Coldcard Mk5, quantum, and AI

Hi everyone, and welcome back to Stephan Livera podcast. Rejoining me on the show is my friend NVK, many of you will know him as the founder of CoinKite, the creator of the Coldcard and various other Bitcoin security, products. NVK, welcome back to the show.

Hey man, thanks for having me.

So, NVK, I know you guys recently put out the MK5, so, tell us what's new with the MK5.

it was time to refresh that line, and, we have like a nice screen

On it, we've upgraded our industrial design capabilities, so the case is a lot harder, it's way more refined. we've, we've built a keyboard from scratch so it's a really nice keyboard. We moved the USB to the bottom so that it's more aligned with, like how people use it, and, we've improved the NFC as well, so the NFC is on par with QLine now. It's very nice NFC. so that you can do your push TX and broadcast transactions without needing a computer. what else? and we added weights so that it holds nice on the table, when you're using it. And, you know, it's got the, the, the slide cover like MK4, it's the same security as the Q. the device is essentially, you know, nobody has, produced to us a key extraction yet of MK4. Or Q line, same as for Mark V, and, we're pretty happy with it.

Cool. And so, yeah, so I'm, as I'm understanding it, then it's, more like a, a visual and a feel refresh to the product. but the idea is you're still gonna keep them running as separate, you know, you'll have the MK5 on one side and you'll have the Q, and are you planning to keep them both running parallel?

Yes,

yes.

there, there is like a lot of people. I, I'd say

percent of people, of our market prefers a smaller device that they can conceal easily, right? And they like the form factor of the MK line, they just wanted a bigger screen, you know, and a few sort of improvements. So that's essentially what we did, we were listening to our users, right? and then the other sort of, you know, forty, forty-five percent of the market prefers the Q with the, with the full keyboard and, you know, QRT and, you know, the scanner and all that stuff.

Yeah, yeah. I think, the scanner feature is so cool for me with the, with the Q. but, yeah, so let's talk about the firmware side of it also. I know you're-- We're still on firmware five point five point zero. what are some of the latest things there? I see you've got, BIP three twenty-two proof of reserves signing.

Yes. Yeah. So that was super cool. there's a lot of,

Devices that needed to do proof of reserves internally for their own compliance. so we added that, and it's kind of cool because you can just, scan an, QR to prove reserves too. Super easy to do. so, so that, that's there. we also have, got it. And sorry,

just on the proof of reserves, as, as I understand it, it's kind of like a, it's like a PSBT signing kind of process. That's right. And then that, in doing that, you are proving the reserves, and, you know, that might be useful, like you said, for compliance or audit reasons or, I know of, of maybe even in some cases where people are doing like You know, they're a customer of a Bitcoin Neo bank or something like that, and they wanna prove ownership of coins. So that's like another use case where maybe the user wants to be able to prove that he controls these coins without actually moving them.

That's right. It, it's, it's used a lot, and I think there is more, like Bitcoin features coming that we will need some proof of reserves, not, not like base layer features, but like people, things that people are building, for contracts and for other things of, of such nature.

Got it. And so, a-anything new on the Miniscript or, Miniscript side of things?

Yeah, I mean, we, we've been sort of like keeping on par, with support, pretty much, it's just like, essentially like a ledger and Coldcard support the- The, the, let's call it the frontier of Miniscript, we, we, we--

Right, on the edge firmware, to be clear, instead of the standard firmware. That's

right. We just keep on sort of, you know, following the features and trying to make it available to the developers. you know, we, we also support it because of Anchor Watch, and, and Liana. so yeah, folks are playing with it. it, it is starting to see more adoption, but you know, like, old school, multi-sig still king. That's what most people use, that and single sig plus passphrase. So, that's interesting there. One thing that we've been putting a lot of attention in the recent year is making sure that we have a rock-solid setup for spending policies. And a lot of people use our devices that way now. It's, it's a very common use case where you essentially have maybe one or two devices, and where you have one device that's available to you, right? Say on business premises or, or home office premise, where the device itself, may have the, you know, has the keys to sign your coins, but, it is locked down to a spending policy. So, you know, you, you can't spend more than whatever you define. Say, for example, you know, you define that you can only spend a hundred thousand dollars or something like that, right? that means that, you know, like if you're under duress or, or something happens, you know, and you're forced to spend, you can only spend that limit, right? They can't take the rest of the coins of that, of those keys using that device. And what's cool is You know, in good Coldcard fashion, there is a million ways for you to conceal and, and, and sort of like, confuse attackers. You know, like the-- we don't say what the spending policy is at the time of signing, right? So you have to know that, so you can make it fail many times. we also have a bunch of sort of like different ways in which you can force, you know, fake fails. you can have it so that it can- Only sent to whitelist. So for example, one use case people are using Coldcard a lot for now is to top up their Bitcoin back loans when the Bitcoin price drops, and that device is set so they can only send to the whitelist addresses of their- Lone provider. So if the price drops, they can sign a transaction, send it over, save their loan from getting called, but they can't send it to any other address. And that device is set to that.

