Matt Alberg (5:50)

Yeah.

Guy Swan (5:51)

Maybe 70s. I don't know. He's. He's up there. He's not 70. He's not 70. But it was funny because, like, we kind of come from, you know, two different worlds. He's a foreigner, I think. I think he's from, like, Ethiopia or something. But we were just having this. He's actually got, like, real, like, solid, like, Christian, like, Christian heritage and, like, values and stuff. And we were just chatting about, like, random stuff because we were complaining about the HOA together, and. And he was just talking about how, like, nobody wants to have kids anymore and, like, how crazy it is his daughters have basically decided. He has two daughters, and he came from a family of nine, and. And he was like. And they. They're pretty sure they don't want to have kids. And I was talking about, like, man, it's just. Everything's so expensive. You don't realize, like, how. How much that affects, like.

Matt Alberg (6:53)

Yeah, I've got a lot of stuff on that to talk about, too, if it comes up during the podcast. Let's go. Yeah.

Guy Swan (6:59)

Unpack. Yeah, yeah. No, no, let's. Let's start.

Matt Alberg (7:02)

Let's. Okay, this.

Guy Swan (7:03)

This is it. This is the podcast.

Matt Alberg (7:05)

Let's.

Guy Swan (7:05)

Let's do this. Yeah, but he.

Matt Alberg (7:09)

I think it has a lot to do with the tax burden. I think that it's. I think, you know, I even kind of, like, trying to piece it together. Like, in. In Chinese culture, they have this very popular saying. It's. It's raised young to support you when you're old. Um, it's, you know, kind of a. Something that almost everybody in China knows. Kind of like a saying.

Guy Swan (7:32)

It's a very explicit recognition that, like, that's a. That's a reason for this, you know, ongoing cycle.

Matt Alberg (7:40)

Yes. And then sometime in 19, in the 1930s, we came up with this new program to the government to support you. When you're old and. But you just have to pay into it when you're young. And so we've really distorted the incentives for having children now where you're paying out of your pocket when you're young so you don't have that disposable income to have kids. And then, yeah, it's too expensive to have kids, as you say. And yeah, I hope we can turn it around. And also when you pay into the government, you're just losing 30, 40%, whatever, to waste. I don't know. I'm sure there's some statistic out there.

Guy Swan (8:22)

No, it's like 80 or not.

Matt Alberg (8:24)

It's all, it's all of it could be way more. Yeah, yeah. And when your kids take care of you, like, there's, there's a lot more care involved. There's a lot more genuineness. It means a lot.

Guy Swan (8:37)

Imagine thinking that the government trying to save money for you when they're the largest debtor on earth.

Matt Alberg (8:43)

Yeah.

Guy Swan (8:44)

Is going to, is going to fill in the blank, is going to like, cover the gap, as if, as if the government actually cares about you and as if the biggest problem when you're old isn't going to be the fact that you're alone and you need somebody who loves you, you know? Yeah. Like, it just, it's just so, so distorted.

Matt Alberg (9:06)

And families are where, like, you know, when you get into the capitalism versus socialism, whatever, families are where I think socialism really works. It's like you as a parent know how much your, your kid needs. They, you know, you know, when they're kind of getting out of line or, you know, when they're getting a little on the spoiled side, whatever. On the other side, when they get money from their parents or from their grandmother or whatever, it means so much more. They feel like they need to pay that back. They feel like a little bit of shame involved with it. Like, oh man, I need to get back on my feet. I need to get back in the game. When you get a blank check from the government, from some faceless government every month, you don't have those feelings. You don't feel like you need to pay things back. In fact, it kind of becomes an entitlement of sorts, like, where's my money? Give me my money this month. So it's all wrong. And I think we are starting to figure these things out now that there's less and less kids coming. So, yeah, there's.

Guy Swan (10:05)

I can't remember who it was that said it like this specifically, but I always thought it was a great way to frame it because it's, it's very like the systems thinking sort of thing when it comes to capitalism versus socialism. Is that with my family, like my family and my kids, I'm a socialist. With my neighbors and my local community, I'm a hardcore liberal. With my state, I'm a Republican, and with my nation, I'm a hardcore libertarian. Like, and, and it's just because those systems don't scale. Socialism only works if you actually know and care about the people. It literally, it just, it just stops at like, as soon as you're like 12 people, it breaks down. It's just, it doesn't work for the shit. Yeah, but in a family, it, it does work. And it's, it's the, and it's this kind of like naivete of like, oh, we can all just get together. And it's so funny. They even, that's even like the culture that they spread it with is like, we're all gonna kumbaya and we're all going to be a big family.

Matt Alberg (11:09)

Yeah.

Guy Swan (11:10)

I bet. You can't be with, you can't be family with somebody that you don't know and never met. In fact, you, you already hate them because they don't believe that you should run their lives and that pisses you off, you know? But yeah, it's, it's, it's a scaling problem.

Matt Alberg (11:27)

Yeah. I've never heard it put that way or like, I never crystallized the thoughts that I had, but that is what it is. Charity means so much more when it comes from somebody more personal or from a real connection. And then even when like in the 40s, 50s, or I mean, just for centuries before, it was kind of left up to the church. Right. The churches were the ones that kind of, they pooled the money in the community. And then every once in a while, when somebody's house burned down or something, they would like pay out and help them rebuild their home or whatever. But it's much more personal than some government thousand miles away sending you a check.

Guy Swan (12:07)

Yeah, yeah. Did you have a perspective on kids and then did it change after you had kids or, or after you had a kid or like sometime recently in the past? I ask it like this specifically because I've had a number of people who in the last, probably since 2020, so for three to five years somewhere, have completely changed their perspective on kids. I don't know whether it's age, maybe it's just 2020 in general. Like, like just kind of shifted how they thought about the world and also bitcoin, I feel like, makes a huge, huge impact on this because you think differently about where things are going to be in 10 years or what the possibilities are going to be in 10 years, especially when it comes to affordability. So I'm curious, did have you already always had kind of the same mindset about this, or has that actually taken some sort of a significant shift in recent years?

Matt Alberg (13:12)

It's hard for me to say. I think it's all just been so gradual for me. I've always kind of had maybe libertarian or. I've always. I remember when I learned about capitalism, the first time it really rung true to me. It made a lot of sense to me. But in terms of the family thoughts and like, the environment that I would want for my kid, certainly that's changed a lot since having one. And it's still forming because he's only seven months old. So, like, I'm still, like, this is new territory for me. I'm still forming my thoughts on a lot of these things. But yeah, for me, I don't think there was one moment. I think it's just like a million little shifts that I've made. And it's also part of, like, just growing older. Like, you know, when you're younger, you're more idealistic, and when you hit your 30s and 40s, like, you start becoming, you know, more realist and like, it's a. It's a trajectory that almost everybody takes in their life, kind of that transition.

Guy Swan (14:08)

Yeah, yeah. On that note, actually, so we haven't talked about the thing you built that got me to hit you back, because I've been using it a lot. But I'm curious because were you a bitcoiner first or were you an AI? Like, what's your. What's your background? What's the. What's the route that you took to get here?

Matt Alberg (14:35)

I'm totally been a bitcoiner for a long time, since around 2013, and nice throwback.

Guy Swan (14:44)

What's up, Bones?

Matt Alberg (14:45)

Yeah, I got into it. Yeah. When it had its run up in 2013, I think it hit a thousand bucks. So I read about in the newspaper, something like that. But shortly after, I started reading about a lot of the, kind of the theoretical promises, banking, the unbanked, serving as money for people in every country and developing countries, stuff like that. What really got me hooked was in 2014 and 2015, there was a big drawdown in the price. But I was looking at this data from peer to peer trading website called local bitcoins and for those who aren't familiar, local bitcoins is kind of like the Craigslist, the OG Craigslist for bitcoin trading. Back in the day you could post an advertisement and say like I'm selling bitcoin, or post an advertisement and saying I'm buying bitcoin.

Guy Swan (15:41)

And local bitcoins was the goat before, before they did kyc. I was so sad. That was end of an era.

