The Fight to Defend Bitcoin From AI | Rob Hamilton

0:00 and the idea that people are losing their life savings with hackers like it it guts me it was a horrifying event if you want to do self-custody i think a very fair lesson we learned from all this was that people were doing things they did not fully understand and they did not understand the risk trade-offs and they their trust exceeded their risk tolerance so like everything needs to be tested like your backups aren't actually backups if you don't verify that they're working right the thing is like i just want people's life savings to be safe and i think an important part of that is having the knowledge and consent of how that money's being stored um and however you want to go down that journey like i'm i just want to help genuinely rob hamilton i'm uh honored to

0:48 have the ceo of bitcoin on the podcast thank you i actually got you a gift oh did you i thought this was this was only appropriate oh yeah this is i see you've been in touch with my lawyer matt keita my requirements this is your rider this is my ride this is the first time of cheers the white monster on the podcast but not the last um you've had a crazy crazy couple of months yes so i spoke to you when all this was kicking off this was early august what what date are we talking here when i talked to you or end of july maybe so july 27th kimmy k3 went open weight july 29th that night overnight was the cold card hack and then by the morning of thursday july 30th people were talking about it in group chats and trying to figure out what was happening and by

1:35 you know in the u.s by like early afternoon late morning it was kind of pointed out so this i called you i had bitcoin on a mark four um and i was like this isn't real is it i'm fine you replied instantly being this is not a test move your bitcoin right now and the whole premise around bitcoin self-custody the thing that everyone thought was true was sort of shattered yeah um i want to take a step back though because bitcoin works yeah this was a vendor specific bug right so i think it's important that like and i think this is maybe part of a conversation and like the fallout of this is having like awareness of the like the different elements of software that allow you to use bitcoin and understanding um that even if there's an issue in one vendor like the core

2:30 library itself is strong and secure no absolutely and worth pointing out but what i mean really at that point is the narrative around everything was kind of in disarray yes and so talk me through what happened over those next few days because you obviously ended up spinning up the red team but tell me about the first few days of this happening.

2:46 Yeah. So I'm going to go slightly back because the story starts, I mean, for the year prior, I had been using AI models to scan our code base at Anchor Watch, right? And around May, they had started doing cyber access programs between Anthropic and OpenAI, which allowed you to get more access into the frontier models and with less blocking right uh if you were using most people never experience this but if you you do use these models and you start asking very specific questions it will turn you away they don't even have to be that specific like you can ask what is mitochondria and you get blocked i saw a lot right like that was my favorite was wise mitochondria the powerhouse of the cell i had a

3:32 tweet of that and i got instantly like like bioteric you know biology like concerns and so That was most egregious with fable 5.0 the original anthropic new generation model um And There's other like their cyber security there is because all this distilling that's happening these open weight models If you say like show your work I want to see your reasoning that will block you because that's how you distill you can see the chain of thought of how the The frontier models are thinking and they take that and use that for training the open weight models, right?

4:04 and so I had been doing this for about a year like i said a year in may they were all these access programs were coming available i was scanning my own code base i was scanning other code bases i was poking around and uh july 27th that monday specifically is important because kimmy k3 was the first open weight model that had significant ability to do cyber scanning stuff it like really powerful and i could talk a little bit about the technology and why that is but uh that monday i started scanning my code again and things that had been very diligently sitting on top of were finding new things i had never seen before and it was a moment of like an inversion where i had toyed around the open weight models prior to that they weren't really up to snuff you could always use

4:56 the frontier models and be able to get what you need but this was the first time that i was paying good money scanning my code base with the leading AI companies models and this was wiping it like all over the place like it was like absolutely dominating it and so around this time too just to set up the timeline like i think it was tuesday or may have been wednesday morning like you know maybe 12 hours before the hack um like seeing people talking about kimmy k3 and cyber abilities like how this is really good and this is kind of like a little like precursor like uh cali had a joke about uh the tweet was uh kimmy Kimmy Kimmy can't you see like he was doing like the Biggie song. Yeah, and I was like meaning back. I was like a model so nice. They named it thrice

5:40 because it's Kimmy K3 right and we were like anyone who was touching these tools realized immediately something was like big happening. I think Callie was scanning Cashew stuff. I was scanning anchor watch stuff like it was very evident and then that Thursday morning. I want to give a huge shout out to Talip who is that LX Dev. He's a former Citro Dev, but chatting with him uh and a couple other people that morning we're like we're seeing like thousands of addresses move like it's very unlikely someone remotely compromised a thousand or like hundreds of people's wallets yeah this had to be an entropy bug how did you even see that because it's not like you're constantly scanning the mempool well the first the first transaction was all to one

6:22 address okay so you see a massive sweep of all of these funds from UTXOs that have existed for months to years and like you know four or five years like all moving at once like the two options Either it was one person who had a very large wallet who did a massive consolidation or it was many people that were being attacked And then I loaded up Claude code Said hey, here's the cold card firmware I want you to evaluate this with fable instant downgrade opus 4.8 and even that it wouldn't even talk to me They were just like kind of turn away saying it can't help you with this Then I loaded up Kimmy k3 and I spent a dollar and I got the full answer in like three five minutes and that was like another like oh shit moment of wait a second i had been diligently doing this

7:07 internally and it was the rest of maybe other parts of the ecosystem hadn't been doing this level of like rigorous like beating up on it to try and find out what's happening right and so that's where everything kind of went off to the races and for the first day or two i was doing more intense scans looking at my code base like anchor watch was not impacted totally fine and uh We had some customers with cold cards. They've already all migrated off with no losses. I don't know of a single multi signature loss that's happened related to any of this and the I start scanning other code bases in the ecosystem. I just start throwing darts against while my car. Let me just you know very naively Get go to the code base clone into my local computer open up an open code session and open code for those who don't know is an

7:53 An open source harness think like a Claude code or a codex if you've used these tools But it's fully open which means you can customize it however you want and additionally you can plug in any model you want It's actually great even if you don't want to use the open weight models because you can be like I'm gonna talk to chat GPT for a bit and then without copy and pasting into a new session You can just change the model midstream and talk to something else right And so I was doing that and I started getting some really weird nasty hits on random code bases and I was like, okay I don't want to be alarmist i'm gonna do a responsible disclosure where most code bases have like a means of if you find a serious bug they don't want you to open it up on a github page because then anyone who's

8:33 looking at it can then see the vulnerability right so you do like a pgp encrypted email and i did this for like maybe a half dozen code bases and most of them came back with like confirmations of oh yeah this is a serious bug how did you find this it's like i use kimi k3 right and they just weren't like super on top of like the because it was only relevant as of like three days yeah right so it It was like breaking bleeding and it was that point.

8:55 I was like, okay, this is something that's systemic. The cold card was hit like first within the Bitcoin ecosystem. I would even say anyone was paying attention in June. There was a Z cash inflation bug. Yeah, that I think was AI enabled going through the low level cryptography. So that made that actually probably was the industry like starting gun of like systems and softwares have bugs and issues and things are going to break.

9:19 so uh i start going uh i start spinning up servers i start just loading up like 100 like a bunch of like hundreds of code bases and just start like churning through and parallelizing just initial scans um this is maybe after a day this is all initially my personal credit card and from there i'll never forget this um like the next like starting the second day cali messages me and says What's going on?

9:49 How like what are you doing? How can I help you? And I'm like I could anything like literally do anything. I've been doing security research on other code bases for about a day and a half. So like any help like please like I'm trying to I've been using security stuff on my own stuff but not interfacing with other developers and other people in other code bases.

10:05 And that's the other thing too is you scan a code base. You don't have the deep domain expertise of how that code works. So something that may get flagged as an issue may not really be an issue. Right. So this is it's a conversation we can follow up on like if you're the maintainer of these projects how to like manage it uh and cali tells me like are you running servers i gonna do i like yeah i have some servers he like can i help i was like of course and he just replies with his SSH public key meaning like he could then log into my search so I whitelist it he in the servers it works out beautifully because I in America he overseas we just have like a follow the Sun thing he brings a couple of the cashew devs to

10:36 jump in and start helping I start tagging a couple friends like Talip Stu Portland hodl that means yard there's a whole like I could go through the whole list but it was like 20 people that we all just started like hitting stuff hard and that went on for like three weeks and at this point you're spending like ten thousand dollars a day that's always we ramped up at the peak it was like 10 to 20 like it got um it's like one or two days like it got really crazy just like going through um at large scale like really the machine was humming i was uh handing out api keys like a like a drug dealer like here you go here's free like go go go like right and like everyone's just like hitting code um and a huge shout out to um daniela who

11:20 basically started also jumping in to help out a lot with uh the operational side of it right we're like okay we have hundreds of findings we need to start disclosing these right and so just like um like we set up like a whole service to make it really easy to hand off all these findings and consolidating everything and organizing it became like for a month or two like i was working with like two dozen developers and this was just like we're around the clock just hitting it from all angles um and it was you know the busiest two months of my life going through all this so in total how many code bases did you review it was just about 650 damn and of those 650 how many serious issues did you find so we found a total of i'm calling them findings 10 000 findings you

12:08 could say vulnerabilities but like i want to be err on the side of caution of you find something maybe flagged it's more of a conversation with the maintainer of said project to cross review and confirm stuff right uh but we found over 10 000 of those 1500 were critical or high in like the top end of the tier list and critical there means money's at risk typically yeah like critical is usually like a money losing issue so was there a point where you were looking at this you were scanning the code base, finding all these vulnerabilities and like, holy shit, we can't, like, this is unsustainable. Uh, there was no way, the only way, uh, to get the other side was go through it. So it was just, we're going to, we're going to get to the other side of this.