Interesting. And so in u-using that, just for clarity, they could be having it in their office or home office, let's say, and then if they wanna spend the larger amount above that threshold, that's where they've gotta go to another device or like the multisig. How are they setting that part?

Well, it see, it-it'll be like single sig or, or multisig, depending, 'cause we, we support spending policies on both modes. Right? The, we have a setup where in multi-sig, the device can have two keys, and they're treated securely differently internally. So that they can enforce that spending policy. I mean, I, you know, I'm not gonna get into the weeds in, in a call like this, but the point is the functionality is there, right? So if you have multi-sig or single-sig, the device can enforce spending policies, so that it can be operationally, like, you know, in a good spot if the price drops, for example. for your Bitcoin loans.

Any standout user feedback or surprises since launch? Honestly, no. I

mean, like people, people like get it, like we, we have, we're very lucky to have a very dedicated user base. You know, we're not dealing with, with the full DeGen folks. you know, doing treasure ledger for, you know, their slew of shikokens. And, you know, the people who gravitate to our devices tend to care a little bit more about understanding what's going on. So they truly leverage like the features that we built to the max, right? The people really use the stuff we do, y-you know, in the way that we intend. so we end up getting like very good feedback that way. And, you know, it's, it's rare that we, that you're gonna find a Coldcard user that gave a Weigh their seeds on a DM, right? Like, this isn't our user base.

Yeah, there's a selection effect at play, like it tends to be the more intermediate and advanced level. It's more

people who have some money, you know, like it, it ends up being people who actually have coins, not so people who just started collecting a few, like Sats, right? I always say like, if you, if you don't have enough Bitcoin To justify the purchase of a hardware wallet, just stick around with a phone, like a phone wallet is fine, you know? And, and then you upgrade from there as-- because as, as your stack grows, your interest and investment, like knowledge investment into understanding self-custody is gonna grow too, right?

Yeah. I think that's the, the common pathway for a lot of people. I also wanted to get your thoughts on the Bitkey, because I know there's been a lot of conversation back and forth with that and Bitkey. Guys recently put out a new version with a touchscreen, and that was one of the big, criticisms people had, but I know there are some other maybe philosophical differences that you might have with the bitkey style of approach where it's like a two of three with blocks server key and the seed phrase is not exposed to the user.

Yeah, I mean, I think it's great that you have a pub code trying to create self custody. I just find it very disappointing that the marketing doesn't sort of like, is not honest enough. because the reality is like, you know, this is a genuine, honest company and they're gonna do a lot of like marketing, they hopefully are gonna get a lot of people into Bitcoin. But, you know, they're removing the, the unilateral exit capacity, which is Bitcoin's biggest fe- biggest feature for self-custody, right? The reality is, you can't export your keys out of a big key, you can't ex- export your keys out of the app either, and you can't export the keys out of their servers, right? So that means, you know, if they want to, they can make people stock, right? they maintain an APK for Android, for people to, to load things if they still have the hardware and if the hardware doesn't have any malicious Firmware, which they only control. We're

talking about like a sovereign recovery, I think, or it's called emergency exit kit, I believe. That's right. Yeah, the break the glass. You put the PDF in your cloud, and the idea is like if block goes down, you get this PDF and some other, and the Android- It's

theater. Like, I, I mean, the, the, the reality is it's theater. Like, because the, the problem is like you have a device that has auto-update, the user doesn't control the software updates, right? And they control the whole stack, so if they wanted to be malicious and push, a bad firmware, right, that doesn't allow people to send coins to certain addresses, they could, right? And then the user is stuck trying to send with their APK, but the APK is controlled by them. You know, like, i-i-it's not ideal, especially when you're a pubco, right? Because if you're a pubco, you're more scrutinized, you're more vulnerable to, to sort of like government, concerns, right? so in that sense, I just find it disappointing that there is no true break-the-glass unilateral exit, which is, again, it's like that's Bitcoin's biggest feature if you're gonna call it a self-custody. I-if, if it's like advertised differently, it's like, "Hey, use this, it's better than a phone app, it's better than keeping on the exchange," right? But, you know- We can't, we, we can prevent you from, you know, exiting, you know, maybe, maybe there is like, you know, it's a better sort of path that way. I just wish it was sort of like, be-- better marketed, a little bit more honestly marketed.

I see, yeah. so, as, as I'm trying to re-- trying to understand a little bit about how that emergency exit kit works, I'm just kinda quickly looking, I don't know the detail of it. I guess you could have downloaded that app beforehand and Do that, especially the target market. Yeah, I see. Yeah, but I mean, the-- but the, I guess the counter-argument would be, hey, they're gonna bring more people into Bitcoin, right? That they are going to,

Well, so is MicroStrategy, right? Like, I mean, MicroStrategy's gonna bring more people to Bitcoin. Nobody brought more people to Bitcoin than Coinbase. I, I think it's like, y-y-you know, like again, if you're gonna have the do-gooder sort of like, "We are, we care about Bitcoin," sort of narrative, right? Well, then, then be honest about it. I, I, I think, you know, like, it's, you know, Coinbase is honest about, like, you know, what they do and how they do it, right? They just, just be honest about it, don't, don't confuse people. I, I think that's, that's the main beef I've had with that product for a long time, because I was one of the biggest supporters.

