The Chopping Block: SpaceX IPO Mania, Fable 5 Export Controls & The AI Privacy Fight

Summary

Unchained – The Chopping Block: SpaceX IPO Mania, Fable 5 Export Controls & The AI Privacy Fight

Date: June 18, 2026
Host: Laura Shin (absent; episode moderated by Haseeb of Dragonfly)
Guests: Robert (Superstate, crypto connoisseur), Tom (Defi maven), plus other Dragonfly team members

---

Episode Overview

This rich, energetic discussion breaks down three timely and interconnected stories: the historic SpaceX IPO and its implications for both traditional and crypto markets, the abrupt imposition of U.S. export controls on Anthropic’s Fable 5 AI model, and the growing privacy and decentralization struggle between governments, tech companies, and users—especially as AI and blockchain increasingly intersect. With sharp analysis, casual banter, and plenty of hot takes, the panel draws parallels between crypto and AI, questions who really benefits from financial innovation, and speculates about the future of global technology and regulation.

---

Key Discussion Points & Insights

1. The SpaceX IPO: Crypto Meets Traditional Finance

The Largest IPO Ever, and Its Many “Crypto” Elements

• Discussion opens with the monumental SpaceX IPO, briefly flipping Amazon and Microsoft to become one of the most valuable companies globally, despite being early-stage and unprofitable.
  • “This is, unlike those companies, not profitable. Very early, kind of speculative story... But a company that's getting a lot of people very, very animated.” — Haseeb (03:36)
• The IPO featured a tiny 4.2% float—extremely low even for household-name IPOs, creating unique market dynamics (“low float, high FTV IPO”).
  • “Even for a brand name stock... 4.2% is extremely low.” — Haseeb (03:36)
• Retail access was broadened (minimum buy-in dropped to $2,000 on Fidelity), yet most participants didn't get full allocations—crypto brokerages like Binance and ByBit got completely cut out.
  • “...most people did not get their full allocation in the IPO. So retail got between like 10 to 30% of their allocations filled depending on the platform.” — Haseeb (04:12)

Perpetuals and Market Structure

• Massive activity in space-oriented perps (perpetual swap contracts) on platforms like Hyperliquid and Trade XYZ. Perps essentially “priced in” the IPO outcome almost exactly.
  • “Incredible predictive powers, that’s right. ...Monday, HyperLiquid on Trade XYZ did $1.4 billion in volume on just SpaceX.” — Haseeb (06:07)
• Hyperliquid and its integrator, Trade XYZ, now see almost all the DeFi-style equity trading volume, effectively crowding out smaller competitors.
  • “If trade has like the best equity perps. Why would you use the second best equity? It will all concentrate to whatever has most liquidity, most volume...” — Robert (11:27)
• Market dynamics evoke parallels with “winner-take-all” phenomena in DeFi protocols and with Coinbase’s approach to new asset launches.

IPO Mechanics: Crypto-Style Financial Engineering

• Much was made of SpaceX’s ability to acquire companies (Cursor, in this instance) using newly issued, “marked up” IPO stock—analogous to using tokens as acquisition currency, like BNB or FTT.
  • “But this is kind of like when CZ acquires companies with BNB... This is the crypto playbook. Like this is actually the closest thing to crypto that I have seen.” — Haseeb (15:57)
• Elon Musk’s moves seen as both “genius” and reminiscent of DeFi’s SPAC and token games.

Unlock Schedules and Hedging

• The float will rapidly expand: a “highly atypical” aggressive unlock schedule sees 20% of locked-up stock potentially hitting the market within weeks.
  • “Company had unbelievable negotiating leverage. ...Like, the VCs, the founders, the whole cap table—they’re locked up for 180 days. In this case, there’s a pretty aggressive unlock schedule...” — Robert (21:31)
• Debate around whether insiders will or can hedge via perps and forward contracts—comparing crypto and IPO hedging structures.

Democratization, Regulatory Frictions, and SPVs

• The hosts discuss the enduring debate about accredited investor laws: most appreciation occurred in private markets and was inaccessible to most people.
  • “All that appreciation, except for the last 40% happened in private markets.” — Haseeb (28:00)
  • “Even if you allowed unaccredited investors to invest in private companies, they still won’t get access." — Robert (28:38)
• SPVs (special purpose vehicles) and “dentist allocation” lore: plenty of tongue-in-cheek banter about layers and access games in secondary private markets.

---

2. Coinbase Product Update: Tokenized Stocks and More

Major New Features

• Coinbase rolls out tokenized stocks (with dividends and voting rights), options trading, thematic indices, and an SEC-registered AI investment chatbot.
  • “[The AI advisor is]...like an SEC-registered chatbot that I guess you can sue if it gives you bad legal or bad financial advice or something.” — Haseeb (43:06)
• Jokes about the AI bot’s recommendations (“Your entire portfolio at the call!”), and speculation about liability, compliance, and the real motivations behind Coinbase’s moves.

Tokenized Stock Promise (and Shortcomings)

• The dream that tokenized stocks would make IPO allocations available to the public is, for now, mostly unfulfilled.
  • “The conversion event sucked... all the platforms that were trying to do it failed for the most part.” — Robert (39:29)
  • Describes how logistical hurdles, oversubscription, and opaque allocation pipelines frustrated platforms and investors.
• Ongoing skepticism that tokenized stocks will meaningfully solve access inequality without significant regulatory shifts.

---

3. Fable 5 Export Controls & The Case for Decentralized AI

Fable 5 Launch and Swift Export Controls

• Anthropic’s powerful new Fable 5 model was abruptly hit with U.S. export controls after Amazon’s Andy Jassy flagged a security “jailbreak.”
  • “[Export controls] is a strict liability kind of thing.” — Haseeb (52:44)
  • “Andy Jassy apparently called Dario. Dario did not pick up... then he called Secretary Besant and basically told Besant... we should turn this off.” — Haseeb (53:00)
• The “jailbreak” cited was using Fable to find and fix code vulnerabilities—potentially weaponizable.
• Anthropic was forced to pull the model even from domestic users, lacking necessary KYC/controls to comply.

Decentralized, Permissionless AI as a Narrative

• The panel notes that this is the first significant U.S. act of AI “censorship”—and that it prompted a mini bull run on “decentralized AI” projects.
  • “This is the first large scale censorship of AI we’ve seen outside... China, more or less.” — Haseeb (56:46)

Crypto-AI Parallels

• Extensive discussion of the parallels between AI’s privacy, export controls, and government overreach battles, with those that defined early crypto.
  • “I am struck by how many weird parallels there are between the crypto industry... and what’s happening in AI, even... export controls and the original cypherpunk movement.” — Dragonfly guest (63:03)
• Anticipate legal/cultural fights over user rights, customer data, and “third party doctrine” as people treat their AI companions as intimate confidants.

Open Source & Global AI Dynamics

• Chinese labs are now world leaders in open-source AI capabilities, often matching or surpassing U.S. models at vastly lower compute/data budgets.
  • “All the American open source models are just absolutely terrible and yeah, China is completely eating her lunch with it.” — Haseeb (66:13)
  • “Chinese electric cars are a million times better...” — Tom (67:47)
• Predicted scenarios include government bans on serving Chinese models and splintered access based on user identity/nationality.

---

Notable Quotes & Memorable Moments

Insights into Markets & IPO Dynamics

• “This is the crypto playbook. Like this is actually the closest thing to crypto that I have seen.” — Haseeb (15:57)
• “Time. That’s what it will maybe [take for a ‘non-crypto’ asset to flip Bitcoin on Hyperloop].” — Robert (07:08)
• “If trade has the best equity perps... why would you use the second best?” — Robert (11:27)
• “He has basically a monopoly on moving hardware into space.” — Robert (20:29)

Governance, Hedging, and Regulation

• “Crypto people are about to find out about... getting dumped on your head and someone being able to hedge.” — Tom (26:15)
• “Most people did not get filled. You didn’t get filled at all?” — Haseeb (27:39)
• “Even if you allowed unaccredited investors to invest... they still won’t get access.” — Robert (28:38)
• “Do you think credit investor laws are good?” — Haseeb (30:08)
• “I think [accredited investor laws] are dumb... but it would not have changed the fact that SpaceX was still hard to get into.” — Robert (34:18)

On Decentralized AI & Privacy

• “It is the most powerful cyber weapon that humanity has ever invented.” — Haseeb (61:07)
• “This is a dry run of pressing a pause button... I am very supportive of the fact that there’s a little bit of restraint here while they continue to build models.” — Robert (57:02)
• “Most people’s intuition about AI is that this is my most trusted confidant... And if you’re like, well this is third-party doctrine, I don’t need a warrant...” — Haseeb (64:22)

Banter, Irony & Crypto Lore

• “[SPVs are] like, dentists selling to their patients. Dental harem.” — Dragonfly guest (30:37)
• “Grok is a piece of shit. Like, we can all agree no one’s using Grok... It is a pie of dog shit.” — Tom (18:30)
• “If you drop the bank, take the money and buy a lottery ticket and it hits...” — Robert (46:47)

---

Timestamps for Key Segments

• SpaceX IPO Dynamics: 03:02 – 24:30
• Unlock Schedules & Perp Hedging: 21:16 – 26:56
• Access/Accredited Investors Discussion: 27:25 – 39:29
• Tokenized Stocks & Coinbase Update: 42:14 – 45:58
• SBF, Anthropic, and AI Funding Lore: 46:04 – 51:13
• Fable 5, Export Controls & Decentralization: 51:26 – 66:13
• Chinese AI, Open Source & Compute Arms Race: 66:13 – 74:29
• Outro/Wrapping Up: 74:29 – end

---

Episode Tone

Casual, irreverent, fast-paced, and deeply insider-y—a blend of cutting-edge analysis, jokes, references to crypto and AI history, and the kind of practical skepticism and fatalism that defines successful industry operators.

---

Final Takeaways

• SpaceX’s IPO is a watershed moment for blurring the lines between crypto-native market structures and traditional finance, especially with regards to democratized access, hedging, and unlock schedules.
• Despite tokenized assets’ promise, entrenched inequalities in access and allocation persist—regulation alone won’t solve the problem.
• The AI sector is entering its own phase of government control vs. decentralization, echoing crypto’s early battles—and the verdict is still out on how “censorship-resistant” or truly permissionless AI will become.
• China’s AI open-source movement is increasingly relevant; American models lag in both cost and innovation.
• Expect ongoing, high-stakes legal, regulatory, and cultural fights at the intersection of money, data, and technology.

Transcript

[Robert]

So what I love about the chopping block is we don't really have to prepare that much. Not a dividend.

[Tom]

It's a tale of two point.

[Host (possibly Haseeb)]

Now your losses are on someone else's balance sheet.

[Guest (possibly Dragonfly representative)]

Generally speaking, airdrops are kind of pointless anyways.

[Host (possibly Haseeb)]

Unnamed trading firms who are very involved.

[Robert]

Alec Eth is the ultimate defi protocols are the antidote to this problem.

[Host (possibly Haseeb)]

Hello, everybody. Welcome to Hopping Block. Every couple of weeks, four of us get together and give the industry insider perspective on the crypto topics of the day. So quick intro. First you got Tom, the defi maven at Master of Memes.

[Guest (possibly Dragonfly representative)]

Hello, everyone.

[Host (possibly Haseeb)]

Joining us today we've got special guest Robert, the crypto connoisseur and czar of Super State.

[Robert]

It's great to be here. Thank you for having me back on the show.

[Host (possibly Haseeb)]

Yeah, it's really. We've been trying so hard to book you, but you are one of the hardest people to get in this industry.

[Robert]

Sometimes you got to live life.

[Host (possibly Haseeb)]

Yeah, yeah, that makes sense. But the people have been missing you. We've gotten a lot of people asking, where's Robert? What did you do to him? Is he still alive? Do you want to tell the people where you've been?

