Yeah, thanks, Nik. Well, very, very tough few days, I think, for the Bitcoin community, obviously, for, you know, many people affected directly that have lost funds in this Coldcard key generation vulnerability. You know, I think just starting sort of with some of the facts as I know them, and then I'll talk about some of the tracing. So, on March 17th, 2021, Coinkite upgraded their firmware on their Coldcard devices to change the key generation method to something they called libng, which apparently was good. But the software that, you know, wired in that number generator failed and was not connecting to the number generator and started defaulting to a different, much weaker number generator on the device, but it was failing silently. So it wasn't throwing an error to users when they were generating keys or anything like that. So shockingly, and I think this is a huge place for the Bitcoin and security community to learn, shockingly this went unfound for five years. And this is not the first key generation weak entropy bug in crypto wallet history. There have been several others. So, I think one of the biggest lessons and takeaways here is that the crypto industry and the Bitcoin security industry needs to be substantially more proactive in verifying the software, right? Remember, don't trust, verify. Well, nobody verified this software sufficiently. There are some intricacies in terms of the history of Coinkite and free and open source software that I won't get into. But suffice to say, I know the whole Coinkite team is devastated, rightly so. This is a very, a truly, truly devastating bug. One of the things that makes it so devastating is that, and, you know, there are hacks of exchanges, you know, across crypto all the time. Most of the time those are hacks of somewhat centralized intermediaries where users at least, you know, assumed that they were taking some risk when they put assets there, or in the broader crypto world, you know, across DeFi, smart contract risk. Again, these are pools of assets typically, and then one bug exploits the whole pool of assets. This unfortunately is the exact opposite. These are people with the Coldcard, not just because they're self-custodying their own Bitcoin, which is what every Bitcoiner has said is the goal, but also even the cultural demographic that owns Coldcards is particularly Bitcoin native, right? These are not likely to be people that are trading and speculating on exchanges or seeking yield in DeFi. This is the common self-custody sat stacker. They, you know, bought their sats in some way and then sent them into what they thought was deep cold storage. They never leaked their keys. They never dropped their hardware wallet on the ground. They didn't store their passphrases or seed phrases in an unencrypted, you know, web store or something accidentally, right? They didn't make a mistake. The vast majority of these people, they did exactly what they were supposed to do. Unfortunately, the keys that they generated were weak. So weak, in fact, I've seen some that could be broken in five to 10 seconds with a standard CPU, if, of course, you could point the CPU to that address if you knew where to look. I think this is the other sort of big part of why now the release of Kimi and other more powerful open-source models, in particular, that lack the cyber safeguards that the frontier models have been enacting, makes it relatively easy if you have even, you know, a couple GPUs to start scanning the entire Bitcoin address and UTXO history to find a potential vulnerable one. Now, you still have to grind on those addresses to try to derive the private key, but, you know, if you move on after five, 10 seconds, and again, I'm not deep on this part of it, but I'm in contact with a lot of people who are, right? It's, it's, it's frankly at this point now that the vulnerability is known, it's relatively trivial to exploit. You need some LLM access that's not going to safeguard you, and you need some compute, some, some decent compute. You don't need a whole data center. But to actually grind across the whole set of addresses and, and find and compile a bunch of vulnerable ones, you do, you do need some decent compute. So anyway, that's been happening since it looks like Thursday, early Thursday morning UTC time. And we've been tracking it, and we've been able to, we, we have a very powerful Bitcoin analytics stack that we've built that is not just the TX index, one, the whole history of blocks and transactions, it's the entire history of every single UTXO. So it's a, you know, 3.88 billion row database of all blocks, all transactions, and then all individual pieces of coin ever at every state, every block in the chain. It's like, and, and this allows us to trace pretty much anything. I mean, I, I, you know, I, I wish it wasn't as, remember there was some people sent Bitcoin a couple months ago to a burn address. Oh, wow. Fresh Bitcoin address for me to run through our stack and see what I can figure out. Well, it turns out it's been very useful for us to trace this attack, the Coldcard attack.