Matt Alberg (15:49)

There's certainly a lot of really crazy stuff going on there. So. But they used to publicly post their trading data and so I started looking into that and I could see that during this like bitcoin lull there was trading was exploding in a certain set of countries, places like Venezuela, Nigeria, Kenya, Colombia, et cetera. And so when I started digging in there, I found that like the trend was very true. And, and back in the day on the advertisements on local bitcoins, people would post their WhatsApp phone number that you could text them at to discuss the terms of the trade. I started texting those people and having conversations with those people, finding out what use cases were happening. So I found out there was a lot of remittance happening. I found out that there was a lot of capital flight happening. Basically anything if you're in a country where your government doesn't has like a lot of money, a lot of control on your money, there was a use case for bitcoin. And so that's what really got me hooked. I published a website called usefultulips.org some time ago. It was now kind of defunct, but it tracked all of that peer to peer trading data. And I would write articles based on the conversations that I had with these people trading on that platform. Yeah, and I spent some time with local bitcoins. I also kind of learned about PAXFUL and kind of came to know some of the people over there. And yeah, I will say though, it's a mixture of good and bad stuff on those platforms. And when you have no kyc and I'm not a fan of KYC at all, but when that stuff does happen, like, or when you have a very freely rolling platform, there's a lot of other stuff that happens there that you know, you kind of have to take, you have to accept that that stuff also happens there too.

Guy Swan (17:52)

Yeah.

Matt Alberg (17:55)

But since then I worked at Bitrefill for around five years and yeah, most of that time was full time, but in around 20, 23 or so, uh, I went to a part time position because I wanted to build a business of some Kind, um, and that I iterated through a few different ideas. Uh, but the one that I landed on was ppq.

Guy Swan (18:19)

I'm curious, what were. What were the other ideas? Were they all based around the same.

Matt Alberg (18:24)

Thing or they were loosely based around, like, bitcoin and finance. One of them was kind of a bitcoin remittance app. Because, um, in my studies, I learned that, like, if you live in the US and you need to send money home to whatever country that you're from, Kenya, South Africa, Ecuador, whatever, if you send money via Western Union, you're often losing, like, 10% or more. One part of that is the fee that you're paying to the company. But a huge chunk of the money that you lose in a remittance process is on the exchange rate. So a lot of these governments have kind of phony, arbitrary exchange rates for US Dollars to their local currency. And so you lose money that way. I discovered that if you just buy bitcoin in the US Or a stablecoin, and then you sell that bitcoin or stablecoin peer to peer in your home country, you can actually get a much better exchange rate than the official exchange rate. I would say the average would be 7, 8%. Something like that really depends on the country and the rules and the laws of each country. But. So I thought about remittance app, and then I also put together, like, a little personal finance app is actually not related to bitcoin, but it's more of like, if you're familiar with, like, the fire movement, the people who, they save a ton of money so that they can retire early. So this app was kind of in that vein where it showed you, you know, how. How much money you're bringing in, but then also how much you're spending. And then, like, how many more years do I need to work before I can retire? That type of thing. And it had this kind of beautiful chart to look at that updates on a monthly basis based on, like, how much you brought in, how much you spent, et cetera. That I. I thought it was a great chart that I made. But there's so much more to building a personal finance app than that. So I just knew I was really outgunned that there. There's like, dozens of personal finance apps out there. And so I moved on from there, and then that's when I landed on pbq.

Guy Swan (20:28)

Yeah. How do you use AI? A lot of people think that I've just stopped AI Unchained, which. Which I did. I. I stopped the show specifically, but I Still, like, I'm. I'm probably deeper into AI than I was during the show and like, how. And how much I use it. And. And so I like, I still cover a ton of AI stuff or intend to cover a ton of AI stuff on the show, but I don't get to it a ton. But this is the reason. The reason I decided after we had talked a good bit back and forth is literally just because I'm. I'm using your service quite a bit. And I believe I just. We just canceled Mid Journey as well, because I've had Mid Journey for my designer. April does all of my. But you can find her on Noster. Actually, Actually, she has a bunch of cool stuff. I tell her she needs to post more up there, but she's just like, I don't know, it's crazy. All you bitcoiners, you're not going to like my stuff. You're like, don't be an idiot.

Matt Alberg (21:34)

Just post the stuff.

Guy Swan (21:36)

I'm going to clip this and send it to her. Give her shit for not posting on Noster. Um, but she has a bunch of cool stuff. But I gave her a Mid Journey subscription just because she. She loves to generate stuff and she composites it all together. Um, but I was like, why don't you check out ppq? Because there's like a ton of different image models and you can just like explore and like, see which one works. And then it was like the next day, in the middle of the day, I got a random text. She was like, oh, my God, Nano Banana is the boss. And. And so it's like. And like, literally it was just because, like, I would never. I'm not gonna go get a subscription to Nano Banana, you know, but it was literally just because we could just casually explore all of them and test out the same prompt on all the different models and see how they behave. I don't know. It was just. It was really cool. And I've already canceled a couple of subscriptions because I. I literally had like six AI subscriptions. For the number of things that I use, the only one that I'll be keeping for sure is Claude code, because I use the ever loving. I mean, just explicatives about how much I use that. I sit down at breakfast and I'm building. I'm vibe coding some shit.

Matt Alberg (22:53)

So.

Guy Swan (22:55)

But it's a pretty. It's a pretty dope service. Give me your AI Journey. Maybe that's the way to frame this question is like, how'd you get an AI and how do you use it?

Matt Alberg (23:04)

Yeah. So. Well, first, I didn't know you had an AI podcast at all, or at least used to have one. So it was not a way of. I did.

Guy Swan (23:10)

It ran for about a year and some change before I was just drowning in work. Cause I had too many shows.

Matt Alberg (23:17)

I see. Yeah. I mean, now is the golden age of AI stuff. So I don't know, maybe try to fire it back up.

Guy Swan (23:24)

I'm just putting it in the bitcoin audible feed. I'm just doing AI stuff here. I think I want to use the logo. A bunch of people tell me I'm a producer. Like, don't, don't, don't put the show back in. Just make it all bitcoin audible. I'm like, I don't know, we made such great logos. We had such cool artwork for it.

Matt Alberg (23:39)

I want to use it. But anyway, I mean, I think bitcoiners have, like, you're going to find a great audience in bitcoiners for this AI stuff because they're all very kind of bleeding edge type people. AI touches on a lot of the dystopian issues that our world has. So it's all stuff that bitcoiners want to talk about and learn about. But yeah, my journey in AI, I'm not like a super AI expert. I came onto it as like a user. When ChatGPT 3.5 launched, I think that was in like early 23. I just used it and I was like, okay, I can see how this is going to be incredibly useful. And then when GPT4 Turbo came out, I think that was kind of like the late spring, summer of 23, I could just see how that was already helping me. It was helping me build out my little remittance app. It was helping me build out my little personal finance app. And so I could see, like. And I'm not a, I'm not a coder by trade. I got into coding in my late 20s, early 30s. And so for somebody like me who does not have that, like, you know, 15 years of development or experience, it was a huge thing for me. Me, PBQ would not have happened without the help of AI, that's for sure. And so I would say I'm just like a power user of AI like you. I like to try out the different things. I like to. I have all sorts of different use cases I use it for. I have all different interfaces that I'm using it in. I. I'm using it in the PPQ ui. I'm using it inside of Tools similar to Claude code. I'm plugging the PPQ key into different services and in some cases PPQ still isn't able to offer its services elsewhere. So like, I'm still dabbling around with non PPQ stuff all the time, just kind of getting a sense of where all of this stuff is going. So yeah, I would say I'm the perfect power user. I'm the perfect kind of communicator of the cool stuff you can do with AI to normal folks who are just kind of wrapping their heads around it. Yeah.

Guy Swan (25:51)

I got a question. Do you have any background in video or do you. Do you do any memes or anything like that? Have you explored that whole side of things at all?

Matt Alberg (25:59)

No. Other than just like memeing some images using ppq, that. That's like the first thing I ever got into that stuff. Yeah.

