One cracked key and every bank site looks real

1:05Eva Galant: Welcome everyone to the next episode of Blond & Quantum. Today my guest is Konstantinos Karagiannis the Senior Director of Quantum Computing Services at Protiviti. Konstantinos, thank you so much for making the time and being here today with us.

1:21Konstantinos Karagiannis: It's great to be here. As you can see, I am not blonde.

1:24Eva Galant: The blonde is me, you are the quantum, so everything is correct. Okay, Konstantinos, I would like you to start, before we go to all the questions and concepts and so on, I would love you to just introduce yourself and tell our audience where you come from. You have very fascinating background and I'm super happy that you are here because of the reason that we spoke off the records. So please tell us a little bit more. How did you land in quantum?

1:53Konstantinos Karagiannis: Sure. I mean, I was always interested in quantum since I was very young. And then I I went to Polytechnic in Brooklyn and and studied there. And I mean, this was a while ago and quantum computers didn't really exist. You know, Shor's algorithm existed, but the machines didn't. So I knew it would be a while before there was any way to do anything practical with them. and my interests were always in other math and hacking and cryptography as well. So I ended up spending a lot of years of my career doing things more related to cryptography and hacking. I ended up doing a lot of like financial hacking, attacking banking systems, and I got into blockchain very early too as a result, because that's a 100% cryptography, right? So yeah, I was involved. I was early in that and I actually was the first to give a talk at DEF CON about attacking Ethereum smart contracts. So that was years ago. Yeah. So I was always involved.

2:40Eva Galant: Wow. What years was that? Because I remember you mentioned as far back as 2012.

2:46Konstantinos Karagiannis: Yeah. So in 2012 I was working on various aspects of Post-Quantum cryptography and including quantum key distribution. I gave my first keynote that year on quantum key distribution. but for for DEF CON 25 which now we just had 34, so it was like nine years ago, that's where I talked about hacking smart contracts and the attacking the Solidity programming language and everything that was used at the time. So pretty technical. it's still online if you wanna see it. It's still pretty nerdy and still pretty watchable.

3:14Eva Galant: 100% You can count on me watching it. As I mentioned to you before, yeah, I mean, my dream was to go to DEF CON myself, which never happened. And I really feel I should catch up on this. Because the thing is, but there are two more things that I think our audience should know about you. First of all, you are running a podcast yourself.

3:35Konstantinos Karagiannis: Yeah, yeah. It's called The Post-Quantum World. And it's actually you can fact check me on this, but I'm pretty sure it's the longest continuously running quantum podcast in the world. Yeah. Yeah, it's almost five and a half years. Yeah, almost

3:43Eva Galant: I was gonna say, you have much more experience than I do. I counted five years at least.

3:49Konstantinos Karagiannis: five and a half years with no breaks. so there was a show like before me, and then he eventually stopped running it. So I was like, Yay! I went. Yeah.

3:59Eva Galant: Knowing how much effort and time this requires, congratulations, all my respect really. And there is one more thing which is even more interesting about your upcoming book. So tell us all.

4:13Konstantinos Karagiannis: Yeah. Yeah. So tying back to DEF CON, a year ago I spoke at DEF CON and did a talk called Post-Quantum Panic. And it got something like 150,000 views in a month on YouTube. And it was it went pretty viral. And Wiley reached out to me and they were like, Hey, let's make that a book. So I just yeah, so that that's what I've been working on. at the time of this recording, I have less than a month to hand over my manuscript and it comes out, it should come out on World Quantum Day. So like April 14, 2027. And the working title is also Post-Quantum Panic. So hopefully that sticks because I kinda like that name. I think it's awesome. Yeah. Mm-hmm.

4:48Eva Galant: I definitely love the title. Post-Quantum Panic. So beautiful, it's genius. Please do not change that. So, okay. So when did you say, can we buy it?

4:57Konstantinos Karagiannis: yeah, it'll be everywhere because Wiley's putting it out. So it's gonna be like a hardcover, you know, like front of the bookstore, airports, wherever. So it's yeah, like if you've seen the AI book, If Anyone Builds It, Everyone Dies, it's really that bad kind of book. So that's kinda yeah.

5:11Eva Galant: So first of all, congratulations and second of all, we're looking forward to it. Okay, so with that, now we know your background. Let's speak a little bit about the topic itself. We're going to talk about cryptography or security and also enterprise case study, a little bit of blockchain, I hope. So it would be a mix of different topics, at least to mention each of them. this one hour interview. But let's come back back all to 2012. You know, in 2012, when we were thinking about how many qubits are needed to break RSA 2048, the number given was 1 billion. By 2019, it was already 20 million. Last year, it was a million. In February this year, it was already 100,000 So, you know, I wanted to ask you, where is the floor? Do we expect to read a paper and I don't know, two or three years where they say, hey, actually 10,000 is enough?

6:17Konstantinos Karagiannis: Yeah, so yeah, those those numbers are correct. yeah, when I started that billion was basically an assessment of the original Shor's paper, like what it would take to do modular exponentiation for all these qubits. So with modular exponentiation, what you're really doing is you're assigning to the superposition state that it's some aspect of finding the factors of these numbers. And it's really hard to do that in your head. It's impossible in your head, obviously, or in a classical computer. But the hope is that with superposition, you can have all these states represented and you need a lot of qubits. Well, the Gidney–Ekerå paper in in 2019 lowered that with improvements to the architecture. And then Gidney built on that work by himself for 2025 when he said that we only need around 897,000 physical qubits to crack RSA. And I want the audience to remember that that's running for five days. Okay. We're gonna talk about that later in the episode, I'm sure, but there's a real big impact to it taking around five days. and we'll get into that later. so then other things came out. There was the famous Pinnacle Architecture paper that came out just last February. And I had those guys on the show and they talked about how yes, it's around a 100,000 qubits or around a 160,000 if you want it to run for about a week. But they admitted that that's not the floor, that they can get that to be improved even better. what by improvements to their qLDPC, which is a type of error correction, and other architecture, they think they could get it all the way down to 10,000. And that's pretty horrifying. I mean, really think about that, getting it down to 10,000. So we've ended up in this weird, like, crisscross state, like the way I think of it. If you've got the requirements for qubits plummeting, you know, they they just keep coming, coming down. And you've got Hardware manufacturers claiming roadmaps with higher and higher qubit counts, and those lines are increasing. And at some point, those two lines cross. And that's Q Day. You know, Q Day could come almost retroactively. It could be like a machine is on the market for a month, and then a paper comes out and says, yeah, that machine you've had for a month, you could crack encryption with it. Yikes. You know, like that's a possibility. So that could happen. And I need everyone to understand that that those two lines crisscross. Now, to be said, there's some caveats. One big one is that a lot of these papers that are coming out and and plummeting the numbers, they're also sort of giving and taking away. Like they're requiring more robustness of the hardware. They're requiring like real-time qLDPC coding, month-long runtimes for some of them, but some of them are shorter. Non-local connectivity, this I be this ability for a qubit to talk to any other qubit in the system at any time. If you look at the early systems, it was very much like, these two are next to each other, they can talk. Like so, so these other things that have introduced, you know, just to really simplify this for listeners. all these things are being added, right? So it's not like they're making it truly easier. It's not like they're saying, hey, vendor, now you only have to give us 10,000 qubits. They're saying you have to give us 10,000 qubits that can talk to each other better, that can stay alive longer. You know, that so we we are, yeah, it's it's a little bit of a superposition of give and take, right? To make a bad joke there.

