DSS 2026 — Top hacks since Devcon 7 - what did we learn? Matthias Egli Source video: https://www.youtube.com/watch?v=bVE6SDYsnX0 Automatic captions — not human-reviewed. Complete spoken-audio coverage is unverified. Source language tag: en (English). Source export SHA256: 03c44056cbcd6e5b36c43f1ca33d643dc9dcc161c67540a8f86e817b5c6f7d19 [0:01] Thank you. So welcome everyone. [0:04] Um let's see. Yes. So this is the first [0:08] talk of today. I'm happy to give [0:10] this one to start of the day giving a [0:13] rundown of the top hacks we had since [0:17] like last Devcon, which actually is a [0:19] long time ago. So it's a really huge [0:22] list. [0:23] Let's see how far we get. [0:25] Quickly [0:27] about myself. So I'm Matthias. I'm from [0:29] Chain Security. We are one of the audit [0:32] companies reviewing a lot of the [0:34] different mostly complex DeFi projects. [0:38] We also do a lot of research [0:40] and and also like to present it. [0:42] So [0:44] among those things you might have had a [0:46] talk yesterday on like smart contract [0:49] quirks [0:50] by Lotaire. [0:52] There is another one right after by [0:54] Doris [0:56] also on tricky things around Solidity [0:58] and Dirty Bytes. [1:01] There is [1:02] one at Devcon by Dominic and then [1:06] I see this [1:08] animations don't work. There is one [1:11] at the end like kind of closing this day [1:14] on [1:16] a new kind of vulnerability found which [1:20] gets elevated by Glamsterdam. So this is [1:23] something I can highly recommend all of [1:24] you to check out on. [1:27] There there will be a bunch of talks on [1:28] this on Devcon too, but the first talk [1:31] will be here this evening by Janis. [1:35] But let's get started. [1:37] So [1:38] the first hack [1:41] I want to talk about is Bitget. [1:44] Bitget [1:46] is a centralized exchange [1:48] and [1:50] they got [1:51] compromised for 388 million. [1:56] So [1:58] how did this [1:59] The attackers here, they [2:03] compromised [2:05] initially a security appliance [2:09] in their [2:11] investigative report which was published [2:13] it's just mentioned as product A. [2:16] So we don't know exactly [2:18] what kind of security appliance it is. [2:21] There was also mention of a management [2:23] software [2:24] referred to as project B [2:27] where we also don't know exactly what it [2:28] is. [2:31] But what happened here is that these [2:36] initial compromises allowed the attacker [2:37] to move towards the signing [2:40] infrastructure of the hot wallets. [2:43] And [2:45] at that point of course [2:47] you don't need a human to make an error [2:49] anymore because you can make the system [2:52] sign off on those transactions which you [2:55] want to do. So [2:58] the [3:00] story here is that like what what did we [3:04] learn basically from Bitget at this [3:07] stage right? So the private keys were [3:10] very well protected inside an HSM and so [3:13] on right? So [3:14] industry best practice you could say but [3:18] the additional protection software they [3:20] put in place to prevent hacks was [3:23] actually in this case the one which [3:25] allowed the attackers initially to [3:27] enter. [3:29] We can only speculate on why didn't they [3:33] say which software they used? Was it [3:37] some NDA and they couldn't say it's that [3:39] and that vendor [3:41] or was it something where [3:44] they say like yeah it's [3:46] it might compromise others if they'd [3:48] known that the system is vulnerable. [3:52] So sadly from some of these attacks we [3:55] don't learn enough to really be able to [3:57] protect others. [3:59] The main thing we can take away is [4:02] any [4:04] piece of software, any component you put [4:06] in your stack [4:08] on the way to signing is potentially one [4:12] which is vulnerable including your EDR, [4:17] including your like web application [4:21] firewall, stuff like this, right? So um [4:26] in this case [4:29] what else can we do to limit the damage, [4:31] right? It's like limiting [4:33] the [4:35] single transaction damage, right? [4:37] Basically, how much is in a hot wallet? [4:39] Because in these cases, sometimes you [4:42] have circuit breakers in place but then [4:45] the question is how much damage is [4:47] already done? [4:48] Um [4:50] what was special also here, it took [4:53] really quite some time until they [4:55] realized that their hot wallet signing [4:58] infrastructure is the problem because [5:01] suddenly things move, right? And you [5:02] don't know [5:04] has someone the keys? Is something else [5:06] compromised? What's going on, right? And [5:08] it took them quite some time actually [5:10] to shut down their hot wallet [5:14] uh signing infrastructure. So