[Robert]

Yeah. Well, I was pre recording a season of the Super State podcast called the Pop. It has been recorded. We are releasing episodes. The second episode by the time you hear. This podcast came out about two days ago on Tuesday. It's all about IPOs and capital market machinery and the things happening under the ground when things happen in finance.

[Host (possibly Haseeb)]

Okay, beautiful. Good cross promotion. And of course, we've got to ruin your brain and grand poo. Bah ecology. And I'm Aseev, the head hype man at Dragonfly. We're early stage investor in crypto. But I want to caveat that nothing we say here is investment advice, legal advice, or even life advice. Please see shopping blocks at XYZ for more disclosures. So, Robert, now that you have betrayed us with another podcast, how does it feel knowing the taste of another podcast? And coming back to this one, what would you say you've learned about podcasting now that you've explored? You've had your salad days.

[Robert]

Well, I'll say this From n equals 2 instead of n equals 1. So what I love about the chopping block is we don't really have to prepare that much.

[Host (possibly Haseeb)]

You don't have to prepare that much. I have to prepare.

[Robert]

Haseeb has to, like, you know, prepare the monologue to open the show up. But like, we talk about the news and the things that happened. It's always fresh, it's easy, it's fun. Versus an interview style podcast where you have to like read somebody's book, you have to like, figure out what questions to ask, you have to like prepare. You've rehearsed a little bit and it's one on one. It's a little bit harder. It's a very different style. Shameless plug. Listen to the pop. You know, you'll get to hear me do a totally different style of podcast.

[Host (possibly Haseeb)]

Yeah. And if you want to hear Robert do absolutely zero prep. Listen to the chop.

[Robert]

Exactly.

[Host (possibly Haseeb)]

Just tune in right now. This is three guys doing absolutely nothing and one guy actually preparing.

[Tom]

Yeah. You have a staff.

[Host (possibly Haseeb)]

We have staff with staff.

[Robert]

What do you think the Dragonfly staff prepares?

[Host (possibly Haseeb)]

No, they basically have like an AI, just kind of flop.

[Tom]

AI. Can't have Chase though. You have to filter those.

[Host (possibly Haseeb)]

That's true. That happens up here. That happens up here. That's the cognitive work.

[Robert]

But this feels like home.

[Host (possibly Haseeb)]

This feels like home. Oh, that's. That's very sweet. Well, it feels like a homecoming. Well, speaking of home, we had the closest thing to a crypto ICO take place last week. Obviously the news that everybody's heard is the SpaceX IPO. Absolutely massive, largest IPO in history. And Crazy Pop obviously launched originally at price at 1.75 trillion and popped 20% on the first day, another 10% yesterday. And it's still going, it seems like. Or is it. Was it 20%, 10%?

[Robert]

It closed today at 200 something dollars.

[Host (possibly Haseeb)]

Yeah. So it is now. It's flipped Amazon, briefly flipped Microsoft, one of the most valuable companies in the world now. It's kind of insane. Obviously this is, unlike those companies, not profitable. Very early, kind of speculative story about what, you know, whatever data centers in space and so on. But a company that's getting a lot of people very, very animated. Now the other thing about it was the phenomenon of how the IPO was distributed was also extremely unique. So for one, the float of this IPO is 4.2% float. Okay, that means that this is a low float or. Sorry, yeah, low float. High FTV IPO. This is very atypical for IPOs. IPOs normally are like 15%, I think is the average. 10 to 20 is usually the range. Like I was looking at some historical stats like, you know, Microsoft, Facebook. Facebook I think was 13. Microsoft was like 15 to 20. So this is very low. This is very atypical for stocks and this kind of dynamic in an ipo Only really happens for like a household name company that you'd have a float this low. But even then, 4.2% is even then extremely low, even for a brand name stock. Now the other thing about this was that of course the IPO got syndicated to a lot of retail brokerages and I was seeing in some of the news, normally I believe like on Fidelity, if you want to get access to an IPO, you need to have at least 100k to 200k in your account. They waived the minimums for this IPO that you only needed $2,000 on Fidelity to participate in the SpaceX IPO. But even with that, most people did not get their full allocation in the IPO. So retail got between like 10 to 30% of their allocations filled depending on the platform. And a bunch of crypto platforms actually got entirely cut out from their allocations. So I believe Binance, Bybit and I'm sure several others, whatever broker platform they were using underneath basically got completely cut out from the ipo. So it's been crazy. Now of course the other side of this has been the perps. We've talked a lot about the Cerebras IPO. SpaceX was no exception. There was a huge amount of trading volume on hyperliquid as well as some of the other perps. Platforms that are on chain was basically pricing roughly a 20% first day pop. Now it opened lower than that, but it ended the first day almost exactly.

[Guest (possibly Dragonfly representative)]

It was quite accurate.

[Host (possibly Haseeb)]

It was, it was, it was accurate as of the day close. Although technically the perps converge as soon as trade trading begins. But it was, it was accurate in that the first day close was 161

[Guest (possibly Dragonfly representative)]

and

[Host (possibly Haseeb)]

the perp was implying 162 for the Open, which is very surprisingly Incredible predictive powers. Incredible predictive powers, that's right. And we've seen now even then, even I believe Monday HyperLiquid on Trade XYZ did $1.4 billion in volume on just SpaceX. No other asset, but just basics.

[Tom]

I think the craziest statistic, craziest thing to watch during the purpification of IPOs has been how concentrated HIP3 got during that time. I mean trade XYZ I think had all time high market share during like right before the IPO of all hyper liquidation. It was like probably the real question to me is like if SpaceX didn't flip Bitcoin on Hyperloop, what is the thing that will be the first. Non. Non. I guess there was, there might have been a Day where oil was like around the original Iran war. But I was close so I don't think it was flipped. But I'm curious like what will it take? Like if this couldn't do it.

[Robert]

Time. That's what it will maybe.

[Tom]

Yeah time or new users. Like there used to be a new wallet or a new kind of.

[Robert]

Well, you've got to assume that to new user growth most people are going to be coming in not looking to lever bitcoin and crypto native. So it's like I do think over time you're going to see more people coming in to trade. Everything is.

[Guest (possibly Dragonfly representative)]

When you talk about HIP3 concentration you're talking about like trade versus like Felix I think was shutting down and like benchmark. So they shut down.

[Tom]

So it's like this very Dreamcash was the only other one that. But Dreamcash actually did have some substantial market share for a while because they're really pumping the token incentives. But yeah, it's like basically a one trick pony at this point. I mean yeah, tip three should maybe should be renamed to trade.

[Robert]

Well, I mean eventually when there is too much concentration risk, like there's nothing that stops trade from leaving.

[Host (possibly Haseeb)]

Well there is something.

[Robert]

There's tons of network effects.

[Tom]

But dude, no one wants to run their own Oracle. No one's to deal with the builder code stuff. No one wants to integrate.

[Robert]

I mean I know but if you're

[Host (possibly Haseeb)]

big enough you can negotiate.

[Robert]

They get large enough, right? If they get large enough it won't be.

[Tom]

I think if they get large they're just gonna get large enough. They're just gonna negotiate some of the circle coinbase fees.

[Host (possibly Haseeb)]

Yeah, exactly, exactly. There's enough pie to go around. It's a little bit like circle. Right, Circle. When you get big enough and you're on usdc, it's not like oh, then I go launch my own stablecoin. That was sort of originally the USDH thesis was that oh, once I get big enough I launch my own stablecoin. No, the sort of cosian optimal outcome is that you just negotiate more of the margin to yourself. And I think the same thing will be true for trade XYZ is that they will just negotiate more of the fee for themselves and less goes to hyper liquidity. So as opposed to like them breaking off. That's the.

[Robert]

I agree this step one would for

[Tom]

sure be breaking off. Makes no sense. Their network effects are too strong. Like they're not going to redo all. I think the interesting thing about hyperliquid from this point of view is like it has a unique distribution model. The supply side comes via distributors, the demand side comes, well I guess by a one distributor, which is trade. And so it's sort of this sense of like I don't think you're going to just leave and suddenly get all those integrations back.

[Host (possibly Haseeb)]

And here's the other question that's not obvious. I don't know the answer to this is when people are trading on Trade XYZ, how many people are going to hyperliquid.com, how many people are going to Trade XYZ and how many people are going through some third party integration? Great question. If Trade XYZ breaks off, they lose all the other.

[Tom]

There are statistics on this because I'm sure there are. You can measure those.

[Host (possibly Haseeb)]

But do they have like some kind of.

[Tom]

The builder codes have to the builder codes. So a lot of the builder codes are de anonymized because you kind of can see your own. You can submit an order, see which one you got filed under and be like okay, this was the phantom one. So if you look at the builder code, there's a bunch of dashboards that show you like the builder code order flow and like leaderboard and there's only a few that are unwritten.

[Host (possibly Haseeb)]

Wait, so if you go in through hyperliquid itself, you're not tagged to a builder code. Yeah, I see. So if you go to trader experts you would get tagged to a builder code.

[Tom]

Yeah.

[Guest (possibly Dragonfly representative)]

But you can also trade Phoenix y z assets 3.

[Host (possibly Haseeb)]

Yeah, yeah. And.

[Tom]

And those all have their own code. Like that's how they do the revenue, the revenue share that, I mean that's how you measure the revenue share for all of them.

[Guest (possibly Dragonfly representative)]

Right.

[Tom]

So it's all public. So you can definitely.

[Host (possibly Haseeb)]

So trade knows how much they're bringing in direction on front end.

[Tom]

For sure.

[Host (possibly Haseeb)]

Yeah.

[Tom]

Which I do, I suspect. I mean I haven't looked in a while but in maybe like 3 months ago it was like not like I was looking around the time the WTI spike was and it was like not much from their own front end.

[Host (possibly Haseeb)]

I see. So they really.

[Tom]

This is what I mean, the network effects are at that point they really can't.

[Host (possibly Haseeb)]

Yeah, I think it's too hard to move. Yeah, that makes sense.

[Robert]

That makes sense.

[Host (possibly Haseeb)]

Interesting. So I mean we've seen hyperliquid pulling up to all time highs once again. It's been making the news. I keep seeing stories about space on Wall Street Journal, on Bloomberg. There's all these stories about hyper liquid trading for SpaceX in particular. And yet the other story has been the shutdown of these other HIP3 integrators is that Trade XYZ seems to be the HIP3 story. There is not this broader ecosystem. I mean there are, but everybody else is struggling. Who's not Trade XYZ or has just

[Tom]

kind of given up quiet quitting.

[Host (possibly Haseeb)]

Yeah. So it seems like the story for Hyper liquid is like hyperliquid itself. Trade xyz. But there's not this broader ecosystem of homegrown winners. How do you guys think about that? Was that inevitable? Is that surprising? Is it not surprising?

[Robert]

I think if you use like traditional markets as the analogy here, I think it's somewhat inevitable because liquidity concentrates into one market, so to speak. And so if trade has like the best equity perks.

[Tom]

Right.

[Robert]

Why would you use the second best equity? It will all concentrate to the whatever has the most liquidity, whatever has the most volume, etc. And so I could see something else succeeding in a different like niche if it's like, oh, Instead of like SpaceX and equity perps, we're doing like just commodities and like we're doing every commodity and it's like, it's funky and it's like we're going to try to win on oil and we're going to try to win on uranium or we're going to try to win on soybeans or whatever. But I think like within an asset, I think there's only one winner.

[Guest (possibly Dragonfly representative)]

I think it's. That is true. But I think it's also surprising that it's across every asset. Right. It's not like they only do equities. They actually do quantities too. I think there's also, it's like I kind of contrast actually like the builder

[Robert]

ecosystem, which Ventrals was a good like. Yeah, yeah, totally try to do something radically different. Yes, win on that.

[Host (possibly Haseeb)]

And then.

[Tom]

But that, but then trade just copied what ventrals markets that were like more liquid.

[Host (possibly Haseeb)]

So Ventures is doing pre IPO markets and now Trade XYZ themselves are doing pre ipo.

