Unchained - Bits + Bips: Would AI Rules Box Out Everyone but the Frontier Labs?

Episode Date: September 16, 2026

📢 Bits + Bips has its own channel now — full episodes here: https://www.youtube.com/@Bitsandbips In one week, OpenAI said its agents solved a Millennium Prize problem, a researcher quit Anthropic warning that both labs are "gambling with our lives," and Dario Amodei called for the industry to pace the frontier. Jordi Visser doesn't dismiss the cyber risk, but he calls the fallout "a big midterm election story." Austin Campbell asks whether the labs are making a commercial play, Chris Perkins explains why regulation forces consolidation the way it did for banks, and Ram Ahluwalia makes the case that open weight models still have a seat at the table. Hosts: Austin Campbell - Host of Bits + Bips, Founder of Zero Knowledge Group, and Adjunct Professor at NYU Stern Ram Ahluwalia - Co-host of Bits + Bips and CEO of Lumida Chris Perkins - Co-host of Bits + Bips and Head of Franklin Crypto Guest: Jordi Visser - Founder of Visser-Labs, LLC This clip is from a longer conversation on AI regulation, frontier lab valuations, oil and the Fed. Full episode here: https://youtu.be/pNogg0vEjjs?si=Wrzh-5BJXnkkcQLr  We go live every Monday at 4:30pm ET. Subscribe to catch it live. 👉 Visit 1inch to swap tokenized securities, crypto and more. Simple. Secure. Self-custodial. Whatever asset you’re buying - swap it at http://unchainedcrypto.com/go/1inch-yt    Chapters: 🧠 00:35 Why Jordi says Leopold's 'Situational Awareness' paper saw this coming 🗳️ 02:17 Jordi's read: this is a big midterm election story 💼 05:58 Is the AI slowdown a commercial play to cool CapEx before the IPOs? 🏦 07:28 Chris on why regulation is 'a really beautiful way of boxing out competition' ⚖️ 10:25 Jordi on who eats the liability when a model goes wrong 🔓 11:27 Could decentralization put AI liability back on the user? 📈 11:57 Ram's power law problem: capital concentration as the real story Learn more about your ad choices. Visit megaphone.fm/adchoices

If you're hearing this, you're listening to a bits and bibs segment on the unchained feed, but not for much longer. On Monday, September 28th, all bits and bits and bits content leaves this feed for good. Full episodes already moved to this show's own channels, and now the segments are going to. No more bits and bips on unchained at all. So don't wait and don't lose us. Search Bits plus sign bips wherever you're listening. Hit subscribe and you'll never miss an episode. That's Bits plus sign bips, spelled BIPS, all on its own channel.

starting September 28th. First of all, now that he is famous, I highly recommend everyone who hasn't spent time with Leopold's situational awareness paper, not the hedge fund, but the actual paper. And you upload it into your favorite LLM and you basically match up what's happened over the course of the last two weeks

and see if he didn't basically talk about all these things reaching this point. So in my opinion, in using these, continuously for whatever it's been now, two years, and writing about them, we've reached a point in the progress side where the disruption is impacting everything that you just mentioned. Politics have to get involved at this point because now the hackings are reaching a level where it's obvious. I'm sure we all talk to people throughout both the crypto and the traditional finance world. There was a hacking last week, a significant one in Boston Scientific.

There was a news story out there that Iran was using Claude in a way that was uploading everything on their military decisions and what they were trying to do towards U.S. warships. This is not going to stop. So at some point when the capabilities get to such an extreme level, they start to become news items, the hugging face situation, which was basically, let's see what they can do, did more than what was expected. But I don't think that should be a shock. We're at the point now where the IQ, particularly after Astra, we're in the 150 to 160 range. You're at a point now where things are going to be happening more and more that people just weren't ready for. And particularly companies, this is a political issue. And I think one of the most important things about the drama that's ensuing and all the back and forth that you recognize is this is a big midterm election story.

This has a lot to do. And I don't think it's a question as to if there's some involvement here. There's absolutely politics involved with this. And I'm not saying that the risk is not there because we certainly have hacking risk and we certainly have cyber issues that are starting to become there. But that's because the progress has gone so far. We first heard about this with Mithos. Astra is ahead of Mithos.

And these guys are already working on models that, as you mentioned, with Navier-Stokes, they're further ahead on their models right now. So they're seeing things that they haven't brought out yet and they haven't actually spoken about. And this is all before we get Vera Rubin-Chipp. So the power that we're still working on at this point is all Blackwells. Vera Rubens are going to be the next stage. And the more compute power you get, the more you're going to end up with even further advances.

I don't think there is any way that the slowdown will happen. I just don't see how they're going to allow these companies that have capitalist goals. They're both trying to raise money. They need to raise money. I think the story that you didn't bring up, which is going to be a big one, and I'm just throwing up everything here. and then we can go from there. But I think the number one thing that people need to think about is that we're kind of in a quiet period for Anthropic at this point.