12:52 And that was just, uh, and it also helps when you have a team of like two dozen people. Like, that's the thing, like the red team, this emergent thing that happened. Um, I mean, even more broadly though in the sense of like is all bitcoin infrastructure broken i think all software is broken it's not just bitcoin um we have found and disclosed bugs to the linux kernel that were part of like if you saw like a couple days ago there was like 1300 vulnerabilities um red team found some stuff there we found some stuff in other non-core bitcoin infrastructure because you're just scanning dependencies and finding other things and you report and hand those off it's all software and like software will always have bugs and you have this dynamic where

13:40 we have decades of humans writing software um there's actually alex b from uh our clubs had this really funny tweet he uh he tagged me in it and it was a message of him talking to someone saying like our children will never know that like the days of bliss where we just shipped software and had dozens of zero days and no one ever knew about it right but these uh llms are just masters of refining context looking through all possible permutations and deeply understanding everything to where uh the long-term hope is in the software development life cycle you're going to leverage these tools so they never get merged into production to begin with right i think that's i'm an optimist in thinking that's ultimately going to end fine and well but you know uh this

14:24 This is like paying off a bunch of human tech debt. You just got to get in there. You got to, you know, do some elbow grease, burn some tokens and get to the other side. Bitcoin is a hard asset. It took discipline to build your position and the platform you use for Bitcoin back loan should have the same discipline.

14:40 That's Ledin, the lead sponsor of what Bitcoin did. When the next opportunity comes along, a Ledin loan puts liquidity in your hands without selling what you spent years building. Loans are funded in hours, not days. There's no monthly payments. There's no penalty for repaying early and all the rates are published up front.

14:55 The larger the loan, the lower the rate. Eight years of protecting client assets with a perfect track record and over $11 billion in loans have taught Ledin what really matters. Your Bitcoin stays custodied. It's never let out to generate interest and Ledin doesn't ask you to take its word for it. Every month, Ledin publishes its open book report with all metrics independently reported.

15:14 It's proof over promises. Hard assets, higher standards. Check your rate at leden.io forward slash WBD, which is L-E-D-N dot I-O forward slash WBD. Bitcoin-backed loans involve risk. Do your own research before taking out a loan. APR eligibility and terms apply and may vary by jurisdiction In the world of Bitcoin security isn just a feature It the entire foundation And that where my new sponsor BitGo come in For over a decade BitGo have been the backbone of the entire industry securing billions of dollars of Bitcoin for major funds exchanges and institutions. Following their 2026 public listing on the New York Stock Exchange under the ticker BTGO, BitGo are now bringing that exact same enterprise infrastructure to individual

15:55 Bitcoiners. Self-custody is still the gold standard, but as your stack grows, it's worth thinking about whether diversifying your custody can make sense. BitGo pioneered commercial multi-sig and their cold storage combines distributed keys, offline key generation, and geographically separated vaults. Head to bitgo.com to get started with institutional grade custody. Your Bitcoin secured like an institution's. That's BitGo. As Bitcoiners, we spend a lot of time thinking about taking control of our money. But what about where we live and the options we have if the rules change? Having a plan B means giving yourself and your family choices somewhere else you could live or a different system you can access if you need it

16:32 and this is the stuff you want to have in place before that happens that's where expat money comes in they help give freedom-minded people optionality so you're not dependent on the whims of any one government or any one system their founder and CEO Mikel has been an expat for over 25 years he's built his own life this way and brings that personal experience to helping others do the same if you've ever thought about building your own plan b download their free report at expatmoney.com forward slash WBD. That's expatmoney.com slash WBD for your free report.

17:03 Is it, you know, you've seen these disclosures where OpenAI have come out and said, we've solved this math problem, like the Millennium Prize problems and things like this. And I've seen a lot of conversation around this and it's basically with enough compute, you can kind of figure something out. Whether it's novel maths or not, it's just enough brainpower on one thing. Is it the same thing here?

17:23 well you can find issues for sure um if you've spent enough compute i think a good example of this is there's the core library that bitcoin uses for all signatures it's called libsec p256k1 that's the name of the elliptic curve that satoshi picked uh we uh we probably put 20 25 000 dollars scanning just that and like at first like trying to beat it up as hard as possible because that's that's actually like the mothership of bitcoin is the cryptography library there it's what every it's what bitcoin core uses it's what every hardware wallet imports that and just loads that code any company that does any bitcoin custody with elliptic ecdsa like elliptic curves for Bitcoin, uses that code. And the good news is very resilient. Nothing like dock changes,

18:20 small little, but like nothing that's actually a problem. Right. And this goes to the point that all software will have bugs. The question is, is what the severity and the risk is of those bugs and managing to make sure that, you know, you can, you know, get a couple hits in your code base and everything's still resilient. So most of the vulnerabilities you found were in like Bitcoin projects not bitcoin itself was there anything serious in bitcoin itself no no there was nothing serious in bitcoin itself either the signature library or um the actual bitcoin core software itself uh i found small things i've now gotten like four bitcoin pr's merged in over the past two months i've you know been in bitcoin for 13 years never actually contributed to the code

19:04 it was actually a funny anecdote where i went to boot my note up one morning and it crashed when i tried opening a wallet and i just took the error logs dumped it into at this point i think it may have been five six or astra and it was like oh it's a one line code change in c plus plus and so i found the issue sunday morning wrote a pr an hour later it was merged by tuesday right because people like it's a one line change of code it's a very obvious like but like that's not that's a bug but that's not like no one's losing money over that it was just like a weird setting like edge case so like uh there are bugs like you can look at bitcoin github repo and see every day people are making code changes either their feature requests or they're

19:47 cleaning up code or optimizations what have you but uh nothing that was bad or critical i think it's a real testament to the bitcoin engineering culture and bitcoin core of the amount of effort and time and human review and the thought that goes into this project that's been compounding for 15 17 years and you know occasionally bitcoin also does have bugs we've had two down times we've had a couple uh what happens is is every six months bitcoin has a release and the supported life is the past three versions so when you get to the end of that if there are any bugs from those older versions that have already been patched those get updated in the bitcoin core like security block you can see like the vulnerabilities that get found uh so like this

20:28 things do happen, but like, no, like, uh, I think it goes to the strength of just the developer culture and the care and attention that goes into the Bitcoin core, you know, monument, like the cathedral that we're building that like the developers are doing a fantastic job maintaining that. So what would your like general state of the Bitcoin software world be now? Is it in good place after all this work you've done i think it's in a better place i think over the next couple of months maybe a year there's just gonna be a constantly just like monitoring triaging um the beautiful thing is with this first wave of reporting and disclosing things and getting them patched you're greatly reducing the surface area for future bugs yeah right so any bugs that get

21:10 maybe a new model comes out maybe it finds something new um you're gonna have way less relevant surface area because oftentimes what you're doing is you're chaining bugs you have like this was a ben carmen who was like helping out with the red team stuff like you know taking two lows in a medium and making it a crit so you just start daisy chaining these things so you start just like cleaning the cruft and then you just start um increasing the complexity required to execute a serious vulnerability you start removing all those stepping stones you get into a better spot so i i think we're going to get to a place where especially for system critical hardware we're going to be able to do a lot of work to clean that up so that this phase can be behind us you know

21:47 You know, you say that all software has bugs. Software will always have bugs. Is that, does that have to be the case? Like if you get something that is completely robust today using, you know, the top AI models, why would a future one be able to break it? So that's a good question. So there's a couple of ways of thinking about this. One for the AI models themselves, it's way easier to talk to these and say, find me a bug versus is my system secure? Why is that? Attacking and finding vulnerability is an objective verifiable state of i am good the like i i am good in the sense of i have achieved my goal i have found a bug versus like how do you prove the negative right how those are not a bug so there is an emerging field of computer science um you've probably seen some of

22:32 this with the open ai stuff with like lean proofs and formalization in theory you could do formalization of code paths and show that it only executes the exact way you're intending it which is a level higher uh but that's not like you know it's it's still emergent and whatever you don't verify becomes the new area that could possibly introduce an issue or if your assumptions are wrong so uh i think we get to a better place it's not going to be perfect though i just don't think that at least for the next couple of years we're not going to get to a state where like cheap verifiable like third like like proofs are going to be able to exist for every single code path in your software maybe we get to a point five years from now ten years from now