Yeah, I, I recall, maybe some of the earlier conversations, maybe you were- maybe so part of your criticism was that they were, let's say, flooding seeds and so the marketing, I think maybe that was part of it. I think, look, what we're getting at here, and look, I'm happy to, you know, host someone from Bitkey on and Block to kind of talk about, from their perspective, but, as I- I, I, I guess I'm somewhere in the middle. I'm okay with like, Normy's kind of, using Bitkey as an example, but for me personally, I'm gonna use Coldcard, but, and for my family, I'll teach them how to use a Coldcard. But, if it's someone who maybe you don't have the time to spend with them, then, yeah, there's just gonna be this fundamental trade-off. And I think, to be honest to you as well, or to be fair to you, I think there's maybe Maybe you and perhaps even Nunchok, the Nunchok guys are also kind of commenting in a similar direction of saying, no, part of the way you use the Nunchok app is that you first create a key and then you create a wallet so that you can understand what's going on under the hood, whereas maybe with Bitkey, the idea is the user doesn't have to understand really much of what's happening under the hood.

Yeah, I mean, ag-again, like, I think like the best thing that we can have in Bitcoin is like honesty. Right? So like if everybody's just clear about their trade-offs, right? I, I think it's a big win for the con- for the consumer. I, I, it's always been like that. Y-you know, a-and it really is that simple. It's like, hey guys, you know what? Like, we have this super easy to use, beautiful new device. You know, like that screen is gorgeous. y-you know, like the only problem that we have is that, like, you know, to make this this easy, you don't have any lot of

exits Because it's like they've got the emergency exit kit, so there is, is that, is that not a unilateral exit, but it's just a bit of a, it's a, it's a harder to access and it's in a different way to the typical like, it's not a, it's like, it's kind of more, you maybe you could argue it's more like a vendor lock-in esque kind of approach as opposed to the- Well, the main problem

is that like it can be, it can be rugged, right? Like that, that unilateral exit is but that's just like splitting hairs, like listen, you know, I hope they do well, I hope they, they onboard. Like, on the other hand, I mean, like what they're doing with, with like, with, merchants is incredible

Right, I think I saw Miles announced they had like over eight hundred thousand merchants.

It's incredible, absolutely incredible, right? Like they should focus on that. I mean, they're very good at that.

Yeah. But I think, I guess, I see them as kind of, they're doing the ecosystem, they're doing everything, right? They're doing the mer- the merchant processing side. Yes. They're, you know, I mean, who knows? They're probably gonna do loans, you know, they're probably gonna, you know, they're doing the

hardware side of it, they This is the problem, right? This is a public traded company, you know, I'm sure Jack has the best in his heart for it, right? But it's not his company, right? It's the shareholders' company. But

when the government gets in, yeah, it's- You know, like look at Twitter, right? I mean, like,

you know, eventually, like, you know, he's, he's gonna get bored of it and move on with his life, right? Like, a-and then, like, who takes over next? You know, may-- they may not it, it's just not the right vehicle for self custody.

Yeah, I see. well, I mean, I'm somewhere in the middle. I'd be okay with like, you know, people just learning on that, and, you know, maybe over time, then, you know, figure out exactly how you wanna do your self custody and if you want it. But I think really, to me, it just comes back to there's like this fundamental philosophical difference of like what, what should be the approach for, let's say, for self custody for people who

Maybe the answer from the Bitcoin side or the block side would be, look, this isn't meant for people who are, you know, ho-holding large, large, large amounts of coins. I don't know, I, I'd have to see. but anyway, let's, let's move on. I think there's some other interesting things around Bitcoin security that are worth to-talking about. I'm sure you might have seen in the industry, Bisc just recently got hit yesterday, there was a big exploit, I think it was about eleven BTC, it was, protocols and what it means for us in hardware security.

I, I, I didn't follow that one, but like if I remember right, isn't Bisk using some xCoin or something like that to, to make things happen? I, I lost track of that project.

Yeah, so they have like, I think they have like a colored coin B-S-Q, I don't know the details, this is maybe, slightly out of date information, and I believe they do have some altcoin trading, so maybe it relates to that, I'm not 100% sure though We are now, we don't know for sure, but they were saying the BISK, announcement thread mentioned that it is likely an AI-powered attack. So maybe people using, you know, is it some North Korean group using Claude or Cursor or Mythos or something?