also this [5:16] is like for emergency [5:18] actions that you realize, okay, what [5:22] what do you need to do to pause your [5:25] system holistically? They very quickly [5:27] paused actually withdrawals and deposits [5:31] as the usual place where they were [5:33] expecting to get hacked, but not the [5:35] internal system which signed off on hot [5:37] wallet uh um [5:39] transactions. [5:42] Now, this was some time ago already um [5:46] but of course, we are here also because [5:49] of a very recent one which is very [5:51] similar, which is the Binance hack. [5:54] So Binance as I guess all of you know [5:57] sitting in here was compromised just [6:00] four days ago by now [6:02] um and for for a record-breaking [6:05] um [6:07] of more than six billion [6:11] USD equivalent. Half of that money was [6:15] recoverable by freezing the stable coins [6:18] USDT and USDC mostly and a few other [6:21] real world assets. [6:24] First a big disclaimer, right? I mean [6:26] what I'm presenting here is [6:28] based on the few things we already know. [6:32] Binance communicates of what's going on, [6:34] focuses currently on enabling [6:37] withdrawals again and and has seen a lot [6:40] of [6:41] like action in this regard. So [6:44] luckily [6:46] we could say it's one of the very rich [6:48] exchanges which can [6:50] as it looks like afford to [6:54] use its treasury to compensate for this [6:56] loss. [6:58] So what happened here, what we know [7:00] already [7:01] a laptop allegedly, right? But this is [7:04] really like preliminary information, [7:06] right? And this is something I I want [7:07] you [7:09] to know that while Binance said this [7:11] it's in the middle of an investigation. [7:14] We will get to know if they share a lot [7:17] more and all of this might be wrong [7:19] after they realized what really [7:21] happened, right? So but what they said [7:23] so far is DevOps laptop got compromised. [7:26] They saw some other compromises across [7:28] their network. In the end [7:30] a service container as they describe it [7:33] was compromised this way which was [7:35] allowed to sign [7:38] and [7:40] usually the way this works is that you [7:41] have some container which has some [7:44] privileged access to the HSM which is [7:46] trusted, usually some certificates on it [7:48] which can talk to the HSM, right? And if [7:50] you get access to that container you can [7:51] sign [7:53] the the HSM trusts you to sign. [7:57] What then happened next is we saw around [7:59] like [8:01] for past eight the first outflows. [8:04] Actually Ripple was one of the first [8:06] chains they emptied 15 minutes later [8:09] roughly the very big transactions [8:11] against Bitcoin and Ethereum happened [8:15] here [8:18] for Ethereum for Binance 14 their big [8:22] hot wallet they replaced the [8:26] wallet with a like [8:29] smart wallet. I'll quickly go into that [8:31] afterwards. They swept that across all [8:34] the tokens [8:36] and then actually the sweeper stayed [8:37] active for three more hours and kept [8:41] sweeping because there were incoming [8:43] transactions actually during that time [8:44] which then obviously also got drained [8:48] and [8:49] some of them were also like refilling [8:52] because an automated system saw that the [8:54] wallet the hot wallet was empty so it [8:56] needed to be refilled, right? [9:02] What did we learn here, [9:04] right? So [9:06] first I'd say we didn't really [9:08] we will learn from this, right? This is [9:11] not something where we can say we did [9:13] learn anything yet. It's just happening. [9:16] It's a very recent incident, okay? It's [9:18] one of the largest right hacks [9:21] ever perhaps I mean you probably can say [9:23] it's the largest hack ever, right? Which [9:25] really was able to pull off. There is [9:27] that one the one that got away which is [9:29] also very interesting case but this one [9:34] I think will have impact beyond [9:39] our bubble, right? This is one where [9:43] people notice Bitget was very big of [9:45] course it was very big but [9:48] Bybit we will talk about soon too very [9:51] big hack, right? But no one knew Bybit [9:53] but Binance is almost equivalent with [9:56] like centralized exchanges [10:00] for [10:01] for cryptocurrencies, right? So we can [10:04] expect really to see some like [10:08] real world impact. Regulation is one of [10:11] the big things, [10:14] Contagion through the ecosystem seems to [10:17] be very limited currently, right? But [10:20] it's one of these puzzle pieces like [10:22] which fell to some degree. Binance [10:24] doesn't now have [10:26] the ability to react as freely anymore [10:29] because they need to invest all that [10:31] money to recompensate users. So [10:34] one of our buffers is gone with that, [10:36] right? [10:38] We'll learn more what can make this [10:40] attack harder. [10:41] Very preliminary things here. What did [10:43] we see like again, [10:46] um, [10:47] it took quite some time until, for [10:49] example, other hot wallets they they [10:52] plugged in something to sweep them to [10:54] cold wallets because in this fog of war [10:57] you don't really know what's going on. [10:59] And if you look into how they swept the [11:01] other hot wallets afterwards, you saw [11:03] that it's a manual thing. It's like [11:06] block after block. It took some time and [11:08] so on, right? So so someone reacted here [11:11] to some degree, I'd say manually. [11:13] Um, [11:14] the of course the size of a hot wallet [11:17] is something we can discuss. Faster [11:20] freezing across, um, the stable coins is [11:23] also something we can discuss here, [11:25] right? We will learn more. All of this [11:27] is very preliminary. [11:31] So quickly for you, because this is [11:33] actually important, if you ever also get [11:37] fished or so, what will happen to your [11:39] wallet is you will get a sweeper on your [11:42] wallet. This is what happens today. The [11:44] same happened to Binance here on their [11:46] Ethereum wallet. [11:48] It will get it will get replaced with an [11:51] EIP 7702, [11:53] um, which which is a very efficient way [11:56] to to interact with Ethereum, right? So [11:58] uh, I want also emphasize it's not like [12:01] the fault of 7702 in this case, right? [12:04] And and the other hot wallet gets [12:05] drained traditionally with normal [12:07] transactions. [12:09] But if you get hacked, you will get a [12:12] A sweeper is [12:14] um a relatively simple smart contract. [12:16] What it does, [12:18] when you send in ETH, it will [12:19] immediately send it somewhere else. [12:21] Um when you uh [12:24] approve um [12:26] or if if you have ERC 20s deposited, [12:30] um it will allow the attacker to [12:32] withdraw them even though you still have [12:35] the private keys and the attackers [12:36] doesn't, right? So they replace with one [12:39] transaction the code to gain control. [12:41] You can regain control like Binance did [12:44] a few hours later, right? But until you [12:46] do, [12:48] everything you deposit will be swept [12:50] away to the attackers. [12:52] Behaving very similar as if your private [12:56] key was compromised. [12:59] To you, you will not notice. The small [13:01] example I have here is kind of almost [13:04] self-explanatory code-wise, right? But [13:07] um [13:08] the real ones we can study also the [13:12] sweeper wallets out there. They are [13:16] well engineered. Um of course people [13:18] just buy them on the black market, [13:20] right? So it's not something they need [13:21] to develop for themselves, but it's [13:23] really this um industry where you [13:26] actually even give part of the profits [13:28] to those who develop that and so on. [13:32] Um [13:35] so summarizing the very first attack [13:37] vector here is like we have [13:41] the [13:43] hot wallet infrastructure of centralized [13:46] exchanges [13:47] and they are being [13:49] attacked, very targeted attacks, and [13:51] then these [13:53] hot wallet infrastructure is being [13:54] compromised. [13:57] Um [13:58] let's talk about an old one by now, [14:01] right? But because it was after Defcon, [14:03] it's still here. Um [14:05] Bybit. Bybit was hacked for uh [14:09] 1,430 [14:11] million. Um [14:13] and this was a case where a supply chain [14:16] vulnerability [14:18] fundamentally was the issue. So, [14:22] and the attacker managed to get to a [14:27] developer of Safe, [14:29] which was used by Bybit, [14:32] managed to compromise that laptop and [14:36] get [14:37] some AWS keys which allowed them to [14:41] deploy a JavaScript version, like a [14:44] small change in the JavaScript of the [14:47] Safe [14:48] centrally hosted [14:50] UI, which then in the end caused them [14:55] caused Bybit to get hacked. And and the [14:57] way this was done like you you had this [15:01] Bybit multisig [15:03] signing off on a regular transfer which [15:06] looked benign, but behind this curtain [15:09] what really was signed off was to change [15:12] the master copy of that Safe to an [15:13] attacker controlled one. And with that [15:17] they obviously had control over the this [15:20] treasury [15:21] and were able to [15:23] extract all the money. [15:25] So, this is one of the cases. [15:28] It's also some time ago where we learned [15:30] a lot. [15:31] And you will see a lot of projects out [15:34] there [15:35] which try to help [15:37] to