So engineers at Block Inc., great engineers there, they identified a pattern which I call wave one. And this was a variety of markers that looked like they were being attacked. Again, we can't just, you know, if a DeFi protocol is exploited, we just, you know, point our analysis at the smart contract and watch where the funds flow. These are all dispersed, right? We can't really tell whether an address is a Coldcard address or any other type of address, let alone a vulnerable one, right? So, the way that Block identified this pattern was burst transactions, full sweeps, dormant coins that for many years that plausibly look like they were in self-custody, cold storage, but none created before March 17th, 2021. And a few other markers in there like how the RBF and other flags are set in the transactions. I took those markers and ran it across our whole stack and developed what we call wave one. These coins are sitting in attacker addresses now, inert. So they have not been sent to exchanges or anything, and in fact waves one, two, and three are all inert. So like none have been exfiltrated to, you know, a DeFi protocol over a bridge or sent to a known exchange address. In fact, they haven't moved since they arrived at the attacker addresses. So wave one, blocks 960183 to 960191, drained 1,195 addresses and gathered 1,082 bitcoin. So that is still the biggest clear confirmed attack that's also later been confirmed by victim reports that have shared their transaction IDs and drained addresses with me. And they appear in that, we can confirm that those are hacked, that pattern Block found was right and, and real, and there are victims behind it that we know. Then the next day, so that was 41 minutes, by the way, those blocks 960183 to 960191, that was 41 minutes early in the morning UTC time Thursday, July 30th. So it's, you know, that's a pretty easy pattern to see something is up because you don't see, you know, 1,200 addresses and 1,000 Bitcoin that have been dormant all with these patterns all of a sudden move in bulk in like, you know, eight blocks. So then other victims started reporting that they had been drained and they were not in that wave, especially the next day. Through victim reports, we were able to establish a second pattern we call wave two. That's since been widely corroborated by additional victim reports that went down over several hours, three hours and 42 minutes on July 31st, which I believe was Friday. 1,400 addresses, 76 bitcoin. So again, now we're talking about much less Bitcoin, right? In fact, all of these, I think the average is less than one Bitcoin that's being stolen, which is another sad part of the story. I know it doesn't, it may not make a difference necessarily. You know, if a big, deep-pocketed whale is hit, then like, you know, that's terrible, but at least it's one person, right? This appears to be many smaller holders of Bitcoin in self-custody. And then I'll just flag also over the next, over Friday and Saturday, another pattern, again identified by victims making the reports to me on X, and then seeing multiple reports that look similar in pattern, extrapolating the pattern, finding all the other transactions that look like it, and then getting further corroboration by victims. 1,900 addresses siphoned 28 bitcoin. Waves one and two are very similar. These were siphoned into collector addresses and then moved in bulk into one or two attacker holding addresses. Wave three is topographically totally different. It is many individual transactions gathered, usually entire wallets drained in whole. So multiple addresses from victims in one input side of a transaction going into what we call a park address. It's not a collector because it's not many people, but it's an intermediate address, and then being moved into a script hash vault. So 293 script hash vaults are holding those 28 bitcoin from wave three, individual ones. So rather than pooled into one attacker holding address like waves one and two, this one's more sophisticated in that sense. The coins are segregated by attacked victim because of the victim reports. In wave three, we think that every draining transaction is one person, one person's seed, right? So you have 293 vaults, many victims on the theft transactions for wave three. I can't know by looking just at the blockchain, especially when we know seeds are compromised, that every input belongs to the same person. There have been many instances where the attackers are using the, are putting inputs from many different people on the input side of a transaction, almost like a coinjoin, right? That common input heuristic that firms like Chainalysis use. Well, it does work here because it turns out it's the attacker that has the keys to all the inputs. But in wave three, they sort of hit every victim individually, move their coins to a holding address, and then their own multi, well, we, I assume probably a multisig vault. We don't know, the scripts have not been revealed because they haven't spent. So those three waves account for 1,350 bitcoin, which is like $88 million. There, there are more. I think we've identified a fourth wave, meaning a fourth unique pattern. And at this time, I'm receiving and waiting to verify additional victim reports. That's sort of how this is working, right? Like if victims report theft and I can say, 'Oh, you're part of wave one,' then I'm helping and providing information to that victim about the status of their coins that they can use to report to authorities. We're simultaneously reporting and have reported the victim and attacker addresses for waves one, two, and three to federal authorities, crypto exchanges, crypto compliance firms, the SEAL team, like anyone who wants them can label them. We want them labeled as hacked so that if they appear on centralized exchanges, for example, they can be frozen. Wave four, though, is still sort of under investigation. I'm seeking more victim corroboration to confirm that they are stolen. I think