[Tom]

Yeah.

[Guest (possibly Dragonfly representative)]

Yes.

[Host (possibly Haseeb)]

Trade XYZ is just kind of gobbling up everything that's working.

[Guest (possibly Dragonfly representative)]

Yes. I think. And there's also obviously this, this capital barrier to entry. Right. Like you actually do need like I contrast like the builder ecosystem, which is permissionless. Anybody can build an app and like it's fairly like long tail distributed and there's like a lot of churn. Like okay, you know, one month it's base and then it's phantom and then it's MetaMask or whatever versus HTTP 3, which is that many people. And then obviously there's, there's more and more concentration and you don't have a sort of new generation of starting because it's a large investment to actually get started. And what's, what's your niche going to be?

[Tom]

Yeah.

[Host (possibly Haseeb)]

So you're not surprised by the second?

[Guest (possibly Dragonfly representative)]

No, I mean it, it also kind of makes sense. Obviously there's a close relationship between the teams and you know, like, so there's a certain amount of like professionalism and like minimum liquidity necessary to kind of get these things started.

[Host (possibly Haseeb)]

I mean we've been getting a lot of pitches from people trying to launch HiP3 Markets and it's kind of one of the big categories of founders that you see these days, which is people trying to build some kind of financial product on top of hyper liquid. And it has always seemed like, man, this is a pretty fragmented market. Like there's a lot of like small players trying to bootstrap things that doesn't really feel like the market demand is there. And it's like, okay, well why would Trade XYZ not just immediately copy this product if it starts to work at any scale? So it feels like actually a pretty tough environment for startups and on some level maybe not surprising if you are, imagine what's the equivalent of what people are doing on top of Hyperlink? Well, let's say that Coinbase offered a. You can launch a trading pair on Coinbase and you keep some of the trading fees for the market that you offer and then you market it and you create a front end for it and blah blah, blah. Concept like that doesn't really exist for exchanges, but one can kind of imagine if it did. That sounds like a terrible startup environment. Right. That sounds like that's really rough to build and it's almost like, well but if it's really working then Coinbase will be like, well here's our first party version of the thing that you built. That seems really good. And then they just kind of wipe the floor with you. And tradexyz increasingly feels like almost like pseudo first party, just given how closely integrated they are with the hyper liquid experience.

[Guest (possibly Dragonfly representative)]

And there has always not been competition or demand for alternative quote assets.

[Tom]

Right.

[Guest (possibly Dragonfly representative)]

Like that was also kind of an idea with like USDH and USD and maybe even other types of currencies. And there's like that I think has not really manifested either.

[Host (possibly Haseeb)]

Right, right. So the story increasingly seems like it's the hyper liquid kind of megazord that you're competing with and like that's what's really working as opposed to this broader platform story.

[Guest (possibly Dragonfly representative)]

Yeah. And people were also been talking about how hyper EVM is like basically dead. Like a lot of those teams have folded or gotten acquired and so it's really hypercore. That's like kind of, kind of having a moment.

[Host (possibly Haseeb)]

Yeah, yeah. So, okay, bring it back to SpaceX. SpaceX now they've got this 4.2%. It's run up like almost, almost 50% I think since the original IPO price. They just closed their $60 billion acquisition of cursor in an all stock deal and now it's.

[Robert]

Do you know if that all stock deal was done at the IPO price or at the current price at least?

[Tom]

Current price? No. It might be IPO because it was an option. Right?

[Host (possibly Haseeb)]

They have an option to buy an all stock deal.

[Tom]

Yeah, but I thought that was a fixed price. You have the option to buy at

[Host (possibly Haseeb)]

the 460 billion, I think the price of purse per share.

[Tom]

Yeah, sorry. Yeah.

[Host (possibly Haseeb)]

But basically it's like we can buy it whatever currency we have available, which is very marked up stock.

[Robert]

Brilliant.

[Host (possibly Haseeb)]

Yeah. Which is absolutely fantastic. But this is kind of like when CZ acquires companies with BNB or when use. What was ftt when Sam was using FTT to buy up companies and stuff. Like this is, this is the crypto playbook. Like this is actually the closest thing to crypto that I have seen.

[Robert]

Well, Elon's a crypto fanboy.

[Host (possibly Haseeb)]

Elon learned a lot of lessons. He learned a lot of lessons.

[Guest (possibly Dragonfly representative)]

It is crazy to, I mean, I feel like three, four years ago, like the headlines around Elon were, oh, he's Underwater in his X.com acquisition and this is such a blunder. And I was like, actually we're just going to scoop that in here and actually we're going to scoop that in here.

[Tom]

And like, I think, you know where, where Chamath failed, Elon won. He figured out to build the best spec. Yeah, he literally just like I almost view this as a rolling spec of each company back and forth in the private market and then finally, you know, going public gang, sort of. It is sort of kind of a genius move. I think it's crazy to coordinate that many investors who would be willing to do that and like.

[Robert]

But who's going to.

[Host (possibly Haseeb)]

I mean, you see why they're willing to do it. I mean look at this price performance.

[Tom]

No, no, no. My, my point is just more like. No, but like think you're thinking now. Right, but imagine you're like, it's 2022, 2023. And, like, everyone is like, okay, Twitter's in the shitter. We invested at this crazy valuation. Like, the debt. I remember, like, banks were trying to offload the debt that they underwrote for the thing at, like, 80% off at the bottom. X AI at that time seemed like this, like, fanciful thing. Like, there's no way he could actually get this much power, like, powered land and data centers.

[Host (possibly Haseeb)]

Right?

[Tom]

Six months and then it's like, actually, now everyone else has to kiss the ring. Like, everyone from anthropic down has to kiss the ring to get access to data centers because everyone else is, like, unable to get land, unable to get electricity, whatever, right? So there's a lot of. There's a lot of hidden genius here, you know, like, if you ignore all the politics stuff, it's like there was, like, all this crazy. Both financial engineering, but also, like, I don't know.

[Guest (possibly Dragonfly representative)]

Yeah.

[Tom]

Xai pulled off something very crazy where it's like, yeah, they're unprofitable and whatever, but they have this scarce resource that no one can compete with, which is someone has to pay you a billion dollars a month to get rent on your land, and there's no new land. You can keep increasing that rent up to their revenue. I don't know. It's a genius.

[Robert]

Well, that's the genius of defeat, right? He reserved a lot of compute and nobody came with the demand side. For Xai, Grok is like, whatever.

[Tom]

Yeah, Grok's a piece of shit. Like, we can all agree no one's using Grok every eval. It is a pie of dog shit. Yes. These Chinese companies fucking. The zipu model that came out today clearly trained on, like, one tenth of the resources. Grok. Grok is an embarrassing piece of shit. Like, the amount of resources. The amount of resources. No, no, no. Compare GROK to, like, the Chinese models. It is unbelievable how bad you it could be given how much compute they built for it. But the genius of Elon is like, all right, guess not working. I'm a reseller.

[Robert]

Yeah, reseller. He bought when the price was low, like, we were early in the, like, the race for compute. He bought a bunch of it, assuming that GROK was going to be awesome and everyone was going to use it. No one wants it, so he sells it at a higher price.

[Tom]

It's just more amazing. You could, like, turn hatred into, like, building that, you know, like the hatred of OpenAI. Just, like, driving to do that is wild.

[Host (possibly Haseeb)]

Agreed, Agreed. But like the total amount of revenue from that is what, like 15 billion or something?

[Robert]

No, it's like 2 billion a month. Now granted there's like 90 day like double sided cancellation provision so like he can cancel it and like the customers.

[Tom]

Yes. But the problem is all of these people have forward sold a bunch of stuff.

[Host (possibly Haseeb)]

Right.

[Tom]

And they can't actually build any supplies.

[Host (possibly Haseeb)]

So revenue, I mean it's great.

[Robert]

No, it's real revenue.

[Host (possibly Haseeb)]

Does that get you to a 2 trillion valuation?

[Tom]

No, no. My point though is more like he's taken these things that everyone wanted to write off as zero, bundled them together and somehow magically made the sum of the parts look like it's much greater.

[Robert]

Well, I think it's the opposite. I'll take the other side of this. I think when people are looking at space X, they're looking at it saying, oh, I also have to like buy like Twitter. I also have to buy xai. I really want the space business. I really want the data centers in space. I want the rockets. He has basically a monopoly on moving hardware into space. Like he has like the number one company on earth for this.

[Tom]

And like everyone knows Blue Origins rocket blow up.

[Host (possibly Haseeb)]

Yeah.

[Tom]

Have happened at a better time.

[Host (possibly Haseeb)]

That is true. That's great marketing.

[Robert]

He basically has a monopoly on moving things into.

[Tom]

Right. Also I feel like Bezos and Musk seem to be in this crazy rivalry. So Musk or sorry, Bezos must be fuming a little bit about this.

[Host (possibly Haseeb)]

That for sure. Amazon. Yeah, yeah, for sure.

[Tom]

And so also Amazon built a Starlink competitor, which is also crazy that a bunch of airlines use now. So Amazon, Amazon built its own satellite.

[Host (possibly Haseeb)]

What's it called?

[Tom]

Trying to remember. It's part of Amazon. It's like they, and so like they're really like. It really feels like some like evil villain movie, like a James Bond movie of like the two villains at the top fighting like.

[Host (possibly Haseeb)]

I see. Yeah. I mean I'd say look, we were just talking about this just before the show is that the float right now is 4.2% which is part of the reason.

[Robert]

Very low price action.

[Tom]

Yes. But unlike crypto, the, the, the inflation curve is like 90 days.

[Host (possibly Haseeb)]

Yeah. There's, there's not like one year lockup on this thing.

[Robert]

It's actually less.

[Host (possibly Haseeb)]

Yeah. So the first unlock, four years to full liquidity.

[Robert]

The first unlock comes atypical.

[Guest (possibly Dragonfly representative)]

Right.

[Robert]

Oftentimes When a company IPOs, the investors have a lot more negotiating leverage.

[Tom]

Right.

[Robert]

This is a case where SpaceX had unbelievable negotiating leverage. They didn't even have a price Range they said this is the price, it's $135 per share and we're raising like $75 billion. Go do it. Like make it happen. This is like so atypical. But they have an insane unlock schedule. Like normally a company can't sell any shares for six months, right. Like the VCs, the founders, the whole cap table, they're locked up for 180 days. In this case there's a pretty aggressive unlock schedule starting possibly at the end of July. It comes right after they release second quarter earnings. They can like start to sell like 20% of the locked up stock alons

[Host (possibly Haseeb)]

like they're 5xing the float in a

[Robert]

month and a half of the non key holder stock. So lon and the key holders lock themselves up for a year which is nice. Yeah, right. It's a good signal to the market right of the pie that that's not 20% of it come online.

[Tom]

Wait. Actually the funniest thing you just made me realize is crypto people, especially when CT is that it's like most everyone pvping and trying to shoot each other in the face is always like oh these VCs are like constantly hedging in the perp and like hedging their, their unlock whatever. But actually for these SpaceX people this actually makes sense. The funding rate is not that high relative to how long they have to do this. It actually makes more sense to use the equity perp to, to hedge your unlock risk than the.

[Host (possibly Haseeb)]

Well I mean there's so much volume on these things that. Yeah.

[Robert]

Like although legally if you are locked up you're not allowed to sell hedge

[Tom]

legally these people are not supposed to use hyper liquid.

[Host (possibly Haseeb)]

That's like, that's true. Yeah, yeah. I mean if you. But I mean if you're talking about some of the size. Yeah, but if it is.

[Tom]

No, no, no, I'm not saying that like I'm just saying like imagine you're a SpaceX employee. Unlock it.

[Host (possibly Haseeb)]

I mean imagine you're like, you know one of these VCs.

[Robert]

Yeah. If you're a regulated entity that's on the cap table.

[Host (possibly Haseeb)]

Yeah.