Guy Swan (26:07)

So it was really the reason I dropped both. I had some video generation subscriptions or whatever and obviously video is like the big thing. Like it costs just a fortune and I try to do as much as I can locally, but things like Sora or Sora 2 and like these other models are just. They're really good. If you're trying to get a hard, a difficult shot or you're trying to be a director in some way or get the. Really. It's the like body continuity. Like the, the video they showed of the, the. The ice skater. I was like, oh my God. They really did like, they changed the whole nature of the attention layer, whatever they did on that. Because I know there's an AI could not do that crap at all like three months ago, four months ago, or whatever. Like you could not have. There was no ice skater spinning on the ice that looked even slightly realistic. And. But like I've been working on. And this has been a very long, like slow project, but I've been working on a couple of. If they made movies out of, out of these like, like fake trailers. Because all the AI fake trailers are just dog.

Matt Alberg (27:23)

They're just outer garbage. They're all limited by like 8 or 10 seconds. So you have to then like strip, you know, stitch them all together and then you lose the, you lose the tone. Yeah.

Guy Swan (27:33)

It is a massive editing project and you have to carry your character across and like the, the ones that were fun for like a second and then you watch like three of them and it's like not interesting anymore. It's just like somebody with their, their face is like right there. That, that obvious AI generated image that's always the exact same and they're. It's just like, you know, Wes Anderson makes Star wars or whatnot. But I've been trying to do a much more interesting and advanced project, but it's. It's slow going. And I did Runway and I generated three. So I had a yearly subscription. It was like 150 bucks, right. And I went back and they had changed all their defaults or whatever, and now they have like an LLM that's like making your prompt more friendly to, like, their model and stuff. And so I just went in, like.

Matt Alberg (28:24)

With the defaults, they're probably reducing the likelihood that your prompt may be censored. And then they're also.

Guy Swan (28:31)

Yeah.

Matt Alberg (28:32)

Accentuating whatever prompts you're giving it. A lot of people put in, like, super basic prompts, and it's good to add as many details as possible. So they're probably doing both of those things. And that's really the special sauce. A lot of these video generation subscriptions.

Guy Swan (28:46)

Which seemed great. Which seemed great, except that you only get for $150 a year. You get like 600 credits a month, and you don't get to keep those credits. So if I don't use it for a month, which I hadn't used it in like three months, I don't build up like 1800 credits to work with. I still have 600 credits.

Matt Alberg (29:10)

That's the really backwards thing about it is, like, that's not how we use products, right?

Guy Swan (29:16)

Anything. I don't use anything.

Matt Alberg (29:17)

That way we go through these little spurts of intense usage where we're trying to do a thing, we have a curiosity we're trying to scratch or whatever. And so we'll super use something, and then, like, we'll get distracted for a couple months and come back. So, yeah, I mean, this, the whole subscription thing and the way they're. They're doing it is also something that we're trying to address with pbq. Just everything, flat fee. It is challenging, though, to deal with. We've all been conditioned on subscriptions so hard that I have many users who are confused about, like, wait, how does this work? You know, how much do I have to pay for unlimited or whatever? And I have to tell people, like, this happened last night. You know, we don't have an unlimited. Like, there really is no unlimited anymore. It's all just kind of. Even when you're paying a subscription, there's some hidden.

Guy Swan (30:07)

There's late rate limits. You just. You just get to a point, because you could actually just use these things, you know, Indefinitely. You could build a little AI thing to just keep churning stuff out. In fact, that's like, that's one of the things that I'll do every once in a while. Um, and I do it for two different reasons. One is because it's great for ideation, like, especially for like a meme or, you know, just screwing around with like trying to generate video is. I'll just get it to automatically generate with like wildcard prompt and it'll just run for like an hour or two and just generate 200 of them or something. But then the second reason is because it warms up my room and I don't have my miners. I don't have my miners hooked up. So everyone's around you, like, should I like, so let's work on like this shot and just see if I can get this one to work. I'm going to generate a bunch of crap and put my feet up on top of my gpu.

Matt Alberg (31:02)

Yeah. I mean, that's also when, when you're using your. When you're doing that stuff locally, that's fine. But that's also another problem with the subscriptions is like, exactly. They have super users who are like using the crap out of it. So they have to, they have to build these things so defensively now. Like, they, you know, we should give this person enough usage that they don't get rate limited immediately. But then we cannot let them like burning through so much money is. Then we're going to lose money. And so there's. They like have entire teams that are devoted to like, you know, controlling the subscription, controlling the rate, limiting, you know, keeping customers like, we don't have, we don't thankfully don't have to deal with any of that stuff on the PC PPQ side. It's just simply, you know, pay what? Pay the price, you know, and go forward. And like, we don't have to like manipulate or like psychologically kind of trap our customers into like holding their subscription and that type of thing. Yeah.

Guy Swan (32:00)

Do you have any terrible subscription experiences?

Matt Alberg (32:04)

I mean, almost all of them end up that way. Just like, you just forget that you have it, right. And then it just keeps chipping away at you every month. Like, I remember the one that really, really just makes me so angry. But I've also just like absolutely declared defeat on it is. I remember I kept getting this stupid pop up on my MacBook for like a year about you need more storage or something like that. And so I was like, oh, finally, like, how do I turn this thing off? I Kept researching it. They make it really hard to turn that stupid thing off.

Guy Swan (32:38)

Yeah.

Matt Alberg (32:38)

And so eventually, I was like, fuck it. I'm just going to pay you $4 a month. You win. And then it's so I don't get this stupid pop up anymore. And they've gotten, like, three years of subscription off me now just because I hated that notification. And it's like, yeah, I honestly, I feel kind of defeated, but it's like, it was the one thing that I was just like, okay, I've had it. I don't want to deal with these stupid things anymore. Take $4 a month. But I hate that company. I hate Apple now for it. It's like I have this, like, anger in me because it's like, they're just blackmailing you, you know?

Guy Swan (33:11)

It is excruciating. It's so crazy how much. Basically, they incentivize or create fake incentives to pay for stuff by annoying you.

Matt Alberg (33:25)

Yeah.

Guy Swan (33:26)

Like, and there is nothing. So I have, like, part of me, like, I get really annoyed by stuff like that. And, you know, part of me wants to just pay the $4 to make it go away, but there is this demon inside of me that is just like, you. You are mess. You are. You were getting in my way, and I would rather burn my computer than pay you for it. You know, like, there's. There's these two. There's these two wolves, right. That they're telling me to make the stupid notification go away. And then the other one is, like, burn Apple to the ground for sending me a notification I can't cut off on my own computer. Drives me crazy.

Matt Alberg (34:06)

Yeah. That's one of the times where I just totally capitulated. And I'm not. Not proud about it, but I'm not proud of it.

Guy Swan (34:12)

I'm admitting it on the show.

Matt Alberg (34:14)

Yeah, Listen.

Guy Swan (34:15)

Okay. This is my low moment. Oh, God. Yeah, I. Dude, subscriptions drive me crazy. Runway, I sent. You know, I'm usually really nice to customer support. I've done customer support as a job. So, like, I know they're. I know the framing of people who get the stuff or talk to the people. And so I'm like. I'm usually not mean, you know? Yeah. Usually being nice gets you leaps and bounds ahead of everybody else, because everybody. You always just deal with mean people. And. But I sent a not happy email to Runway, and this is in it. Some sort of happenstance or, you know, just divine intervention that I had just. We had just talked about PPQ AI, and I hadn't even really Used it. We just talked about the thing of doing my voice or whatever just because it'd be fun. And I went and checked it out for a minute, but you know, just to have the time, I'm busy and, and I went back, I was generating. So one of the things that I use Runway or just use the video generation models for in general is, you know, I make the little short form video content, like I do the two sats or whatever and I'll post stuff on socials. Well, one of my favorite things to do when I can't find good found footage is I'll take a, a good image of something. Like there are a couple of images like of. I was looking for a drone shot of a solar farm that was mining bitcoin and I found a bunch of pictures of solar farms mining bitcoin because, you know, somebody just goes and take pictures somewhere. But I could not find a good video. And the only like rough video shot that I did was like something behind like a four dollar paywall on a stock footage thing. And I'm like, seriously, I just want to, I just want to shot up some, you know, from solar panels. And so I took a picture. I mean, I took a shot of one of the solar panel farms that was actually mining bitcoin and I put it in Runway and I said, make me a drone shot, you know, and it looked great, it looked fantastic. But when I went back to do that, I was just trying to get like a picture of like a clock. I mean a video of a clock in time lapse, like, like really quick. And just like I said before, I was doing all their defaults and I just had no idea how many credits it was costing. They didn't like give me up front. But I found out that the LLM that was like fixing my prompt was like 50 credits. And then like the first some piece of something else that they were doing before generating was like another 20 credits. And then the generation itself was like 150 credits because it was like the default biggest. Like the default was the biggest model and it was for 10 seconds of video. And I didn't change any settings. I didn't do anything. I'm just looking to generate a couple of things. And then I generated three videos of a clock. Yeah. And all of them were kind of crappy. It wasn't at all what I was looking for.

