The Breakdown - The US Just Made the Case for Open-Source AI

Episode Date: June 15, 2026

Anthropic released Fable V — a wrapper around its Mythos model with strict guardrails. Within 24 hours, the White House forced it offline over a jailbreak that exposed Mythos's full cyber capabilities. David walks through David Sack’s thread, the export control, and why this is the strongest case yet for running your own open-source AI at home. Plus: Blockworks just acquired Messari, and the SpaceX IPO turned into a mess on crypto exchanges. Enjoy! TIMESTAMPS: (00:00) Intro (01:24) Anthropic (11:57) Blockworks Acquires Messari (14:38) SpaceX IPO FOLLOW THE SHOW › David — https://x.com/dcanellis › The Breakdown — https://x.com/TheBreakdownBW › The Breakdown Newsletter — https://blockworks.com/newsletter/the-breakdown Get top market insights and the latest in crypto news. Subscribe to the Blockworks Daily Newsletter: https://blockworks.co/newsletter/ DISCLAIMER As always, remember this podcast is for informational purposes only, and any views expressed by anyone on the show are solely their opinions, not financial advice.

It is Monday, June the 15th. Welcome back to The Breakdown. I'm your host, David Canales, as always. A few quick things for you this morning to get you ready for the week ahead. We're going to chat briefly about Blockworks are crying Masari, two of the biggest data and intelligence providers in the business coming together to form one giant beast of a platform. We're going to take a look at that. We're going to be looking at Anthropic, the Anthropic Rugpole.

It released Fable. The US government said, no, you've got to take it offline. And it's still offline. We'll take a look at the reason why and what it means for just AI in general moving forward and what you can do in order to protect yourself against government overreach. We're also going to be looking at pre-IPO stocks also effectively a rug pull. The SpaceX IPO, at least the on-chain experience for the SpaceX IPO, was not exactly optimal for many people.

We're going to be taking a look at that. So enough jibba-jibba. without further ado, whatever else I like to say before we head into the episode. This is the breakdown. Let's get to it. Nothing said on the breakdown is a recommendation to buy or sell securities or tokens. This podcast is for informational purposes only and any views expressed by anyone on the show are opinions, not financial advice.

Host and guests may hold positions in the company's funds or projects discussed. All right, let's kick things off with this Anthropic story about Fable 5 because everyone was excited about Fable 5. Anthropic had styled it as a mythos-like model, the most powerful potential. the most powerful ever released to the public to date. But it had strict safety rails so that it wasn't misused because it is effectively so powerful. So you didn't do anything cybersecurity related, you couldn't do anything related to biology or anything like that, I guess, at over concerns that people used to make bio weapons and so on. It was online for, I think, I mean, less than a day

before the US government stepped in and effectively forced anthropic to remove access to it over fear, is that people outside the US could use it to gain some kind of either economic advantage or security advantage over the US. So essentially a national security concern, it became. And I'll leave it to you to decide whether or not that is because the US government had bought into the marketing hype surrounding mythos and anthropic and so on. But from what I have pieced together from reports on Reddit and how can use and so on, is that the vibe is that Fable is actually very quite powerful.

And Ben Thompson, in his latest piece about all of this, he kind of goes into it, is that whether or not you believe that this is an overhyped model and so on, it's kind of beside the point, because if it is not this model that is going to be as powerful as anthropic claims, it's going to be the next one. And if it's not that one, it's going to be the one after that. It's almost a good thing that we have had this blow up, at least in the meantime, because we get to see exactly what the US government response is going to be in relation to super powerful AI models that are at least available to the public,

because we know there's Project Glasswing that has given access to, you know, hyper-powerful institutions and governments and so on, all around the world banks, so that they can prepare for the eventual release of mythos and its full capacity. But in any case, what we have today is that at least yesterday, David Sacks, the co-chair of the President's Council of Advisors on Space and Technology, and he used to be the AI Tsar before he stepped back. He's come out with a thread, and I'm pulling this from Tom's hardware. He goes into detail about why this has all happened,