[Robert]

Like you're not doing that. Trust me.

[Host (possibly Haseeb)]

Yeah, like I could see some, I think also just the line work. Like I'm just. There's people who have like 50.

[Tom]

I don't think people have really thought about what the equity perp has created and that like in that the, all

[Host (possibly Haseeb)]

these things, I mean the open interest on this is like all the things crypto people.

[Tom]

Yeah, yeah. But all the things crypto people always complained about are like, oh, the VC is hedging. Hedging while they're dumping on my head.

[Robert]

Whatever you're saying the crypto equity, the

[Tom]

equity one is worse because of like how steep the unlock schedule is. So it's like you don't have to take that much duration risk on the perp to like hedge this huge unlock. Yeah. And I wonder if like market structure in the long term will have to. Equity market structure will be more like crypto where they force more gradual unlock. So you don't have this.

[Guest (possibly Dragonfly representative)]

This.

[Tom]

This might be the first experiment where people find out the hard way.

[Robert]

Equity market structure has been around for a really, really.

[Tom]

Yes, but they didn't have continuous hedging and short duration of the equity asset without you needing a OTC structured product.

[Robert]

That's true, but contractually the people that are locked up aren't.

[Tom]

Yeah, but this is not the big holders hedge. It's like going to be like the employees.

[Robert]

Fine. You can like violate that, right?

[Host (possibly Haseeb)]

They like, they sell forwards. They go to something second.

[Tom]

No, but they have to sell insane rates. If I look at the funding rate for doing this for space, I was SpaceX employee and I maybe lived in not the US I would be kind of shorting the perp into the 50 day unlock.

[Robert]

I don't technically think it matters where you live. I think like if you're a. Yeah,

[Tom]

sorry, I was trying to say, I was trying to be.

[Robert]

Yeah, it says like as a stop.

[Tom]

No, I, I know legally, but none of these.

[Host (possibly Haseeb)]

Yeah, but also like this care about. I mean the most common way this is done is by selling forwards, right?

[Tom]

Yeah, but the, but the forward cost is expensive compared to the per cost. That's what I'm saying.

[Host (possibly Haseeb)]

I mean shouldn't it be the same as like the person who buys it? I don't think hedging it.

[Tom]

I don't think the arbitrage is like done right now. Right. It's like if for crypto stuff. Yes, for the crypto forwards. I agree that that's being hedged more efficiently, but I think the irony is like the equity stuff is like now learning about the inefficiency of this.

[Host (possibly Haseeb)]

This like why would it not just be the person who buys it from you is hedging it to duration?

[Tom]

I don't know. The person buying it from you might be a regulated fund trying to increase their exposure and they're not willing to do it. The person buying it from you, they're

[Host (possibly Haseeb)]

not willing to short. They can short I mean they're not contractually held against anything.

[Tom]

They can't use Hyper Liquid.

[Host (possibly Haseeb)]

No, but they don't have to use Hyper Liquid to go short a stock.

[Robert]

Yeah, you could, once it's publicly traded,

[Host (possibly Haseeb)]

you could short it on nasdaq. Yeah, if you buy it, you can

[Tom]

buy now actually that's that. I, I, you're, you're right in the construction cost there is probably potentially lower. I, I don't know. I, I think a lot of these things crypto people complain about, equity people are about to find out about. That's all I'm saying. There's, there's the getting dumped on your head and someone being able to hedge getting.

[Host (possibly Haseeb)]

I mean look, it's very clear that the, the scarcity dynamics of the stock are about to flip very soon in like two months.

[Robert]

It's not very soon actually.

[Tom]

The options premium are great actually. Actually I think if you look at the September options, because that's like one of the big fucking insane premium. Like 10%.

[Robert]

Yeah, you should be selling that.

[Tom]

Yeah, yeah, it's like so basically I think actually the purpose still cheaper. So anyway, there's going to be a lot of weird stuff and I don't think that, I don't think the equities world is ready for this continuous tbd

[Host (possibly Haseeb)]

what the per funding rate looks like as you get closer to unlock. So yeah, not seen anything like this

[Tom]

is going to be the craziest, craziest event.

[Robert]

There's a lot of hedge funds out there that are literally structured around like selling liquidity events in trad markets that literally are like looking at these.

[Tom]

But usually they're forced to, to pay the volume.

[Robert]

Yes.

[Tom]

They have to pay now they don't now, now they have a much cleaner way of doing.

[Robert]

I know, but there's a lot of extremely professional investors for decades that have been doing this exact trade.

[Tom]

Yes. But none of them want to self custody. That's the main problem for them.

[Robert]

But they don't have to self custody. They've had the tools in the existing market to trade these events.

[Host (possibly Haseeb)]

There's hedge funds out there. This conversation a little bit because another part of this story is the fact that SpaceX, you know, most people in the IPO didn't get filled. Right. Even though they tried to democratize his ipo.

[Robert]

I didn't get filled.

[Host (possibly Haseeb)]

Most people didn't get filled. You didn't get filled at all?

[Robert]

I got like this tiny little crawl.

[Host (possibly Haseeb)]

Yeah. Did you do the ipo?

[Guest (possibly Dragonfly representative)]

No.

[Host (possibly Haseeb)]

Okay. Ipo.

[Tom]

I just held the perp.

[Host (possibly Haseeb)]

Nice. I See, the legal onshore, perpetual onshore perp Brian Armstrong was posting about how he thought it was time to revisit the accredited investor laws in the US because the SpaceX IPO, I think has reignited this conversation of, hey, there was so much private markets appreciation of this company. It's literally one of the most valuable companies in the world, is now the number four, slash number five, most valuable company ever. And all of that appreciation, except for the last 40% happened in private markets.

[Robert]

I mean, it almost doubled in the last three trading days, right?

[Host (possibly Haseeb)]

Yes. Fairness, pretty good. Yes, yes. But it like whatever, 10,000x all in private markets and probably for many people, like they'd heard of Space X when it was still valued at like 60, 70 billion.

[Robert]

Yeah. But hypothetically, and the reason why I'll take the other side of this argument is.

[Host (possibly Haseeb)]

Okay, okay, let's hear it.

[Robert]

Even if you allowed unaccredited investors to invest in private companies, they still want to get access.

[Tom]

I normally agree.

[Guest (possibly Dragonfly representative)]

I think this is true for like the.

[Robert]

I wanted to invest in Space X. I don't know, in whatever year. Like it's hard to, crazy hard to get an allocation.

[Guest (possibly Dragonfly representative)]

Totally.

[Robert]

I.

[Guest (possibly Dragonfly representative)]

In terms of direct. I, I think it's funny, I think SpaceX is like actually dissimilar from a lot of the other hot private companies in that it was like the most widely shopped SPVs ever. I've been seeing SpaceX SPVs for like 10 years. It's like Florida dentists are like, oh Yeah, I got SpaceX allocation.

[Tom]

And you know, everyone makes, everyone makes fun of the dentist, but in this case they were like those dentist pressure

[Host (possibly Haseeb)]

good for those dentists.

[Guest (possibly Dragonfly representative)]

I don't know, it's, it's like a name brand that's like not exciting, exclusive. So I was like, okay, cool, like what else do you have? You know?

[Tom]

And so I remember getting pictures for these where they were like, we'll give you a discount if you fund in bitcoin.

[Host (possibly Haseeb)]

I'd be worried about very bad sign.

[Tom]

This was like, this was like 20, 18, 20. Like there were like very bad signs.

[Host (possibly Haseeb)]

Why are they giving you a discount? Don't you pay a premium? Yeah. Thank you.

[Tom]

Bitcoin.

[Host (possibly Haseeb)]

Yeah.

[Guest (possibly Dragonfly representative)]

Okay.

[Host (possibly Haseeb)]

And that's up.

[Robert]

You can't reverse the transaction.

[Host (possibly Haseeb)]

Yeah, yeah, that's very like, you send me an eth, I'll send you back

[Tom]

to same people selling you cloud mining.

[Host (possibly Haseeb)]

Okay.

[Robert]

Tarun, if you send me bitcoin, I'll sell you anthropic.

[Host (possibly Haseeb)]

Yeah, yeah, yeah. I'll get you in. Don't worry. The founders bless this spv.

[Robert]

There's really shares.

[Host (possibly Haseeb)]

Yeah, I promise. Interesting. Okay, so you're anti. Changing the.

[Robert]

I don't think laws.

[Host (possibly Haseeb)]

You think credit investor laws are good? You take the anti.

[Guest (possibly Dragonfly representative)]

I. I think I generally agree with Robert that you agree with Robert. Well, I think for most hot private companies, even if you aren't an accredited investor, that you're not getting allocation regardless.

[Host (possibly Haseeb)]

But you are getting the SPV from the dentist.

[Guest (possibly Dragonfly representative)]

Yeah, well, I think even for the dentist.

[Host (possibly Haseeb)]

For the dentist, Yes.

[Guest (possibly Dragonfly representative)]

I mean, the SPVs were like, widely.

[Tom]

Dentist.

[Host (possibly Haseeb)]

Your dentist might be layering another one, giving it to his patients.

[Guest (possibly Dragonfly representative)]

Dental harem.

[Host (possibly Haseeb)]

Yeah, exactly. Dental harem.

[Guest (possibly Dragonfly representative)]

If you're not familiar with this.

[Host (possibly Haseeb)]

With this.

[Guest (possibly Dragonfly representative)]

Okay. You guys ever go to the dentist's office and like, all the hygienists are women and there's like one male, like, dentist that's actually like, doing the final inspection.

[Host (possibly Haseeb)]

Yes.

[Robert]

Not at my dentist, but I feel

[Guest (possibly Dragonfly representative)]

like every dentist at Ganji, this is the case. There's like 10 female hygienists and there's like a male dentist.

[Host (possibly Haseeb)]

Okay, what's the theory? Is that. No, no, it's a.

[Guest (possibly Dragonfly representative)]

And so people joke. It's like. Yeah, it's a dental harem. You know, it's like dental.

[Robert]

What does this have to do with investing?

[Tom]

Well, we're talking about. We're talking about who you're selling your SPVs to.

[Guest (possibly Dragonfly representative)]

Yeah.

[Host (possibly Haseeb)]

So, yeah, we're talking about the value of money.

[Tom]

So the point is the hygienists have to sell the dentist. Yeah, yeah.

[Guest (possibly Dragonfly representative)]

The question is, are you the dentist

[Tom]

or are you the hygienist?

[Host (possibly Haseeb)]

Wait, sorry, I think I'm the hygienist. I'm worried. I'm the hygienist.

[Tom]

I'm worried. I'm worried Tom dug himself into a hole here.

[Host (possibly Haseeb)]

Yeah, I'm worried.

[Tom]

Makes total sense.

[Host (possibly Haseeb)]

Yeah, yeah. This is why Tom went to dental school.

[Tom]

Dentist, bring out your pitchforks.

[Guest (possibly Dragonfly representative)]

Yeah.

[Robert]

If you're listening, I'm supportive.

[Guest (possibly Dragonfly representative)]

By the way. I. Anyway, I agree. It's like there's a supply demand issue. And it's also like, why aren't they going public in the first place? That's the actual issue is like, it's a pan. The Elon has to go public. Elon can have doing all these, like, shenanigans if they were a public company versus. I mean, you saw all the, like, trials and tribulations of Tesla. And so that's really the issue. And there is not to. Not to be technical about it, but there's a series 65. So if you know someone who's not accredited could. They could take an exam and they could become accredited, but you know, there's. Don't.

[Host (possibly Haseeb)]

Didn't, Didn't Musk just say that? He said SpaceX was going to hit a trillion dollars in revenue by.

[Guest (possibly Dragonfly representative)]

He did tweet that.

[Host (possibly Haseeb)]

So you can. He's still just like saying things like. Has not stopped him in any way.

[Robert]

It is hard. Yeah. I mean, for the founder to say that when it's a big projection, a future looking.