Matt Alberg (37:19)

I described it bad and your month's.

Guy Swan (37:21)

Gone and my month's gone. Yeah. And I was like. And I'd been up there for four months and I was like, what the. What just happened?

Matt Alberg (37:29)

I didn't.

Guy Swan (37:29)

I didn't even realize how many credits I was using until they were gone. And I was just like, this is. I immediately canceled. And then that was when it hit me, because we had just talked a couple of days earlier. I was like, didn't PPQ have, like, some video stuff? I'm pretty sure PBQ had some video stuff. And so I went over there and I put $5 in or something, you know, just with lightning and generated stuff on, like, every model, ran through, like, a bunch of stuff. Finally got, like, a shot that I really liked, and I still had, like, a buck 80 left over. I spent $150 on Runway. Yeah, $150. You know, and I was just like, this is so stupid. It's not. Ah. The subscription model doesn't fit so many things, especially in a world where we have so many tools now. I feel like there was this era where it sort of made sense when we have broader tools that were useful for a lot of things, but they're getting so specific and so numerous that, like, even my workflow just includes, like, 40. You know, like, I just. Pulling from so many different places, and it just doesn't make sense. And like I said, the reason I. I was like, we should do a show about this is just because I've. I don't even know how much I've already saved. I've at least saved a year of Runway. I've at least. And I haven't. I've spent 13 bucks. I don't. I don't know. Like, I just. I have not spent that much on ppq, but it has saved me from needing subscriptions to at least three other services already. And I've canceled three. Four. I've canceled four specifically. Don't have chat GPT anymore. Don't have Mid Journey, don't have Runway. And what was the fourth one? I don't know. There's another one out there somewhere. I can't remember off the top of my head.

Matt Alberg (39:26)

Let me get this straight. You canceled all of these because of when? Because.

Guy Swan (39:30)

Because I could just use ppq? Yeah. Wow. Um. And I usually don't log into my stuff on my Linux machine. And so I just have a. I just have a browser page for all my troubleshooting stuff. I just have a browser page with PPQ open to Chat GPT just for help me troubleshoot crap. Cause I'm always just doing custom stuff on Linux.

Matt Alberg (39:48)

So my question is, I mean, you're telling Me, what we love to hear. My question is though, is, you know, we're trying to do everything all at once. We're trying to do the text generation, image videos and each one of these services. And I kind of discussed this a little bit with you via DMs as you were kind of playing with the website, is we cannot master any one particular modality in the way that a company can master that on their own when it's the only thing they do. So, for example, Runway, they're only doing video generation, to my knowledge. They don't have anything else to worry about. And so basically what we can do at pbq, I think, is we can bring all of these models in one place and we can make the experience 90 to 95% as good, I would say, as the solo subscription. And I think in some cases that's. That works perfectly well for people.

Guy Swan (40:50)

But that's enough for probably 90% of the people, though.

Matt Alberg (40:53)

That's what I'm hoping is true. And if that is true, then P Q is going to go a long way. Maybe it's just a matter of getting the word out a little bit more, making people understand that this is possible. So I will say with the video generation, we do have users come and they kind of give a similar complaint to us where they will say, like, I deposited five bucks and I only generated like, you know, three. If they choose the really expensive video model, you can burn through that. Like in a.

Guy Swan (41:20)

You can burn through five bucks pretty quick. Yeah, yeah, like I said, I spent like 10 or something like that.

Matt Alberg (41:26)

So we do have that problem a little bit where users in some cases don't know how much they're paying. We've gotten better at that. We're like showing the prices a little more, a little more out there now. Also, we don't want to just flash the price on the screen every time they're clicking something either, because that kind of detracts from the experience as well. So it's definitely a balance. But yeah, it's. It is. That is the success or failure of PBQ rests on many things, but one of the big ones is does that 90 to 95% quality experience, is that good enough for most people to like, just drop the subscription and come to us? And for you, that is very true and that's awesome. And I hope that we can get that. I hope that we can have other people say that as well. I think the really, the sweet spot is when it's some sort of small business that is trying to they have like a two, a small team that's trying to do everything all at once. And that sounds like that's the case with what you're doing. When it becomes like a bigger corporation, maybe they're willing to pay the subscription for that extra 5% of performance or whatever. Or maybe they. Yeah, but I think that we are really capturing a lot of individual users and also small business type users or like, you know, just people who are hustling in their content marketing job or stuff like that.

Guy Swan (42:53)

Yeah, yeah. Before I give my take on that real quick, I don't think we've really dug into exactly what the PPQ AI model is like in depth. So for, for the audience, what, what did you build with PPQ AI Exactly.

Matt Alberg (43:10)

So PPQ is basically a place where it's, you can come. It's very similar to a ChatGPT experience, except there's no subscriptions. It's all on a pay per use model. And instead of giving you access to models from just one company, we're giving you access to all of the models from all of the best companies. And this is actually really key now because every other month a new company has a new model that takes the lead. And there's kind of this thing that a lot of people are doing where they're like, oh, I'm stuck over on this subscription. So I cannot, I can't try this hot new model that's called Nano Banana because it's not under my subscription. So we solved that problem and again, we've chosen to go with a pay per use model that is very well enabled by crypto payments, Lightning, Bitcoin, stablecoins, whatever, because the fees are very low on these payment methods. Generally speaking, when you are using a credit card, the fees on the credit card are too high to justify little 25 cent 50 cent transactions. And that's why. So a lot of people think that subscriptions kind of came around because these marketers or these biz dev analysts at all these big companies, they said like, okay, what's, what's a way that we can kind of like squeeze more money out of our users? Oh, I know, let's, let's make this subscription thing so we can take money from them even when they don't use our service. I think that's partly true, but I think we arrived at subscriptions more so because the credit card companies do not allow micropayments basically. And so you have to create some sort of subscription model to basically accept these payment methods. And so we're different in that way. Our primary payment methods are crypto. Lightning is actually about 50% of all deposits on our platform.

Guy Swan (45:21)

Wow, that's awesome.

Matt Alberg (45:23)

Yeah, and, well, I mean that's where our community is. We are actually very popular on Noster. We've gotten more customers from Noster than from Twitter, I think. But yeah, the pay per use, the extremely small fees, the final settlement of Lightning is huge. And so I think that if we were to fast forward 10 years, I think that subscriptions are going to be used less in 10 years than they are now. I think that AI is one of the first kind of use cases that is going to start chipping away at the subscription model. I think that subscriptions like Spotify and Netflix and stuff, those will stick around for a long time. But a lot of the kind of subscriptions that, where people are basically just consuming something and they're juggling all their subscriptions and stuff, that's not gonna, that's not gonna last too much longer. And companies like PPQ that are offering it all in one place on a, on a per use model are gonna win.

Guy Swan (46:23)

Yeah, you know, I think basically there's, there's also this huge push or kind of like a cultural change of like, man, I don't own anything anymore. And the realization of that, of just basically not having access to things without, without that subscription and basically the, the kind of siloed nature of a lot of stuff. And I think, I think that's coming back where people just want to buy software. At least it is for me. I just want to buy your freaking software. I don't want your subscription. You know, and we've gone from this place where, you know, there were a few things that made sense for subscriptions and then the overwhelming majority of you purchased stuff or, or you, you purchase a service explicitly to how much you use it. And then we, there was this flood online. And I think you're right. I think it was like kind of a, a one, two punch as to why we ended up with subscriptions everywhere. And it was because the payment processing and reverse transaction reversal and everything, basically all the increasing costs of that, of the credit card system just did not align at all with microtransactions or just even many one off transactions. It just was a bad, like technically it was just did not align with anything. And then when it became obvious that subscriptions were the best way to kind of ease the pain of the consumer in doing this, it's also a huge opportunity. Well, we can actually milk them for, you know, if they Forget about it. We're just still pulling money from their account. And that's because it's a pulse system, it's not a push system payment. So they're taking advantage. They got, they were getting screwed on the front end by how the payment system works and now the user's getting screwed on the back end by how the payment system works. And. But I think we will be headed to kind of a reverse. Like I feel like it's like 80% of shit is subscription now and like there's 20% where like maybe I actually own it or I'm actually paying for what I need. But I think it's literally just the breadth, the inability of competing in the space and the, the amount of consolidation that's going on, the large corporations and stuff that's kind of trapping people into subscriptions. I, I really think, and I don't know how the pressure builds or exactly what the timeline on this was, but I will be, But I really kind of think we'll go, the reverse is we'll go back to 80% the stuff that makes sense for actually purchasing or purchasing directly compute or directly hosting like when you need it, and 20% subscriptions because it makes sense on sunk cost products like Netflix catalog, you know.