9:19Eva Galant: And the race continue in that sense. Yeah. Okay. That's, that's make actually a lot of sense. And then speaking of, I wanted to ask you about the case with Google where they work out how to attack elliptic curve cryptography in minutes with under what they claim half a million qubits. And on the end, they choose not to publish the secret. So what's your opinion about it? Was that a responsible disclosure or not? the situation when the research labs are actually deciding, you know, what we're gonna see as a public or not.

9:52Konstantinos Karagiannis: Well, th there's a little more to the story that makes it kind of funny, but I think it's responsible because we didn't used to have this zero knowledge proof, right? This isn't newer. Like there would be some talks years ago, let's say at DEF CON, and people would be like, Just trust us, this part works. We can't show it to you. Yeah. And that everyone in the audience was like, Yeah, we don't like that. You know, we don't Yeah.

10:08Eva Galant: Yeah, exactly. That's what I wanted to ask you because the DEF CON world is very famous from the full disclosure, right? And like very clear position,

10:20Konstantinos Karagiannis: Yeah. So we want more, but zero knowledge is cool because you can mathematically prove it. You know, you can prove it without revealing something. That's why it's called zero knowledge. It's publicly checkable. but within two months of that Google paper, a French researcher, André Schrottenloher. He independently reconstructed and published an open circuit. Of Google's resource claims. And that opened some interesting work because he kind of improved it slightly. And then the world had access to it. So what did they do with it? Well, if I can make a slight digression. At that talk a year ago, Post-Quantum Panic talk, on stage, I said that I think AI will help quantum algorithm research. And I gave the example of how already Google had improved on matrix multiplication, something we thought was done for 60 years, right? And some people pushed back. They were like, that's crazy, whatever. And I'm like, okay, I'm telling you it's gonna happen. Well, this open sourced work, essentially, it opened the possibility for a project to appear online that your listeners could check out. It's called ECDSA-Fail And it's it's fascinating. What they did was they made a challenge to folks to Take any AI model they want and to optimize one piece of this whole Google claim, okay? The elliptic curve point addition quantum circuit inside Shor's algorithm. That sounds terrible. And if you're not into math, I guess it is, but but to simplify, what they do is they want you to reduce the Toffoli count and reduce the logical qubit count. And then there's like a multiplication factor of when you consider those two things, what percentage better is it than Google's claim? Okay. So you and you're always working at Frontier. Like whoever the leader is at ECDSA fail, you can log in, take their work, and build on that. So currently, I mean, I should have checked this morning, but it was something like 50% better than Google Solution, which is crazy. And this is like AI models competing. It's like, and again, that's a tricky number. It doesn't mean it's now 250,000 qubits. It means that a little bit of a qubit reduction plus a little bit of a gate reduction, and then that's where you get the number. So this is AI plummeting those numbers, you know? So I think that's fascinating. So Google did the right thing with their zero knowledge claim, but you know, folks kind of ran with it. And and you're always gonna have clever folks sort of you know developing their own. Yeah.

12:39Eva Galant: Yeah, and that's actually fascinating what you said in that case that even if something is like not the circuits are not disclosed, there will be someone else not only replicating it from backwards engineering, but also improving it. Fascinating.

12:52Konstantinos Karagiannis: Yeah, AI is having tons of impact right now. Yeah, absolutely. Mm-hmm.

12:54Eva Galant: Yeah. And we're going to talk a little bit more even later about that, right? But coming back to cryptography, I wanted to also ask you because on the Protiviti website, I found the cryptographic apocalypse could arrive in three to five years. And I heard that claim before as well, some other people. And obviously, you know, when I'm thinking about the company you work for, you guys are selling the readiness assessment. Right? So you're going to this big company saying, Hey, this is going to happen. It may happen three to five years. You need to be ready. Let's start working on it. So what I want you to do is to make that three to five years case to me as a technical claim rather than a commercial one.

13:38Konstantinos Karagiannis: Sure. And there's a few ways to do this. one of them is just through Mosca's theorem or Mosca's inequality. Okay, if your primary fear is harvest now decrypt later, then you have to consider a few things. You have to consider the lifetime a secret needs to remain secret, the number of years it takes you to migrate, and you have to add those two numbers together. And then you have to compare that to how many years away you think a quantum computer will be cryptographically relevant. And basically, no matter what your business is, you're gonna fail on some metric there. You just will. You know, if you have a secret that has to be 10 years or more, and it'll take you three to five years to migrate. You're looking at a number of like 13 to 15 years, and there's no way we have 13 to 15 years by any of the papers, right? So that's one way you could look at it, right? But the actual three to five by itself, technically, is very defensible. we just discussed those plummeting qubit counts. needed and the rising qubit counts and systems. And I'm not alone in this time frame, right? so Scott Aaronson was a huge, huge quantum skeptic. And just recently, I think it was May, he admitted that people he trusts more than himself who talk about error correction and things like that, they've said that 2029 is the first year a cryptographically relevant quantum computer is possible. I've publicly said I think that's the first year it's possible. IonQ has said many times that they might even have it by 2028. And they were just recently it hasn't aired yet at the time of this recording, but IonQ was on again on my podcast and they did not back down. They still claim 2029, roughly, they'll have

15:08Eva Galant: They said 2028. Wow.

15:10Konstantinos Karagiannis: as enough qubits. and if you read the Pinnacle Architecture paper, they will. Their their roadmap matches it perfectly. so there is a lot of of that like. Comparing and drawing those crisscross lines that we mentioned earlier. So I stand by that as being the first year that it's possible. I'm not saying 100% that's when it's going to be, but yeah, but it's the first year it's possible. And then when you layer on the whole Mosca thing, we're already too late, right? Like people should have been doing this for a while now. so I'm comfortable making that claim. And and other really big companies back me on this too, right? Like Google has said, nope, we're gonna migrate by 2029, Microsoft, 2029, Cloudflare, 2029. Like all of a sudden. all ducks are lining up saying, Yep, 2029. Yeah.

15:48Eva Galant: Everyone. This actually, think, there was a bill introduced in European Union asking companies as well to be post quantum ready until the end of this year. Is this over-accelerate or is just

16:04Konstantinos Karagiannis: Yeah, that's not exactly companies. the EU, what they're doing there is they're talking about member states initiating national strategies by the end of this year. So that's not companies. So like if you you know, if you're Bosch or whatever company, you know, and you're you're you're not the one. It's it's more like, hey, Germany, you have to have this strategy by the end of the year. Yeah. But if you look at their other stuff, it's like high risk use cases by end of 2030. medium risk use cases by 20, 35. So it actually matches the US a lot. It's very similar to us. Yeah.