validate that when you sign on a Safe [15:41] that you have different means to check [15:44] what you are doing. [15:46] I talked to [15:48] few people here and what what they [15:49] actually are often doing now is build [15:51] their own thing there, right? I mean [15:54] have Claude build their own thing, but [15:58] this is something where we really [15:59] learned [16:01] that wait, let me get to the [16:03] learning step that we can improve things [16:08] by validating with independent [16:11] ways, right? And that that we can [16:13] develop that software and it's actually [16:14] easier today than ever, right? [16:17] So, [16:18] learnings remove single point of [16:20] failures, be very [16:23] holistic about this, think deeply about [16:25] where are your single point of failure. [16:27] Limit again, limit the damage which can [16:31] be done with one transaction, which to [16:33] some degree you can also think of a [16:34] single point of failure, one [16:35] transaction, right? Then [16:38] specifically here, provide multiple [16:40] front ends. If you provide something [16:42] like a service which others rely on, [16:44] then ideally you'd give them some [16:46] independent way which is harder for [16:49] attackers to [16:52] compromise at the same time as your main [16:55] way. [16:58] Then [17:00] start to [17:02] like [17:03] put some pressure on to incentivize [17:06] people to prove to you that they did [17:09] this, right? So demand custodians to [17:11] prove and [17:14] we like Bitget, Binance and so on. Of [17:17] course they had ISO 27001, of course [17:20] they had SOC 2, right? [17:22] But of course these things don't focus [17:24] on the attack vectors we care about, [17:26] right? They are for different measures, [17:29] right? So they're still important, it [17:31] still makes sense to get them, but it's [17:34] far from sufficient. [17:38] A lot of this also was actually what we [17:41] talked about [17:43] two years ago in Mumbai [17:46] with like also a list of all the things [17:48] you should think about [17:50] at a presentation at Devcon and we'll [17:52] have a like the same presentation [17:54] actually two years again in a few days [17:57] at Devcon again together again with [18:00] muted from Polygon and with Nick from [18:02] Safe. [18:04] I'm skipping a few. Drift protocol, a [18:07] perp dex was hacked for almost 300 [18:10] million. [18:11] This is much smaller here than in my [18:14] version, sorry. [18:16] So just very quickly what happened here [18:19] like they compromised Sinus [18:21] by actually approaching them at a [18:23] conference [18:25] for half a year pretending to be like a [18:30] what was it like a [18:32] I think a market maker. [18:35] They managed to then by compromising two [18:39] keys or basically gaining the the [18:43] access to the laptops where where those [18:45] two keys were on to become protocol [18:47] admins. [18:49] They [18:50] added fake collateral to this perp decks [18:54] and they borrowed against it and and [18:56] then they bridged quickly back to [18:58] Ethereum. [19:01] So this is a little bit similar to one [19:04] of the incidents which really [19:07] caused [19:08] really big problems across the industry, [19:11] right? So which is Kelp DAO or layer [19:13] zero getting compromised for almost 300 [19:16] million. [19:18] So [19:19] roughly half a year ago they got [19:21] compromised. The [19:24] attack here was [19:26] a pretty good one from like [19:30] engineering perspective, right? What [19:31] they had to do to compromise layer zero [19:34] was impressive. [19:36] So [19:38] first social engineering again as so [19:40] often to get [19:44] some access to layer zero to to again a [19:47] developer machine to the Terraform [19:50] project [19:52] access the cloud infrastructure, [19:54] modified two of layer zero's internal [19:57] RPC nodes but just in memory to trust [20:02] fake blockchain data, right? Because [20:03] layer zero needs to listen to what's [20:05] happens on the blockchain and so on and [20:06] react to it and then now those ones [20:09] reacted to things which weren't real. [20:11] But that's actually not enough, right? [20:13] Because infrastructure was set up to [20:15] need independent third-party validators. [20:18] But in this case they DDoS'd the [20:22] external party [20:24] and then Zero fell back to say, ah, we [20:26] still have a trusted one, which is us. [20:28] In this case, also, of course, Kelp DAO [20:30] had it configured that this was allowed [20:32] and so on, but it was a risk back then [20:35] which was like many did. It wasn't [20:38] highlighted. Afterwards, of course, [20:40] finger pointing is always easier and you [20:41] should never do this and [20:44] we had cases where where like people [20:46] looked into this before Kelp DAO, which [20:49] we reviewed and where it was [20:51] highlighted. So it wasn't totally [20:52] unknown that you shouldn't do this, [20:53] right? But still it was widely done. [20:57] Um, they [20:59] um forged that uh [21:03] bridge transaction basically and then [21:05] there were like that attestation and in [21:07] the end they were able to mint um [21:09] unbacked RSEs. [21:12] And the problem with that is that there [21:14] was a lot like this was basically [21:16] hardcoded one-to-one on Aave and allowed [21:18] the attackers to liquidate two ETH by [21:22] depositing to Aave and causing um and [21:26] and other protocols, but Aave was the [21:27] biggest one affected here with the [21:29] biggest liquidity, causing [21:31] what we now call DeFi United and so on, [21:34] right? Requiring really a backstop, many [21:37] people chiming in and so on to try to [21:40] not have basically Aave having to um [21:45] not be able to service like their debt [21:48] anymore. [21:50] And and we are still paying off that [21:51] debt. [21:53] Um, [21:56] what did we learn here? [21:58] There [21:59] there were a lot of learnings on this. [22:01] Like the obvious one is like have Layer [22:04] Zero with multiple independent [22:05] forcefully validator, but actually a lot [22:07] of people are moving away to different [22:09] infrastructures. Also those we'll have [22:10] to see how well they hold up. [22:13] Um, [22:14] try better to isolate developer [22:16] environment from production environment. [22:19] This is a repeating thing across many of [22:21] the incidents we have seen, [22:23] but it's technically very hard and from [22:25] we come, we build this stuff [22:28] as developers and we grew as developers, [22:31] right? So we tend to have a lot of [22:33] access and power as developers. [22:36] While the traditional industry kind of [22:39] doesn't trust their devs, right? But but [22:41] we are the devs who build this, so [22:42] obviously we trust ourselves. [22:45] Yeah, put more limits in place, the [22:47] bridge outflows [22:50] like safety guards of to which degree do [22:53] you trust certain systems? We see this [22:56] finally happening a little bit more and [22:58] more, right? But [23:00] the the limits in place in many of the [23:04] decentralized exchanges and so on, they [23:06] are none. It's like as much liquidity as [23:09] there is, it's okay, right? And and that [23:11] can be a lot. [23:14] Um [23:16] yeah, what else is there? [23:18] Again, think holistically. All these [23:22] pieces you have in your stack can be and [23:24] will be attacked and many of these [23:26] attacks show [23:27] where different pieces have been [23:29] attacked. So let's look [23:32] quickly after this into some of the [23:34] other hacks, right? So here attack [23:37] vector was really [23:39] the users signing malicious transaction. [23:44] Um [23:46] ledger ledger itself wasn't hacked, but [23:50] what was hacked, right? Was their supply [23:53] chain allowing someone [23:56] to [23:58] like put [24:00] inside a ledger [24:02] chip which would [24:04] capture the seed phrase when it was [24:07] generated and shown on the ledger and [24:10] and basically phone home, right? Just [24:12] send an [24:13] a message via cellular network and and [24:16] and that's it. [24:18] So [24:20] this likely happened over quite some [24:22] time [24:24] and [24:26] over a few days then these funds were um [24:29] like stolen until people realized what's [24:31] really going on. [24:33] Um [24:35] here [24:36] some pictures. Credit goes to Nick from [24:39] Safe who provided those of how it looks [24:42] like inside the ledger, right? So you [24:44] have an additional board which shouldn't [24:47] be there. You have this small battery in [24:49] there to give it power. You have a a [24:52] main board of course where it's [24:53] connected to and you have the antenna. [24:55] You can actually open your ledger and [24:56] it's very easy to see, right? And I [24:58] advise like if [25:01] it really doesn't hurt to do this. [25:03] Um [25:05] what did we learn here? [25:08] Uh [25:08] a little bit tricky. Um I like try to [25:12] limit your exposure when buying smart [25:14] wallets, right? When buying when buying [25:15] hardware wallets. Of course buy them [25:17] directly from Ledger is not something I [25:20] am here to say yes that is the right way [25:23] because I [25:24] myself decided not to do this directly [25:26] due to their history of losing customer [25:29] data when you buy directly from them [25:32] which is also annoying, right? So it's a [25:34] lot