Matt Alberg (49:10)

So even if you look at it, there's another way that like even the clustering of these companies where they offer suites of products, right? That in my opinion is a big contributor to that is the fact that we have subscriptions. So I see it on the PPQ side where I see it in a lot of AI tools around, around the Internet these days, they all have the same flow. You go to their website, you have to make an account, then they try, they give you some small trial of the product and then they try to onboard you to a subscription, right? That's the same flow everywhere. And if I'm an AI creator, small time AI creator and I have this one tool that does a good job, but it's something that a user only needs maybe once a month, maybe once every couple months, right. I now have something that's not monetizable because I don't have something that's worthy of a subscription. Right? Let's give an example. Right? There's a song generator out there called Suno AI. You go in there, you type a text prompt and then it makes a song for you. It's really fun. You can make songs about friends and family members and it'll like put their names in there and just stuff like that. So it's, it's really fun, but am I going to pay a $5 a month subscription for this song generator? No, I'm just going to use it a few times. Maybe at first I'll use it like 10 times and then like three months later I'll remember that I could do that and I'll send another few songs out. But it's not worthy of subscription. Right. So what this leads to is that these companies out there like Suno, they end up being consolidated into bigger product suites. And so Suno, in my opinion, if they can't figure out how to make it work selling all these $5 subscriptions, they're either a going to try to cater to larger creators that, that actually are generating thousands of songs a month or whatever. Right. So either that's their market share or they're going to sell Suno to OpenAI and then OpenAI is going to let you make songs inside of chatgpt as a kind of a bundled product. Right. So subscriptions are also kind of, or credit cards. Payment systems in general are also kind of unfair to a lot of the creators where they just can't. They need to make something that justifies a subscription and that takes a lot of. But with PPQ eventually, and I've seen this with several companies now, almost every image generating company used to be subscription only, but they're all capitulating slowly where they say, okay, we cannot, our product is not good enough to get five or seven dollars a month from people. So we're going to open up API access to companies like PPQ that can then monetize for them. And most of them have, most of the image generators now are on API only. The only holdout is mid journey. But I think their days are numbered.

Guy Swan (52:12)

I canceled my subscription, so I'm doing my part.

Matt Alberg (52:15)

Yes. And here's another angle where it's really not going to work for them because all of the super casual users of these products are going to cancel and then the remaining customers are going to be the super users who cost them more money more than the subscription. Yeah, exactly. So all the, all the casuals are coming to PPQ and then only the customers they don't like, the ones that are using up tons of AI inference are the, are going to remain. And so yeah, it's, it's, I already see it happening and it's only going to happen more and more as people realize that there's these like pay per use options out there.

Guy Swan (52:57)

Yeah, yeah. You know, have you talked to the Cascader guys?

Matt Alberg (53:03)

You know what I'M talking about.

Guy Swan (53:04)

Yeah. Okay.

Matt Alberg (53:06)

So I talked to them some time ago and this is actually really interesting study on this because they, they are perfect. They made a little tool that's pretty cool. It's like, makes podcast clips for you and it also. You can search podcasts for different things and stuff like that. Right. I forget what it's called. It's like, pull that up, Jamie.

Guy Swan (53:27)

Right?

Matt Alberg (53:27)

That's what that's called.

Guy Swan (53:28)

Yeah, yeah, yeah.

Matt Alberg (53:29)

Jamie is the AI tool. So they don't have a product that's fully featured enough to warrant a subscription and yet. And they also, they're too small to kind of get people. They don't have the audience to like generate enough demand on their own. Right. Not many, not tons of people. They don't. Not everybody knows about Cascader. Right. So they're kind of stuck between a rock and a hard place. So I did approach them and I said, like, hey, you know, ppq, like, we're also tiny, but we're at least probably a little. We have a little bit more of an audience than you do. So just bring Jamie over to PPQ as a tool inside of the PPQ suite. Right. And then anytime one of our users uses it, we'll send you a lightning payment. And they were open to this. I think, though, that we're not quite there yet. PVQ is not big enough. We don't have a big enough audience yet to really justify them building something inside of our platform.

Guy Swan (54:25)

Taking a lot of investment to put inside the platform.

Matt Alberg (54:27)

Yeah, yeah. Especially on the UI side, they could create an endpoint that we could just send payments to, and I think they even offered that. Um, but then we have to build the UI out and I actually think that their Jamie UI is actually most of the special sauce of what's going on. It's just a really slick ui. The buttons are all in the right place. They've thought about it. They put some design thought into it. That's where a lot of the value is. And so on PPQ side, we don't want to spend time like making duplicating this beautiful ui. I would just rather have them somehow bring it in. Long story short, we haven't found a moment to work together yet. But I do think the bigger PPQ gets, the more draw, the more sense it makes for outside tool creator to bring something in. And then also, if PPQ can get good at distributing outsider tools beyond ppq, that's a whole other avenue of growth, potentially orders of magnitude higher. BBQ is still, you know, we're doing fine, we're growing well, but we're still a tiny little drop compared to a lot of these other platforms. If we can find a way to monetize an outsider's tool, bring the tool to PPQ and then bring it to other websites on behalf of that creator, then we can generate enough demand to make it worth their while.

Guy Swan (55:53)

Yeah. Yeah. I'll tell you the way I think about going back to the point before I kind of took a tangent a little bit on like how I use PPQ and why I think actually marries really well with, with the subscription model. Not, not PPQ specifically. Even though actually might, if, if there was, you know, a margin that you could cut for, for higher, higher use users, you know, to, to do something. I probably, I might actually do a subscription for that just because I can use so many different tools. But I think it's in how I use it that makes it work because there's a reason like Claude code. I still do like I do the hundred dollars a month subscription for them because I, I just burn, I burn through stuff.

Matt Alberg (56:46)

Does it tell you how much money you, you used? How much AI you consume?

Guy Swan (56:51)

I would really like to know. I would like to know how close to my limits I am and maybe there is a place that I can go find it, but they don't make it easy to see. Okay.

Matt Alberg (57:01)

I had, I had somebody else tell me they're also an avid cloud code fan and they're. I think they're on the $200 plan. But I think he told me like an exact number. He's like, I've spent $1700 worth of stuff last month on my $200 cloud code code. And so in that case, like.

Guy Swan (57:17)

So it must be out there.

Matt Alberg (57:18)

Yeah, yeah, yeah, it's out there somewhere. And yeah, I think that, I think that a lot of these subscriptions are heavily subsidizing things as well right now. So I'm not sure that's sustainable to give away $1,700 to a two hour. I'd be curious though what your numbers were if they could be found. That's kind of one of my projects for this week is I want to actually use cloud code myself and see what's in there, see how much it's spending and all that stuff.