16:33Eva Galant: I see. And did you ever have like the opposite case when you talk, tell your client, for example, no, this is too early. Like you don't need this yet. Come back in three years or something.

16:44Konstantinos Karagiannis: No, and and it's funny. So every company needs to do PQC, period, right? Like this isn't, just your sector. No, everyone needs to do it. we've only had the finalists for two years. So it's kind of a strange time differential, you know. So the work before we've had these finalists, I was helping companies begin the process of crypto agility. And that's a major point I want to make. That regardless of when Q Day actually arrives. crypto agility is the first thing that everyone needs to address because cryptography was not built into IT as like a core architectural principle. And some people are gonna hate me for this, but it's it's absolutely true. Like, you

17:21Eva Galant: I think you're right.

17:23Konstantinos Karagiannis: know, if you if you are a Microsoft shop, let's say, you know, you know that you could push out Windows updates easily. You know, you can you can update Office easily, whatever. You can update Teams. it's trivial. Right. That kind of stuff is just baked in. But you know, all of a sudden we're having this discussion about how do we migrate to PQC? What does that tell you? That means we have an entire cryptographic industry that is built to not know how to migrate anything. You know, and it's certainly not migrate as I and I mean those words, migrate everything. it's just not a thing we know how to do because cryptography is not just like, ooh, this one setting on a server, it's everything. It's keys, it's certificates, it's signing.

18:01Eva Galant: And speaking of that, that reminds me because I read that you said somewhere, you know, everyone is talking about, harvest now, decrypt later And you actually came with the new phrase, harvest now, forge later. And when you give the example with, over Microsoft, that was actually, that has to remind me about it because I was trying to also understand what's coming behind this statement. And one of the example I, I have been, given is that basically once you have those keys you can imitate that you are the company itself doing those updates for example. We've obviously adds on like you know some malware or yeah can you explain yes can you because I'm pretty sure you explain that better than me yeah.

18:43Konstantinos Karagiannis: Yeah, I'll give you an ex I'll give you an example of how the two are very different. Sure. Absolutely. And and this is like a core argument in my book. because I want to make it very clear. Harvest now decrypt later is gonna be a threat one day. It will. but what everyone forgets, remember earlier in that conversation we said that it takes five days to run the Gidney approach to crack RSA? Why is that significant? Well, if you run something for five days. That gives you 73 attacks a year. If you're China and you have a quantum computer and you can run 73 attacks a year, are you going to read 73 emails and say that that's what you did your first year with a quantum computer? That's insane. Why would you do that? What a waste of time. Like most of those messages will probably be nothing. Even if they're from the NSA or whatever, there's still a good chance it'll be. Hey, meet me at the drop. And what drop? I don't know. You know, like it might be a meaningless message. And you just wasted five days running the world's leading computer. so I don't think for the first, maybe even two years after Q Day, that Harvest now decrypt later will be the attack. I think that'll come a little later when we lower the amount of time it takes to run. Like when we're cracking keys every few minutes instead of five days, then you can start cracking packets like that. So I think the first attack will be Harvest now, Forge later. And what that means is in those five days, wouldn't you rather? Let's say you want to attack the a government agency like the NSA. I keep saying their name, now they're gonna like stop our recording right now. But any agency. Why would you want to read a few of their messages when instead you would figure out what laptop they use? You would pick a driver, let's say the audio driver, something innocuous. And you would create malware that still works as an audio driver, but is signed by that hardware manufacturer because you're able to crack that key that the hardware manufacturer uses. And this isn't like you have to break into the hardware manufacturer. No, no, no. That's not how public key cryptography works. By nature, when you have a public key, it's public. So you could take the public key, use a quantum computer to reverse the private key, sign your own package, push it out. And now every laptop in whatever company or organization doesn't have to be government, right? Every laptop says green check mark, we're running a valid ASUS, HP, whatever, signed driver. Yep.

20:58Eva Galant: So basically nobody's alerted and you just collect all of the data and you don't even need a quantum computer anymore. Correct?

21:04Konstantinos Karagiannis: It just did that one thing, right? and then at that point, you're gonna probably wanna use something like Frontier AI to find the most efficient delivery method, the easiest way to get that package on. Maybe they have a local caching server for their update packages, in which case you would use that to attack that and deposit the file, whatever you do. There's there's numerous attack vectors, but yeah, you only needed that one run and now you own the entire organization.

21:29Eva Galant: Exactly. So just to visualize it, as you said, the 73 attacks per year now multiply to millions or hundreds of millions in the same time. Yeah.

21:39Konstantinos Karagiannis: Yep. Yeah. So that could be 73 different code signing attacks. Or even worse, can you believe there's something worse than what I just said? Even worse would be what if you were to attack a root CA - certificate authority? Now all of a sudden you can own domains, countries, like you know, like you can authorize fake bank sites, fake whatever, and people will log in and it'll be green, everything's fine on their browser and they're they're you know, authenticating the IRS or whatever. It's It's it's really awful, you know.

22:07Eva Galant: horrible. You know, I'm really glad you mentioned that example because nobody so far gave me that example in the show. And that's really, really scary. And I also justify why the biggest country are in this race to put as much money on R&D for quantum technologies. Wow. That's terrifying. Because I think as you mentioned just before we start talking about the the crypto later concept you mentioned that nobody basically is ready in terms of the compliance, right? And I have read somewhere before my cat is joining us Hey, say hi This is this is the real cat from the from the intro.

22:45Konstantinos Karagiannis: Hi. This is like almost an internet, you know, meme, cat joining.

22:50Eva Galant: Yeah, and so I have read in one of the report the basically and it's terrifying me that 91% of the businesses have no formal PQC roadmap or strategy or basically anything. And if we look at the large organizations, only 30% of them have completed the cryptographic inventory. So, and even out of these, think only there was what 24 that they did something meaningful in it. So I wanted to ask you because you deal with those companies every day, like you talk to them. Why do you think it's like that? Is that the negligence? Is that the fact that nobody care? Because as you said, the hardware was not built with the cryptography and security systems in mind from the first day. Or it is because we as an industry kind of playing this role of crying wolf since a very, very long time, since basically Y2K. So. So which one of this is it? Do you think they don't really feel yet that the real threat or this is because we talked too much about it and nobody care?

23:57Konstantinos Karagiannis: Yeah, Y2K is a funny example because everyone compares it with this, like Y2Q, right? It's like the same thing. well Y2K wasn't really crying wolf, right? Like if if we didn't do the work, it would have been a very bad day. I I still remember that night kind of watching on TV, like, is everything gonna shut off? Like, what's gonna happen? Yeah. I wasn't confident. And back then that we had orders of magnitude fewer systems that needed any kind of touching. but now this is like, this is a massive scale. So is it negligence? I mean, I wouldn't call it negligence. There there's a lot that's at play here. Th those numbers you cite are correct. They're from like different data sources. but because they're large pools of participants, we could still view them as roughly the same, right? You know, this thousand companies, that thousand group, whatever. it's still large numbers. So DigiCert did their own survey, they asked a thousand IT decision makers, and 87% of them were planning, but only 7% had PQC on just half of their servers. So that's a really small number. But when they ask, this is why I'm bringing this study up, like when they asked what the biggest challenge was, you know, it was legacy system complexity, 26%, performance impact, 19%, and budget 19%. So like there, there are things that are sort of out of your control. And and companies, it's not that they really deeply don't want to do it, but they're kind of stuck, I feel there, there's a lot of, they need a lot of help. You know, this is complicated.