and I'm just reading from Tom's hardware here. Sachs claims that a trusted partner of both Anthropic and the U.S. government, testing Fable, came forward with a jailbreak of the guard rails that separate the consumer model from the unrestricted cyphor capabilities of mythos, the model it's built on. He said the administration asked Amity to fix the bypass or de-deploy the model and that Amity declined. Anthropic instead prioritized keeping its consumer model live over safety, Sachs wrote, calling that inconsistent with the company's position as a safety first lab

that it had self-lobied for mythos to be regulated as a cyber weapon. What's interesting is that someone had told semaphore that Amazon had originally flagged that the model was able to be jailbroken. And I think there was also claims that someone in China had effectively been able to access mythos through the Fable model, which has been the impetus for the US government to say, hey, we need to turn this thing off because we don't want just anybody in the world to be able to access mythos. So, I mean, even as I'm saying, this is such a brilliant marketing for Anthropic that this is going on. And personally, if I was anthropic, I would want this situation to drag out as long as possible,

because it only creates a certain mythos around mythos that this thing is really essentially a cyber weapon at this point. I'm going to bring up the David Sachs thread here as well. I mean, let's just go into it. He says, Fable is mythos with guardrails. But if those guardrails fail, then you've exposed mythos and its advanced cyber capabilities to people who shouldn't. have them. Keep in mind that Anthropic itself widely promoted the idea that Mythos was a cyber weapon and needed to be regulated as such. They asked for government regulation of Mythos and championed the guard rails on Fable. If there's a vulnerability, big or small, it is Anthropics' responsibility

to patch. I mean, I can kind of understand that, but we already know that Mythos had already leaked, I think, within the first day or two, when some people in a Discord server had figured out the URL to be able to access mythos itself already. So, I mean, this is a little bit moot. It's a little bit in my mind. So the situation right now is that the White House has issued an export control on Fable 5, which is essentially keeping Anthropic from returning Fable 5 to be online. The logical conclusion here is that even if the export controls are lifted,

I can imagine, and other people have said this too, that we will soon need to have some kind of passport upload or KYC in order to access models that are built in the US because the US government is effectively going to say we only want US citizens to be able to access this. But of course, we already know how difficult it is to actually police geo-blocking on a fundamental level because the crypto industry has become very good at operating even with IP-related geoblocks in place. So TBD on how effective those controls are actually going to be if Fable actually does come online at some point.

So all of that aside, what I found interesting is that Venice, the VVV token that allows access to AI models, a variety of AI models completely anonymously. It did briefly pump on the news that the US government had effectively rug pulled Anthropics' latest model. but it essentially already walked some of that back. And to me, in my mind, I think that all of this is an advertisement for open source AI models, even if they are less powerful. And I mean, I want that to be true. The idealist in me would prefer if open source AI models were the norm and were as powerful

as proprietary closed source models. But I'm not sure how realistic that is. But in any case, we do have some stats about venus.a.i. And unfortunately, I mean, I understand there's some speculation about the token price to increase because it would, you know, speculation about an increase in users. But in terms of the total amount of VVV state in order to access the platform, it hasn't exactly grown into my soul of this. And in fact, over the past 30 days, it is actually down somewhat.

You can see the stats on my screen here in that the total VVV state on the third of June, so about a week and a half ago, was 32.5 million VVV. And right now that's down to under 32 million. So it's not like all of this news is pushing people towards using open source AI models, at least through the crypto lens. But I will say at least on Google trends, there is some evidence that we are seeing a peak in interest for open source AI. Here I have the Google trends.

And this is topped out at the end of May. He can see it was an all-time high for Google search interest across the world. world for open source AI in May. The same thing for the Google search for best open source AI. It topped out at the end of May as well. So it's not like we have returned to those highs since all of this stuff with the US government effectively flexing its regulatory muscle against the front, at least one frontier lab. And this is of course happening alongside a string of investigations open from state attorney generals into open AI's business practices. So it looks like the regulations are coming for the frontier labs to say the least, which in my view only

strengthens the case to start exploring open source AI models. For those who have been listening to this show for a while, my ranting, raving about AI, I mean, I'm not a software developer. I'm not, I'm not exactly super interested in vibe coding and so on. I mean, I do use it for this project and that project and so on. But what I do find particularly interesting, and this is not a new thought also, is that we are clearly going to see in the next five years at least, maybe even the next 10 years, in that what happened with running your own Bitcoin node, running your own crypto node for whatever your favorite blockchain might be,

running your own system to train your own open source AI model on and use it completely offline within your own home, this is going to be the next frontier in personal computing, especially for the more self-sovereign-minded individuals in the audience and at large. Because, I mean, I know that me personally, I'm becoming very interested in. I'm not exactly interested in the software engineering part. I mean, I'm not a software engineer. I don't really find it particularly interesting.