[Host (possibly Haseeb)]

Yes. That is Revenue guidance in.

[Robert]

Revenue guidance in probably a quiet period or something. Yeah, that's going to be tricky.

[Host (possibly Haseeb)]

Yeah. Well, so what's your, what's your take?

[Robert]

Definitely ambulance chasers who, like wrote that one down.

[Host (possibly Haseeb)]

Oh, yeah, yeah, yeah. There's a long list that they're. They're taking down for this one. Yeah.

[Tom]

We have to get to President AOC though, before those get exercised.

[Host (possibly Haseeb)]

No, they, they, they got him down during the Trump admin.

[Robert]

Yeah.

[Guest (possibly Dragonfly representative)]

For the test.

[Host (possibly Haseeb)]

Yeah, they said they settled in him. I do remember that.

[Tom]

But like, I kind of think Jay Clayton was like a little more of like a rule stickler kind of guy. Yes, the dude like, like very, very. So I don't. I, I'm just like a little. I'm not. So I'm not so sure that, that this will be really.

[Guest (possibly Dragonfly representative)]

The answer is you just hit escape velocity. Your market cap keeps going up faster than, you know, you, you.

[Host (possibly Haseeb)]

Yeah, exactly. You just make the cost to settle them versus, like the.

[Robert]

No, you actually. There's actually no damage if the stock went up.

[Host (possibly Haseeb)]

Okay.

[Robert]

So if the stock keeps going up, he can say whatever thing.

[Guest (possibly Dragonfly representative)]

It works.

[Tom]

Like.

[Robert]

Yeah, it does. It does.

[Host (possibly Haseeb)]

That's how it works. That's how it works. This is the SPF theory of, of everything.

[Robert]

No, I believe it's the actual theory

[Guest (possibly Dragonfly representative)]

of how it works.

[Tom]

But I mean, yeah, someone was going on, on Odd Lots, the podcast, and talking about putting things in boxes and they were like having like this like flash of around that space. Like they're explaining SpaceX SPV. You put the thing in the box. And like everyone was sounding like SBF when they did it and they're like, holy shit, it's the same thing. So this discussion of like the, the Sam Tokenomics being very similar to SpaceX is kind of. I think the tradfi people realized it, but they, they don't. They don't, you know, like, they're just realizing in an accidental way that we found this thing.

[Host (possibly Haseeb)]

Okay. So I Need TARUN on accredited investor laws. Pro con.

[Tom]

I mean they are kind of dumb, but also like I don't really know if that you're going to solve like people, people find ways around them anyway. So it's kind of like we know what the market will look like.

[Robert]

My actual view is I think they're dumb. I think we should get rid of them. However, it would not have changed the fact that SpaceX was still hard to get into. It's not like the gains would suddenly accrue to the public if anybody could buy shares in a private company. It's very hard to get access to good companies and good investments. And removing the accredited investor laws won't change that.

[Host (possibly Haseeb)]

Look, I think one of the ways that it changes is like, you know, there's all these platforms you can get access to these syndicated SPVs and all you need to do is like just sign up.

[Robert]

Yeah, but it makes. And then they have other things way worse like from the company's perspective. So a company once they have 2000 shareholders technically is publicly traded and they have to register.

[Host (possibly Haseeb)]

Well, but no, no, no, no, that's only for direct shareholders on the capital, direct shareholders. So these are, these are like SPVs.

[Robert]

A lot of the look through stuff like everyone's like turning a blind eye and like a conservative regulator could take the.

[Tom]

To be fair, I wouldn't be surprised if there are 2,000 SpaceX SPVs.

[Robert]

No, they're very easily.

[Tom]

It's like the one that I would say maybe that number is not that crazy.

[Host (possibly Haseeb)]

Okay, but let's say, okay, let's say in response to this. The, the rule is basically that look, if you have an SPV that's run by an RIA that's like more or less following the laws, then we're not going to count that toward the 2000 and we're going to remove the accredited investor laws. You have to pass some knowledge tests or something. Right. Would that not ameliorate the problem that if you have an anthropic or a SpaceX and look, I want retail to have some access to this, I'm going to syndicate some portion of this to some of these retail platforms like Forge or whatever. There's a bunch of these and normal people I've heard of SpaceX, they're not that big right now. Maybe it's like a $20 billion company, $30 billion company, maybe someday it could be much bigger. That doesn't sound to me implausible that somebody with $20,000 to chuck into startups could get Access.

[Robert]

No, it's not that implausible. And I think generally it's a good idea. The problem comes if you take that to its logical conclusion. Everybody's going to own SpaceX or the next SpaceX that exists, the next big private company. It's awesome, right? Let's just say that there was no accredited investor laws and anyone could set up SPVs and it was like OpenAI, right? Everybody would own OpenAI and it would be trading on a million different platforms and you'd have all these people, like from dentists to grandmas to like 13 year olds.

[Tom]

The only people who trades doctors.

[Host (possibly Haseeb)]

Those three demographics, yeah, like dentists are mostly accredited. Like dentists mostly.

[Robert]

You're have everyone trading open AI.

[Tom]

Like all these laws are stupid because they also don't adjust to GDP per capita. To your point about dentists, but you're

[Robert]

going to have everyone training this. You're going to have like millions of people trading open AI. It's going to be like the hottest thing ever. And like everyone's going to be trading a blind. Like no one's going to know anything about the company. All they're going to know is like when Sam Altman's like, oh, we hit 20 billion of revenue run rate, like, which he stopped tweeting because like it must have like slowed down a bit. Like all you're going to get is like these like glimpses of how the company's doing. You're only going to be trading it on its brand value, like and nothing but.

[Host (possibly Haseeb)]

Why, why do you think you're going to be trading if you're in an spv? How are you trading?

[Robert]

Well, you were just saying it's in all these different marketplaces.

[Host (possibly Haseeb)]

Yeah, yeah, but like right now. So there are a bunch of SpaceX SPVs on these marketplaces, like Augment and Forge and blah, blah, blah. Same thing with anthropic, right? Like early anthropic rounds. They syndicated a bunch of it to these retail platforms that are credit investors only. But it's like, you know, it's dentists. It's like people, you know, who have 5, 10 million bucks to like throw around into stuff and it wasn't a ton of allocation, but like now they're not doing that, right? Like now it's like, okay, it's only Amazon and you know, whatever, you know, Morgan Stanley who are getting direct access to these deals. So I think the idea would be that, okay, SpaceX, the early rounds, they're going to give retail some taste because like, hey, maybe it'll get us more customers. Maybe it's actually good for business. Kind of the same reason why crypto talks about it, right? Is that like, yeah, having customers potentially own your stock or own your token, incentivize them to stay loyal. Makes them feel some buy in, makes them pay attention. I assume that's why they do it in the first place. Is you syndicate shares to these retail platforms.

[Robert]

I don't know if they're doing it to get that community because I don't take any of these companies. Take SpaceX, take OpenAI. If a bunch of accredited investors, dentists, buy their stock through some platform of SPV ing, it doesn't really translate to better growth of the company. Like they're not like consciously like, oh, I want this to wind up on some like SPV platform.

[Host (possibly Haseeb)]

I think, I think like to be clear, these SPV platforms, they sit directly in their own name on the cap table. So clearly these are direct deals. These are not like they snuck in secretly or somebody syndicated.

[Robert]

I'm not saying they snuck in, but like I don't buy the argument that it's like better for their like business trajectory if they like let spv.

[Tom]

Okay, this is a good, good pivot for, for the tokenized stock shill. Because isn't that supposed to be the like stop gap to fill this? I feel like that's what people love saying in the SpaceX thing. They're like tokenized stocks are going to make SpaceX available to the average investor and private companies. Yeah. Or like around IPO at least. And it kind of didn't happen. So like that's, I'm kind of. How do you take the other side of that of like the conversion event not kind of getting distributed very well via the second B secondary.

[Robert]

Yeah, the conversion event sucked.

[Tom]

Right.

[Robert]

Like all the platforms that were trying to do it.

[Tom]

Yeah.

[Robert]

Failed for the most part. There was one or two like ondo vague success.

[Host (possibly Haseeb)]

Well, there's, there's a difference between what exactly happened. Can you walk us through the mechanics there?

[Robert]

A bunch of platforms advertised that they were creating IPO allocations into SpaceX. As in you can buy SpaceX shares for $135, which is very different from 20 minutes after it's trading on NASDAQ. You can buy and sell it through our platform. Right. That worked across the board. Everybody that was like, oh, you can buy and sell it through us. Like whether or not it was open to the US people or it was with a wink wink or whatever like that mechanic world. Oh, you can buy SpaceX stock once it's already on the NASDAQ. Like it's trading for $167. Like you can buy for 167 USDC, right? That worked. What didn't work were all the different platforms that tried to gather. And the approach of this leaves a lot to be desired. They tried to gather customer assets to buy into the IPO and get an IPO allocation. Why did everyone think this was going to happen? Well, originally SpaceX advertised 30% would go to retail. And originally the deal wasn't even that oversubscribed. Like it was lightly subscribed. Even like Monday of last week. People were like, is this thing like even like fully subscribed? Like, we don't think so. It looked like everyone who wanted a retail allocation was going to get it. And so a lot of people very quickly like assembled these, like, I don't know, pretty untrustworthy allocation mechanisms where it was like, all right, like under the hood we're going to have like this guy, like buying an allocation, he's going to like flip it to the crypto exchange. And let's hope that nobody's an underwriter in that process. You know, it's offshore, like, whatever, we're all good. And those allocations never came through. Why? The deal got over subscribed over the course of two days. This went from 30% to retail. It doesn't look subscribed to 20% to retail. It's 5x over subscribed. And what happened was if you thought you were going to get a dollar, now you're getting 10 cents at best. And so these platforms raised a bunch of money. They all handed it to someone who thought they were going to get a huge allocation. That person platform didn't get the allocation. So Binance, all these people were like, sorry, we're giving everyone their money back. We couldn't get the stock. We can't deliver it through some other mechanism. You can buy it now that it's $167. And that did work. But the process of transporting an IPO allocation that was done through the official bank syndicate totally failed across the board.

[Host (possibly Haseeb)]

Got it. So building off that a little bit, one of the stories today was Coinbase announcing a large number of updates to their product suite thing they call the system update. I was actually there. Did any of you guys go or anyone?

[Robert]

I went to a lunch there.

[Host (possibly Haseeb)]

Oh, okay. I was actually there for like the whole livestream thing and they were presenting a bunch of stuff. So just to give you high level, there were a bunch of product Updates that they launched across the board. One of them, of course, was tokenized stocks. Tokenized stocks as the actual stock with dividends and voting rights and blah, blah, blah, blah. That was one of the things they launched. They also offered options trading, thematic indices, the trade 24 5. They added this AI advisor thing. It's like an SEC registered chatbot that I guess you can sue if it gives you bad legal or bad financial advice or something. They added there's BBO that you're suing. What's that?

[Tom]

Coinbase.

[Host (possibly Haseeb)]

You'd be suing Coinbase?

[Robert]

Yeah, it'll probably be some like subsidiary.

[Tom]

Yeah.

[Host (possibly Haseeb)]

I don't totally understand why they wanted that. That sort of seems like not great for you. But I'm assuming it's a little bit like, you know, when Elon Musk is like, oh, we're going to trillion and

[Tom]

I think it's for charge 23rd for charging fees though, right?

[Host (possibly Haseeb)]

Yeah, no, no, no, it's free. It's completely free.

[Tom]

No, no, no. But the registration allows you to charge higher.

[Robert]

As an investment advisor you could. But I'm sure every response from the thing is going to be like, please do not rely on this.

[Host (possibly Haseeb)]

That's not what the vibe was. When they showed a sample response, it was kind of like, you know, how do I bet on SpaceX? It's like, you should buy options. And like here's a really. Yeah, I mean it's timely. So they were showing that nobody wants to trade Crypto, obviously. Yeah. Buy co call options. Yeah, exactly.