23:16 where like you can just code something up and built into all of that is all the formal security proofs maybe uh but talking about the software landscape today i just don't see that happening quite yet so i know a load of businesses will get like external pen testers to audit their code try and attack it yeah did you have any did you review any code that had gone through those kind of audits and still find vulnerabilities yes um i can i guess i could talk like i'm trying to err on the side of caution but if a maintainer tweeted about the feedback afterwards i feel pretty good about it um there's a a great uh cryptographer who writes a lot of libraries um paul miller he has like skewer and like it's like a bunch of hashing libraries and a bunch

24:00 of cryptography libraries and he did a tweet uh that was like a trail of bits and open ai audit our stuff we have no serious issues and then i said okay let me take a look and i took kimmy k3 and i went through the whole code base and i found a couple things and then he patched it and it's all good now right um this was like a month ago and he tweeted about how i helped him like fix things he issued a patch so like it's out there um but it was fixed um so that was an example so like at anchorage we've easily spent over 100 grand 150 grand like on code audit since we started pen testing looking at the bitcoin specific code looking at our application code um very regular like sock two uh testing like all of these things and what's wild now is what i used to have to wait

24:44 two to three weeks for and pay 25 50 000 for i can get for like 25 bucks to maybe if i want to get really spicy like 500 or a thousand dollars and i get an insane additional level of detail and it's not that the security pen testing firms aren't doing their job it's that this is just so much more efficient at doing it and i can spin up multiple agents and i can have conversations as it finds the bugs and i can explain the context of like architecture of code and all these things it's just so fast and it's so much cheaper and like that's amazing in the sense of the democratization of the ability to access this information um is amazing right does that mean it's over for these external auditors no i think they're going to start leveraging the ai and and

25:28 start building their own harnesses and start kind of leveraging their domain expertise to do more like these models while you generically can just be a professional podcaster so to speak and say is there a bug in this code base and you could find things that people would not have been able to find for years.

25:45 Yeah. If you actually have deep domain expertise in a code base or understanding of cryptography or security, what have you, your ability to go further with these models is insane. A great example of this is I was scanning a very large important code base and I did a quick sweep with kimmy k3 and i found like eight things okay and i went and talked to a cryptographer uh and long time like bitcoin expert i was like hey i found these things and they were like you know they weren't formally part of red team or anything they were i was just asking as a friend like can you take a look at this for me and they i got like three half sentences back and one line of code to like look at and i copied that and i pasted into the uh the model and if you've ever had a

26:29 conversation with someone and someone says something you have like a cascading wave of realization like many things click into place at once i saw the model do that in real time it was like oh wait you're right oh wait you're right oh wait and it started like looping and like and then it like cascaded and found like three more issues and like upgraded the severity right uh which was wild to see and this is like part of the story of is like it's not that like software development's over or uh you know like pen testing and security research is done it's um the the productivity these tools unlock for the most competent hands it is an accelerator on their existing skill and productivity in such a way that it's kind of inspiring in the sense of like if you're super

27:14 passionate about bitcoin security all these things you're gonna be all right like whatever marginal like um average person who could touch these tools like like average person who's using ai and how far they can get versus the people have deep domain expertise how much further ahead they can get now i think that we probably will be in a spot soon if not now already where you it's the gaps even further now which is inspiring like that's an optimistic take on it yeah and it's very obvious why bitcoin was first in these attacks um like you get the bitcoin you have the money there's no there's no recourse there right and the first attacker in the first few days got like over a hundred million dollars of bitcoin um but did this actually reach like a

27:54 broader world because i remember seeing a tweet from some like ai guy on twitter and he was basically saying there was so much fuss when kimmy came out that it was gonna destroy security on the internet i've not seen anything and i think cali had just responded being like lol um like did this reach broader than just bitcoin uh i think bitcoin makes it way more public let's say a hospital or or a bank gets hacked, you don't see a bunch of money move on a globally transparent ledger.

28:21 So I think everyone's getting popped all the time now. People just aren't talking about it. And if you look at the frontier AI companies like Anthropic and OpenAI, they are leaning heavily into the power of cybersecurity these tools have. And I think it's because a lot of CISOs and heads of security and IT managers are seeing these issues knowing that ai is causing the problem and they're looking for a way to fix it and now they're realizing that's a great market to sell into uh so that's i think like yeah but do you think we're going to see waves of massive disruption as these models just keep getting better uh probably it wouldn't surprise me if like utility companies banks like there's going to be more things that you're not going to be able to obscure

29:13 and hide behind like to the point like you hack bitcoin you see a thousand bitcoin move in one transaction on chain everyone's looking at it um banking is different because i can't just like wire myself a billion dollars like that's just how you get the fbi to knock on your door and get thrown into prison and you don't get any money yeah um so they're not doing that kind of stuff but i think they're weaseling their way in to try and but if they hit a power grid and everyone's lights go out that's not easy to hide no no that's my point though utility company or water you know you know, so utility companies, uh, bank, like, like there, it wouldn't surprise me for by the end of the year. There's a couple of like big stories, the follow-ups around this stuff. I

29:46 think it's just inevitable. You know, the guy that did, so the attack on cold card, there was one initial attacker got the vast majority of funds. And then we think there was a few others after that three or four waves. Yeah. Is there any idea of like, has anything been done there in terms of trying to find who that attacker was? Have we got any clarity over that? I have heard, I have not talked to, but I have heard that there was a FBI team that's investigating it um i don't i'm not in touch with the law enforcement doing all that stuff um huge shout out to alex thorn who uh was doing a lot of that tracking and kind of coordinating and seeing the addresses move and stuff and kind of tagging providence of coins and to people and

30:26 stuff uh but i don't have the particular details on that so there's a chance we will get some coins back at some point maybe um while i'm an optimist in many things i think you also have to be a realist and your base case has to be the answer is no and then if something were to appear later then that's gravy right just like a mount gox hack and getting your claim back and 10 years later like you maybe the funds haven't moved yet it's non-zero chance the recovery is going to be messy um but yeah so you said before that uh one of the things that got sort of the narrative that got hit was self-custody's dead clearly not true maybe people need to reassess but clearly not true one of the other narratives i think took a real beating is the don't trust verify thing

31:05 For sure. This had been said in Bitcoin for such a long time. I always felt like in some ways I was left out of that narrative because I can verify like I can read a line of code Yeah So so I had to trust I had to trust in people that could go and review those code bases right and it not that anyone did anything wrong specifically but obviously there was stuff that missed absolutely now are we in a world where everyone actually can verify and should be verifying these things or do we need to rely on some level of trust i think the level of verification has been more democratized for sure um and what's amazing with these tools is you can have long socratic dialogues to get to place of your understanding right you could walk and say like i'm not knowledgeable on this code

31:42 base i'm not a cryptographer i need you to hold my hand through this whole thing and it'll do that for you right so that democratization is incredible whether it's verifying you know software um asking questions trying to educate yourself enabling yourself to do other things like those are all things that i think are amazing for these tools can do um i think that you're gonna want a web of trust of people that you could rely on and have multiple people um i think that was part of the red team effort was to try and provide a sense of stability and reassurance that other people had looked at these code bases um but yeah like i think the proliferation of these tools is gonna if it's don't trust verify you're gonna have to do the work to verify right so the question is

32:34 is where on that spectrum you're falling are you manually recompiling the binaries and checking the hashes and looking through all the code paths and making sure everything works exactly as it said to or do you go back to falling back to a layer of trust of other people and you know that's based on each person's willingness and risk management um i think the idea of multi-signature has been made stronger because of this like it was a single point of failure um it was a very particular single point of failure which was is poor risk management to have that and i think that's a general positive learning from this you know you could have easily had it's insane like you could have had a two of three where like two phones and a cold card and

33:16 maybe a bit more you could have one phone and a cold card and a ledger right or you could use costar unchained or anchor watch like or the big like there were just so many ways that you could have dodged this bullet completely you could have rolled dice um you could have had a pass phrase right um but it was just a single concentrated point of failure that i think that is the lesson to take from here and when people ask me about this stuff like when's enough bitcoin that i should go to multi-seager when's enough bitcoin that i should use anchor watch or when should i you know withdraw from an exchange and i think that most of these conversations are very misguided the reason why is they pick a number it's ten thousand dollars it's one bitcoin whatever it is

33:53 that's not how risk management works it is based on your personal life and risk profile and what is a meaningful amount of money to you right um and the way i've been trying to guide this is like listen whatever that number is cut it in half you should already be well off and migrated to that point yeah because bitcoin doubles bitcoin rips you don't want to be on vacation the price rips and all of a sudden like you feel like you're in a really awkward spot uh and i think that's just a good meant because there is no one answer you can't have one answer like the the diverse different experiences and backgrounds that we all have in life whether it's financial cultural um your own trust network of your friends or your