What I do find particularly interesting is that I live in Europe, but my AI use is soon it looks like going to be regulated directly. by the US government. And this is not something that I want to entertain. And even those of you living within the US must also feel something similar. So at this point in time, I'm researching what the best hardware is to run an open source AI model is. I want to protect my family from having to worry about what the US government is going to think about, how we are using AI, what we're using it for. Is it government approved? What we are conversing with our our LLM of choice and so on.

And so clearly, I mean, they just seem to go together hand in hand in that if you want to run your own Bitax, your own Bitcoin node, why not also run your own localized offline AI model as well? This ability to have your own technological moat within your own home to disconnect yourself from the wants and needs of world governments when we have this flouraging technology for intelligence and so on, I think it's going to be imperative moving forward. So I would encourage you to do the same thing.

But don't forget to also run your own nodes if you are an avid daily crypto user as well. I really think the two go hand in hand. So that's enough out of me for this one. What's next? All right, super quickly about Blockworks acquiring Masari. And perhaps you already caught a podcast episode that went out through this channel

and every other Blockworks channel as well that featured Blockworks co-founders Mike Ippolito and Jason Yanowitz that broke down the philosophy and the intent behind acquiring Rosari. I have the blog post up on here. If you miss that episode, please go back and check it out because you're going to hear directly from the co-founders exactly why and what's happening. The most important thing I guess for you as a listener to the breakdown in that is that nothing is going to change with the breakdown. We're still going to be giving this to you every week, three times a week for now. So nothing gets interrupted in terms of this particular podcast and show.

But what it does mean is that Blockworks is growing, absolutely for sure. And it also means that the industry is about to get perhaps the most cohesive all-in data platform that it has ever seen. So I'm going to read from the Blockworks blog post. And this kind of encapsulates the whole vibe, basically. what we're building. We have two key customers, issuers of on-chain assets and underwriters of those on-chain assets. For asset issuers, these are protocols, chains, foundations, applications, stable coins and RWA issuers, prediction markets and on-chain businesses that need to earn trust, to disclose

information, explain performance and manage investors. So asset underwriters, they are servicing the investors and regulators and the platforms that analyze, review and list those assets. These are funds, policymakers, exchanges, custodians, fintech, brokerages, and increasingly,

AI agents. So up until now, Blockworks has been building for the issue aside. Up until now, Blockworks has been building for the

issue aside. We're putting together a framework for standardized disclosures through the token transparency framework, and we also

have trusted research and institutional relationships and staled distribution. Masari, meanwhile, has,

for the same time that Blockworks has been building this, going right back to 2018, Masari has been building

for the underwriters through their market intelligence and data coverage and API and AI tools and so on. We are now bringing those two worlds together. So I'm going to leave it to you to go back and check out that podcast episode featuring Mike and Jason that goes into this a little bit further. But just for now, know that nothing is changing. I'm still going to be hosting the breakdown. The breakdown is still going to be coming to you multiple times per week.

So business as usual, but be on the lookout for big changes at Blockworks in terms of the platform and the product that we serve you and everyone else in the crypto space. All right. That's enough out of me. On to the next one. Okay. So it turns out that the tokenized stock experience for the SpaceX IPO was a little bit of a disaster, particularly for crypto exchanges, for centralized crypto exchanges.