25:16Eva Galant: Can I ask you about, because this is very interesting what you said, 19, 19 and 19. Okay, I can understand the legacy systems because it's the problem like, hey, is still someone here who understand what's inside. I can understand the budget restriction, but you said 19 % is performance. What do you mean by that?

25:33Konstantinos Karagiannis: So, yeah, definitely. I'm gonna talk about that for a long time, but well we have to remember that PQC isn't performance PQC isn't free, right? Like it's math. That's great. PQC is not a product, but you're so you're not buying anything really, but what you do have to buy sometimes is hardware that can handle it. the signature sizes are larger, the key sizes are larger. it's all much, much larger. And for most modern hardware, that's okay, but there is a huge amount of devices out there that can't be upgraded. So performance wise, like they're just not they're gonna choke, they won't be able to run it. so this is a serious thing. Like there are billions of IoT devices in the world, right? And I think like 90% of them don't have the horsepower to run full PQC for signatures.

26:24Eva Galant: So it basically means that everything will need to be replaced and then migrate.

26:29Konstantinos Karagiannis: The rip and replace story is terrifying and people don't talk about it. it is no one wants to talk about it, and it's

26:33Eva Galant: Nobody wants to talk about.

26:36Konstantinos Karagiannis: so much worse than you think. And Microsoft did a great job, sort of. I don't know if this is they knew they were doing it, but they really started preparing the industry for this. And do you remember the backlash Microsoft took when they said that certain machines won't run Windows 11? Do you remember? It wasn't long, it was like, yeah. And boy, everyone was crucifying them like as if they're money grabbing or something. It was it was crazy. Meanwhile, all Microsoft was saying is that Windows 11's not going to run on these machines because they don't have a certain level of trusted platform module TPM, you know, of performance, right? So they're literally saying that it would not be safe to run. You could it'll run, but it wouldn't be safe to certify. Yeah, it's a security thing.

27:12Eva Galant: Yeah, to do that. It's not possible, it's not safe and we don't want to, we choose not to do it.

27:18Konstantinos Karagiannis: You know, everyone pretended it was Microsoft trying to get you to buy a machine. It's not at all true. They didn't have TPM 2.0 chips. Windows 11 wouldn't be able to use its cryptographic primitives, long story short, to run. And I was like baffled by how people didn't understand that. They just thought it was something else. So Microsoft sort of started preparing the industry for this rip and replace thing. And we saw the backlash. So now I have to be in this uncomfortable situation of telling customers. Yeah, there's certain servers that are gonna have to go. Like in theory, they can run and you can run things on top of them, but the reality is PQC has to go all the way down. You know, it's turtles all the way down in in the old creation myth. And it has to be PQC all the way down. You know, the example I gave you before of that audio driver, if you don't have PQC all the way down, that audio driver could then install a bootkit, which could become a rootkit.

28:10Eva Galant: Something like that can happen.

28:11Konstantinos Karagiannis: And now you're on the firmware of the hardware. And you can reboot it as many times as you want. You can reinstall Windows, you can reinstall Linux, whatever. It doesn't matter. It's it's right on the firmware. So like those servers have

28:17Eva Galant: It doesn't matter. Yeah. And it's something trivial like that.

28:21Konstantinos Karagiannis: to be replaced. And and that's that's the reality.

28:23Eva Galant: And I also have read somewhere that you mentioned that the hard part isn't actually the new algorithm. It's finding where the encryption is embedded or buried, maybe hard coded into application and hardware. this is it. Sometimes it's the hardware that nobody has opened in 20 years. So from your experience in your daily work, What's the worst thing you personally find inside a real institution? You don't need to obviously give that a name of the

28:55Konstantinos Karagiannis: yeah. I definitely can't name customers. and it's not even just customers. I mean, I have friends who work in IT and all types of companies and all levels and things. So yeah, yeah. And sure. And

29:03Eva Galant: Yeah, so you probably also chat about those cases. Please give us a couple of the most...

29:08Konstantinos Karagiannis: at DEF CON I chat with folks about this type of stuff. Yeah, too. So I mean, yeah, a challenging thing is hard coded keys in compiled binaries with no source available. That's that's a nightmare. That means that there was no agility in place and Crypto agility again, that keeps coming up. This will mean that you have to rewrite things instead of having a modular approach where you could just swap out a cipher. That's bad. Vendor appliances where crypto can't be enumerated without the vendor's cooperation, that's bad. Third parties are always a challenge, you know. when we're doing a PQC migration and crypto agility assessment, like we always have to talk about who your third party vendors are. We need to know what they're already doing. Some of them are doing a great job. I mean A great example is like Palo Alto. they've had PQC in PAN-OS since 12.1. So like they're ahead of the curve. But some they just kind of vaguely have a plan. And you know, you gotta you gotta deal with that. certificate chains where the root CA's private-key custody is undocumented. Yikes, you know, documentation is everything with something like that. We talked about rip and replace, right? Embedded devices. I keep seeing devices that will never have a path and they're gonna have to be migrated. And in the past, we were able to seen companies absorb a lot of that by like air gapping and all this. But I don't know. Air gapping is it's not always a real thing, you know. I've seen I've seen companies where they think something's air gapped, and I promise you it's not. Updates are still delivered and things like that still happen. and I'm just not convinced that they're truly air gapped if you can, you know, if you can get something to them. so there are Companies that have decentralization of cryptography to the point where it's like an Excel spreadsheet tracking stuff and then there it references another Excel spreadsheet that I can't find or whatever. Like I've seen that kind of stuff. You know, so then that's why I'm saying cryptography was built very much as like a let's add it to each little thing, and none of it ever has to talk to anything else. And even though we've done cryptography migrations before, you know, no one uses SSL, they use TLS 1.3, hopefully, but not, but hopefully. like we haven't done a thing where it's like, no, no, no, no. Every single cryptographic thing you have now has to change in some way, shape, or form. We've never done that. And that's why we're revealing these problems.

31:17Eva Galant: There is a lot of work ahead of you.

31:20Konstantinos Karagiannis: yeah, I'm here for it. I mean, you know, I we're building a safer world. I love it. It's great, you know. And and even without PQC, like pretend let's say imagine a quantum computer never becomes cryptographically relevant. We will have built a world where cryptography is ready to like that, well, not like that, but almost like that, cryptographically improve, right? Like it there are other things. There are side channel attacks other things happen to cryptography besides quantum. and one day you'll wanna be able to just say, we can't use this cipher anymore. And you don't have to initiate a multi-million dollar project to do it. You just have to kick off your processes that you set up, your new methodologies and software and whatever you have in place to do the migration, you know. So that's why crypto agility is so important.