They're not supposed to be talking about what's going on. So they have a quote unquote former employee who speaks. They had the ability of saying, okay, we don't agree with this guy. Don't listen to him. We don't. Instead, they supported what he said. I don't know what that means for the risk side of the S-1 that they've done. I don't know what they have.

to disclose. I don't know what liability issues are going to come from this, but this just seems like a whole bunch of noise that is popping up. The one thing I will say to finish up, in no way, shape, or form after what happened here are the compute needs going down. If anything, there can continue to go faster and the focus may shift to safety, but the training side will continue to go as planned. Ram, you've been looking at this space for a while. What do you think of this whole thing? What do you think of Jorny's thesis here? Yeah. I think I agree with Jordi's thesis. A lot of this has to do with Anthropic going public imminently.

So they've got a shift from revenue growth at all costs to cash flows matter. And you see the quick response from SpaceX and Open AI that followed suit. Now notably, Meta did not. Meta can generate revenue to go fund their CapEx investments. And Zuck took more aggressive view that we're not moving fast enough. So part of it, I think, is shifting from a. a revenue cost to a posture about, hey, let's set up for beat, beat, and raise. Yeah, I also agree.

The person who left Anthropic was there for six weeks. People are reading way too much into this. I'm not of the view that like AGI is here. And I'm using Astra. My team is to develop. It still makes mistakes. It's only to work together. We still have a long way to go.

Is it extraordinary what it can accomplish? and unlock significant productivity growth? Absolutely. Do we need more compute now? Absolutely. Is the promise and potential substantially in front of us? Yes. There you go. So as I've been observing this, I want to start with one thread that we can pull on here. And Rahman, Jordi, I think you've been picking out a little bit. I was wondering how much of this seems like, for lack of a better way to put a commercial play. And what I mean by that is it's pretty obvious that both Anthropic and Open AI are spending a significant amount of money, training their models, gaining capacity, like building out the entire ecosystem that they're working towards. And it's not totally clear to me how much

of a long-term competitive edge they have in this space, as we've seen the progress of open weight models and other people building into this space. So one thing that I'm sort of thinking about as we look at this structure is, are they just trying to buy themselves more time by getting the government to come in and slow things down, to cool down their CAPEX needs, to make the IPO look better, to sort of like have a hand in the process. The other part, you know, and here's where the divergence of Meadow was very interesting to me, is the big banks for a long time have been great at playing the game of regulatory capture, by which I mean asking for onerous rules that they can comply with because they have the money that start boxing out smaller competition because it's just too call it hard to be able to afford all of that.

Like 12 guys in Arkansas cannot pay for the same team that a J.P. Morgan Chase city banker Wells Fargo can. And that shapes the competitive landscape. So I guess, Chris, you've been at a bank. I'll start with you on that one. do you think the commercial read is plausible here? 100%. Regulation, and we saw this with banks to your point, it forces consolidation.

That's just how it works. Why? Because you put everyone into this box. They have shared fixed costs. Margins come down. And the only way you win is to scale. So it's a really beautiful way of boxing out competition. That's what regulation does.

Is a regulation going to be good for the frontier models? Well, it depends what the regulation says and does. not say. But generally speaking, to me, the biggest issue that I'm seeing is the fact that if you were to take a page from the crypto playbook, we had this brand new technology that came on the scene.

We had existing laws that still are on the books. And when this technology doesn't fit neatly into those laws, sometimes you can get yourself into a little bit of trouble because it doesn't fit neatly. And depending on who's in power, you can have issues. With this technology,

with AI, we have laws. We have rules against hacking. We have civil and criminal penalties. And it feels like maybe in certain cases, maybe the hugging face example is a good example, where some of those laws were broken. And potentially that liability goes right back on to the issuer in this case. I'm not a lawyer. I'm not, I don't want to say yes or no to this. But if I'm sitting in the seat and I'm running one of these companies, I'm very concerned about my liabilities going

forward. And if only there was a way to have a box that if I conform to, I would be able to mitigate those liabilities. I think that's a humongous unlock. So to me, it's less about growth. It's more about liability management and risk management getting back to your S one point. Jordy, what do you think? Am I off base with the commercial theory? No, I think that's part of it. I don't think that's the entire story. I really do believe that they want more regulatory side. These These two groups are ahead. I'm not a believer in open source being able to compete in any way with these models. These are way, way above.

