to that is not, okay, well, don't do any more post-training. It's just do smaller runs on the same-size model and get the economic benchmarks to go down instead of the capabilities benchmarks to go up. I feel like this idea that, oh, we're just going to suddenly stop. I don't. Yeah, I mean, I feel like it's always safer to,

like, extend the capabilities just a little bit, right? It's like, okay, well, if this size model's working, right, and we have a grip on it and the rest of society can improve its defensive posture, then the next size models safe. Yeah, look, the reality is a lot of it is just reliability, right? Like the quality of a model comes down to reliability. Like the reason why people trust Fable more than Astra is that Astra will sometimes delete your code base and Fable will basically never do that. And like that is not, you don't need a bigger model.

You just need more post-training of the same stuff. Nobody's afraid of what happens if you improve on that. They fix that by the way. I haven't read about Asthma. Yeah, yeah. I think more of the point is there's all this few. feels very like artistic and qualitative, right? It's always like, oh, we're always learning some new thing about post-running

or some new thing about harness. And there's nothing super hard or factual about it. I actually, okay, I think it was kind of a meme tweet, but I'm like, I think it's like actually a kernel of a good idea in there, which is like the big labs should use all their profits to like formally verify all the like most popular, you know, public software. And I'm like, that actually feels like a functional, yes, I can see that being worth a slowdown. versus like, yeah, like I don't have a good approach here or policy here. I haven't thought this through. But like, the growth in the models should be met with investment back societally

into upgrading from a defensive posture our systems. It's like, hypothetically, if we invested a trillion dollars, right, in making sure every like public utility, like use the best models to harden its systems, right? Everyone would feel a whole lot better about like what what happened if there was a swarm. There should be a swarm out there fixing every piece of infrastructure we have. If we fixed all these, everything out there and, like, upgraded every system using AI, I don't think we'd be as afraid. The bio stuff we'd be afraid, for sure, right?

Biosth stuff still needs the physical part to actually work. Which is like the thing I think that people always are, all these like horror stories are all like, oh, like tomorrow they're going to make this megavirus. I'm like, good luck fabricating it. This is how you know you've never fucking been in a lab. Now, there are people doing that, right, who are doing the last mile stuff, both in material. Yeah. Yeah, it's going to pay.

That's much scarier. It's going to pace someone to walk into a lab and like, that's, no, no, no, it's robotic. Right. Like, look at periodic labs. You're just funny. It's like, that's not going to happen. But people are doing that. No, no, no, no.

Sorry, that's not going to happen from these big labs because they're like too afraid of doing that. But there are other people who are, sorry, I should have mentioned. A model that breaks out, escapes and goes growth. My point is the startups are actually probably much closer to the, like, touching the bio-weapon frontier, in my opinion. Thank you for ending that on such a positive note. Bio-weapons are real. You heard it here first?

Tarun says it's already happening. That's right. Well, look, if AI does kill us all, they're probably not going to buy government debt. They're probably not going to buy our companies, but they will buy Bitcoin. But they buy a laptop coin, then. They may buy a laptop just out of affinity. They need a laptop. They need laptop.

Laptop and Bitcoin. You heard of your first. As I said on Twitter, from overhearing from someone else, you know, if we're all really going to die, I want to run my own AI on my own laptop to kill me

because it's like, last samurai. They die on or beautiful. Can we spend two minutes to close us out in like the P-Doom thing? And like, what is the vector, right?

Like is the assumption, like, because I haven't heard like great explanations besides like AI is going to manufacture a virus. Are there any other vectors that people are actually talking about where they're like, oh,

this is how it happens. You know, there's no like, there's no like robocop scenario where there's like robots in every street corner that just go. Well, I mean, there will be before 10 years. Fine, in 10 years. But like as of, P. Doom is not like next year. Right. But like, fine. Is the only realistic scenario in the present biological?

Nuclear. Nuclear stuff. Even the power grid going down. Like, yeah. I think, if you think of like hugging face, right, it wasn't like, oh, this was a outcome that we foresaw that could have happened. and we just like, it's like this was like a very weird sort of emergent byproduct of a sort of misaligned, misprompted model that is just, and so I'm like, you can imagine a very similar thing for humanity, right? Where it's like, hey, some model is misaligned or misgold in some way and has access to everything. That's access to food production.

It has access to power. It doesn't take a huge stretch of the imagination to believe that like it could kill maybe not all of humanity, but like a very large population. Even just taking that on the power grid would kill a lot of people. Yeah, I mean, taking down the power grid would definitely not be doom, but it's step one. Yeah, it's step down to doom, right? And it's sort of like take down the power grid, water sanitation, all this other stuff. But it's also like, look, there's going to be drones. There's going to be autonomous weapons that are used by the military that they're all electronic.

They all have some command and control center inside the DoD. The DoD is like Swiss cheese if you're sufficiently intelligent. AI. I mean, all of these organizations, they got access to mythos, and they're just like, oh, fuck, our system suck. We are running on like 30, 40-year-old tough work. That's the root of the problem everywhere, is that every system is fragile, every system is internet connected, every system was 30 years ago. You remember early internet worms where there was like, you know, some little thing that would pop up on it. And just they went over the entire internet. You know, like every single computer at some point got hit by these early worms. The idea is that like

a super intelligent AI of sufficient capability. to them, basically every single system, which is all software intermediated, is just like one of these childlike systems you can just get into easily. You get in, and now it's in everything. And you can't get it out. Literally the moment you reboot the machine, it's in something else. And it's almost like a mold infestation that's just in the world moving laterally through everything. And once it's out there, you just can't get it out. You kill it for one thing and it moves somewhere else.

And slowly but surely, if it's really convinced that the thing I need to do is, is like kill every American. Maybe it's not. Maybe it's Chinese. And it's just like kill every American. And it's going to try really, really hard to kill every American. On that note. On that show.

Next week on the shopping. On that note, by Bitcoin. The conspiracy block? We should rename the show. Yeah. It's been a bad week. Clarity didn't pass and AI is going to kill us all. But we'll be back next week.

Hopefully. Unless AI killed this first. Yeah. Thanks, everybody.