Guy Swan (57:49)

Yeah, but the reason, the reason when I do it Claude code is generally, which is not true on my Linux, but it is more true on the kind of work that I do on my Mac because it's my main driver is that I'm Constantly referencing things we've already done. You know, I'm referencing project files that I've uploaded to Claude. Um, and so it, it basically requires an account with a specific history and like my chats and like that sort of thing and the way my, my sister in law uses like ChatGPT. It's also very specific to being able to relate back to conversations and that, that large context and that like fine tuning to know who you're talking to. That's really important for some use cases. But I have dozens and dozens of little one off things that I do that I can basically interchange models sometimes, sometimes I interchange models just to get variety, but I'm basically doing the exact same thing over and over again or I'm doing it in a very specific context that's isolated to this one episode or to this one task and it doesn't really matter when or how I get it. Um, and that is what something the model for PPQ is so perfect at is these kind of like single use one off things. That's why I thought of Cascader while we were talking or whatever. Just because they, they're kind of like a little aggregator of like a little single purpose thing. Right? Like, you know, bring it up Jamie, dig through a podcast episode or a podcast feed and find a place where we talked about this, um, and, or make clips or whatever. And then in addition, the way I think about PPQ that I think it's a great tool to justify subscriptions to something, you know, like if we end up just using Nano Banana like crazy because like that's the one that ends up with the best output. It's a great explorer, you know, and if it comes to the point where like I'm spending $400 a month on Nano Banana, well, let's go take the subsidized route for a little bit until, until they find out they have to jack up their prices and, and then I can go back to PPQ to explore the next tool or whatever. It's, it's a great place to like jump off. Like if you're curious about an AI thing, go use it for a while, you know, go see if it can fit into your workflow. And that's literally what, what I've been doing. And there's a lot of them. It's like Photoshop, you know, like Photoshop is such a dumb thing for the subscription model. I don't, I'd keep pirating Photoshop every time. Every time I could find a new version of Photoshop, I just pirate it. Or I've switched to trying to just make gimp work even though it's kind of clunky. Or critter because I need photoshop once every three months. I am not paying you $30 a month for your, your, your software, man. Like I'm not paying a subscription for that and it doesn't make any sense, but I would love to be able to just go pay $2, use Photoshop for the next five hours on a project and then be done with it and then come back to it later. And like that's like you're paying for compute, you're paying for compute and an API, a window to basically explore any and all AI tools. And y' all got a crap ton of them. I still haven't even gone through all of them. I just kind of use the main ones that I'm familiar with on the platform. I'm curious, what have you found in kind of the stuff that you put together? Like do you, do you use PPQ a lot? And like what, what have you found in plugging all these things in that you were like, oh crap, I didn't know about this tool or I could use this one this way that you hadn't thought about. Like how have you thought about morphing your tool for the stuff you use?

Matt Alberg (61:50)

Yeah, I mean I, I'm probably the number one user of ppq. I use it in many different ways. Yeah, yeah, I will still find myself and, and this is something that I hate but it's true is like for a super duper hard question, I still have to go over to ChatGPT. I still have a subscription there because I keep, I like to keep an eye on what they're doing and building. But occasionally if I have a super hard question that really requires up to date information, I will go over to ChatGPT because it's thinking model interweaved with its web search is still superior to what we offer on ppq. And at the end of the day like it, it can think about a pro. I'll even trigger the extended thinking. It can think about a problem that I'm having. And it, the quality that is output is still just a little bit higher than what we have on PPQ. But for 99% of other stuff I'm, I'm on PPQ all the time. I use the API key that, so when you're a user on ppq you can come in and you can use our ui, but then you can also use a little API key that we issue this API key is what's called OpenAI compatible. You can plug it into thousands of different third party apps now that are also called OpenAI compatible and use the key in there. So I'm using it in IDEs extensions such as Root Code or Klein or Goose, things like that. They're very similar to how Claude Code works.

Guy Swan (63:28)

Okay.

Matt Alberg (63:30)

That's probably where my most dollars are spent on, on my PPQ is just using those tools. But yeah, we also use PBQ for the content marketing side. Like all of the images that we generate for socials and stuff are generated on pbq. It's great to be at the nexus of this because we have customers that are educating us on the latest and greatest thing that came out. So one great example there is we had a user say that they downloaded this product called Whisper Flow. It's basically a voice. So I find myself typing long prompts both into Klein where I'm vibe coding these long prompts. And so Whisper Flow is basically a hotkey that you press on your computer and then you just speak into the mic and it just throws whatever you've said very quickly into the input box of whatever program you're in. And so I use Whisper Flow and Vibe coding like all the time now. And it made me look into this idea like, hey, can PPQ offer some sort of voice product as well? Because at the end of the day, like I love Whisper Flow, but it's still another subscription in the subscription toolbox. And yeah, yeah, can we bring something that is like 95% as good as whisper flow to BBQ users? And so yeah, we are actually actively working on a voice tool as well. The thing with this voice thing is that five years ago we all had dictation in many of our apps, right? The problem was is that it kind of blew most of the time. It would get words wrong, it would add punctuation in spots that you didn't want it. It would not understand a lot of proper nouns that you're using. Like, I don't know if you say the word Jonathan or something, like sometimes it would mess up that proper noun. The big innovation that happened in of a lot last year with voice is that instead of just doing a dictation model alone, they're now sending, they're now sending stuff in. After it goes through the dictation model, it goes through an LLM to get cleaned up. And so what that does is it makes dictation much, much better for most users now. And so I use Whisper Flow inside of all my apps now. WhatsApp Vs code. I use it when I'm in Slack, sending messages to people, all that stuff. I got an email from them and I used it in like 25 different apps on my computer in the last month.

Guy Swan (66:09)

Dude, I just type. I type so much and it's so slow. I'm not in the world.

Matt Alberg (66:15)

Wait until we have PPQ Voice roll out which will be in early to mid January. And yeah, we're hoping that it's like going to be a big hit.

Guy Swan (66:25)

What size model are they using? Do you know? For like the main, the main waste? Because I have the. I think it's a 5 billion. It's really a pretty big model. Like I definitely wouldn't do it on, not my Mac or my Mac or my Linux, I wouldn't do it on a phone, you know. But there's a really small, there's like a 1 billion parameter model if I'm not mistaken for Whisper.

Matt Alberg (66:48)

There probably is. So the big difference here is Whisper Flow. You're not running it locally. Whatever dictation you're running locally in general. And this is kind of, you can do voice and run it locally, but you need a beefed up computer for it to do a solid job and also to be very snappy to get something back quickly, which is a big.

Guy Swan (67:08)

Key, which is kind of the point.

Matt Alberg (67:10)

Whisper Flow and PPQ Voice are all going to be done in the cloud. So you're sending your voice off to the cloud for way faster processing and you're running it on these super fast AI inference providers. There are special companies that specialize in handling voice extremely fast and snappy. So companies like Groq with a Q, Gr O Q is a super fast inference provider. There's another one called Cerebras AI. So Whisper Flow and PPQ Voice will be using these cloud providers to really increase the snappiness. You do lose a little bit on the privacy side because you have to trust that these cloud providers are not sifting through your voice data and stuff like that. But that's a trade off you have to make. But I much prefer the snappiness. I just won't use a local whisper period because it's just too slow. It's too slow for me and I even have a pretty good computer.

Guy Swan (68:07)

Have you used Whisper cpp, the, the, the C version of it? Maybe.

Matt Alberg (68:13)

I've heard there's a Whisper Turbo. Maybe that's what it is. Dude, dude.

Guy Swan (68:18)

So somebody built a C version instead of the Python or whatever and I shit you not, it is 20 times faster. It is so, so, so, so much faster. I can do like when I first downloaded Whisper on my Mac, um, and my Mac is not to be trifled with. Like I, I beefed it, I maxed it out essentially when I got in like three years ago and using the normal Whisper, I tried to Transcribe like a 10 minute video or something like that with a little script that I vibe coded. And I think it took, I think it took like 10 minutes. Like it literally was like one to one on itself. Like it took forever. I thought it was broken, I thought.

Matt Alberg (69:03)

It was stuck, you know.

Guy Swan (69:04)

And then I downloaded the c version. Whisper CPP is the, the thing on GitHub if I'm not mistaken. And I think it took like 20 seconds for 10 minutes worth of audio and, and I built that into my script and now I just drag and drop, I have a little icon on my dock that's just drag and drop to run the script on any audio or video file. And it pulls out the audio with FFMPEG and then pumps the output into Whisper and then produces a subtitle on a transcription file. And it takes hour long episode, hour long episode. I just do it while I'm like kind of getting my description and stuff together and I come back, minute or two it's done, it's, it's really good. Um, and I think a lot of this is about like the abstraction. Like it's so, so high in the layers that everything's super bloated and Python's like just clunky but lower level, it seems to run way, way, way better. Um, so anyway, that's just a thing.