Yeah, but that's like, you know, that's, that's like saying that they just encountered an adversary that's a little bit better than they are. I mean, like, you know, the reality is people are trying to break stuff all the time, and if you have like Bitcoin to be taken, you You know, bed

security. Yeah,

but I guess the point would

be, is there-- Has the game changed, right? Is there a, you know, now this is a big new threat that people need to start thinking about, which is basically AI-assisted hacking, and in this case, it's like a decentralized protocol, right? So obviously the keys aren't offline, it's a different thing, like it relates to like a smart contract, or if they find a bug or an exploit, whereas obviously when we're dealing with like, you know, your cold card and your long-term security Kept offline, it's a very different thing. But by the same token, I mean, who knows, maybe in the future, if people start doing more, quote unquote, DeFi or lending, using like, you know, Bitcoin on chain, then maybe there's kind of, an angle there. But I guess, do you think the AI assisted hacking is changing the game in some way, or is it more like, well, there's AI assisted hacking, and now we need AI assisted defense, and you need to fight, you need, you need to defend,

Review of the Coldcard Cyber, whatever they call it, the KYC NDA version of their, security assessment tools. you know, the first thing he did was run after, run it on the Coldcard repo. And, you know, honestly, like we, we saw the, the bug reports, they're all like, you know, extremely mediocre stuff. everything was like high, right? Like they, they said everything is like high, high, so it's like super dangerous, and everything was like completely false reporting. The tools are still abhorrent, the quality of this, this, this hacking, AI hacking is still ultra, ultra- you know, a-again, like if you're gonna point that against, you know, WordPress and a bunch of JavaScript stuff, you know, of course you're gonna find a bunch of stuff, but like if you point like a, a, a, an Indian, company that does like, you know, like those like, assessment, threat assessment things, you know, against these JavaScript tools, they're gonna find a bunch of bugs too, right?

I just, you know, it's, it's gonna definitely get better, it's not to dismiss it. I, I think it's gonna definitely get better. I, I think, you know, 'cause the reality is these machines can, can try a lot more stuff concurrently than a person can, right? But what we notice is that like these machines, they've been trained on very poor code, right? So most, most of the code they've been trained on is like crap online, right? Like, it, it's not, you know? All the closed source missile like guided systems and, you know, like, and the, you know, like, th-th-there's a lot of amazing stuff out there that just isn't public, right? And, and these machines, they can't test against hardware yet either. So all the embedded code, they can find like obvious bugs, right? But they can't Test their bugs because they don't have access to the embedded hardware in an, in a, in a, in a lab, right, on their desk, because that's how you test embedded hardware. You can't just like connect to it or just run a, a virtualized version of a, an MCU on your computer So, anyways, I, I think it's gonna get a lot better. I think we're gonna see a lot more interesting bug reports in the future. it's gonna be very helpful. I'm super optimistic about the stuff. I think it just makes everything better. but it's-- we're definitely not there yet. It's like it's only things that are very poorly made that are getting cracked. So like, if you're Coca-Cola and you're Procter and Gamble or whatever, like, you know, you have a bunch of cobbled together garbage running on your enterprise servers, right? of course, these tools are gonna find a bunch of holes. But, you know, finding holes against like hardened Bitcoin security things, I think is gonna be a lot harder.

Let's talk about AI. I know you've been doing a lot with that and, also the Bitcoin quantum debates, you've been weighing in on some of that. We've seen, some, FUD from the, quantum types who are panicking about it, and you've been responding back to them using, some AI as well. Do you wanna talk to us a little bit about the quantum side of it before we then get into the AI stuff

Yeah, I mean, like the, the quantum stuff, I mean, I have a pedestrian, you know, like understanding of a lot of the quantum stuff. I, you know, I don't claim to be an expert on it. You know, I do, I do understand some of the hardware related to that and, you know, the limitations of it, like We are so beyond far anything that is usable, for actual applied, cracking. you know, we still need-- where, where we might have-- I, I, I actually like, I am skeptical enough that like, I don't even know if we get to proper quantum computing, by the way. You, you know, we may just not find a way to scale the, the hardware, like, and to scale the computation. There is a very hard

It's, we're at a point now, I think the best analogy is like, we have essentially like, you know, time machine science, like time machine math, right? But we can't build the time machines. Think about a Delve, okay? No, really, it's like, it's like a Google

paper talking about the quantum circuit, yeah.

It's all crazy shit. Like when you actually dive in, it's like, you know, like, it's, it's like, it's kind of like similar to talking about like black holes, you know? It's like we We have some math, you know, and it's also math that we can't prove either direction either. Like, so like, we just, you know, we can describe what we think is describable in certain ways, but You know, and, and like after you sort of like dissect the realities of the math itself, which is still not like as solid as people think it is, including like the scientists, like the people working on the shit say it themselves, it's not me. Like, you, you know what I mean? Like once you get out of Twitter and you actually go research, like, you, you look at the size of the preambles or, or the, the assumptions that people put on the, on their papers to be taken seriously, right? It's ginormous. Like they're Listen, you know, like assuming this, this, this, and this and that, which isn't very solid yet, you know what I mean? We can't calculate this, you know? That's where we are with that stuff. So, and that's not even touching on this insane absolute, like, you know, s- space, time continuum changing hardware that you need, right? Like, it, it, you know, it, it's just unrealistic to, to be freaking out on Twitter. Like, I mean, you know, I'm sure somebody had some shorts on Bitcoin, somebody made some money, you know, some people have some, some quantum, some PQ, some post quantum math to sell, you know? I, I would say like at this specific time, like I would just like put it into the ignore bucket.

