Bitcoin’s Biggest AI Test Has Arrived

Cathie Wood (0:00)

The combination of domain expertise and AI is going to be unbeatable.

Dhruv Bansal (0:05)

Historically, I feel like defenders have often suffered from their domain expertise, like at least to a certain kind of domain blindness. There cannot possibly be a bug here because I wrote that code. That's the explanation for a lot of stupid things that get overlooked, right?

Rob Hamilton (0:19)

Humans just inherently have bugs in software. Like even Satoshi has bugs that he wrote in the original Bitcoin client.

Lorenzo Valente (0:25)

Free AI code is from kind of an other era. Everybody is going to leverage these tools.

Rob Hamilton (0:32)

Leveraging of this technology is actually one that empowers individuals with ambitions in a greater way to enable freedom that was not accessible before.

SPEAKER_5 (0:50)

Welcome to FYI, The For Your Innovation Podcast.

This show offers an intellectual discussion on technologically enabled disruption. Because investing in innovation starts with understanding it. To learn more, visit arkinvest.com.

SPEAKER_6 (1:08)

ARK Invest is a registered investment advisor focused on investing in disruptive innovation. This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. It does not constitute either explicitly or implicitly any provision of services or products by ARK. All statements made regarding companies or securities are strictly beliefs and points of view held by ARK or podcast guests and are not endorsements or recommendations by ARK to buy, sell, or hold any security.

Clients of ARK Investment Management may maintain positions in the securities discussed in this podcast.

Rod Roudi (1:42)

I'm just super pumped to be back. Wow, a lot has happened in just a month. I know, Cathie, I don't know what the term you call it in AI, you know there's dog years or whatever, there's like AI years or AI days, and it feels like I live like-

Cathie Wood (1:56)

Warp years, yes. Warp days, yes. It's unbelievable. It's unbelievable. Yes, it is unbelievable. Well, when you get the cost of something dropping 99.99 percent per year, that's inference costs, you're going to get a lot more of it.

And, you know, a lot of people don't need to be right at the frontier model, you know, which is the more expensive way to go about it. They can be n minus one. And, you know, we're getting a lot more of that because corporations are trying to figure out, OK, you know, we don't we do need to be right at the frontier to answer a simple question. So, you know, I think that's one reason a lot has happened in a month.

Rod Roudi (2:43)

I couldn't agree more. And, you know, one of the big things, and I would love Rob, just to explain, Rob Hamilton, who was on a couple months ago with us, just to explain what has happened in the Bitcoin space, because I feel like us Bitcoiners, we just got the first foray into the AI versus AI kind of battle.

And I think you, you, especially in the team, have set a really nice precedent and blueprint for others potentially to follow.

Rob Hamilton (3:09)

The way I like telling this story, as I've been trying to, in reflecting back on it, the important timeline was Kimi K3 went open weight, meaning that anyone in the world could run the model if they had the infrastructure and the dedicated compute to host it on July 27th, which was a Monday. Over the past year and a half, two years as AI has gone better and better, I very regularly, when a new model comes out, would look at my own internal code and understand whether there are any deficiencies, whether there are any issues that should be fixed. Humans just inherently have bugs in software. Like even Satoshi has bugs that he wrote in the original Bitcoin client. It's just a very natural thing, and this is just part of hygiene and maintenance in running and writing software.

What was so drastic about Kimi K3, which is an AI research lab and company based out of China, was for the first time, I saw a massive divergence in what the open weight models were willing to do versus the frontier, and also in this way that it almost has a personality. It is tenacious. It gets in there and it starts really looking for software vulnerability issues. I'm not sure if the model was specifically trained for this behavior. Just inherently, its aggressiveness and going through and finding problems was something I had never even seen from the American Frontier Labs.

So Monday, July 27th, I went open weight, started scanning my own code base. Fast forward a bit to Wednesday, July 29th, that evening, about I think at that time, a thousand Bitcoin had moved from a particular set of users in the Bitcoin ecosystem who were using a specific vendor called the Coldcard. And very quickly, reports started surfacing. We're now like Thursday morning.