Now, I've been showing you the markets on hyperliquid, which was always going to be cash settled perps. So there was no real implication, at least from my understanding, that. traders on Hyper Liquid would have received actual shares in return for their pre-IPO positions on Hyper Liquid. This is more related to crypto exchanges and particularly X-Stox, which is a tokenized equity provider behind tokenized SpaceX offerings that had deals with with crypto exchanges like Binance, Wallet, Bybit and Big Get. So as it turns out, X-Stox did not receive the amount of SpaceX IPO shares that it had believed it would get. And I'm reading from Dismodo here, that CoinDest reported that X-Stox and his partners gathered more than $1 billion in customer orders tied to SpaceX access. Binances campaign alone attracted $557 million in on-chain subscriptions before being unwound with.

with no allocation distributed, which is quite brutal. That is quite a lot of money being tied up in the SpaceX IPO that effectively you wouldn't get anything for. And many of these crypto strangers, basically all the crypto strangers are now in the process of awarding refunds and also extras. And apparently, Cracken did actually get its hands on some actual SpaceX shares, but reports are that Cracken was only able to give out four shares worth of tokenized. SpaceX exposure to those who had subscribed. So not exactly a lot, but in X-Stox's defense,

it's not like everyone else in the off-chain space got their allocations either. Reuters reported that retail investors chase the offering in huge numbers after SpaceX reserved to record 20% of the deal for individual buyers. But even then, many customers at traditional finance institutions receive fewer shares than they requested through Fidelity, Charles Schwab, and so-fi. So, yeah, I mean, it just looks quite difficult to get your hands on an IPO allocation due to how big this IPO actually was. All of this said, I mean, I just think that it is important to be realistic about what could come of all this. Because, I mean, we know Anthropic has really made moves to stamp out even the possibility of pre-IPO shares being legitimate after the fact. So we already have Anthropic really trying to stamp this thing out.

I wouldn't surprise me if Open AI is going to do the same thing. And whatever other AI company, whatever other Frontier Lab comes along that wants to go public, they're probably also going to do the same thing. So I would advise, I mean, I don't just want to dictate what the crypto space should do, but I really think that this idea that crypto exchanges are going to always be able to get the allocations that they advertise, I just can't see it happening in real life. I think that we need to taper our expectations to really have this just about cash settled perps and just treat it as price exposure, pre-IPO price exposure and nothing more.

You're going to have to have this line separate between the on-chain pre-IPO economy and the off-chain post-IPO economy. These two things are going to be separate for quite some time. And I think we're going to have to wait for this to play out a few more times. And what we're going to have to see is that these big companies want to incorporate on-chain rails into their initial distribution properly and really look after the crypto space. It really be proactive about including the crypto space into the IPO process. Right now, that is not exactly what is happening. And it will take that kind of communication and relationship between the trad-fi stock market and the on-chain world.

We're going to have to see these two ecosystems completely merge before investors are going to be able to trust that what the crypto exchanges are selling them is actually what they get in the end. And I understand a lot of this is really out of the hands of the stock providers because they are essentially middlemen and they make applications and bids for these certain amounts of shares that they can then pass on to their users via the crypto exchanges. but this just seems a very wishy-washy relationship right now. So the SpaceX IPO might not have been a success in terms of what the user experience was, particularly for the centralized crypto exchanges. Over on Hyper Liquid, it seems that it all panned out how it was implied to pan out in that you were just getting price exposure before the fact, and after the fact you're not going to get any more shares.

You're not going to get any shares for that price exposure. You just pre-related to take your gains or losses and walk away at the end of the day. But I wouldn't count on crypto exchanges not really hamming up the next one, whether it's anthropic or open AI that is the next target for this kind of stuff, especially when we see that there was over $1 billion in demand, on-chain demand through crypto exchanges for pre-IPO, post-IPO, tokenized stocks. that's such a big number and especially in a crypto space that is apparently going through a bare market, what else a crypto exchange is going to do except trying to attract that volume once more for whatever

the next IPO is? So just a word of caution, I mean this is another advertisement for decentralized exchanges and also just minimal promises and understanding what you're actually getting and just being okay with the fact that you can get semi-liquid price exposure for something that you couldn't get price exposure before. We need to operate in that space and own it before we start figuring out whether crypto is big enough and important enough to be guaranteed allocations from not just the companies themselves like OpenAI and Anthropic, but their underwriters and their investment banks that are handling the IPO process too. So TBD on whether or not this converts to something that is really material change for how market dynamics play out for IPOs.

Right now, it looks like the best we can get is price exposure through cash general perps, especially great by going by how rocky this particular one was. But you never know, something might change for the next one, but TBD on that. So this is about covers it for what I'm looking at on Monday morning, June the 15th. Looking forward to seeing what happens over the next couple of days. And we'll see you again on Wednesday. In the meantime, as always, look after yourselves. Goodbye.