34:44 family that can help you with like all of these things is so person to person dependent that you can't have a one-size-fit-all option i want to give one shout out to because this person had this cold like 10 years ago it was michael flaxman the multi-vendor multi he called it's the bitcoin 10x security guy um and the whole general premise was having multi like multiple sources of entropy generation of your keys that do the signing do them coordination and not having a single point point of failure and that is just that's good risk management yeah right recent hardware wallet vulnerabilities have definitely made me rethink my entire security setup as i'm sure it has for a lot of bitcoiners and the uncomfortable truth is that no single security product can promise there'll

35:27 never be a problem which is exactly why the design matters bitkey is a self-custody hardware wallet that uses three keys created in three separate environments there's one on your phone one on the bitkey hardware itself and one on a secure bitkey server two of those three keys are required to sign any transaction so a problem isolated to just one key is never enough to move your bitcoin so if you are rethinking your setup right now as i'm sure a lot of people are you should really think about the bitkey the hardware looks great the app and the user experience is second to none backing up and restoring your wallet is seamless and they even have inheritance built in so i really don't think you'd be disappointed so download bitkey today use the code wbd for 10

36:06 off the second generation bitkey which is this device right here with the screen download the app, download BitKey, use code WBD and get 10% off. Every Bitcoiner eventually has to answer one question. If something happened to me, could my family access my Bitcoin? Would my children inherit what I spent years stacking That why Anchor Watch builds Bitcoin custody solutions designed to protect you and your family against real accidents errors kidnappings and even death Every solution includes their inheritance protocol designed to help your Bitcoin reach the people you intend it for. And US-based customers can also add Bitcoin insurance backed by Lloyds of London.

36:43 Whether you're a self-custody expert or want multi-institutional support, your Bitcoin estate plan should not be an afterthought. Bitcoin is only generational wealth if it can be passed down. So make sure your family can access tomorrow what you've built today. Anchor Watch is your custody, your way.

36:58 So visit anchorwatch.com to get started. That's anchorwatch.com. So the things like you're saying multiple entropy generation. I think for everyone that's knowledgeable enough and wants to go and do it, you should be able to go roll dice. This is great. But it's also never going to scale to everyone.

37:15 If we want Bitcoin to be a big thing, people aren't going to be rolling dice. And if they don't even understand exactly why they're rolling dice, maybe they shouldn't. Like I do think that's a reasonable thing. No, that I agree with. If you don't understand what you're doing at any point, you shouldn't be doing it.

37:27 So, and random number generators are very good. Like this wasn't a problem with a random number generator. Like, should we be able to trust that? Because it just feels like to me, especially like trying to represent more normally Bitcoiners rather than like the developer Bitcoiners. Like I feel like there has to be elements of trust somewhere.

37:45 So my first thought, just going through that is Bitcoin can't scale to everyone. like the technology actually like you a lot more people then understood i just want to like um it is a question of the level of seriousness you want to take the issue um that's like the starting place of like and how much is this of your life savings like you're telling me that people maybe have multiple bitcoin and they can't spend the time and it's like a majority of their life uh like their net worth they don't want to take two three hours just to understand how this all works it's not even that they they couldn't like they here's a great could i'm just i'm just thinking about the people that won't so here's a great answer to that i think

38:26 the etfs are great if like i think a really really fair trade-off is if this stuff overwhelms you and you like it doesn't make sense go to the etfs like you because you you have to pick your like you have to pick your poison um i in the fallout of all of this i posted that famous photo from reservoir dogs near the end where everyone's like bloodied and they're all pointing guns at each other's heads that's bitcoin security your options are you're going to do the work and secure it yourself or you're going to trust someone else there is no getting around it in the etfs you're trusting um mainly you're trusting the custodian holding the coins right because it's not like uh if secure like the stock exchange is down your bitcoin's not gone right so like your actual

39:08 concentrated risk is within a handful of custody providers that do this like large-scale institutional custody for the etfs i think that's a totally fair answer like if if um if you want to do self-custody i think a very fair lesson we learned from all this was that people were doing things they did not fully understand and they did not understand the risk trade-offs and they their trust exceeded their risk tolerance yeah like like i think that's just like part of it like um i mean the way things work at anchor watch since we're a required cosigner people are able to roll like get a hardware wallet have default entropy and they're fine because like we have to countersign right it's more advanced bitcoin native smart contract or using a multi-single

39:48 like constant chain bicky like all these other players like there are ways that you can mitigate that risk but to fully self-custody and delegate key parts of the security to someone else is it really self-custody at that point you're really you're trusting someone else to handle that end to end execution for you but it's also self custody yeah um there is no free lunch here yeah and i think a path to scaling through this is either you know etf adoption or uh using collaborative custody services people that can like have your hand and like help you reach out along the way um but like you're gonna have you have like there is like it's not it's not a funny answer like this is not like i think i would i would love nothing more to be like no rah rah like

40:33 we're gonna like no like we have to have a real honest conversation yeah over what's happened over the past two months yeah like and that is like that's my blunt opinion like if you want to be fully sovereign self-custody you can use an air gapped bitcoin core computer if you want to you can use you can roll your dice you can do a bunch of the codex 32 is awesome allows you to do shamir sharing with a pen and paper right and you just you roll dice and like it takes a couple hours but like then you have a single key that you can shard out never touching a computer right like that's awesome like all that but you have to put in the time yeah you have to put in the time there is like there is no way around this i think it's i mean it's really clear that people were

41:13 doing things that they were not comfortable with i i had only a couple i know people like sessions i'm sure yourself had a load of people reaching out being like help me move this money oh i had dozens i had i had two people reach out that wanted help moving bitcoin off a cold card and if you need help moving the bitcoin off a cold card you should never have had that bitcoin on a A thousand percent agreed people like being in a position where you're not comfortable using your device that you're holding your life savings.

41:36 Like this is like maybe the best analogous example. is firearms right like firearms are this amazing tool for protecting your property and your life you don't just buy a gun though and leave it like tucked away and never use it you have to be comfortable fire you have to go to the fire range right so like doing test transactions right doing dry runs testing your backups and your recovery systems right oh this failed like uh or an inheritance right uh my co-founder becca like had this story where um this is pre anchor watch like she had bitcoin in self-custody and she explained like had the recovery to like the friends and family and then like six months later she's like oh hey by the way let's pretend

42:23 i'm not here how do you get my bitcoin and they were like dumbfounded yeah right like they couldn't answer the question so like everything needs to be tested like your backups aren't actually backups if you don't verify that they're working right and so this is all just part of general security um verifying and being fully self-sovereign right there are many ways that you can have that self-custody because like i think what's important with self-custody is you are removing your correlation with the rest of the bitcoin ecosystem if you leave it at an etf you're putting your full faith and confidence and the security down like the risk downside on making sure the custodians can properly manage those funds and you're like i don't want to trust anyone

43:05 amazing you can do that you got to put in the work though there is no there is no wave your hand get a free lunch and everything's just perfectly secure so pre-call card i had like multiple sets for my bitcoin i'd have bitcoin on like a lightning wallet on my phone it would be you know not that much money i sometimes wouldn't even back it up like it was just it was spending money yeah then i would have a single sig which i liked using just for the simplicity sure never have that much bitcoin on there and then there's like the full multi-sig set up behind the scenes is the single sig part of that dead um no single sig is not dead i think single sig is um single sig is still good you just have to know what you're doing people

43:46 who did single sig with a passphrase were not impacted by this yeah right they probably at the time were not thinking oh wait what if the original seed has bad entropy this is now like my added encryption yeah um but people also use passphrases that were only one word or like the number 21 which those trivially got popped in the hack right um shout out to james o'burn um who helped out some of the red team stuff and also made the ck tripwire site right so he he took a bunch of insecure entropy and started dusting it with money and all right let's say all right i'm gonna add five dice rolls or i'm gonna do a passphrase with like two words i'm gonna do this that and you can actually see over time attackers are increasing you know the rainbow tables like

44:30 basically pre-computing all the possible options to see what happens um you know like you have to know what you're doing right it goes back to fully self-sovereign you want to use single sig you're fully self-sovereign you're not working with anyone else um you have to do that now like there's cool things with like co-signing and taproot where you could do like a music cosigner where like you have a single stick but then someone else like cosigns it with you and you You could have it decay where if the cosigner goes away or turns malicious after some amount of time, you can sort like these things technically exist.

45:01 But now, like what I just explained, most people probably don't understand the nuance of how that actually works. So now, like even if you have like a sovereign, like you have to understand this stuff, right? Yeah. All this to say, like, I think the single sig, like if I'm being blunt, single sig with a passphrase is a very great mountain man, sovereign individual.

45:20 I want to travel the world. I want to be able to like have my money accessible if needed, or I can have a way of like pretty good disaster recovery, but it's a two of two, the original seed and the passphrase. Right. So I think it has a place. I think it's a power user thing and there's nothing wrong with being a power user.