32:01Eva Galant: And speaking of, because you mentioned that, when do we know if... So here's the concept. If a nation state or certain countries had a cryptographically relevant machine tomorrow to break some stuff and so on. They obviously, they would not talk about it. They would not announce it. There would be no press release. Nothing like that, right? Because why would you do that if there's advantage? So what is the observable signature? What would you as an expert look for personally by observing outside the world? When would you know that, maybe that country already has something that everyone else has no idea about?

32:46Konstantinos Karagiannis: Sure. So yeah, you're right that they wouldn't announce it. it depends on what their motives are, and we'll talk about that with Bitcoin. But we have a great example of this with Alan Turing, right? When when he cracked Enigma, he didn't send a message to the Germans, ha ha, we know how to read your messages. You know, he did not do that. you know, he definitely didn't do that. and then Harvest Now Forge later is terrifying because what if you could then not he was reading messages, right? But imagine if Turing was able to tell the German submarine, hey, attack this other thing or whatever. That'd be forgery. So that's why we're talking about exciting stuff with forgery compared to Turing But what what should we be looking for? Right. so I would watch out for is there something happening on the blockchain? That'd be a great, you know, a great indicator. That's one. another is unexplained mass procurement of cryogenic infrastructure. Like, what do you do if all of a sudden some country tries to scrape up every dilution refrigerator on planet Earth at the same time. It's like, why do they need these overnight? Like there's something, you know. No, no, that's just one example. Yeah, it's one example.

33:45Eva Galant: Not all of them need cryogenic devices, right? Yeah, okay.

33:50Konstantinos Karagiannis: and well, currently the ones with the highest qubit counts do though. So you know. then a sudden national security classification of previously open error correction. Like let's say there's like a work path that all of a sudden it's like nope, we're not talking about this anymore publicly. It's like yikes, Why is this secret? certificate authorities, doing some kind of like rekeying ahead of schedule. any kind of like signs the forgery has occurred, you know, that can't be explained otherwise. Like we've had supply chain attacks before. You know, we've had situations where software like SolarWinds was a great example. Like, but for SolarWinds to work, they had to hack into the organization and essentially insert bad code into the product lifecycle that will then be delivered. Like if you could prove that never happened, but all of a sudden we've got some malware running that was validly signed, maybe that's an indicator. You know, so there's like subtle things that we can watch for. But I don't think it'll be such a big surprise because I think that it'll be a period of time where we can reasonably guess it's possible. Like let's say everything I've said so far comes to pass, that the algorithmic developments keep getting better regularly. And the qubit counts keep rising regularly. We're gonna enter some period, let's say it's around 2029, where we're like everyone's gonna just be super uneasy. Cause we'll know that even the public information is that, okay, we're just kind of almost to that point. And then you gotta wonder, did someone tip the scales and then stay quiet about it. So, like that's kind of all the stuff that I would consider. hopefully though, you know, we're all in such a great place before that happens. that that that doesn't matter. And Do I believe that? No, but I don't think I don't think the whole world is going to be ready, quite ready for Q Day. Unfortunately, I I don't I don't think so.

35:30Eva Galant: Okay. And you mentioned the Bitcoin thing and actually the blockchain movement and so on. So let's talk about this because I was waiting for that moment where it's very exciting thing for me. I was introduced to Bitcoin in 2011 and I was very skeptical about it. So, yep, I could be retired, but that never happened. But what I noticed being in the quantum industry for the last two years is that the question about Breaking blockchain is the most common question that actually pop up whenever there's a discussion about quantum by business and not technical people. So I think there is a huge need in the society to understand how quantum can affect Bitcoin, can affect blockchain. So let's talk about it a little bit. Yeah.

36:23Konstantinos Karagiannis: Yeah, I can explain the attacks. Yeah, definitely. So there's a couple of types. So the one of the dangers is that the earliest Bitcoins, the first 1.7 million or so, those did a pay-to-public-key. and as a result, permanently forever, their public keys are exposed, and you can at your own leisure download the blockchain, go through them, find interesting wallets. And reverse the public key to private key, and then you can spend that money and it's yours. So that that's using Shor's algorithm, obviously. it turns out that so far, cracking them that way, cracking ECC, might be doable before RSA, which is terrifying in its own right. So the Oratomic paper is a great example. 26,000 neutral atoms running for 10 days cracks ECC. Yes. So that could be Q Day. Like that could be how Q Day begins. It could be attacking Bitcoin or ECC is used elsewhere as well because it has low, low resource requirements. So it's used in other places. But Q Day could not be, it might not be RSA. It might be when we first do this. And to me, that's still a valid Q Day. so that's one attack. And of course, that attack or any Bitcoin attack is devastating because I don't believe cryptocurrency survives such an attack. Personally. Some people think I'm wrong to say this, but we've never seen an industry that's so volatile based on mood. Right? Elon Musk tweeted once, we're not gonna use Bitcoin at Tesla because it's bad for the environment. And Bitcoin plummeted 30% the day after. Like, that's just a tweet. Like, imagine if someone who's inside knowledge says, I think the quantum computer can actually attack Bitcoin this week. Yikes not alone, you know. But if we have

38:07Eva Galant: Yeah. it can kill it. Yeah. Or lose 90% of the value or so.

38:09Konstantinos Karagiannis: Yes. Like no one Yeah, when I did a show of hands I I spoke at DEF CON again this year. When I did I did I asked for a show of hands in the audience, I said, Hey, who among you would buy Bitcoin the day after it's been proven that a quantum computer moved a coin? No one raised their hands. So what's the value of Bitcoin

38:23Eva Galant: Yeah, nobody, of course.

38:26Konstantinos Karagiannis: point? Zero. And then I mean I don't know if you want me to go through the other attack too, but yeah, that's one.

38:30Eva Galant: Yeah, I would love to certainly, but I just wanted to also emphasize because I think maybe some people are not aware. You said there is 6.9 million bitcoins out there, right? So that's.

38:41Konstantinos Karagiannis: That are vulnerable. Yes. So so that's an amalgamation. That's a com

38:44Eva Galant: So that's like $75 billion.

38:47Konstantinos Karagiannis: Yeah, yeah. That's a combination of reused addresses. there's a thing, there was a thing called Taproot introduced a few years ago, and and it has a default key disclosure. and then there's the original 1.7 million, including Satoshi's 1.1 million. So that's a lot of money, you know. and that could be moved that could be done slowly, right? And because it could be done slowly, you can move all of it at once, in theory, right? I could, I could spend a few months. And get like Satoshi's keys and like a few others and then move enough, yeah, quickly

39:17Eva Galant: all at the same time. Yeah.