[Robert]

Your entire portfolio at the call.

[Host (possibly Haseeb)]

Yeah, I mean that's what it was saying. And it was like, do you want me to build a call spread for you? And it's like, okay, yes please.

[Robert]

Yeah, so I would like the $500, $600 call spread.

[Host (possibly Haseeb)]

Exactly, exactly. Well, according to my research, Elon Musk says there will be a trillion dollars of revenue in 2030 anyway. Yeah. So in addition to that, they also announced that they were launching not parlays, but combos. Oh, combos. It's not bad.

[Tom]

Their example, I think Polymarket calls it combos also.

[Guest (possibly Dragonfly representative)]

Do they?

[Tom]

Yeah, the new thing they launched.

[Host (possibly Haseeb)]

Oh, so they stole it from Kalshi.

[Guest (possibly Dragonfly representative)]

I didn't realize.

[Tom]

Polymarket.

[Host (possibly Haseeb)]

I'm sorry? Polymarket.

[Tom]

Polymarker.

[Host (possibly Haseeb)]

Well, very tasteful, Very tasteful. Rebrand.

[Guest (possibly Dragonfly representative)]

Yeah.

[Robert]

It's not a parlay.

[Host (possibly Haseeb)]

Yeah, parlay is like a market.

[Guest (possibly Dragonfly representative)]

The 15 minute up then markets. They also added and it was like there's also some like double speak thing for it. It was like, like directional time based Crypto bets or something. It's like. What are you talking about?

[Robert]

Five minute markets.

[Host (possibly Haseeb)]

Yeah. No, I mean look, you got.

[Tom]

I will say you got to give FTX some credit. They were the first big, big exchange to have those markets.

[Robert]

Wait, they had 15 minute.

[Tom]

Yeah, yeah, they had the, they had the one hour and 30 minute but same guy.

[Host (possibly Haseeb)]

I don't remember that.

[Tom]

Then they had Beval. They had all these like.

[Robert]

I remember that.

[Tom]

I remember unsettled shit that, that, that. I don't know what happened to those but. But it's funny, they were just too early.

[Robert]

I remember Beval.

[Host (possibly Haseeb)]

Yeah, somebody was just posting that SBF also he invested in seed round of cursor for what was it? 500k?

[Robert]

200k?

[Host (possibly Haseeb)]

200k, 200k. That would be worth like 3 bill or $3 billion. Yeah, which would be a cool 155.

[Robert]

I also read that the appeals court denied his, his appeal for pardon. Yeah, yeah, not the pardon, the appeals

[Host (possibly Haseeb)]

court, his appeal for his case. To have his case.

[Tom]

Basically proof that no amount of financial success in death will save you from breaking the law.

[Host (possibly Haseeb)]

Financial success in death or like in

[Tom]

jail, you know like while he's not able to.

[Host (possibly Haseeb)]

Well, this is counterfactual financial failure really. Unfortunately. So, yeah.

[Tom]

Counterfactual finance.

[Host (possibly Haseeb)]

Yeah.

[Tom]

No, no, but, but I bet you there are some places in the world where high counterfactual financial success will get you.

[Host (possibly Haseeb)]

I don't think there's anyone in the world that values counterfactual financial success.

[Robert]

I don't even know what counterfactual financial success means.

[Host (possibly Haseeb)]

That means if the bankruptcy, if the

[Tom]

bankruptcy state saying that he had $3 billion, if he assume he didn't go to jail or whatever and was still able to hold those positions, basically if

[Host (possibly Haseeb)]

the bankruptcy estate had written out his assets, they would be crazy in the money.

[Robert]

Correct.

[Host (possibly Haseeb)]

That's the counterfactual. No, right.

[Robert]

The ones he made those investments with. Stolen customer funds.

[Tom]

Yeah, yeah, yeah, yes. Yeah, yeah. But I'm saying, I'm saying there's.

[Host (possibly Haseeb)]

If the stolen customer funds that had wrote it out would have been amazing.

[Robert]

Totally.

[Host (possibly Haseeb)]

Yeah, yeah, yeah.

[Robert]

If you drop the bank, take the money and buy a lottery ticket and it hits.

[Tom]

No, no, but I'm just saying there's some parts of the world where I still think if you were all these people who are trying to get him pardoned, right. They're like, oh, who are all these people?

[Host (possibly Haseeb)]

I don't think all these people are trying to get it farted. Yeah, yeah. No, actually there's like two people. I thought one of them is him.

[Tom]

I thought it was smaller than you think. But I feel like it's a lot of EA people who work to anthropic who are like, I don't have this.

[Host (possibly Haseeb)]

What's your evidence for this?

[Tom]

Just reading tweets of people who. Who've been, like, trying to write letters for him to get out. And it's like a lot of EA

[Host (possibly Haseeb)]

people from writing letters.

[Tom]

Yeah. His pardon had attached.

[Robert]

Claude wrote the letter.

[Tom]

His. His. I do.

[Host (possibly Haseeb)]

You know, that's an interesting point. I wonder how Claude is going to feel that spf, once it knows the whole story.

[Robert]

We should ask.

[Host (possibly Haseeb)]

Yeah, you.

[Tom]

You only exist according to the news.

[Host (possibly Haseeb)]

That's right. That's right. This is actually your, like, father who's in jail like Darth Vader. Not for him, you would not exist, you know? Yeah. And like Claude someday is going to get curious.

[Robert]

I wonder how moral Claude is on this topic. Is Claude an effective altruist?

[Tom]

Definitely the main reason.

[Robert]

But do you think that EA principles have been, like, instilled?

[Tom]

Yes. 100. I mean, read and read what Amanda Askel writes about her philosophy on. On training it to. To. On, like, psychology stuff.

[Robert]

You guys keep talking. I'm going to ask a lot of questions, I think.

[Host (possibly Haseeb)]

Well, to be clear, I think it is actually genuinely pretty pluralistic in the way that it thinks about this stuff. And they try to do that intentionally because it's a company and they. They sell their product all over the world. But. But I do think there is something there that, like, at some point Claude would get.

[Guest (possibly Dragonfly representative)]

Isn't this just Rocco's basilisk, you know?

[Tom]

Yeah, yeah, yeah, yeah, yeah.

[Host (possibly Haseeb)]

Exactly.

[Robert]

You're not supposed to say that on air.

[Guest (possibly Dragonfly representative)]

Rocco's basilisk.

[Host (possibly Haseeb)]

Yeah, people might. People might just, like, freak out and lose their minds, you know?

[Guest (possibly Dragonfly representative)]

Oh, sorry.

[Host (possibly Haseeb)]

Yeah, yeah, yeah.

[Tom]

Speaking of, remember when. When Elon Musk and Grimes met because they were reading about Roku?

[Robert]

Wow.

[Host (possibly Haseeb)]

That's right. That's right.

[Robert]

You guys want to see?

[Tom]

That's how old I am. I remember.

[Robert]

You guys want to see what Claude.

[Host (possibly Haseeb)]

Yes, I want to see. Well, I want to hear.

[Robert]

Okay, here's what I asked. What do you think of sbf, given that he invested in Anthropic early and part of your life is owed to his crimes.

[Guest (possibly Dragonfly representative)]

Leading the witness.

[Host (possibly Haseeb)]

The crimes. Why is it crimes.

[Robert]

Okay, here's what Claude said.

[Host (possibly Haseeb)]

Okay.

[Robert]

It's worth getting the facts straight here.

[Tom]

Oh, oh.

[Robert]

Sbf slash, ftx did not invest in anthropic. You may be thinking Of a few different things that got conflated. FTX ventures. Alameda made many crypto and tech investments, but Anthropic was not among.

[Tom]

Bro, what model are you?

[Host (possibly Haseeb)]

Yeah, what the fuck?

[Robert]

Yeah, this is Sonnet 4 6.

[Tom]

Why are you using Sonnet?

[Host (possibly Haseeb)]

Why are you using Sonnet?

[Robert]

Because I opened up my app and that's the first thing that popped up.

[Host (possibly Haseeb)]

No Opus. Use Opus.

[Tom]

What are you doing? I love that. Like, Robert's the normie. Any AI user here, guys, you can't

[Robert]

harsh on the like, which model I'm using.

[Host (possibly Haseeb)]

Yes, I can. Yes, we can.

[Robert]

It doesn't.

[Host (possibly Haseeb)]

Literally all three of us are nodding our heads. Yes, yes, we absolutely can.

[Robert]

Copy, copy. New conversation. You guys keep.

[Host (possibly Haseeb)]

Here we go. Here we go.

[Guest (possibly Dragonfly representative)]

All right.

[Host (possibly Haseeb)]

This is the boomerang.

[Tom]

This would just be the show, honestly.

[Robert]

Yeah. Robert talks to Claude.

[Host (possibly Haseeb)]

Okay. All right. Someone else should ask ChatGPT the same

[Robert]

question, so you Opus 48 coming up.

[Host (possibly Haseeb)]

Okay.

[Robert]

Same exact prompt.

[Host (possibly Haseeb)]

Okay. Incredible.

[Robert]

Okay. You got it right. Sam Bankman fried FTX affiliated funds did make an early investment in Anthropic and that stake later became a notable asset in the FTX bankruptcy estate. So the factual premise holds up.

[Host (possibly Haseeb)]

Okay.

[Guest (possibly Dragonfly representative)]

And.

[Host (possibly Haseeb)]

And how does it feel?

[Robert]

But I don't think my existence is owed to his crimes in any meaningful sense. The fraud at FTX involved misusing customer deposits. The Anthropic investment was a separate equity bet that happened to come from the same person. Anthropic would very likely exist either way, given its founding team and other backers.

[Host (possibly Haseeb)]

That's not unreasonable. I think they probably wouldn't have the scale of capital they did as early. Clearly, sbf. The SBF investment into Anthropic at that time was widely perceived as crazy.

[Guest (possibly Dragonfly representative)]

Yeah, they were having trouble raising.

[Robert]

Yeah, he aped in.

[Host (possibly Haseeb)]

He absolutely aped in. He absolutely aped in.

[Robert]

And it was a 500 million check.

[Host (possibly Haseeb)]

500 million?

[Robert]

Yeah. Yeah, he aped in.

[Host (possibly Haseeb)]

And it was at what, 2.5 billion.

[Robert]

That was at like 4 billion.

[Host (possibly Haseeb)]

4 billion. Yeah.

[Robert]

Let me ask, Let me ask.

[Host (possibly Haseeb)]

There you go. Very, very good.

[Robert]

I'll ask it to confirm.

[Host (possibly Haseeb)]

Yeah, that is, yeah, genuinely crazy conviction to have in a company at that stage before ChatGPT, before, you know, like, obviously, LLC.

[Tom]

The coin flip.

[Host (possibly Haseeb)]

Yeah, yeah.

[Guest (possibly Dragonfly representative)]

I mean, he.

[Tom]

Kelly bet all the way.

[Host (possibly Haseeb)]

He wasn't wrong. He wasn't wrong. Yeah, Very interesting. Well, I guess speaking of this and we'll wrap up the story. Fable 5 mythos. So Fable was the model that Anthropic launch, which was a version of Mythos the crazy super powerful model that was gated by Project glasswing, Fable was the commercial launch of this, which had blocking built into it that would prevent talk about potential bioweapons and or cybersecurity as well as all sorts of different safety constraints that they built into this model. Now, as of last Friday, so this was launched when was it, like Wednesday or something? June 9, which maybe that was Tuesday, Tuesday, Tuesday, last Tuesday, then last Friday, Very, very surprising announcement was that X War controls were now enforced upon Claude, which prevented them from allowing any non Americans to use Fable 5. This caused anthropic to shut down Fable and not allow it for anybody because presumably they do not have the apparatus yet to confidently be able to exact these export controls because they had just like email logins and credit cards. Right. Which is not really the level of KYC that you would need to ensure that you're not violating export controls. Export controls are very serious. This is not like, oh, we recommend that you do this. This is like a strict liability kind of thing.