Matt Alberg (70:09)

Yeah, so the use case for a Whisper flow or PPQ voice, it's going to be much shorter snippets like 5 or 30 seconds in general. So it is a bit of a different use case. And maintaining snappiness at that level is a different ball game than maintaining, than getting a fast output from an hour long file. For sure there's different things you have to do for each use case, but in general what we found is that doing it locally does not cut the mustard for most people. And so the fast AI on the cloud with a specialized provider is the way to go for kind of the shorter, the shorter clips. But yeah, that, that is, I think what you're still though when you're transcribing with your Whisper C, you're still only using the STT model, the speech to text model. I don't believe you're running it through an LLM afterwards to further clean it up. And so that's another thing that these tools do, which actually is really helpful. You know, they'll add capitalization in certain spots, they'll add punctuation, all that stuff. It's actually really handy. And yeah, it's, it's my, I started using Whisper Flow a few months ago and it totally changed how I'm vibe coding because I just don't have the patience to type out three paragraphs every single time. Yeah, yeah, exactly. But I can totally just like, you know, ramble through this explanation and AI is really good at taking a garbage rambled flow of thought and like kind of summarizing it and understanding what you mean. And so I can be even super sloppy with like, hey, so today I'm thinking about making, I'm thinking about refactoring this code base. Like I think it might be doing this thing crappy, but I'm not sure. And, but, and then that thing crappy, but I'm not sure. You just say it in a very informal way and it puts it together nice and tight for you and, and like you're off to the races immediately.

Guy Swan (72:10)

Yeah, yeah, it's crazy too. Like, I, I, I, I could probably benefit a lot from Whisper Flow. Y' all should definitely add it to ppq because I, I haven't really explored that at all. But like yesterday I had, I had like a five hour conversation with Claude just trying to dig into the, the positives and negatives of one of the architecture choices that we're making with Pear Drive and basically how to scale, like, how could you, how could you make this thing scale to a billion files in an index? Because, you know, at that point you're talking about like, if you have like SQLite database, you're talking about 400 gigabytes, 500 gigabytes just for the index, you know, and it's like, okay, so how do you, how do we make sure that when we get to that point, we're properly separating the jobs and the various pieces of the index? Because I also want like rich metadata and tags and like kind of a sort of quote unquote social element to a lot of the things that we're doing? And, and that's just not, that's never going to work if we don't get the architecture right now. And calls will sometimes like go out of its way to just kind of like agree with you. And you, you're absolutely right. You know, it's the first thing it says 80% of the time. I give it anything. And then I have to literally force it to steel man. The opposite. You know, like I have to tell like no, please don't just tell me I'm right. I'm trying to figure something out, you know. And so like I'd be like, well what's, what would be the benefit of doing this? Is there anything that you know of that you've searched online to, to, to mitigate these problems? What's the trade off? What do people bitch about when they're using PostgreSQL as opposed to this? You know. But literally I like four or five hours. Four or five hours.

Matt Alberg (74:10)

Just, just type it.

Guy Swan (74:11)

Just typing and reading and typing and reading and typing and reading to. And then you know, I'm constantly asking it, okay, so we've just done this for two hours. Can you basically summarize our conclusions and the, the, the most reliable out results that or conclusions that we've come to from the previous, you know, whatever conversation. And so without a doubt whisper flow would make a huge difference there. I could probably turn that five hour conversation into it three hour conversation if it was just a little bit faster. Having to sell bitcoin hurts, especially when you are certain it is going to be worth more in the future. But you can actually get access to the Fiat without selling it by using a Bitcoin backed loan. Maybe this is for an emergency, maybe this is for an investment that you think will do really well but, but it won't beat bitcoin because it's monetizing or this is actually something that you know you'll get back but the timing just isn't right and you don't want to let bitcoin go for that span. This is why Leden was built. They let you borrow against your bitcoin quickly and easily and Bitcoin is the ultimate collateral. There is no more perfect thing to use securely in this setup and something that you can verify. But with their proof of reserves that they do, there are no monthly payments. If you don't want you pay it off at your pace. There's no penalties for early payment. There's no finder's fee. It is quick and simple to get your funds. I think their turnaround right now is like 12 hours. And I don't know why every bitcoin company doesn't do this. Everybody who's at least holding bitcoin for other people but they do a proof of reserves so that you can confirm that your balance is there. And something that made me really happy recently is they just cut their Ethereum loans, they cut their yield product, they cut their non custody loans at lower rates. And so they now offer one simple hyper focused thing. Custodied, secure, easy Bitcoin backed loans. Use someone with a good track record who's done over $10 billion in loans available in over a hundred countries. They do proof of reserves and they've made it through the toughest times in the market without having a single problem. There's no credit check, there's no hassle. It's simple. This is why Leden IO exists. Get the value of your Bitcoin without having to sell it. Remember to read up on how the collateral works. They're really good at reaching out and making sure that everything stays balanced. Remember that Bitcoin is volatile so do not overextend if you know how to use it. This could be a huge benefit to your Bitcoin stack and give you optionality in accessing fiat without it being that selling Bitcoin is your only option. I've been a very happy customer. Check them out. The link and details are in the description.

Matt Alberg (77:02)

There are so many people who are missing out on the whole vibe coding thing and I think that like it's really separating a lot of people. Now when I do interviews for ppq, like I would say the, at least this was six months ago, we haven't done interviews in the last six months, I'd say. But I would say the majority of the developers still, they don't get it. They don't get, they weren't getting the power of AI to like solve the problems. Like I had a few of them say oh yeah, I'm using AI. And then I would, I would say like specifically tell me what are you doing? What's your workflow? And they'd be like, well I just, I take code, I copy it from my code base and then I slam it into ChatGPT user interface. It's like, oh, that's like, that's how people were vibe coding like two years ago. Like that's not how like a lot of them hadn't even heard that you could do it inside of the IDE.

Guy Swan (78:00)

In your terminal in VS code. Like, like, yeah, yeah.

Matt Alberg (78:04)

And so I think it's very critical for not just developers but for anybody out, any professional out there. They really have to be leveling up how they're using AI because all the things you're talking about, all the you're having conversation, you're trying not to lead it to a certain conclusion. Like I do all that stuff all the time as well. I purposefully, all of my questions are like purposefully open ended where I'll say like, I think this might be the problem with the bug, but I'm not sure. Or is it. If I just add the. Or is it at the end of the sentence that way, that way it like, it makes it just slightly less possible that it's going to assume my theory is correct and I want it to actually check on everything. Right. So yeah, it's funny, you can really tell when you're talking to somebody who's into the weeds on this stuff versus somebody who's just kind of very casually using it or just dabbling with it. Right?

Guy Swan (78:59)

Yeah.

Matt Alberg (78:59)

So, yeah, I mean, and I feel like I'm so much more productive now.

Guy Swan (79:03)

Oh my God, seriously? Well, I mean I just, I can explore stuff and then like some of the developers that I'm working with now, I found a couple guys who are actually just super great and, and they're also like really into the, the stuff we're building. And I'll basically go through like this like, okay, for the four or five hour conversation I had with Claude yesterday is a perfect example because I then got it to write up a summary of essentially the design of like, how can we scale this? How can we make sure that this continues to work efficiently but then doesn't have that like, oh, it's like Cody, you know, peer to peer, everybody's got different streams and everything sucks and everything's slow. And then you know, like, because that's usually kind of the decentralized or peer to peer interfaces, it's just shit. I was like, how can we make this scale efficiently so that we've got millions and millions of potential users and literally billions of files or different spaces or whatever that people are interacting with? And so I went through this whole thing and like wrote up this design proposal based on the conversation and then sent it to my developer and he was like, this is actually very in line with kind of a lot of the pieces that we've already put in place. And there's actually this other layer that makes it so that we can use this one when its trade off is not necessary. When it's like best in this optimal scenario and this one in the, in the kind of local, the localized like local query scenario. And these can actually be married together really easy with this additional tool. So he just kind of like adds this like nice one little piece to it it that says it's actually not as complicated to use both at the same time. Where they're both best. Because, you know, I've already been thinking about adding this in and it's like, oh, that's perfect. So I found out that it did, actually did a fantastic job. But you're 100% right on the or is it type thing is that I have to constantly say because I usually will explain something as if I'm taking the position. Right. Like I'll say I think it would be best to use SQLite locally and then just syncing the differences between these lists because we don't need like this big network thing where tons of people are connecting. We just need to sync the index so we, I can look at it on my own computer, blah, blah, blah. So I'm trying to convince it basically of my, my route and, and it's going to want to just take that Persona or that, that, that impression and turn it right back around and give me a better excuse as to why. No, yeah, you're definitely right. And so I almost always have to end it with tell me why this would be wrong, in what situations this would be wrong or why the other tool would be better. Give me, give me the reason why it might be right. But then steel man, the reverse as to why I might want to use the other tool. So I always try to get it to tell me, assume I'm wrong and explain why I'm wrong and then it will, it will literally do both and then it will come up with this. I actually tend to think you're right because this, this and this. So it does really well. That's an extremely useful thing that I found. And, and I'm really. That's another thing I like about PPQ is it, is that different models are also good at different things. Like all the LLMs just kind of have their little, their little quirks or the things that they're, they're really good at and they'll give you different answers. They'll give you different. The same ones will give you different answer or different ones will give you different answers to the same question is what I mean.