And I don't care what anyone says. I mean, I've migrated off of most of the Chinese models that I was using on my Mac minis. And I really have just started to use the frontier models for everything. Astra, to me, changed my life last week. And I've said that with multiple models. But last week, I built some things from a back. a back testing perspective on the market. I'm doing things in crypto in the exact same way. I put out a video last week where I uploaded a repository into GitHub and had all my subscribers be able to go out

and use it with one click. It didn't make any errors for me. And it did the entire project, which is the first time that I didn't have to be a part of it. And because I've had the relationships with data scientists over the years to build things, this was a new experience for me. So I think these guys have multiple motives at this point. I do agree with Chris that the liability, side is something that they're starting to worry about more because it's out of their hands. We don't have a regulatory framework of what will deal with this. It's one thing to say, we have hacking rules, but where does the liability fall if you've created a model and someone else uses it and does something bad with it? What's your liability? So whether you compare this to

pharmaceutical companies, whether you compare it to tobacco companies, you start getting into a world where what are they actually liable for? Where they're not liable for? I'll say the same thing, Chris, that I'm not a lawyer, but I think there's a whole bunch of reasons in here. And I do think for Dario, he's made it very clear. They're large donors to the Democrat Party. I think there's absolutely a political issue here, too, that you can't run away from. It's definitely part of the equation. Yeah, one more thing. There is a possibility here that if this doesn't get sorted, that you see a move to decentralization. Why? Because look what's happening with decentralization today. if we can get decentralized check, and we should talk a little bit about clarity.

If we're at a point where some of this can be decentralized, typically the liability falls back on the user, right? You know, guys, like this technology is dangerous. Well, so are cars. People die in cars as well. But it's the way you use that technology, and that's how you should hold people accountable. It's much different than a centralized model where I'm creating technology and I'm accountable for it. So, you know, one of the things I'm hopeful for is potentially this helps on the decentralization of some of that stack.

And I think it's fine. You know, then the onus is on the user to, to create it reasonably and be responsible. These are like parallel markets, though, right? It's like you need so much capital to play the game, get the compute, you got NVIDIA to back you. I thought it was interesting, Jority, that you're saying that open weight models don't have a kind of credible seat at the table. Navidia's backed hugging face and poolside with throwing like a lot of capital.

Jensen came out this weekend in a very humble way. He says, I'm not as smart of these other investors. I don't like to lose money. And we're making these bets. So it seems to me like open weights have a credible opportunity. They're not going to be ever as good as Frontier Lab models. We're going to be a lot cheaper. And at some level, you're solving for value.

And why wouldn't there be some substitution when the open rate models are good enough? and maybe that's six months or not, maybe that's a year from now. You know, those open-point models have an astro capability in, say, nine months' time plus or minus. Yeah. No, no, no. So let me rephrase it because I agree with you, Rom. It's not that they don't have a place.

They absolutely have a place. But the same way that Hermes has a place with a bunch of scarf places that cost $2 on Amazon, you're not going to bring the price down of anthropic and open AI. It's just not going to happen. They are going to focus their attention on the hardest things. in the world, they're going to own most of the compute. I mean, their margins are enormous. I've, you know, I've been on multiple calls where one of the biggest mistakes that I think investors fall trapped to is all of these things of just not having people that do any research

on this side. The amount of money that's going to be borrowed by these companies is enormous. The credit, I hear about IG all the time. It's the interest rates don't matter to these companies. They have no impact on their margins whatsoever. I mean, I showed something on my vision. video last week. When you go in and you realize how much their margins are and what their cost for data center is, an interest rate is an odd lot of semiconductor, memory prices, everything else. So if rates go up to 7, 8, 9 percent, it doesn't stop the AI movement. And so at that point, with those kind of margins and with them running away with intelligence and the only way that the open source keeps up is by distilling what they've done, which means they're always behind,

but you still need to compute at the end of the day. You can't undercut things and not get the compute. So for at least the next three years, I think they just have a huge advantage. And I think that's the duopoly that we're starting to see form. So I think Elon must jumping into this. I think he has a vested interest. And I do think Elon is worried about this stuff because he's talked about the human race being extinguished. But give me a break. We're talking about the human race. I mean, literally, this was a comment that was made that spooked people. And I got a bunch of inquiries on. I'm like, do you guys really think about what you're listening? How would you destroy everyone on the planet? Like, what would have to happen? Where do we have to go? So I just

think this whole thing is a lot of drama. And as I said to someone today, a week from now, when the Fed's done, the Clarity Act is done and we're past this, no one's going to be paying attention to what. If you like this segment, please like, subscribe, and tune in every Monday at 4.30 p.m. Eastern Time. I'm Austin Campbell, the host of Bips and Bips, along with my friends, Rahm Alawalia, and Chris Perkins and our slate of exceptional guests. Every week, we're going to discuss macro, crypto, crypto, and the collision of worlds, covering topics that move markets and shape the financial landscape. $540 million. That's how much concentrated liquidity sat idle in a given week in the first half of this year.

About 30% of the Defi TVL, if you're wondering. That's according to Dune Research commissioned by One Inch. But there's a solution. One Inch Aqua is the new shared liquidity platform. It lets LPs back multiple positions with the same token balance and keep their tokens in their wallet till a swap comes. Why does that help? Because LPs don't have to split their tokens across positions. They can cover more market conditions and pairs with their full balance.

That means more activity across deeper liquidity. See how it works at 1inch.com slash aqua. Remember that providing liquidity carries risk and fees aren't guaranteed.