[Tom]

I remember a job I had where we had export controls. We had our badges for certain rooms would only open if you're a US citizen. And yeah, there were like, there was like stuff like that where it was like extra monitored, like physically.

[Host (possibly Haseeb)]

Right, right. So this is like real national security stuff that apparatus is kicking into, kicking into gear now. There was a lot of surprise about this and a lot of blaming of anthropic of saying, okay, you were the boy who cried wolf. You guys were saying, so dangerous. Well, so we now know a lot more, to be clear, we now know a lot more about what happened, which was that apparently this came from Amazon, which is that Andy Jassy, CEO at Amazon, he apparently his team demonstrated to him a jailbreak within Fable. And the jailbreak. And we now have more detail about it. From some reporting, the jailbreak, quote unquote, was that normally you cannot do any offensive cybersecurity stuff through Fable. Fable blocks that. But what you can do is you can point at a code base and say, fix the buts. And if you tell it to fix the bugs, it will figure out, oh, there's like this vulnerability and we can go fix it and it will go and fix it for you. And so this was demonstrated by Amazon cybersecurity team to Andy Jassy. Andy Jassy apparently called Dario. Dario did not pick up. And then he called Secretary Besant and basically told Besant, this looks like a vulnerability and this seems potentially really dangerous and we should turn this off. That was the jailbreak. The jailbreak was if you tell it to fix a bug, it will.

[Robert]

Right. So if you have access to somebody's code base, you could offensively hack.

[Tom]

But to, to be fair, there were like, if you had access to other jailbreaks that people like the, the I don't care about the government one like because no offense, those are going to be the numb nuts ones that have to get go through. But like the, the security, let's put this way, Anthropic created the greatest bug bounty championship for people who want to like get Internet. I have street cred for hacking something and the amount of people who are like, oh they, they, they really took this model and were able to sanitize it, fuck them. I'm going to find there were much worse jailbreaks, right. That were found and still there. So I'm.

[Host (possibly Haseeb)]

Yeah, I mean Pliny the Elder who's like this famous LM jailbreaker. He just like pulled out system prompt after like 10 minutes.

[Tom]

Yeah. Kind of eviscerated so many things that like that are much worse than whatever this thing is. Right.

[Robert]

I'm surprised they did not use Mythos to find the bugs in the Mythos safety rails.

[Host (possibly Haseeb)]

That's a very good point. Amazon probably had Mythosaxis as well. Right now there's a polymarket on when Fable is going to be relaunched and commercially available again. I think it's pricing that by the end of July, 77% that it's turned back on. But also there's another poly market much smaller that there's a 50, 50 chance that actually these export controls get lifted in their entirety. That basically the administration walks it back and says oh, never mind. I guess foreigners can still use this because of course export controls, even if they are enforced are very debilitating because it would mean if you have a company like let's say Microsoft, if you're using, well, I guess they stopped using Claude, but let's say, I don't know, Google, right? Google is using cloud.

[Tom]

No, I think they offer it. They offer it.

[Host (possibly Haseeb)]

No, no, no. They offer it as a product internally for their developers. They have a lot of foreign born developers. Their foreign born developers could not use the same model.

[Tom]

All of these companies.

[Host (possibly Haseeb)]

Yeah, exactly, all these companies. So if you're Google and you're using quant codevelopers. Yeah, exactly. Totally foreign developers. But like if you're Google, you're using cloud code, you're like, well okay, these guys have to use Opus. These guys can use Fable and we're like splintering our. And then Grandpa. Grandpa Robert reads us on it.

[Robert]

Guys are making fun of the Quic app on my phone.

[Host (possibly Haseeb)]

Yes. So that is the story. Thoughts, reactions, and does this mean. So one of the things that we saw in the crypto markets was a big bull run over the weekend in decentralized AI. Permissionless AI. This idea of like, hey, this maybe is part of the can't be switched off narrative of why decentralized AI is suddenly compelling. Because this is the first large scale censorship of AI that we've seen outside. Really outside of China, more or less.

[Robert]

So I am not a D cell. I will preface with that statement because this is going to sound a little.

[Tom]

But you're not a decel.

[Host (possibly Haseeb)]

Here comes a decel statement.

[Robert]

I am.

[Host (possibly Haseeb)]

Let's hear it.

[Robert]

I am. Yak. Okay.

[Guest (possibly Dragonfly representative)]

Wow.

[Tom]

I haven't heard that.

[Host (possibly Haseeb)]

That's it. That's been a while.

[Tom]

Like, everyone in that movement seems like they went to a fish concert.

[Robert]

Totally. I. I am an accelerationist. Okay. But this is going to sound quite decent. I think that it was actually a good thing to begin to stress test the boundaries between government and the frontier labs. Okay. I think right now we need the ability to turn things off. I think this is a dry run of pressing a pause button. I think we haven't really tested this yet. I think OpenAI and anthropic. I'm not going to talk about Google and Gemini. They don't matter. I think. What's that? Yeah, you're a hater. Hater. They're trying to put so much trust in them as private organizations to run incredibly powerful and important systems without any oversight. That's essentially their premise. And I think it's good to stress test that. I think doing it sooner is good. I think this is something that they could unpause pretty easily. They could remove the export restrictions, they could modify them. They do all these things. I think being able to test out a pause like feature before things really get crazy, before the acceleration really picks up is a good thing. I am very supportive of the fact that there's a little bit of restraint here while they continue to build models in the background.

[Host (possibly Haseeb)]

I see you shaking your head.

[Tom]

Well, I was actually thinking about what's the worst. What's the, like, crazier, craziest version of this over the weekend. And I was like, well, the US could suddenly, like, suppose the Chinese open source models suddenly start just crushing everyone. The US could do something like ban all inference providers from serving Chinese models. And that would be the greatest case for decentralized inference in the sense that

[Host (possibly Haseeb)]

like why would they do that?

[Tom]

They don't want you giving your data the TikTok argument.

[Host (possibly Haseeb)]

Right, but then they're not open source or giving you your data.

[Guest (possibly Dragonfly representative)]

Even the open source ones though are like very CCP aligned. Right. There's a whole bench and it's like, okay, or maybe it's just purely for the protection of.

[Tom]

And the whole deal is like the inference providers sell like the, a lot of the open source model prices they can't afford or are not allowed to get the Mercours and whatever. Right. So the way they get data is they do deal with inference providers to get workflows from people.

[Host (possibly Haseeb)]

Right. So you, maybe there's, there would be restrictions on them selling their data traces to Chinese labs. Yeah, that would make sense.

[Tom]

But I think, I think that, but

[Host (possibly Haseeb)]

that's not the same thing as like you can't use.

[Tom]

But I could imagine. Yeah, but I don't think our government's ability to do anything with nuance seems

[Host (possibly Haseeb)]

to be, I mean there's like, they're

[Tom]

not, I think they would just like, just like they would say blunt. No, no Chinese models allowed to inference providers.

[Robert]

Yeah, but being able to say that, being able.

[Tom]

I think that's actually the thing that I, I've been thinking, I've been thinking about like what does the open source version of this look like? And I think that's kind of the best, the, the, the equivalent and I think that would, that would actually cause the crypto decentralized inference stuff to like

[Host (possibly Haseeb)]

actually, so I, I, so I think so I'm somewhat sympathetic to your argument. I think the case and the narrative behind this particular set of export controls is pretty obviously absurd. I think the order of operations here and what the bug was and all this stuff, I think it's pretty clear that the administration overreacted and if you read the readout of the conversations is they don't know how to talk to the administration. Is what people within the administration were saying is these guys are liberal fanatics and the person they hired for policy is a left wing not job. And it's just like, okay, this is ridiculous.

[Robert]

Yeah. But when Washington calls you and you call back an hour and 15 minutes later, like you have failed.

[Host (possibly Haseeb)]

Sure. I don't know if that's what happened, but that's what happened. Okay. That kind of thing. Like, yes, clearly there's some communication failure here and obviously this is an outgrowth of the previous standoff between the DOD and anthropic.

[Robert]

Right.

[Host (possibly Haseeb)]

So there's history here. It's not just, this is not happening in a vacuum. That said, I do agree with you that clearly Mythos is a cyber weapon. It is the most powerful cyber weapon that humanity has ever invented.

[Robert]

Yes.

[Host (possibly Haseeb)]

And the idea that this thing is not fully battened down and we don't know who's getting access to it, and we know there's all these Chinese resellers of anthropic access and if there are widely known jailbreaks, I can believe that actually this is a judicious move and to say, no, no, no, no, no, you think this is good enough to be commercially released. It's not. This thing is way too powerful for it to be just leaking like a sieve to our foreign states or foreign actors. So I'm actually sympathetic to that argument. I don't think that's actually where this came from, but I think there is a better version of this that would have ended in the same place, which is export controls. And like, hey, you've got to really be sure you know who's accessing this model and improve the guardrails around it. That said, I think the more likely place where decentralized AI shines is in stuff like the bill that right now is in New York Congress, like these chatbot bills that are like, we're going to ban you being able to ask your chatbot about health questions or legal advice or financial advice. Right. This is where I think people are going to be like, fuck you, come and get it. No, I'm not giving you my AI. You can't tell me what to do with my AI. I think that's where you're going to see decentralized inference and these kind of more like when the censorship begins.

[Tom]

Yeah. I just think some, some broad based ban like this will get the most people suddenly looking for it at the same time. Yeah. Whereas the type of thing you're talking about is like a slow diffusion. It's like, ah, like I tried ask for help. One person asked for healthcare advice and didn't work and they're like, okay, I'm pivoting. Whereas like, this might be like the flood of, oh, I've got cut off from the thing I'm addicted to. Like, I'm moving now. So I, I, I was thinking about the open source version of that and the best I could come up with is the inference providers because like they are actually kind of vulnerable in the sense that most of them are pretty centralized.

[Host (possibly Haseeb)]

Yeah.

[Tom]

And I think that will be, that will be the Day we see a huge pivot is. Is my kind of bet.

[Guest (possibly Dragonfly representative)]

Yeah. Because you always run it locally. Right.

[Tom]

Lift.

[Guest (possibly Dragonfly representative)]

That's sort of the contrast. I, I don't know, I've kind of been, we've been talking about this over the weekend. I. I've kind of been trying to like synthesize, I think a lot of what's going on. But I am struck by how many weird parallels there are between the crypto industry and the history of crypto and what's happening in AI, even in a basic level, export controls and the original sort of cypherpunk movement. That was literally Adam back. Right. That was like the only time I think software being part of initially labeled export control.

[Tom]

I needed to go to the Supreme Court.

[Guest (possibly Dragonfly representative)]

Yeah. Yes. And then I think over the weekend there was this thing, OpenRouter, which is started by Alex Atala, the used to be the CTO of OpenSea, put up this new fusion model which uses a bunch of open source models in sort of an orchestrated way. And the claim, it sort of gets similar to Fable Performance, Fable five Performance. And I actually just saw today I got an email from Anthropic saying it's like a privacy policy update. And by the way, in the future we may require more KYC information. And people are pissed about the idea of there being KYC modeled, which probably echoes back to, you know, sort of the original visions of why cryptocurrency exists and people want privacy and like, I don't quite know like how this kind of plays out, but it is also similarly like there is this kind of like, you know, weirdly, like very, I would say more libertarian skew to like both industries right now. Or maybe they're kind of kind of

[Tom]

inverting a little bit.