45:36 I'm, I'm a power user, right? I was a power user before I started anchor watch. Um, I also just consumed hundreds of hours of Bitcoin podcast, Bitcoin talk forum posts, Reddit, like Twitter, like looking and learning and reading through all of these things. So like that option is there. um i'm not going to suggest any of my family to do single sig or single stick with a passphrase though like absolutely not yeah um this is like obviously always been a game of cat and mouse in terms of security sure uh but it feels like the game is like on steroids now yeah there's no stopping this this is just going to continue this is like pandora's box oh yeah this doesn't stop it accelerates from here and my my model of how this projects and goes forward is it's going to

46:17 accelerate for a bit more and it's going to get to a steady state where developers are going to integrate these tools robustly in the software development life cycle and they're going to stop these bugs from getting at it and it's not that all bugs are going to go away forever it's just going to elevate the necessary level of complexity to be able to execute these attacks right like that's where it goes and so a load of people during the cold card hack were moving bitcoin to like exchanges just maybe maybe in the meantime because they they weren't weren't confident of what they should do at that point.

46:46 Fog of war Yeah If someone who did that wants to now move back and go into a self custody or a collaborative custody arrangement what do you think the best absolute best design is for sort of the normal person um multi-sig right um i think a multi-sig of some form anchor watch we use the mini script stuff we also have multi-institution custody like those are all different ways and that's like each of those is like its own conversation on the risk profile right i want to have a segregated asset but i don't want to manage keys myself so i want to delegate multiple institutions to be able to custody manage those funds uh versus i want to have full sovereignty i can spend whatever i want but if i lose a key i can work with someone like a cost or an unchained or a big key right or with

47:34 anchor watch like i want to be able to have the co-signing so i know i'm safe the funds can't move less anchor watch cosine so i can do whitelisted addresses and i could do like velocity control so i only send one bitcoin a month or per year like whatever you can hit all these buttons and dials and then also if anchor watch like disappears you have a way that you can just pull the money out yourself after a year right like so these are all different things in the risk profiles and the different ways you want to manage it um i think any of those are great options if you don't feel confident and i'm going to say if in that fog of war you you did not feel good about your custody set like the storm came and you did not feel safe you have two options you you either have

48:14 to not do self-custody because you were weighing over your head and you got like in one word maybe lucky or you could and you and because of that use a collaborative custody service other was just leave it at the custodian or an etf whatever and the third option is is you sit down and realize that you deeply care about the mission of bitcoin and maybe up to this point you had neglected some of your required homework and your proof of work and you got to sit down and do that work and then migrate your way out i did it may have been over a hundred but like easily dozens of dms of people messaging me and asking about this is my setup am i good or am i not good right and i did a couple posts that went viral about like here's the full threat model if you're using a

48:58 a cold card and like all the decision tree branches and like what i would do if i were in your shoes and like all these different things um seeing all of those stories um i mean to be honest my first thought was like the level of like panic and anxiety of my existing customers versus people that were off on their own like we had a couple people with cold cards they did a quick rotation out clean good to go everything's fine um but people were like you know i i was doing this like i basically like volunteer to be like tech support for bitcoin because and that's what people are asking like why did you start like i just i care about bitcoin right and the idea that people are losing their life savings because hackers like it

49:44 it guts me like it is i think i think there's a lot of like i mean a word is trauma like it's a very um it was a horrifying event and i was going through and answering all these messages because i care and i was like you don't have to use my product like literally just go anywhere and that was my point of like send it to an exchange because most people like oh i don't have the hardware while i need to do something now it's like send it right now right um like hearing also how some people were running their self-custody setups was also scary yeah hearing how some people were managing their life savings was frightening and i think that's like as a reflection point of learnings from all of this is that like you things that i never would do

50:33 and people were holding you know millions of dollars per person that were like holding this stuff um i think this is a wake-up call and kind of reflect and think about these things and finding an answer and uh i have opinions but you should also go forth and think for yourself like there's you know there's not going to be a one-size-fit-all for everyone you got to think through this stuff and if you don't if that cognitive load is too much or it's too scary you believe in the price thesis of bitcoin but you're like i'm like the you could always go like my grandmother who has an estate and she she believes in bitcoin but she's not gonna it's like okay fine like use the etfs and i think that's a totally you know i keep going back to that but like i think

51:12 like we have to just be blunt and honest about like this whole ecosystem and you can there is a side of this where like the code gets so secure and you get so good and you get to this place but if you're not gonna like run shotgun alongside it and understand how it works then like maybe you're gonna have to and it's funny i i was having this conversation with erin and casey on the hell money podcast and erin's counterpoint was like well rob people leave their money in etfs and bonds they have no idea how that works so maybe i'm being overly hard i think that's a fair like maybe i'm being overly hard and risk averse swinging the pendulum to the other side because of what happened but also like just bitcoin is such a unique asset that like it's math yeah and

51:55 like munitions and cryptography and math like is a weapon of sorts like it like in the in the corollary of like firearms training like you have to do your drills you have to understand how it works and if you can do that then like you you relying on other people and uh making sure there isn't some sort of cascading failure that pierces all the way through and you lose your money you know obviously you use miniscript with anchor watch compared to like a cast or an unchained where they're just doing multi-sig yeah how do you trade off the the added complexity there obviously there's an added feature set as well but there's added complexity and like when it comes down to these finding vulnerabilities does that not give you

52:32 another attack surface oh no uh well so let me be direct one directly yes also uh before the incident and even post since then the amount of additional scanning um it sounds complicated the user experience is very cleanly abstractive interacting with a hardware wallet where it's just like signing a multi-sig for all of them it's an identical user experience when you set up the wallet um for the complexity the beautiful thing is is that bitcoin script is so beautifully simple it has rough jagged edges that you have to be careful of but like the way the anchor watch scripts work is says okay anchor watches keys is a two of three did anchor watch sign yes or no if anchor watch didn't sign it says oh okay we're using like the sovereign path anchor watch doesn't

53:20 exist it's like another like signature check and then a time lock very simple i could diagram it out on a white board very in like most of the complexity is explained to someone who's not an engineer how bitcoin stack like the stack works in bitcoin script and all that stuff but the core logic is like if else right it's not rocket science it's also like how the lightning network works where you have one spending path else this like um it's not while there is extra of pieces in there it's beautifully simple um and like the same object multi-sig that secures anyone else's multi-sig is what is the backbone of the manuscript stuff you're just adding an if else statement and the whole original thesis was that like bitcoin uh multi bitcoin custody

54:10 in general self-custody enterprise custody had not evolved since 2014 so 2014 is when we got paid a script hash and like mike belchia bico like really champion pioneered like two of three and three of like all these things and then like nothing happened for a decade and this and people are like oh well you need an upgrade it's like well the beautiful thing about this stuff was that it didn't require an upgrade and it's something that is very trivial to do i think especially if you're a large institution i view it as somewhat of a no-brainer like why would you not want extra redundancies and services and like contingencies that allow you to not have a single point of failure and with our model you as the end customer can lose your keys but not lose your bitcoin

54:45 because we have recovery paths that can return the bitcoin back to you so when you look at it compared to a normal collaborative custody setup it's more secure my opinion yeah oh absolutely i mean and there's like other projects too so like if you want to use the mini script stuff you don't use anchor watch uh there's the on a wallet fully open source and that does degrading multi-sig or expanding multi-sig you can have a two of three that becomes a one of three or two three that comes a two of four right like these are all things that like is that as long as you're able to the introduction the the key piece here is the introduction of time as a security thing as an element of your security uh that the passing of time can change how you spend your

55:22 bitcoin you have one bitcoin address but there's many ways you can spend from it is like the paradigm shift um and it just allows you to have contingencies it allows you to what if this what if that and if you have a static two of three three of five you know four of seven whatever it is your model is like i must always be able to get n number of shares uh this stuff just allows you to have more optionality i think that's awesome and the fact you could do that in bitcoin today in a secure way is like very exciting yeah um i've been i i wouldn't say i've been friends with mvk i've known mvk for a long time um i didn't know him anywhere near as well as you have yeah have you spoken to him since all this happened yeah um i think i can't confirm because

56:05 i don't know who else he talked to i may have been the one to report the vulnerability um after talip got it confirmed i loaded into my own clinker triple confirmed and i sent it over to him um since then i think the red team also did some like because we scanned everything in the bitcoin ecosystem i think there were some follow-up just like disclosures and stuff i haven't got on the phone with him i haven't seen it's like i haven't seen him seen him um yeah what's your feeling as someone who runs like a Bitcoin security infrastructure? Yeah. Are you like, what's your feeling towards him right now? Because like I'm pissed. And one of the things that I'm really annoyed about is I feel like there's been a sort of gaslighting campaign where he's been

56:50 reposting every vulnerability found in anything in Bitcoin. And it's like, look, it wasn't just me, but this didn't come down to like AI exploiting some trivial thing. Like this was the core tenant of what it meant to secure people's Bitcoin. I don't feel like there's comparables there with some of the other stuff that's happened.