I have some. So we're pretty confident that it is one attacker. It's not clear whether waves one, two, three, and four are all the same attacker. I think if I had to guess, maybe waves one and two are the same attacker. Wave three's pattern is totally different. The attacker could have, it could be the same one. They've totally changed up their transaction style. Wave four, we don't know enough yet to say. And I will say one of the other, I mean, at least waves one, two, and three, they appear to have been contained, as in not ongoing at the moment, one, at least each of them, one operation, defined operation. And in all cases, wave one, two, and three coins are still sitting, as we record here on August 3rd around 10 a.m., still sitting inert. Wave four is a little different because we haven't confirmed it. So we're not quite sure, like there could be false positives in there. I do see coins moving downstream that I have, I had identified as possibly wave four, but we still need, those could be false positives. And so we still need more confirmation on that. And then lastly, there are victims who are not part of any of these waves, confirmed victims, which I call footprints. They're not quite big enough for me to establish a full wave. I think I've got footprints A through E so far of just other hacks of Coldcard victims. Those could be opportunistic copycats, smaller operators. They understand the vulnerability. Perhaps they don't have enough compute to find many, and so they're finding some. Those are much trickier. Several of those I've seen are bridged through Thorchain, put on ETH, sent to a casino. There was an incident I tweeted about involving an offshore casino called Duel.com that I'm very confident received like 7 to 10 bitcoin's worth of exfiltrated funds. So those are trickier, and I say that as a word of caution. Even if your funds are not in, you were not hacked yet, people are out here. People have learned from the attack. They are going to attack. And just one last thing, I'm not the expert on this. There is a good, strong group, you've probably seen people talk about it on X, of Bitcoin security engineers and researchers that are doing different work, right? I am focused on tracing the impact of the hack on-chain, working with victims, and submitting evidence to law enforcement and crypto exchanges to hopefully get these coins frozen and one day returned to their holders. There's a giant other group of people who are literally burning, I don't know, millions and millions of tokens of compute, reviewing every known open-source codebase, all other hardware wallet codebases, and then examining the Coldcard MK3, 4, Q, MK5, etc., to understand, and talking to victims to understand, you know, were their drained funds just, funds just a single seed with no entropy added and no passphrase. We're starting to see some evidence that some users who did not roll dice and add entropy but did add a passphrase are being drained. So that would be harder and a bigger technical lift for the attacker. We're still trying to triage that. I think the strongest recommendation here is to just simply migrate out of Coldcard if you can. I think if you did all your own entropy and you are a sec, you know, feel very confident in your security and technical skills, then it really wouldn't matter whether you use Coldcard or any other wallet. But if you don't, if you have a modicum of doubt about whether you did that, I would just recommend moving off Coldcard immediately. The one other thing to say, well, there's many other things to say. One other caveat I'd like to add. When I posted about wave four, I was trying to post as fast as possible because it was occurring in real time. There were transactions in the mempool, there are definitely false positives in what I reported last night, but out of abundance of caution, I wanted to be fast. One of the false positive sets, Nunchuk, the multisig collaborative custody provider, pointed out to me was that it included multisig transactions. I vetted those transactions. There's no evidence. There's zero multisig thefts in waves one, two, or three. There's no reason to believe that a multisig in which a Coldcard cannot achieve quorum, reach the threshold to sign, no reason to believe those have been affected. And that's also logical. That's not so much a technical question, right? Coldcards are the only known hardware wallets that have this vulnerability. The red team has confirmed that the other major hardware wallet vendors do not have this vulnerability. So if you generated your own entropy well, or you're sitting in self-custody that is not related to a Coldcard, you're most likely fine. Almost certainly fine. Or I should say you are not affected by this vulnerability. If you have a multisig with a Coldcard in it, but that Coldcard alone cannot reach the signing threshold, you are fine for now. Do not rush out of panic to redo your whole custody. You could make your own mistakes if you do that. And look, if people are really freaking out, I think it's safe to say move your funds to River, move them to Coinbase. It's totally fine. You can sort of sit this out, build a new self-custody later. That's, that's okay. That's okay to do. I'll, I'll stop there. I'm waiting. I will say wave four, which we're waiting to confirm, if, if all of the addresses that I think could be part of it end up being confirmed as being part of it, then we would be up at about 1,850 bitcoin, worth $115 million now siphoned. We haven't confirmed that yet, but yeah.