39:19Konstantinos Karagiannis: do some massive devastating damage, and then yeah, the value will drop, but who cares? By that point, I've already cashed out, I've done this, I've done that, whatever. so that's one way to go. And if you're a chaos actor, you might want to divest yourself of Bitcoin, do an attack, and then watch the countries that didn't divest themselves have all these sorts of financial issues. I mean, that that's one kind of attack, right? So that's the slow move attack. I don't believe it'll be unnoticed though. Like I think you gotta act quickly. Cause there are nerds out there, and I say because if anyone's a crypto nerd, it's me, but there are nerds out there who do spend all day looking at the blockchain going, ooh, that was an interesting transaction. What was that? Well, well, look at that money that moved. You know, anytime a whale event happens or whatever, it's it's reported instantly. So I laugh when people are like, no, it could be done slowly over months, slow bleed, even BIP 361 Right in it, he talks about like a slow attack, and I I just I don't agree with that. So that's that's one kind of attack.

40:12Eva Galant: And not only nerds, right? From my understanding, there are other whole companies also watching those movements. There was one called... Yes, that's true. I

40:18Konstantinos Karagiannis: Absolutely. Absolutely. Yeah. I'm just saying that you're not gonna get away with it. You're not gonna get away with being subtle. I don't think it's gonna happen.

40:24Eva Galant: think there was a company called Coin Analyst. They were just basically like raised to do only that for, I think, governmental reasons or something like that.

40:34Konstantinos Karagiannis: Yeah. And and the creator of the BIP 361 protocol. And BIP 361, I'm which we'll probably talk about in a moment, but but even the creator, he says it's not ready and he says that you know, it it's not an ideal solution. But I I I'm also worried about the other attack that can happen in real time. so the Google paper you mentioned

40:52Eva Galant: Let's talk about it.

40:53Konstantinos Karagiannis: earlier, yeah. That allows once you break the 10 minute barrier, you're breaking the block creation time barrier. And what happens as a result is no matter what type of wallet you have, you know, you're using coins now as opposed to being Satoshi from the past. When you go to spend money, you are front running the mempool here. You are loading the memory pool with your public address. It's just how it works. It's gonna be in there. So someone can retrieve it, can run. the run the optimized and pre-compiled version of Shor that that Google's using and you can within nine minutes get the private key right then you could put your own transaction out there as a competitor maybe offer higher fees to miners and then they show with their combination of metrics that 41% of the time the attacker will win even though the they put theirs out later. The attacker will win, and 41% of the time, you will get to spend their money, and the valid owner will be accused of trying to double spend. So that's super dangerous. that's a real time attack. But again, that that's I think that'll be a second attack because if you look at the qubit count, that's half a million qubits. So I feel like we might not be able to do that nine minute attack first, right? I think it might be the offline 10 day attack, maybe first. but they're both devastating. And this is another chaos attack. I I shouldn't be giving people these ideas, right? But it yeah, but let's say there's the chance that a quantum computer can do the front running the mempool attack. if you're a chaos actor and you have like half a Bitcoin you don't mind burning for fun, you can just make a transaction and say, Hey, someone, you know, someone just double spent my money. And you can you can like you can try and forge that. You know what I mean? Like you can do could try and do something like that and then, you know, confuse people or or cause public sentiment, you know.

42:48Eva Galant: That was the question that I also asked myself. If someone seriously breaks the algorithm, would they move the coins at all? Or just keep them and see what happens later?

43:03Konstantinos Karagiannis: It's an attack, in my opinion, you could do it once. I I I don't believe it's gonna I don't think it's like, we had another quantum attack today. Ha ha ha Bitcoin will survive. No, I don't think that's how it's gonna be. I just don't believe it. I the whole reason we we buy into something like Bitcoin is this idea that it's, you know, your key, your coins, you know, like it it's this and it's cryptographically irreversible. The original language in the Satoshi paper was that it was supposed to be cryptographically irreversible. And it's just not the case.

43:30Eva Galant: And yeah, and don't you think there's also a question that, you know, usually people don't want to talk about it a lot as well. But don't you think that a lot of people would actually prefer if Satoshi coin were frozen? Because you know, they are there for what 15 years now overhang.

43:48Konstantinos Karagiannis: Yeah, yeah. I I think I think that's that that is like a philosophical thing, right? it would be good if if you got rid of, let's say, even more than that, just the 1.7 million in general. yeah, it's good for every holder. Bitcoin is now more scarce. There's fewer coins. Every coin's gonna be worth more money, you know, and the threat of them ever being touched is now gone. But that of course becomes that whole political thing. That's why Bitcoin's not a company, right? It's it's a bunch of people who have to have ideals and argue about stuff. it's not one person saying, do this. Like with Ethereum, you can argue it's still, you know, community, but but ultimately it has its creator who's very active. Vitalik's a super smart guy. and they've created this path towards 2029 to migrate Ethereum's L1 EVM and everything that goes into it to be PQC. So yeah, there's a lot involved there, but you know, there there's things like BLS signatures, which do proof of stake. if you can crack a BLS signature, you could say, hey, I have 10,000 ETH, listen to my vote, you know, like that's not good. So so they're migrating all that by 2029, which is great. But with Bitcoin, it's not like that. It's a bunch of people have to agree and everyone hates everything. And if 361, everyone hates it. They're like, that's not my keys, my coins. That's my keys, my coins, if I don't do what you say. By a certain date, you know, and and that's just not very popular. It's not gonna be popular. And then BIP 360, which tries to fix what was introduced with the whole smart contract thing, exposing public keys. BIP 360 has its own issue, and that I had on my show Chris Tam from BTQ, they built a testnet of BIP 360, and he admitted on The Post-Quantum World that he thinks it would take Bitcoin seven years to apply something like. So ouch, you know, that's way past way past Q Day. So for all these reasons, I don't know if Bitcoin's gonna get there in time. And they're the ones that everyone should be watching. Like, even if the rest of the entire cryptocurrency community was magically PQC ready and just Bitcoin was vulnerable, I still think it would cause irreparable harm to the entire industry. Like we've never had something like this before. If you think back to the SWIFT hack over a decade ago. Payment systems didn't go away. You know what I mean? Like like we fixed the thing and yeah, sure we lost 80 million, whatever it was, but then payment systems didn't go away. I think with with Bitcoin, we're in a situation where cryptocurrency, everyone's gonna be like, I'm never using that again. I got burned, I'm out, you know, and and that'll be the public perception.

46:14Eva Galant: And that will be on not only for Bitcoin, but for all crypto actually, right? So let's consider the opposite situation. So let's say that the freeze never happened and one day a dormant early wallet moves. So from that moment, probably, as you said, within one hour, maybe a minute even, and the market will have to decide whether Satoshi is back, right? returning or maybe a quantum computer announcing itself and we have Q Day or whatever happened. There's no way to tell that from outside, right? So what should we look for and what will happen to the price in your opinion? Zero completely or lost 90% of value and everything will collapse?

47:00Konstantinos Karagiannis: Yeah, some people think yeah, like I think personally it's zero. That's me. And some people say I'm wrong, but that's what I feel. I just I I just think that if the whole point of this, the whole reason that paper was written, the whole point is to have this thing where cryptographically you could prove it's your money and and if that's the point, that point doesn't exist anymore. So I don't get why it should be anything but zero. That's that's me. That's my hard stance. Yeah.

47:27Eva Galant: So should we all sell before Q Day, you think?