I say, yeah. And, it's interesting, I, I think, one of the articles had, made an interesting point, which is often that experts in a field often- are wrong in their technology predictions, and even domain, domain insiders are systematically twenty years more optimistic than outsiders.

That's right. I, I mean, like, unless we have, you know, some gigantic breakthroughs, in, in, in both the math and also the hardware, like, I don't see us having quantum computing for like fifty years. Remember, like, we've been working on this stuff for, like, you know, more than fifty years. It's not new. and again, it, it's just, it's just too out there. I, I, you know, like it's so out there that we can't quantify and qualify any of the advancements either.

The other thing, obviously, people will wanna know, just because obviously you're the, you know, you're, you're the guy from, from CoinKite, people are might be curious on if we were to do some kind of quantum mitigation and, you know, just for example's sake, let's say it was Jonas, Nick's, Shrinks plus shrimps as an idea, and if we had to have much bigger signatures, you know, what would that mean for us in hardware terms? Like, you know, do you have any kind of speculation on what that might look like? Like, would it be feasible or not? Like, just if we were to assume, like, okay, the quantum thing is real and it's coming in, I don't know, ten or fifteen years, like, do you have any thoughts on what a quantum hardware wallet would look like?

No. I mean, we have, like, there is Twenty different proposals for, I don't know how many anymore, like maybe ten, twenty, whatever, proposals of like new cryptography. You know, for the hardware itself, like, we, we don't know, right? We're gonna essentially just sign what- Bitcoin needs signed. Gotcha. Right? So if, if it looks like we need whatever new curve, whatever new SPAC, you know, it's gonna come to Libsack eventually, right? And then we're gonna support it, you know, there is a non-zero chance that it might require new hardware depending on the, the requisites of the signature, right? 'Cause these are all big signatures. yeah, you know, I, it's, it's impossible to know, and, and it's so out there still that, you know, I just, I just, it's hard- It's too early to

even speculate. It's, it's too

early. I mean, I highly recommend people going to bitcoinquantum dot space. I have three articles there, fully researched. The experts couldn't poke holes in it, and, you know, like it really goes through everything. you know, this isn't my research, even though it's in

person there. a-and all I, I got the bots to do some, some like very, very, very, very well cited and researched, papers. Or like essays or whatever you wanna call it, and just review what's out there, you know? The reality is that, you know, like an abacus and a dog can do better than a quantum computer right now. So let's,

yeah, I mean, I don't know if there's a paper about that, yeah, yeah, yeah. Like,

this, this is, this is true. Like, really, like that, this, this is where we are. An abacus and a dog can do better math than a quantum computer at this moment.

So yeah, so it definitely

You're really into that. Also, you've got, some websites. One of them is called LearnToPrompt dot org, and, you've also been working on this, LLM Wiki. So tell us a little bit about that. Give us an overview on what you're doing there with AI.

Great. so So essentially, w- what's, what's his name's, Stephan Livera, NVK, Coldcard, quantum- Andre

Karpathy, the guy who used to be like a Tesla AI or something, yeah? Yeah.

Yeah. So, so that guy put out a very, very interesting, idea file or, you know, like a, an idea out there, right? Where You know, to use weak keys to, to essentially become the persistent memory for the LLMs, right? For people that don't know, the LLMs, the AI, whatever you wanna call it, these large language models, they don't have real persistent memory, right? They're, they're kind of like they live in a memento, sort of like, remember Memento, the, the movie? I

haven't seen it, but I've heard of it, yeah.

Yeah. So every time they wake up, they have no memory. Okay, so every time To stock, which is the last date they were trained in, right? so, so then how do you, how do you use them? So like you tell them to essentially create, to, to absorb context, right? The context may be the last session's history, it might be the current repository you're doing coding in, it could be like a list of things you tell it to, to sort of start looking into, it could be like search the internet for this, you know, that kind of stuff, right? So, you know, that's very restrictive and that's very, and like, you know, it really doesn't scratch the surface on how, like, usable these models are, right? It's like, it's like having a Ferrari with, with no fuel in it You know, it's like, you, you know, like you put a drop in the tank, you can go very fast for like a second, but then, you know, it dies out, right? So it's not very usable. So, and then AIs, they use vector databases, right? And vector databases are great at some things, say for example, fuzzy search, so they can find things that are related to each other because they find points on a map but they're not very good at finding precise things, right? So that's why, for example, using time series data on, on LLMs like Do My Taxes sucks. Right? so then what people do is they create like plugins and they create like things to try to infer back into the session, right? Like, the, the data and try to guide this, this very, smart, engine, right, into the correct, path, right, in the correct context. So what's interesting about the wikis is that wikis are essentially a graph database, and a graph database is more like a, a, a relational database. So you have a relation between topics, right? So you're no longer looking for math points on, on, on a graph, you're now inferring that, like, you know, like the topic of quantum computing and Bitcoin and, you know, and then like this other topic, they all sort of relate to each other, right? Because you're looking- Fully giving an index of a wiki that points to these things and they each point to each other in different ways, right? so anyway, so, I, I got the bot, you know, the, the LLM to, to go and start working on what would be a good wiki based on this guy's, like idea, and, you know, and then after, like, you know, I've been working on that for, for months now, but, you know, that's why the LLM wiki was born orn l l m dash wiki dot net,

and- And so I guess in simple terms, it's like you can create your own LLM wiki and it can ingest papers and debates and material, and then you can then go back and ask- Good about those things.