of trade-offs here. Um [25:38] a better general trade-off here is try [25:40] to have two or three different um [25:45] hardware wallets for funds where it [25:47] really matters, right? Obviously not for [25:48] your like daily operations but for your [25:53] savings [25:54] having two independent one and a [25:56] multi-sig behind it kind of easy [25:59] nowadays and definitely advisable. [26:03] Um [26:05] cold card. [26:08] Cold card [26:10] similar size estimated a little bit hard [26:13] to say how much it was how much it [26:15] really was but estimated to be around [26:16] 100 million too. [26:19] This [26:20] a long time ago, 2021, right? They [26:24] upgraded their firmware to a version [26:28] where [26:29] like [26:30] they refactored they put the random [26:33] number generator, which was um the one [26:35] to generate the seed into a library. [26:37] This was all well written, was all [26:39] correct, but the library when it was [26:42] actually built was linked to the [26:44] software RNG instead of the hardware RNG [26:47] of the cold card, causing weak [26:49] randomness. And then many years later, [26:52] attacker were able to actually brute [26:54] force through the 32-bit which this seed [26:58] generated to [27:01] generate um the wallets, which [27:05] ultimately at that point obviously held [27:08] funds, right? And then were able to [27:10] drain those. [27:11] Um [27:14] the [27:16] thing here is like I mean the RNG inside [27:19] a wallet is really [27:24] the most most critical thing you can [27:26] have. There's almost not and it's very [27:29] hard to like [27:31] if I think of a random number in my [27:33] head, right? Can you prove that it was [27:34] random? Of course not. It's also very [27:36] hard to prove randomness, right? That's [27:37] just like very hard. You can generate a [27:39] lot, but [27:40] uh you will not find out if the RNG was [27:43] 32 bits by by any external measure. The [27:45] only thing you can do is internally look [27:47] into it. [27:49] What did we learn here? Um [27:54] when you audit this more for the [27:56] auditors, right? You need to take care [27:59] not only of the code, but of the build [28:01] because this was ultimately a build [28:02] failure. [28:04] Um combine independent sources of [28:06] entropy. This would help like many [28:09] things of defense in depth for your RNG [28:12] would have made this harder. [28:16] So attack vector three, we have people [28:19] get your seed phrases, [28:21] novel ways. We haven't seen these [28:23] before, right? These are kind of new [28:25] attacks. [28:27] I'm skipping over Liquid Network, [28:29] another 320 million hack here. TL;DR uh [28:34] again a software bug caching um [28:39] like LBTC [28:41] is is like a yeah Bitcoin derivative and [28:44] they minted that they were able to to [28:47] circumvent that check but they were [28:49] allegedly white hat hackers who returned [28:52] 80% and kept 20% as a bounty or whatever [28:55] right [28:56] um which which is not something we [28:59] should support. [29:00] Um [29:03] Cetus [29:05] let me see oh I have one minute left um [29:08] so Sui's largest DEX was attacked with a [29:11] nice vulnerability in that sense it was [29:14] again a software bug right in in their [29:16] system so um [29:18] yeah [29:20] um [29:22] in a in a library one error right so I [29:25] can quick no I will not walk you through [29:27] this because ah sometimes it's big [29:29] sometimes small okay it's kind of hard [29:31] to read here. [29:32] Um [29:35] I'll skip through this due to time [29:37] constraints how how the bug actually [29:40] happened um as an example um perhaps the [29:45] final vulnerable code you can see here [29:50] the takeaway I have actually on this [29:53] is like anyone who looks into this from [29:55] you guys here to say like hey where's [29:57] the error right [29:58] you you might know then you will be fast [30:00] if you don't know it will really take [30:03] a lot of time until you figure out [30:05] why this code is really broken. [30:09] And a lot I mean [30:11] a few minutes I guess for the best ones [30:14] perhaps an hour right but uh still [30:16] doable still doable [30:19] today [30:20] on the learnings when I tried now [30:24] and I rewatched what um like I wrote [30:28] about this before and I gave it to to AI [30:31] to where's the bug then immediately it [30:32] found the bug and they said I'll give [30:34] the cheapest I'll take the weakest AI I [30:37] can find still can find the bug [30:39] immediately, right? So this one is easy. [30:42] Time's up for me. [30:44] Um so we'll skip balancer V2 which was [30:47] the last one on the list. [30:49] Uh yeah, thank you so much.