57:08 Yeah. You know like the five stages of grief I think those were put on hold for the first week or two as i was going through everything but then going through that whole wave um like the whole like all of the steps i don't i mean to be direct it's an unacceptable bug yeah like catastrophic failure not not acceptable i understand that building things is hard i think about it if i were to start building a hardware wallet like that was like my thing i started autistically spiraling the entropy would have been most of the effort because any other bug in a hardware wallet or in a cryptography system that isn't the entropy, you can issue a patch that fixes it and it doesn't systemically hit everyone at once.

58:18 Yeah. The bug though, and the way it was introduced and the way like all of this complexity, it's just not acceptable. Like it's just it just isn't. I've I don't obviously envy his position. My heart goes out first to everyone who lost their their their Bitcoin. Like that is like that's like the waves of grief and anger.

59:00 My heart first is with all of those people. Right. Many of them my friends. um as for him like in the coin kite team him and doc as someone who's building a company it feels horrible in the sense of like man uh you lost everything you've been building for the past five six years right um i'm i know there's a lot of tinfoil hat conspiracy stuff of this was an intentional backdoor um i think it's way more obvious to point to looking at the structure of it all that was probably you know error you know right like it wasn't something that was maliciously there um i agree with that i don't i don't believe the other side yeah but the thing that i do think is concerning is i know james i've earned posted a chat he had a few years earlier right or at some

1:00:00 point earlier where he had basically asked the question if this was a thing and and it seemed to get rejected from a point of arrogance rather than acceptance and actually going down and seeing if this was a real thing yes um a thousand percent james fingered like and like an engineering term like the code smell like like like basically you look at you're like smells rank like this is shit um and he was obviously right having done these disclosures now large scale their reaction is not unique interesting not unique and it's not coin kite is small so they're arrogant and every big player is good it is a complicated every code base is different i mean this was not a bitcoin related bug but we found a bug in a library reported it the maintainer told us that like this is slop

1:00:59 stop giving him crap to look at that's wasting his time and then 12 hours later the next morning is pushing commits fixing the exact thing and no follow-up like hey sorry about that sorry about that you were right right so they're like i think developers and people who are maintainers of libraries it is an egoic effort to be like i am going to build software and do these things right so i think there's what you need just really strongly compliment that is a security culture that understands that more eyes and work make things stronger and even if it's a false alarm it gives you deeper conviction how everything is set up and i think that's uh what's happening now just in general across maintainer code bases and you know companies is this understanding that like

1:01:46 it's a whole new frontier and everyone has to kind of level reset their their perspective and how to do this stuff yeah there was another conspiracy theory that was going around which i i still don't think i believe but it's more believable which is that they did know this like you know with people like james o'burn coming out and saying this they they potentially had some idea that this might exist but fixing it's kind of impossible like what do you do if you find this bug first and you know it's real like what do you do so i mean that is the conspiracy backdoor Right, because like you could say either it was by accident or on purpose, but they were sitting on this knowledge and knew it.

1:02:21 That means at that point, like they're complicit in a cover up, whether it was from the start or it got introduced later, it wouldn't. surprised me that it was just outright dismissing like arrogance if they knew about it how do you fix it like there is no good solution yeah there is no good solution you have the option of stealing the funds yourself and trying to reimburse people when they do some sort of proof claim update you have um that's really like the one way you try to do your best to get everything back to everyone you have which some people did in the cold card hack i'm not sure if you saw that like i think like 50 bitcoin white hats came in and now i got to like a trust somewhere went to a trust in wyoming yep uh like a digital asset trust that specifically does this which um

1:03:10 is definitely a cleaner way than letting a random hacker take it you're still taking like you're still moving bitcoin around but if you know the game theory is is that i want to try and do harm reduction and get this back to the original holder an hour from like in that moment it was like minutes from now an attacker could find this and just steal it yeah so you just do what you think is harm reduction um you can't even really discreetly tell people how to update because you don't know when people generate waltz you don't know who is using dice who was using seed phrases like default entropy who versus who's using passphrases you you don't know um i would say if they knew about it i'm not saying they did this just stuff that was floating around

1:04:01 well i mean like that's just like i mean that's like criminal negligence if they were sitting here and they knew about the problem that's why i don't think you don't think so i don't think so i don't think so either but um the the only thing that in reality because still like taking everyone's bitcoin even with the intention to give it back that's like i think they're still probably illegal like I don't know how the the law actually works with things neither do I'm not a lawyer and so probably the thing they would have to have done is essentially what happened once the vulnerability got caught anyway is everyone starts screaming from the rooftops being like you need to change what's I mean that you tell two three people they tell two three people they tell two three people

1:04:33 all of a sudden one person posted on Twitter and then you start the warning gun for a hacker to try and steal it and more Bitcoin would have been kept safe in that scenario but it's still yeah terrible absolutely it was an absolute mess um with the trust like with the with the white hat hacks yeah which we've we've obviously seen now how does the legality of that work is what they've done not a lawyer my understanding is there is a trust based out of wyoming um and what they do is they offer an identity agreement to you the white hat saying listen if you find assets and you send them to us we will legally like put the effort into indemnify you and prevent you from getting legally individually hit and the trust takes on that responsibility okay that's the

1:05:20 not a lawyer but that's the high level my high level understanding of how that works okay so the people that actually did that probably are not going to get in trouble no my assumption again not a lawyer don't know the court filings um but you're a mess man absolutely no it's it's like i said it's the worst like the things i would have done if i was building a hardware wallet the level of like spiraling and testing and like generating wallets and going and going and going like the amount of i would have done to like you can run like what are called chi-squared tests which i think are that sample size like you can run it with a dice and prove your dice isn't biased or you could do um you basically can run statistical analyses and tests to make sure the entropy is

1:05:58 good like which is actually cool i found this in the bitkey firmware and some of the randomness that they use there's there's unit tests and chi-squared to show that like you can run it on long time frame and prove that it's like probabilistically show that it's actually using good entropy right like there's lots of ways leveraging math statistics um testing to be able to sure make sure it's good like that is the most critical thing that you have to evaluate when you're doing this stuff like it is the most it's the most important job of a hardware wallet is making sure you can generate clean secrets and like even what they're doing now with these firmware updates i haven't looked at it yet but i've seen like the documentation that like they force user entered

1:06:36 entropy yes which is like you know what probably should have been there from the start as a fail safe yeah not having a single point of failure yeah it's really hard that though isn't it because like again the bug's not in the code the entropy is good um and they want to sell devices their business and if they're selling a device where you have to roll dice they're going myself off your devices yeah absolutely i mean this goes back to um i wouldn't trust my wife with my entropy let alone mbk fat right but that's me yeah that's me that me understanding like all of the pieces right yeah um yeah it's not and there's and dice is just but one way yep um bitcoin core has like four different things it does within its wallet um to be able

1:07:24 to shuffle and do that randomness and entropy like there's you can there's many many many ways the nice thing about physical things where it's dealing cards flipping coins rolling dice like i like what cool is that you physically can verify that the entropy happened whereas like the complexity for you to audit and understand each code path for all of the deterministic operating system calls is insanely difficult and you basically have to run like a stack trace and like see it all happen in real time and like it gets very complicated um like the heap allocations like it gets very complicated i'm not saying it can't be done obviously it's been audited i'm just saying that like for you as a lay person to be able to

1:07:56 understand and verify this stuff that is just trusting and saying oh bitcoin core is reviewed so i'm going to trust that yeah um as a random like shout out have you seen the uh entropy labs project i have i mr hodl and wicked bitcoins like ai psychosis yes yeah they've gone full-blown ai which is fun like it's like it's like knowing mr hodl for years and now he's like coding something like oh this is scary he's taking a break from just screaming everyone on twitter yeah but i have seen it but i don't exactly know what they're building uh i looked at at the start i haven't looked at much more i just think it was funny that like they're they i one i would not use it i just wouldn't use it like there but i think it's like in the effort of for like the whole like

1:08:37 how do you verify yeah mr hodl's path for understanding and verifying this stuff was i'm going to write software that does it i'm going to walk through with an llm and i'm going to build it myself as a way to verify it and like that's that that is not a scalable approach but it is a way of making sure you deeply deeply understand all of the different pieces of the cryptography is building one for yourself and i think that's a you know you you build like a little like a model rocket so you want to understand the physics and like the takeoff and the landing but that doesn't mean that you're going to put your life your family on that like bottle rocket you know i mean anyway so it's like yeah just as a random like for the journey of how you can try and learn and