So the, the LLMs are designed to guess the next word, okay? That's literally what they do, they're just very good at guessing, right? But I don't want them to guess I want them to research and read and then, and then do the conclusion, like do the thinking based on the actual data. So essentially what they-- I want them to do is like, do everything from like, you know, near first principles. So what the, so, so then what the LLM does is, is it, it's agentic, right? So it has multiple agents, right? So for example, on the research side of the LLM wiki, I have five researchers. I have the researcher that looks for, for the stuff, I have a researcher that looks for news related to the, to the topic, I have a researcher that looks for the technical analysis of it, I have a researcher that looks, from a different angle, and then I have a researcher that is a contrarian. Right? And then they all kind of like keep on like talking to each other while they're researching, almost if it was like a research group, right? And then they come up with their research summary and all the stuff that they found and all the, the citation and everything, right? So, for example, let's say you're interested, I don't know, using red light therapy, infrared and near infrared light therapy for, for yourself, right? you go on itadawiki, like, you know, go research infrared and near infrared Light therapy, right? And so the researcher's gonna go out there, it's gonna search the internet, it's gonna find the papers, you know, and it's really the cool thing is you go read the papers, not like, you know Like you find some blog, but then the blog maybe re- like points to paper, then go tries to find more original sourcing, right? So that it has, and, and then it has like a, a confidence sort of scale for the information that it found. so it can infer if it's like, okay, this-- I have high confidence that this is true, I have medium confidence, I have low confidence, right? Which is very important. So, for example, you know, in some topics, it may find things that are just unfortunately low confidence because it couldn't find deeper. Like papers and, and more science on it, right? So then you can tell it, "Well, go try to find it from this other angle or whatever, right? And try to like get to your truth." but what's cool about this is, you know, you can use it for very different topics, you can use it for technology as well, not just like, you can use it for anything really. 'cause that's like how people think and how like things are built. it's, it's kind of like an engineering principle to things. and then once you have the research done, then you can Do things like, for example, you know, like write me a five thousand word, you know, like essay explaining to me this topic, right? And, and, and you use its research, you create an outline plan that you can ask it to present to you before you write, and then you write it, and you can even change, like, you can do research on writing styles. This is something I like. So, I told it to do research on writing styles, and also Also research on how not to sound like an LLM. So then, so then you can say like, "I'll put this research on this thing, also making sure that you read this paper on not sounding like an LLM." And I'll put something for me, right? So it, it be-- it starts to become very useful, usable and useful, for you to do things, and, and, and that's how you- You end up using the LLM Wiki for all your usage of LLM, like you're no longer going the GUI version of, of like ChatGPT or whatever, because it starts to look like a very Crude use of LLMs. It's almost like right now people are using LLMs like toys. People go on the LLM and they use it like if it was Google or something, right? and LLM like fails miserably and it's not great. Now, if you start using like, like, just very powerful ways with multi-agent things, like, it's, it really shows like the potential and, and like how far we are on this stuff, it's remarkable.

Right. And I guess the point here is you've got, so this website l l m dash wiki dot net, and as you've mentioned, it's building on what, Andre, did with LLM Wiki. So what's specific about what you have done? Is it a specific flavor or what's, what's the thing that you've done?

No, yeah, no, I don't have it. Like, I, I really sort of like, I, I, I-- I mean, maybe I'm, like, I'm grateful I don't know the guy. I'm not part of that industry, so I kind of went my own direction. this, this really sort of like takes like his idea into how I think, right? So, you know, I created like, a way for you to plan, things into specific specs or RFCs, or a spec, spec, you know, this, this thing has like a libarian feature where You know, Yup keeps all the data. It has a thesis mode, which I like to, to do it, I, I use it a lot. So, you know, like I think of something, you know, stupid, and I go like, and I say, "Okay, like, this is my thesis," you know, like, I don't know, it's like, "We can, I don't know, put rockets in space using baseballs, whatever, like something completely idiotic, right?" and, you know, the LLM wiki will think it through So you can be very terse, you can be very sort of like, as they call it, retardumaxing on it, and then you will decompose your thesis and you will try to sort of like offer you Back what you understood for you to confirm, and then we'll start research on it to try to like prove or disprove it, right? We'll try to find truth in your thesis. because like most of the time people don't start research just out of like, you know, sure you may need research on something, but oftentimes it's because like you had a thought. Right? And, and your thought is like, you know, I think this, right? And so like, I, I find that a better path to start research from. so, so there is that feature. the libary and feature keeps up the, the library, going and I also introduce this freshness because like, there's a lot of fields where the data changes all the time, right? So AI or Bitcoin or whatever, like, things are changing at all times. So, I support a way for you to qualify the- data in terms of freshness, there is audit, capability on the wiki, so you, you can audit more deeply the research, which is great. lessons is one of my favorite feature. So there, it's essentially learned lessons, the, and the shortcut is L L. So Oftentimes you're using your, your chatbot. I use, either Claude Code or Codex or Py. And You, you, you know, you spent your two, three hours working through a problem, right? You, you know, if you're building something or whatever you're doing. And, and through that process, you know, there were many times the LLMs failed, you failed, but then eventually you come to a conclusion, right? Like you, you fix the thing, right? So you have this wonderful knowledge in the chat history, right? Of, of the things That you've learned, right? So what you do is you go like la- wiki lessons learned, right? You will read the history back and we'll extract all the lessons and all the things that, like, you know, why did you think that and why did you got that and why you got that wrong, and we'll also put that into that topic wiki. So then in the future, like when you're trying to do similar things, and people bump into this problem all the time, you try to do the similar things in a new session of the LLM, and it's failing the same stupid stuff, and it's really pissing you off. So, with this feature, like, it's all in the, in the, in the wiki. So when you go, like, okay, let's, let's go back to setting up this environment that we were doing, with this, with this problem, please review the wiki first. A-and it's like, boom, it's got like this absolute golden quality context for you to use it that, you know, you won't, you won't destroy your stuff, your-- Yeah. You know, you won't waste your time. So,