47:33Konstantinos Karagiannis: I yeah, I like I don't know. I I I used to love trying to find use cases for for blockchain and everything. And and there are all these great L2 use cases with Ethereum offshoots and all that, but if Bitcoin goes down, I just I just don't see how we we go forward from that. I don't, you know, it's just the value it's just too big a difference. You know, there are post quantum There's there's the Quantum Resistant Ledger (QRL), right? They've been at it for a long time. That thing is that's fine. But again, like what's it really gonna be worth in relative factors with if Bitcoin goes down? Like I I just I don't know. So that's why it should be all hands on deck, panic, figure this out. And if they do decide to go with BIP 361, then they should be really careful how they implement it. Like because this idea that, hey, freeze everything, but then we might have an ability for cryptographic proofs in the future to unlock what's frozen. You better prove that's working before you force it on people. You know, that you can't do that. You can't you can't just have this like control. Yes, yes. That better be working first. So I

48:39Eva Galant: especially in that community.

48:43Konstantinos Karagiannis: would never be okay with that. Just, you know, I'd never approve a control whose compensating mitigation is a research project. I wouldn't approve that, you know.

48:51Eva Galant: Okay. because then I would love to try change a little bit of topic. know you're very passionate about the synergy between quantum and AI. And you mentioned that a little bit before. I would like to talk a little bit more about it. cause I know that your team actually raised the AI driven cryptographic discovery and, the remediation tracking as a topic as well. So I wanted to ask you, like, what is the real working capability today? And. or are we about to bolt hallucinating models into a compliance process that nobody can audit or what we're going to do about it.

49:30Konstantinos Karagiannis: Yeah, lots to say about hallucination. So but but let's see. So we'll start with with the basics, right? So discovery, that's a search and classification problem. ML works perfectly there. I mean, there no one's gonna deny that, you know. So scanning binaries, configs, network telemetry, whatever for primitives, and introducing a CBOM chef's kiss, perfect, you know. no human could ever do it. at that level. And then if you want to validate something, you can actually go look. It's not black box. The output is there. You can you can say, this says that this cipher is on this asset. And yes, it is. So you can do things like multi-shot prompting. You can you can do multiple runs to validate things. and and I do some of this myself. Like when we build a cryptographic bill of materials, we use a tool that I built and it's not technically AI, but the but the tool parses outputs I get from the company and it goes through. But then I wrote Python scripts that go through and count the number of ciphers and make sure that the CBOM matches it and make sure that any AI analysis in the future takes into account how many actual certificates are present. So nothing gets missed, nothing like that. So it's it's pretty airtight as far as like you know hallucinations or any worries like that. where it gets dangerous is if you do something like have a model. assert remediation status and and basically generate attestations that no humans were involved in, that's that's not good. AI can't just like propose something and then, you know, you don't have a way to have evidence like dispose it. So if a finding can't be traced, like I said, to a specific thing, you know, it's suspect. And and

51:05Eva Galant: Do you think, stupid question, but do you think AI could potentially read all the literature about quantum and help us to solve the problem that we still struggling to solve?

51:16Konstantinos Karagiannis: Well, I mean, it's already happening in a way, right? Like like that that ECDSA-Fail thing I mentioned, there's one example. You know, it's taking that one little piece and improving it with models gradually. And I'm sure over time AI is gonna find other little pieces of the algorithmic process to improve. I'm also sure AI is gonna continue to improve just quantum in general, you know. Like we've had examples of AI making machines run better, error correction, AlphaQubit, whatever. You know, yeah.

51:41Eva Galant: Exactly. I wanted to ask you actually about AlphaQubit and DeepMind's invention. What do you think about it? Can you first talk a little bit maybe what AlphaQubit is for our audience and then tell us what do you think about it?

51:55Konstantinos Karagiannis: Yeah, so AlphaQubit was a DeepMind and Google Quantum AI thing that and it was like November 2024. so it was a while ago. But using on the Sycamore processor, it made 6% fewer errors than tensor network methods. And that was that was impressive, but it was too slow for for real time decoding at scale. So Google improved it and came out with AlphaQubit 2 one year later, you know. So that was actually fast enough to matter. It was it was It was able to beat the time by a good amount. It was like 10 to the negative 11 instead of 10 to the negative 10 and under one microsecond. So AlphaQubit 2 was able to meaningfully improve errors. And I think it's only going to get better. And there's other hardware that's being introduced, and there's NVIDIA with NVQLink, and there's all these other attempts to have AI sort of help quantum. And then you've got this like feedback loop, hopefully, is what we're hoping that AI might improve. quantum processors a little bit. Then the hope is that those better quantum processors might help with training to make AI run more energy efficiently. And then that better AI will improve quantum. And it could be this wonderful loop. Unfortunately, the reality is yeah, the reality is in the middle there you have to have fab time, engineering issues. Like it's not just so AI improving AI is is the superintelligence thing that Nick Bostrom thought up, right? Like AI is gonna keep improving itself and then the smarter AI will make itself smarter and that's smarter, you know. But when you introduce hardware to the mix, it's slower. It's slower. It'll be like a a slower loop. But AI is actually having an effect at at every level right now.

53:30Eva Galant: One of the use of quantum also improving AI that I read about before is the synthetic data to help train the models. Do you think? Yeah? Yeah, okay.

53:40Konstantinos Karagiannis: Mm-hmm. Yeah, that was like what I meant. Yeah. I think that'll improve. And and it'll it'll minimize like what we call like wet lab time. You know, because like because things like chemistry, for example, that's probably the killer quantum use case outside of cryptography. You know, like if we wanted to summarize the whole use case world right now, I think the first place we're gonna see provable quantum advantage will be in chemistry. And even though right now the use cases are small. It it everyone criticized that, but it they shouldn't, because you are doing the small use cases to match the hardware you have today. And then the logical extrapolation is that you're learning and you will then have more hardware, you know, more qubits that are fault tolerant. And you'll expand and you should.

54:21Eva Galant: Expand, you have the experiments.

54:24Konstantinos Karagiannis: And the way quantum works, I mean, it's an often misunderstood concept, but basically every time you add a qubit, you by some metric. Double the power of a quantum computer is the idea, right? 101 qubits is twice as good as 100 is the idea. And peop our brains don't work that way. You know, it's exponential growth. We don't really think that way. So once a quantum computer is like a little better at, let's say, chemistry than the best classical model, it'll only be a few months before it's orders of magnitude better, possibly. And that's exciting. That's really exciting. But I used to think, and this is one thing I definitely was wrong about. I used to think we would have a few years of like all these examples of advantage before we cracked encryption. And I didn't realize how much research focus people would put into cracking encryption. Like I used to get asked, what will be the ChatGPT moment for quantum computing? And I was yeah, and I would be like, Well, it's gonna be when we have provable advantage. And I mean Willow said they had it, but not really, because random circuit sampling is useless. It's a simulation. then D-Wave came up with a paper where they showed they modeled, they were able to do modeling Ising model magnetic use cases that would apply to physics, and that exceeded what a classical computer would do by if you listen to them, thousands of years. And no one's actually defeated the D-Wave claims, which is kind of cool. It's still out there, it's still, you know, peer-reviewed. So so all that did happen, but still we don't have that like practical business use case that has provable advantage. It's just like what let's say D-Wave would call customer advantage, you know. So I was wrong about the idea that we'd have tons of advantage before we have RSA being cracked. And now the sad reality or terrifying reality, I should say, is that the ChatGPT moment might be Q Day. And that's