1:09:11 educate yourself on this stuff um can we talk about the future of ai and security sure because because one of the most interesting things that came out of this was that obviously you were trying to do this with with anthropic with open ai they were saying they were saying no kimmy let you do whatever you want um and we saw loads of conversation where like the us wanted to regulate these Chinese open source models like what do you think is going to happen from this point I have so many thoughts this is actually the biggest story coming out of all this the big stories are one software is broken not just bitcoin it's all software two the posturing and the seeding of conversations around the policy implications with these open weight models

1:09:53 I'm that's when I got really fired up when I was personally experienced this I would find dead to write exploits in code and i'd be like this is broken and i would try and talk to this frontier labs and they wouldn't talk to me and i would use kimmy and the conversation wasn't these models should be more open and let me do more things it was we should ban open weight models yes and that was i mean that that's that's that was radicalizing for me like seeing this and then you start deeply going in through and leveraging these tools and you start realizing wait a second we are just speed running the cryptography wars in the 90s where they tried making prime numbers illegal for rsa encryption like all this stuff yeah they were trying to make math illegal it failed it was very important that

1:10:35 it failed especially now like this is all just linear algebra right matrices multiplication the attention like like all the things that these like it's all large matrix multiplication and doing many many many many rounds of that and you're what you're doing is you're taking words you turn them into numbers you shove the numbers into this big box of many many rounds of matrix multiplication and spits out numbers on the other side and then you reconvert those numbers back into words that's like my layman like high level understanding of this like tokenizing like that's what tokens are is like you take a word and you convert it into like numerical space and hyper dimensional space and you put it through and you get output tokens on the other side and those get

1:11:08 converted back into words now i get really fired up about this because the there's so many things things to talk about one the idea that and I'm gonna I'm wearing my anchor watch hat but like from like an American like a very proud American believing in the ideals of the American experiment in the Bill of Rights and all these things the idea that you're gonna make math illegal for me sitting in my house it's what it's un-American and it's not what a free society does right and you can see because there's like this um dance that's done but like all these frontier models they're so scary they're so bad and then you have these open weight ones like you know like they're even scarier like you have to like you trust us and let us run it we're gonna do a regulatory body where we basically

1:11:58 pull up the ladder behind us um and we're gonna own this like so many pieces here uh one the way it's being pushed as i'm gonna use the it gets overplayed in like politics discourse but i think it's definitionally true it's fascism you're taking companies and the government and you're merging them together and saying that they're going to call the rules and the shots and you're gonna have to go get a license if you want to be able to think in your house right if i own compute infrastructure in my house in the privacy of my home it's not your any your damn business what i'm doing with that that's not what we do in a free society that and the deep irony of this is that this is what china's doing in proliferating all these models

1:12:41 and I think strategically it makes sense because there's so much capital and money going into these like frontier AI companies that it like kind of like a cornerstone of like GDP growth and like productivity and all this stuff they want to basically take out the capital base as a side note I think selling metered intelligence is a horrible business This idea. I'm just gonna pay and you give me thoughts right like because obviously like this stuff is good enough now where it's finding breakthroughs in math and science and physics and all these things. Why would open AI let any of these leading pharmaceutical companies spend a million dollars in inference tokens and come out with this new groundbreaking drug and then like

1:13:17 the pharmaceutical company gets all the upside no no no no no no no anthropic opened up a wet lab because they're going to own it all inside and there's going to be tiers of models of our supreme intelligence that sits over here and everyone else gets access to this wait do you think so you think there's going to be like an ip ip battle in this in the deepest sense of irony uh ip battle in what sense in the sense of like if someone's using one of their models to come up with the breakthrough they're going to they're just not going to let you well this this is what the guard rails are this is what the guard rails are oh you can't you can't ask about the mitochondria being the powerhouse of the cell you shut off entire waves of intelligence from the model because

1:13:54 they're going to go capitalize that because it's a way better business for anthropic to come out with the new groundbreaking drug than let the pharmaceutical company do it you think that's the real reason for the guard rails i think i think it's two things i think one thing is that i think there's a liability thing since these are companies and their model does something bad they are on the hook liability wise whereas an open weight model does not have that and that's why the open weight models and things like cyber security among other things are able to move much faster because they don't have that uh since it's open it's software you can run the software and do whatever you want like they don't have the same love it's the same level yeah they just they're releasing software

1:14:25 and then people do what they want with that software whereas if i'm i have a vendor agreement with one of these companies uh and some and i'm a terrorist and i use it for bad things it's gonna go bad but like like just point blank it's an attempt to corner the market and cause a fat it's fascism like you're gonna take the companies and then you're gonna do a union with with the government and maybe they'll do an intel stake thing where they get 10 like you're just merging these two things together and like that is that's not america um i i don't think they will succeed even with all their doom and glooming and trying like fear-mongering like i made this comment that like i would rather roll the dice and have humanity end than for it to go into some sort of like

1:15:09 fascistic oligopoly yeah of these companies like it's not acceptable like i i call your bluff and i say like you're gonna have to bring armed men into my house and like take my computers away from me before i'm going to yield on this like it's not it's not acceptable um not in a free society absolutely not um and the thing too is like these models if they're thinking right this is this is what's interesting too with their rhetoric they're like they went and lobbied like the pope on these things being human oh so they have consciousness so do they have rights freedom of speech oh no like no no within our sandbox they have rights like they have a leash so they get to like in here if these these models are proxies for intelligence and freedom of thought for my own ability to have

1:15:50 freedom of speech and freedom of thought and expression i need to be able to have dialogues and conversations like freedom of speech freedom of press freedom assembly are all downstream of the idea of the freedom of my own thoughts and that what's in my mind is sacrosanct and you can't regulate that i'm allowed to think whatever i want if as part of a dialogue a socratic dialogue you want to have conversations to build and form your understanding like this is a very critical part of the feedback loop of developing your own thoughts and ideas and if you start saying like oh that is a it's a thought crime that's what you're doing oh this is not a licensed model it's a thought crime you can't talk to it not in a free country not going to happen and it's not uh i don't

1:16:28 feel that the principles of america are subservient to these companies making sure they can hit their revenue goals and locking things down and banning all the competition it's just 100 yeah and like that's and the deepest irony is what they did is they went and scraped the entire internet and took everyone else's thoughts built into a model and now they want to sell it back to us and it also if you distill and you get that thought out and you just distill what they did they distill all the intelligence they have a model someone distills it and they open that up like that's not fair it's like okay well that isn't really how this work is it like you the the these models are monuments to the total sum of all human experience and knowledge it's not it's not

1:17:09 it's not theirs they didn't design this all themselves they took a bunch of freely open available data and spent megawatts and gigawatts of energy to compress and trick sand to think right because computers used to just do zeros and ones and it would be very deterministic these lms are probabilistic which is a fascinating thing it's a whole new paradigm of like computing where if i ask to i want to make a grilled cheese sandwich and you want to make a real cheese we ask it we'll get like maybe 85 percent same but there's gonna be differences yeah right that's because it's not one size fit all like these are random shufflers and for the ai security thing you're just pulling the slot machine and you're getting different vulnerabilities out each time

1:17:49 right um but all of this like intellectual property stuff like uh it's deeply deeply ironic and i extremely suspect of people telling me that like we are the exalted ones that are gonna be able to figure this out because that encodes their morality Their morality is already encoded into the model to extend from that They trying to prevent any other moral system from existing outside of their own. It's an entire shell game of saying technology, like safety and all of this stuff, but they're in the lead. They want to pull the ladder up behind them and they want their moral values, their alignment, their thought and philosophy of the world to be encoded as law and for you to use anyone else to be illegal it is

1:18:35 like it's bluntly un-american and i think that they can't win for america and human freedom like let's even like even take off my america hat just like humanity we're gonna let oligarchs in the silicon valley be the ones who shot call morality and thoughts and systems no we can't have future it can't happen like it's an unacceptable thing i say uh you know mall on lobby like come and take it like you're gonna have to come and take gum into my house men with it's the whole argument of like you know like what if they ban bitcoin they're gonna have to come into my house you're gonna send armed men with guns because i'm sitting here thinking for myself and using my own computers for whatever the hell i want to do like it's gonna have to take that like they're not they

1:19:21 can't win and the other irony it is it is a hill worth dying on like that is the level of how serious this is it is a hill worth dying on i don't say that cavalierly it's just like to restrict like all the thoughts of man and to compress their moral philosophy and project that onto the world and not be able to escape it it's tyranny yeah can't happen won't happen hopefully fired up i love this um the other irony though is that they're saying that we need to safeguard we need to slow down and then releasing a new model the next week like it's obviously not what they're actually doing right um i i agree with you that we can't let this succeed do you think they could succeed to a degree in the short term no because here's the thing okay you're going to

1:20:05 restrict the models in america you're not the world yeah you're gonna it's the same argument as banning bitcoin now yes it's the same argument as man in bitcoin and you're gonna say the one geopolitical player who is a strategic rival to the united states is just going to let them do that and coordinate with them this is where like the whole safety arguments fall apart okay you every model that isn't one of their licensed models in america's band what about the other 184 countries in the world all of the data centers that are going to go up everywhere else and everything else is going to go run and do these things you won't be able to win like you're just necessarily blowing off america's kneecaps to protect a handful of ai companies and their