so you see the value for a researcher, for, and I guess the idea is-- Everything. People, I guess, pre-AI would do, you know, this idea of like obsidian and this kind of quote-unquote second brain, and they would put all the notes in there. This is kind of like a way to, maybe in a way, loosely, you're, you're autom Or Claude or whoever, whatever, or Gemini or whatever, you are like setting up your own knowledge base based on all these articles, and the AI is kind of collating the material for you and rating it or ranking it based on relevance and things like that, and then you're able to query for an output Based on that. Is that loosely what's happening?

Yes, and it's for, and it's for everyone and everything. Like, the, the wiki, what's cool about the wiki is that like You know, we have like topics of interest, right? and, and it's surprising how many a person will have. And like I'd say like sixty to eighty percent of the time when people are querying Their chat bots, right? They're querying related to topics that they queried before, right? Yeah. Maybe, but it's like probably sort of like new ground on that topic, right? What's cool is that, like for, if you do that using the wiki, now you are, adding to your corpus of understanding, but also you're not repeating the same research, you're not repeating the same mistakes. Right? Because again, the LLMs hallucinate because they're trying to guess the next thing, they're not doing the research. Got it. And that's true for anything you use LLMs for.

So you think it might actually save you tokens?

Oh, absolutely, yes. I mean, it's, it's kind of a wash because, like, you know, the agents use a lot of, a lot of tokens. Like, this isn't a cheap sort of system to use, right? There's-- because the context is big, it, it likes large context models. and that's why it's so good. but, you know, I don't have a concern about that, like things are gonna get like cheaper, faster, better in the next, like, you know, as, they are everyday getting cheaper, faster, better. Got it.

And you can use different AI agents, right? You've got Claude, Code, OpenAI, Open Code, Py, and I guess so. The idea of the time-- Yeah, so I support everything. You might eventually start supporting even those, like the deep seek of, deep seeks of the

Oh, no, no, that, that stuff is already supported. So, the support really is for the, the harnesses, for the holding agents, not for the models. So like, you can use any model you want with this stuff, okay? As long as they're like generally fast and they have larger context windows. But the- The actual support itself is for the harness, so the harness knows how to call the, the, the wiki agents and stuff. So, you know, I really support all the open ones that you can use with any model, so like Open Code, Py, you know, or just any LLM agent with an agents file. that's all there. and, and it works. Like, I mean, we, you know, like the repo has now, I don't know, like forty, fifty, forks and three hundred plus, almost, yeah, three hundred and fifty, stars on it. So people are using it, that, that stuff works, and people are using it with very different sort of agents.

Interesting, yeah. let's go now to your broader one where you've got learn to prompt dot org. So as I'm understanding this, this is more like For using AI coding agents but securely and effectively. Can you elaborate a bit on this? What's the goal with learn to prompt dot org?

So I, you know, like as I started playing around with this stuff, I'm like, like people are just running Like, you, you know, essentially like loose dogs on their like user space that's full of secrets on their computers, especially as devs, right? Like, it's ludicrous. It, it is absolutely ludicrous to run, like, you know, cloud code on your computer, like it's crazy. the thing has like, it, it could go and look at your SSH keys It's, it's really bad. So I started doing research on like how, 'cause you know, the best way to run LLMs like that is to really use a virtualized way, right? So use a VM on your computer or something like that. But you know, sometimes you do want it on your main machine, right? So I started doing some research on like, how can you accomplish that? And, you know, also I bump into no no, which is great little library, like a tool that creates sandboxes, for

sandboxing the, what you're try-trying to work on. And I guess the idea is, also for secure, like very security conscious or security, you know, relevant things, like your SSH and things like that, or I guess your, your passwords and things.