56:13Eva Galant: You're like, no. Yeah, it's not a good version

56:15Konstantinos Karagiannis: That's not what anyone wanted, but it's very it's very possible. I mean, like the pinnacle paper. you know, they said they're gonna improve it, right? And in that pinnacle paper, they didn't just have the cracking encryption, they also had this this like chemistry use case that that dropped the number of qubits down by an order of magnitude. So they also gave us something quote unquote useful, right, in it. But those are two work streams in parallel. I I don't know what's gonna happen. I don't know what everyone's gonna put all their eggs in. And and I'm sure if you ask a nation state, realistically they wanna crack encryption. You know, that that's that's what they wanna do. Yeah.

56:50Eva Galant: Of course. Yeah. Because that's the biggest threat for everyone. But coming back to actually use cases, because I'm really glad you mentioned that and we are running out of time. So try to be quick with that. But there's so much stuff. really enjoyed this conversation. So much stuff I would love to follow up with you. But coming back to the real case studies. So we obviously see a lot of those POC around and all the bigger quantum hardware companies are now. this and that and different corporations with different enterprises. As you said, there's a lot of criticism as well saying, hey, but none of them is actually on live production right now. From my perspective, I'm happy that those POC are happening because as said, we are getting ready there slower. But what I was thinking during the preparation for this conversation is like, you have a very unique experience and also exposure to the clients in your everyday work. And I wanted to ask you, how the enterprise companies look at the, know, quantum problem, well, not quantum problem, the problems that they have right now, okay? So the problems we all know, the drug discovery, the materials, the chemistry, the optimization problems, la la la. From enterprise company, it doesn't matter actually what do you use. They don't care if there will be a quantum computer of which modality. They don't care if that will be, you know, AI system and so on and so on. AI arrived first, it was faster and cheaper. Quantum is still there on the stage of POC. Please tell us, what do you see in your everyday work? How the enterprise look at this solution or implementation of quantum versus AI?

58:29Konstantinos Karagiannis: Yeah, and and I agree with you that companies don't care at the in the end. sure it's nice to have a a marketing blurb that you used quantum on a problem, but but it doesn't really matter ultimately. If you get a few points on a financial application, that could turn into millions of dollars. So that's when you matter that's when it matter. so the cloud model supports that too, right? Like today you go into the cloud and you say, Hey, I need two CPUs, three TPUs, five GPUs, and one day you'll say, and two QPUs, whatever. So that that's how it'll all end up. That's great. and and as I said, you do have to start small with the hardware you have today to prove an approach so you can extrapolate for for future advantage. And I believe that. I believe it's not just for chemistry, I think it'll be for everything. So most companies should, if they have a serious problem that's critical to their business, and they're solving it with some combination of smart people and AI and other things today, they should realistically be doing a POC. With quantum just to get a sense on what it would take to push it a little bit, you know? Yeah. So I think I think that's important. and you could do that in a way that gives you some ROI today. Like, like I'm big on tensor networks. I gave a talk at DEF CON on those too a couple years ago. and tensor networks are cool because they run on classical hardware and they're quantum inspired. and I don't mean they're inspired by quantum computers, they're inspired by quantum physics itself. And and The beauty of a tensor network, which some people don't know, is you can map any quantum circuit one-to-one perfectly to a tensor network. The challenge is then if you make it too big, you have to do what we call contracting and and that becomes challenging for some hardware. So in theory, you can you can take any problem, turn it to a tensor network and run it if as long as you do it right. and it's the same goes the other way. So if you try something on a tensor network, you might be able to map it to a quantum circuit in the future. So you can build a tensor network that does something today and gives you a performance edge, run it on classical hardware, put it into production now, and then when a quantum computer gets good enough, you could say, Hey, I think we can adapt this to a circuit and see how it runs. And that I mean, I'm I'm way oversimplifying for your listeners. There's a lot more involved and there will be intermediary layers of coding involved. The concept is there.

1:00:35Eva Galant: 100%. Yeah, but that's the concept. That's the most important.

1:00:39Konstantinos Karagiannis: So that's one thing I think should be said more often. Then there's there's some other areas of things that are actually real today. There's there's quantum sensing that no one ever talks about. It's not that exciting, I guess. But but it does have some it's exciting if you're a pilot because there are 1,500 GPS denial attacks a day, which is insane, including around like military conflict areas and stuff. So think about like a plane is told, hey, you're really flying this way when you're really flying that way, and that's not good. So something like a quantum sensor can can avoid that. And and you can do other kinds of like synchronization with satellites and things like that. So so there are other quantum technologies that work too. But again, I I think every company has one use case that is super important to them and they should be looking at some way that they might be able to push the envelope in the future. That's or most companies. but every company needs to deal with the threat. That's just a sad reality. You know, no one's immune to the PQC threat. Yeah.

1:01:33Eva Galant: percent. Yeah. And that's a good summary. Konstantinos, we need to wrap up. we are past the time already. So I don't want to keep you longer, but before I let you go, one last question on the record. I would like you to bet on the time, specific year, when we actually will have the cryptographically relevant quantum computer with enough number of qubits, whatever the number is. And that will be existing, maybe not commercially available still, but just existing. When do you think it's going to happen?

1:02:03Konstantinos Karagiannis: Yeah, the terror is that if it's just existing and not commercially available, then we might not know.

1:02:07Eva Galant: You may not know about it.

1:02:09Konstantinos Karagiannis: Yeah, that that's that's scary. and that and publicly on the record, I do believe it's gonna go that way. I think the first relevant machine will be hush hush. So like I said, 2029 is the first year that's possible. And and it's it's a weird situation because it's not about a set qubit count that we have to build towards. It could surprise us. the people working on having X number of qubits by X year could be surprised that two years before that, they have a machine that's powerful enough because of someone's paper that was published on arXiv. So that's that other surprise, you know, variable that that we don't know about.

1:02:45Eva Galant: So basically in the best case scenario, we have three years to get ready.

1:02:49Konstantinos Karagiannis: I think that's first year. as far as harvest now, decrypt later as I mentioned, if you're not ready the first year, well, you know, with Mosca's theorem anyway, your your data's out there. But as far as like short term transmissions, things that are only valid for a few days, those are not going to be cracked with harvest now decrypt later. It's going to be years before anyone turns their attention to that. It'll just be that really long shelf life data that's already been harvested, you know.

1:03:11Eva Galant: Yeah. 100%. Well, thank you so much. It was really, really a big pleasure today. I really enjoyed the discussion. Please let us know once the book will be out already so we can link it in our social media channels as well. And once again, thank you so much. Enjoy the rest of the day.