1:20:54 balance sheets and their ipos and their market value short term you know what's going to happen like the it's just as this continues ramping up i think maybe likely the gap's going to start collapsing on some of these things maybe um it just doesn't work like it just doesn't work and then like you're going to prevent me from acts it's it's like an inverse of like the the great firewall i'm not going to be able to access the outer internet because if i can access the outer internet i'm going to be able to talk to an api endpoint for an open weight model that's illegal in the united states yeah it's like a sick inversion of the tyranny of the systems um that's why it can't succeed i think there's a lot of scare margarine a lot of people who are

1:21:39 concerned when they're like oh there's a we think there's a 10 15 chance that humanity's going to die and it's like wow well it sounds like you have a shit ton of liability on your hands we should heavily discount the value of your company oh no no no because we'll regulate it's like oh no okay so you're saying that there's an overhang that your technology may kill there's a 50% chance it kills humanity but we should value it multiple trillions do you think this is all trying to protect their IPO price though uh oh I think in part I mean like I said metered intelligence is a dog shit business you're just gonna be in like if you just wait a couple months for 10 cents on the dollar you can get what the breaking leading models are today in the privacy of your own home

1:22:13 if you have the computer infrastructure or being able to use these random endpoints like it's it it is where they started and how they had to get revenue in the door um i think also part of this like slowing the frontier stuff also may be the realizing like wait a second i need to we have let's say just to make the math easy we have we have 100 computers that are running the math and at the moment 70 of them are running uh training for the next model and 30 are actually selling the credits maybe we need to make flip that and we need to shut down our research cluster so we can get revenue in the door so we can jack our numbers going into an ipo so we can have more compute available to sell ah interesting right like they continue releasing models absolutely

1:22:53 um they're probably getting more efficient at training the models and post training them and doing all these things uh but you know they're at trillion multi-trillion dollar market caps and now they need to actually show some revenue so they can hit an IPO. So they need to do, they've left the luxury of being a private company with private investors that let them run off on their research experiments and then I need to turn into real reality, which is why I think they're doing things like wet lab with anthropic or i think open ai is also doing now cyber security consulting what's better if you're open ai are you going to do a two-year million dollar cyber security contract for a utility company to secure like a whole region of the united states infrastructure or are you going

1:23:32 to let the it department spend twenty thousand dollars in api credits and fix it for themselves easy answer right so like this is the complicated part but you know this narrative that open source models are we're 12 months behind then six months behind now three months behind whatever it is it was always that they were behind but open source models allowed you to do something that the closed source frontier models could not is there an argument they're actually equal although like differently talented uh no it's the american frontier companies intentionally knee capped their ability to do things like cyber they had the intelligence and this is like an observation that like the code always existed and i took it like a slight spiritual spin of like there's demons in

1:24:13 the code and what was happening for a year not longer the frontier labs were seeing these bugs and they would do post filtering like if you've ever i i saw this happen all the time where i'd start talking to him like you would like hey can you help me find a bug in this area the code and it would go and it'd go and it'd go and it'd go and it'd go and all of a sudden it would stop be like hi i can't help you this is sensitive like this is cyber security stuff and like means it found something means it found something yeah bingo right and so like this was happening for which you know if this is a beautiful business for them they charge me to tell me they can't help me after it spends a bunch of money right and so i think they were training on that data they would

1:24:51 when you get a cyber flag like hey this is sensitive there's also a flag on the back and being like hey we found a bug let's probably look at this yeah start training on it and start looking at it right and um i forgot the exact uh question but like all of this is coming down to It's about how far from air frontier is.

1:25:08 Oh yeah. I think they're gonna pull that like this like breakthrough like math model. We're never gonna get access to that. Yeah, that's my assumption. Yeah, or if we do, it's gonna be heavily heavily lagged. Like just yesterday mythos became available for everyone. I haven't used it yet. I've been traveling.

1:25:26 My thought though is that they now hyper like the game is a foot They understand the distilling game and kind of like the like they gonna just start having models that are never gonna be seen to the public. They're gonna have a biology model. They're gonna have a physics model. They're gonna have math models like you saw this with open AI where the Navy or Stokes like one of the Millennium prize problems.

1:25:46 Yeah, there was a guy who is dumping all of his research into it and then like a month later they find the solution and they're like, oh, well, we found it our own independent it's like well did you kind of springboard off his ideas and then slightly tweak them a bit so people are maybe not even tweak them just throw more compute yeah right to brute force it yeah absolutely yeah so like uh at the moment that's why like they why why are they subsidizing all this cost it's like because they're hoovering up all of the data it's not that like it's facebook early days hoovering up all the data and figuring out later right which is also like a misaligned incentive to run a business on well it's vc subsidized it's not it's not a perpetual business

1:26:21 model so what they're trying to do now is find what the next like they're bootstrapping they're getting revenue in the door they're raising the money they're going for the ipo and they're realizing that the model that got them this far like the business model monetization model that got them this far is not good enough to get to the next rung so they're trying to figure that all out is there anything that i've not asked you that you really want to talk about um let's see i think we went through the high level red teaming free open source ai um yeah like this has been the busiest two and a half months now of my life like by an order of magnitude and i was talking to um someone i was talking to lawrence who's a lawrence is a

1:27:05 big one core contributor and he said well you know my family would always say i was like what's that he said if you want to get something done ask a busy person to do it because you kind of just raised to the level of the level of intensity. You just find ways to spin more plates. Um, I'm trying to from here on the way out find, uh, I can't be doing this forever.

1:27:30 This is not my job. I have my own job and obligations and people are always like, oh, can you scan this? Can you scan that It like at a certain point the warning gun has been fired You have to start being responsible for yourself and looking through code bases i can be there for everyone forever i doing this entirely free so as a shout out like open sats and hrf both gave grants like there's an open sats red team fund if you want to donate specifically for cyber security stuff other people beyond like this initial bitcoin red team are now scanning and looking through code bases looking at through different ways um i want to get to a place where i can enable others to start thinking about this and that's why like i'm very open about these conversations

1:28:08 and what i'm doing in scans and looking at um i want to just get to a spot where maybe more people are just being more proactively security minded and uh figuring out this new era we're in how to navigate it brave new world indeed oh this is the most important part yeah you gotta get that bow tie off yeah see like i i hand tied this it's not a clip on i've never hand tied a bow tie oh well you just got to get to i'm going to borrow after and see if i can do it oh well we should you should bring back the patreon for like the subscriber only content yeah so you could show everyone that because that would be funny wearing a bow tie with my t-shirt um rob thank you man thank you for coming on the show but thank you for all the efforts over the last two months uh it's incredible

1:28:51 appreciate it the ceo bitcoin now that's right you need to get rid of that title and go back to your real job uh that and my family yeah i got a lot of uh you know i gotta get i get to find this new normal and balance um anyone who's working on any of this stuff i'm super approachable i want to help people um like you're able to reach out to me um rob one ham rob the number one ham on twitter rob at anchorwatch.com you know i want to help people wherever they're at even if you're not I'm not using my company or business.

1:29:25 I just want my the thing is like I just want people's life savings to be safe. And I think an important part of that is having the knowledge and consent of how that money is being stored. And however you want to go down that journey, like I just want to help genuinely. So don't hesitate to reach out if you're doing security stuff.

1:29:43 More than happy to talk shop on how I doing things Yeah it a brave new world indeed and we gonna just you know i think bitcoin is worth the fight to get to the other side of this i think on the ai side like freedom is worth the fight it doesn't mean it's going to be easy and i think preserving and making it easier for people in self-sovereign ways to store and secure their bitcoin is worth the fight in the mission and things are getting harder on that security side but the opportunities and the ability to build things is also getting easier so i think um we are truly now in an era of you can just do things as i've been telling people i've been doing security research now for uh two and a half months so i'm somewhat of an expert uh

1:30:31 i just you know found there was a gap in what was not being done and stepped up to help Because I care about Bitcoin and I would just encourage people to. We live in such exciting times like never before has one person and like been able to do so much in general whether it's like you're going to start a company you're going to start scanning a Bitcoin code.

1:30:53 This is and also like I'm just a guy on the red team. The red team is not an organization. It's not a company like there are two dozen developers who have all done amazing work and many hands make light work and uh just a huge shout out and a gratitude appreciation for everyone who along the way has been spending their own time away from their family away from their jobs and just kind of leaning in to try and help bitcoin and um i think it was a really beautiful uplifting thing getting to the other side of that um and seeing that you know just people want to care and they want to do something and so um yeah i think we just got to get to the other side of this and see what comes next. It's incredible, man. Thank you for

1:31:37 all the work. If you're thinking about your Bitcoin security, check out Anchor Watch. Let's go.