That's right. So, so it's- Essentially, you-- what you can do, you can do a few different things, right? So, I'm doing a chain there of tools. So first of all, you pin down the, like the, the tools you're using, so you have a hash of the tools and a hash of their path. and then you have, AMFChain, which is like a, a cool little, tool to inject, to inject the, the secrets into the session. So first of all, you need API keys, right, for this device, for this tool to use, so you can inject it that way. I, I created a fork because that one was not maintained, so it's, AMFChain dash extra.

And so this is like AI, but in a, done in a more security-paranoid way, let's

say. Yeah, pretty But, but like, the, the problem that I arrived at is that like, i-i-it's, it's so bad, like, y-y-you have zero security right now running these things on your computer, zero. Because Claude is known for essentially ignoring the dot env, like instructions. So for example, you go like, "Like, ignore these security files, ignore these security folders, don't read them, right? " You can put that in the, in the Claude sort of, spec file, right, in the config file. But these tools all, once they reach a certain threshold of trying something else and they don't, they, they literally have an instruction to ignore their own configs.

Yeah, that's, concerning for people, and 'cause a lot of people just- Especially with OpenCL, it kind of opened up this idea of just yoloing things on your own machine and like connecting, not only that, but connecting it to your real world, whether that's your email or your Telegram and Signal and whatever So I guess that's, that's an interesting, thing people will have to navigate of like they want the AI tool to be able to make their life efficient, but on the other hand, how do you do it in a secure way?

I mean, like, listen, you have your like Sparrow files on your computer. You know, you have your, maybe you have your Sparrow open, you know, and things are decrypted in memory, and, you know, the,

the

machine is uploading everything. Like, and the other problem too is that like all these LLMs, the way that they function is like they will tokenize whatever context, right, that be your drive, your files, whatever they're working on And all that stuff gets uploaded as tokens. So like you're uploading all your data to OpenAI, to Anthropic, right. These are all

input tokens.

That's right. And, and they can, they can reverse that process and read those files. So it is. So like you have this thing that has like no guardrails on your computer. Haha. With the capacity of uploading everything full

permission, off you go. That's right. Like, it, it is

insane. When you actually start looking into it, it's absolutely insane. So, so anyway, so, so I created a nice little stack there. A-and, and the cool thing is like, you can ask your agent to set up the stack for you. Like, you don't have to understand it. Oh, right,

because you can point it to that side. Yeah, yeah. I,

I literally have, I, I wrote at the end of these articles, in-s-ess It's an instruction like, this was written in a format that the agents can read, so that they can set it up for you. You can say like, "Hey, go to learn to prompt dot org and help me set up my environment just like that."

And so the idea, and you could set this up on a VPS, it could be in a virtualized system, or just if you're yoloing, just straight in your, on your actual physical box.

I mean, if you wanna, if you wanna use a virtual machine, like just like a, a virtual setup,

That then just like on the virtual machine, you can, you can do whatever you want.

Got it. Okay. So I guess you're saying this would be mainly for, if you're, like the idea is you're trying to secure it on your own bare metal at home kind of thing.

So essentially, like, how did you just create some sanity with hashes of? each part of the stack, right? So you have a sort of truth of like who is what, right? Like who's running and what version is running, and which paths are the absolute paths of these tools? Right? And then, and then that gets passed on to Lightning Extra, which essentially is the decider, y-you know, if the, if the, if each two sh-which keys each two should see it, and uses the OS level keychain and, and half signature checks to do that. Which is really cool. And then no-no checks, well, it enforces which, paths of the system each agent should have access to. So, so it's like nice and tidy that everything sort of like gets like very sort of specific access.

You're trying to lock it down or sandbox it. That's right. Okay, interesting. And so, any, any closing thoughts or what should people think about, with, doing, AI coding and AI use?

I, I mean, like, yeah, like, you know, either do it on a separate machine or, or like, lock it down. it's incredibly powerful. Like, these tools are super fun now. You know, like, you, you can build websites, you can build, like, I use LLM Wiki to build the Learn to Code, sorry, Learn to Prompt dot org website. So I essentially told her to keep track of all the things I was doing on the computer, and then, you know, everything got annotated on the wiki, and then I said, "No undoxing my system, create a cool website for it." Learn to promp so other people can take advantage of this, and then it generated this website. And then every time I do it, I just go like, "Now go and update the website. Like, how cool is that, right? Like, w-we can produce this knowledge for other people to have, you know, with minimum effort." so, so things are progressing fast. I, I highly recommend people play around with it, play with LLM Wiki, if you don't like mine, there's a bunch of others around as well. and, but, but it, it does like increase your, your, the quality of the output and the output capacity by, you know, like a hundred x. you know, stop using AI like, like you're trying to Google something or, you know, i-i-it's just you, you're using the tool wrong, right? so, so that's it on the AI stuff. It's just like have fun. It's, it's, it's a fun space right now.

Awesome. Well, yeah, so for listeners, those links are learn to prompt dot org, l l m dash wiki dot net, bitcoin quantum dot space, and of course, people can check out the coldcard over at coincard dot com. NVK, thanks for joining me on the show today.

Hey, thanks for having me.