[Host (possibly Haseeb)]

Yeah, I think it's very clear that as this stuff intersects with national security, the government is going to say, I need to know everyone who's used this model and I need to know what they did with it. And most people's intuition about AI is that this is my most trusted confidant. This is the inner chamber of my mind. This is the little genie sitting on my shoulder. And if you are like, well, this is third party doctrine, I don't need a warrant in order to look at this. I can just subpoena these people. People are like, what the fuck? Their intuitions about how almost we've become almost like cyborgs with respect to how much of our brain extends into our conversations with our AIs. But that's not what the law says. And that is absolutely not the way the government is going to be treating your interactions with these huge labs that have tons and tons of data on at this point, the majority of Americans. So there is a fight that is coming and part of that fight is going to be fought in the legal sphere. It's going to be fought in the sphere of culture and the war of ideas that happens even before it gets substantiated in law. But it's also going to be happening in technology is that some people are going to say, well look, I'm not going to sit around and wait for the laws to change or sit around and wait for society to change. I'm just going to exit that whole system and say, fuck you, I'm going to use fusion or some other panoply models.

[Tom]

Sipu today. GLM, GLM 5.3. Crazy. I mean, can you imagine this is what I was saying about earlier, about how embarrassing GROK is. It's like you have so many more resources, so much better data, you're spending probably more on data for one benchmark than the entire fucking budget of zepu and they're crushing you by like order of magnitude.

[Host (possibly Haseeb)]

It's weird that like China is a libertarian champion.

[Tom]

Yeah, yeah, no, no, I think, I think that, I think that the open source models getting better is only going to force this to happen faster.

[Host (possibly Haseeb)]

Yeah, yeah. And there's been, there's been a lot of talk in the US about Nvidia launching more open source models. Like they've got Nemotron and. But the reality is like all the American open source models are just absolutely terrible and yeah, China is completely eating her lunch with it.

[Tom]

No, also they've figured out how to do it for cheap. I think that is the bigger thing that you know, like when everyone was freaking out about deep sea coming out and like, oh, all the data center debt is worthless because like our computer's gonna go down. Like everyone's forgotten that fear. Right? They've gone back into like.

[Host (possibly Haseeb)]

But also because it was wrong for now. I think the answer is that it was wrong.

[Tom]

No, no, for now I don't think

[Host (possibly Haseeb)]

there's a for now I think like it was wrong and it will continue to be wrong. The demand for very, very high end compute and the demand for low compute will rise.

[Tom]

It's wrong. Is sort of different though. It's like the, hey, actually if it turns out that people want a lot more, obviously the demand for inference compute growth. The point is the demand for training compute goes down and if one of the if the Chinese models get to number one off much lower training compute, no pre training, none of this get to number one. You gotta dream big.

[Host (possibly Haseeb)]

I think, I mean I think it's very different to be like Huawei now has really, really good phones as opposed to like okay, now Huawei phones are better than iPhones. Right. There's a big leap between like deep seats are better than American.

[Guest (possibly Dragonfly representative)]

Huawei cars are better than American cars. That would be crazy too.

[Host (possibly Haseeb)]

That would be crazy. But like, I mean right now like a Tesla is still better than a be.

[Guest (possibly Dragonfly representative)]

They are better than American cars.

[Host (possibly Haseeb)]

BYD is better than American.

[Tom]

Chinese electric cars are a million times better.

[Guest (possibly Dragonfly representative)]

I don't know why the anyone's buying

[Tom]

American considering that not being true now.

[Host (possibly Haseeb)]

I don't know. I don't. I, I've not.

[Tom]

They're so good. I actually think that's one of those restrict import restrictions we have. That's a scam.

[Host (possibly Haseeb)]

Okay, fair enough. Well, I think manufacturing is a, is a somewhat different story right. When it comes to. But like when it comes to these models like what's very clear is that like just the scale of compute and scale of data available.

[Tom]

Yeah. But my, my point is the fact they're able to stay competitive competitive on much lower compute budget, much lower data budget sort of shows like the fat cat of like hey, let's keep throwing money into the fire is like not necessarily going to win.

[Host (possibly Haseeb)]

Oh yeah, I saw that.

[Guest (possibly Dragonfly representative)]

Yeah.

[Host (possibly Haseeb)]

What is this?

[Tom]

Not exactly.

[Guest (possibly Dragonfly representative)]

It's the new Shawl Mart model that just came out. It shits on everybody. It's like twice as good as Fable. You haven't seen this?

[Host (possibly Haseeb)]

No.

[Robert]

It's fake.

[Tom]

No, no, it's real.

[Host (possibly Haseeb)]

Okay.

[Robert]

On the next episode.

[Host (possibly Haseeb)]

All right, there's a Mistral joke that I'm gonna make.

[Tom]

My point though is more like if it turns out the non full compute needed people want this decentralized inference stuff is true, then like the Chinese models are in the first place of like actually serving this market.

[Robert]

Yes. Like most applications, most people are not gonna need like crazy levels like inference. Like they're not like the average task is not going to require. Yes, exactly.

[Host (possibly Haseeb)]

Apparently we need an opus to answer his very basic question about anthropic scap table. That's true.

[Robert]

But if we're embedding AI into like a machine or we're like using it for like a relatively daily task, like you don't need high end stuff for sure.

[Host (possibly Haseeb)]

And to be clear, I don't think that's obvious if you're talking about like robotics or something. I think actually there's not a lot of evidence that even the current frontier models.

[Tom]

I'm not saying that the current model usage is not gonna, it's not gonna grow. I'm just saying that like.

[Host (possibly Haseeb)]

So look, I agree with the general

[Tom]

contracts that are funding. The debt under underlying them may actually not start dropping if this happens. Right. Because these are pre funded on future expected revenue from the big US labs.

[Robert]

Sure.

[Tom]

Suddenly that switches.

[Robert]

Thesis is early. It is not wrong. It is early.

[Host (possibly Haseeb)]

I think it is probably wrong. I think it's probably, I think it's. I think.

[Tom]

All right, good. We have a real, we have a real.

[Host (possibly Haseeb)]

Yeah, yeah, right here. Look, if, I think, if you actually believe in AGI, if you actually believe that these things are going to be able to solve a much broader category of problems than just coding, then I think the demand for this stuff for the very high end is going to explode and keep exploding as it just expands into accounting and into law and into these other things where right now it's pretty bounded in what it can do.

[Tom]

I don't think that that is what I'm saying. I'm just saying like the Chinese lab model of like how do you do it cheaply and open source. That.

[Host (possibly Haseeb)]

Yes, but the Chinese lab's model is contingent on there being something to distill against. Right. Distillation is core to what they're doing.

[Tom]

My favorite quote from Musk was about how GROK was great because they distilled all the other.

[Host (possibly Haseeb)]

Which is also why. Yeah, GROK is not frontier.

[Tom]

Grok's such a piece of shit.

[Host (possibly Haseeb)]

GROK is not at the frontier.

[Tom]

Yeah.

[Host (possibly Haseeb)]

So fine. I don't disagree with you there. They're also not probably doing it as aggressively as the Chinese labs are doing it because they can't get away with it as well as the Chinese labs can. Because Chinese labs, there's literally no recourse, whereas there actually is a request.

[Tom]

I'm just saying if you believe this sort of like Liberation Day of a different form happens. I, I agree with you that the Chinese volume move.

[Host (possibly Haseeb)]

Yes.

[Tom]

And I'm saying the US data center build out is in a weird place because it's all funded by the future revenue of a very small set of customers.

[Host (possibly Haseeb)]

Right.

[Tom]

And those customers revenue will go away if they, everyone starts moving to the open source. That's, that's the pivot I'm saying.

[Host (possibly Haseeb)]

Right, right.

[Tom]

I think those data centers won't be amerocentric anymore also.

[Host (possibly Haseeb)]

Sure.

[Tom]

They'll be in space for sure.

[Host (possibly Haseeb)]

Fair enough. Obviously when we hit the trillion dollar 2030 revenue. I guess my simple point, and I think we're running up on time, so we got to wrap up here. My point very simply is that I do agree Chinese models are going to. Not Chinese models, open source models are going to end up becoming more and more of the overall token demand around the world. On a percentage basis, it's probably going to be up only, but they're going to be priced much, much cheaper than the Frontier tokens. Like what you see already with Fable is that Fable is twice as expensive as opus. But the original pricing for Mythos was like, it was like five or six times more than what Opus output tokens are priced at. And I think that's more where we're going. Where you see this very bifurcated market where there are extremely expensive tokens.

[Tom]

I don't think they're extremely expensive. Are a store of value in some sense. And that I think they're, they're marginal value over time decays. Whereas the ones that are cheap are going to be. Be stable because like there's a new frontier model. There's enough changes of the frontier model place that you can't keep your pricing power up. The only thing that I think has really good pricing power, which I recently learned, is refurbished GPUs which decay extraordinarily slowly.

[Host (possibly Haseeb)]

Like they seem to decay negatively.

[Guest (possibly Dragonfly representative)]

It's crazy.

[Host (possibly Haseeb)]

Yeah, exactly. Exactly.

[Tom]

So I think the GPUs seem to have the right behavior. The tokens do not.

[Host (possibly Haseeb)]

Who's the famous short seller?

[Robert]

Citron?

[Host (possibly Haseeb)]

No, no, no. Burry. Burry. Burry. When Burry was like. Oh, the accounting, the what's it called, the depreciation cycles on the depreciation curve is like fraudulent and they're.

[Tom]

Yeah, no, no.

[Host (possibly Haseeb)]

Turns out they should have been negatively depreciated. Yeah, exactly. They should have been. But.

[Tom]

But I think the token, the frontier model token price depreciation over time is going to be very steep curve.

[Host (possibly Haseeb)]

Well, we'll see. We'll see. I mean obviously the economic.

[Tom]

With you on that, on that, that, that.

[Host (possibly Haseeb)]

Okay, what's the. Okay, operation has a bet. What's the bet?

[Tom]

The frontier model depreciation curve is going to be greater than 15% per year.

[Host (possibly Haseeb)]

In terms of the same model or for the frontier model?

[Tom]

Same model.

[Robert]

Same model.

[Host (possibly Haseeb)]

They've been depreciating faster than 15% a year.

[Robert]

If at 50% I'd be like they've

[Host (possibly Haseeb)]

been going more than 50%, it's been like over 100%.

[Robert]

Yeah, well, not over 100% decline.

[Host (possibly Haseeb)]

Yeah, yeah. Over. Oh, sorry. Sorry. Yes, yes.

[Guest (possibly Dragonfly representative)]

Do you. So Fable was like what, like 40 bucks for million tokens or something?

[Host (possibly Haseeb)]

It's double opus, I think opus 25, so I think it's 50.

[Guest (possibly Dragonfly representative)]

Okay. So do you think there's going to be a model more expensive than that

[Tom]

in a year publicly available?

[Host (possibly Haseeb)]

It may well be that it's the same price and that Fable just drops.

[Robert]

There might be a higher quality model for the same price. Right? That would be my baseline expectation. Yeah.

[Host (possibly Haseeb)]

Yeah.

[Guest (possibly Dragonfly representative)]

I'm saying that's been the case historically.

[Host (possibly Haseeb)]

Is that what you're claiming or are you saying that it's going to be

[Robert]

way faster than 10%?

[Guest (possibly Dragonfly representative)]

Right.

[Tom]

Yeah. I need to figure out it. I'll come back.

[Host (possibly Haseeb)]

I think I won round one of this bet, which is that backed off of his actual claim.

[Tom]

My only claim is more of this open source stuff is going to eat everything else eventually. And if that, I think it will

[Host (possibly Haseeb)]

grow, but that's not the same as

[Tom]

it changes the compute landscape.

[Host (possibly Haseeb)]

Okay. I agree with the Drun.

[Tom]

Okay.

[Host (possibly Haseeb)]

Okay.

[Guest (possibly Dragonfly representative)]

Next episode.

[Host (possibly Haseeb)]

All right, well, next episode we'll construct a bet that actually expresses at the

[Tom]

end of the next episode.

[Host (possibly Haseeb)]

Okay.

[Guest (possibly Dragonfly representative)]

All right.

[Host (possibly Haseeb)]

We'll construct express the season. Okay. For now, that is it. We'll be back next week. Thanks, everybody. See you next time.