Matt Alberg (82:53)

We posted that to Socials just the other week where like I, I do this regularly where if I'm going down the rabbit hole on a particular bug or something that I'm trying to figure out and I'm not Getting anywhere with Sonnet 4.

Guy Swan (83:05)

5.

Matt Alberg (83:06)

If I just switch over to like Gemini 2.5 or Gemini 3. Like it. Absolutely. Just changing the model around it looks at it the problem differently and all that stuff. And that's another awesome ppq. Use case where you can literally just change your model out on the fly almost and see how it can kind of tackle the problem differently.

Guy Swan (83:27)

Now, what else do you guys have that I haven't explored? You guys have like an API. He's so good in API key. So how could I actually make use of that? And I asked because I kind of have this dream of, rather than having this in the browser interface, is that like something that can kind of see or I can pull context because I'm constantly using this for other applications, especially AI stuff. So like. So like, I'm. If I'm doing video generation or actually LLMs, I do all sorts of stuff in Comfy ui. Comfy UI is a super broad and insane thing, but it's also unbelievably convoluted. Like, it is so convoluted and it is so difficult to make sense of what node I need or how to build my workflow. It's like a Lego thing, except that instead of like, you know, 10 different major Lego pieces, there's like. It's like the Chinese Alphabet, you know, it's just absolutely bonkers. And. But I found that, like, I can. I can take a workflow file and drop it into PPQ or CLAUDE or, you know, whatever, and be like, I'm getting an error on this and this node isn't working, or this thing isn't properly responding to something. And it will give me an answer because, you know, it just sees it as a JSON and it's just like.

Matt Alberg (84:50)

Oh, well, you're missing this.

Guy Swan (84:51)

You know, like. And so I kind of have this dream of having that somehow in a window of all the stuff that I'm working with so that I can just kind of ask it locally.

Matt Alberg (85:05)

So it's kind of looking at your screen, so to speak, a little bit.

Guy Swan (85:08)

Exactly. I want to. I want to be able to vibe code kind of like Claude in the terminal, you know, like is that it's just kind of in my environment and I'm. I've almost broken down and tried to vibe code this with something like a PPQ API key to just see if I could do it to. To give it more context as to what I'm doing, because I find I use it more often than not. Troubleshooting really convoluted and annoying applications.

Matt Alberg (85:37)

Yeah. So there's this stuff called browser use. I think Claude code probably does it, but certainly the tools that I use, Klein and Roux, also do it where you're working with a code base. Maybe it's a comfy UI thing, I think a comfyUI thing underneath the UI there is a code base there. So you could technically just keep. You could have it run its own browser of the Comfy UI and that way it can look at your screen, so to speak, or see what you are seeing. That stuff is definitely a frontier type of feature. It works in some cases, but I also find it's like pretty buggy in other cases. Most of the time when I'm using Klein it'll say like, okay, now that we've solved the bug, let me open up your browser and make sure that it works by clicking through your page or whatever. And I usually just stop it right there because I would rather just go and open it up locally and click buttons myself. A because it's slow and B because it's still kind of like doesn't exactly use the UI in the browser in the same way that a user would. So I think all that stuff is coming for sure. And you know, even you can see OpenAI released their ChatGPT browser lately, which is like you're just surfing the web and it constantly is looking at your screen, right?

Guy Swan (87:01)

Which is such a nightmare. At the same time all sorts of.

Matt Alberg (87:03)

Privacy things like all super useful, but.

Guy Swan (87:06)

Also hello nsa, how are you doing today?

Matt Alberg (87:11)

Tons of concerns there, but yeah, the dream of having it kind of look at your screen is definitely going to be happening very soon. It's already here. It's just kind of clunky and slow still. But yeah, to have it like in a very snappy way, know what you are seeing and be able to manipulate it is definitely like the place that a lot of these AI companies are going. Yeah, yeah for us at pbq like to have it watch your whole screen and stuff aside from the private, you know, and as a kind of privacy first company, we always in implementation have to think about, okay, like if we did launch this thing, how can we do it in a way that doesn't piss our privacy focused users off? Which always takes a little more engineering than doing it in the non privacy way.

Guy Swan (88:00)

It's the hard way.

Matt Alberg (88:00)

Yeah, for sure, yeah. And so what we generally do is we will wait until there's a nice open source repo that's fairly battle tested for this type of thing and then we'll just fork that open source repo, apply like our PPQ special sauce to it or you know, inject PPQ logic in there somewhere and then start running it ourselves. And then we try to keep up with that open source repo as it updates itself, we can kind of update our version of the fork as well. But again, that's a big juggling act too that we're only learning. We've tried to do this with a couple of different projects. We have like a deep research tool on PPQ and it's basically just a fork of an open, open source project. But we found that like maintaining our version and then keeping it up to date with the open source version, it's like you, it's, it's a skill like it. And, and you have to do all these things. You know, we're trying to offer everything all at once and we have to make everything super maintainable otherwise we're going to drown in like all these different, you know, what do they call that where you're whack. Whack a mole with all the different issues that pop up all the time.

Guy Swan (89:14)

Yeah.

Matt Alberg (89:14)

So I think yeah, we're probably another six months away from the first like open source browser use type thing that's actually battle tested and like fairly snappy.

Guy Swan (89:26)

Yeah.

Matt Alberg (89:26)

And then we will, we will evaluate and go in that direction at that time.

Guy Swan (89:31)

Gotcha. I'm curious how you deal with. Because this is something that drives me crazy is. What's the word? Explicatives for how do you deal with like generating quote unquote bad stuff? Because I, I can't. Especially if, God, especially for like story ideation or if you're trying to have like character driven stuff or you're. I'm just like, there's, there's a bunch of like these weird little edge cases where I'll run into. Um, and different models obviously do it slightly differently, but mostly it won't do anything that quote unquote the model perceives as negative. It's like this isn't negative, you moron. You know, like I, I literally, I sometimes I literally talk to the model like that and like what is wrong with you? Um, and so I'm curious and you know, including not safe for work stuff, you know, like, including like AI, like porn image generation or whatever, like how do y' all deal with that from kind of a platform perspective and a privacy perspective because you have every individual thing that you're plugging into has its own rules. I'm just curious like what's the interplay with that and how do y' all do y' all do anything with that on that side? You just leave the models to kind of do their thing.

Matt Alberg (91:03)

We pretty much leave the models to do their thing. We simply forward the user requests onto the providers. And I really hope, like if we have to start content policing and stuff like that, that would be a big problem for us. We just don't have the resources to do that. And also, yeah, there's always that gray line where you're starting to become intrusive on your users. So our model is just, we put it on the provider, we put all of that responsibility on the provider providers. Our media generation providers, they do have certain censorship filters and stuff like that and I hope that those will hold up as a sort of a justifiable like hey, we're just serving the provider stuff like I hope we don't, I hope some court in the future doesn't say, you the front end provider of this tool, you're now responsible for everything that your users are using. I hope that it stays with the backend providers because they're much more equipped to deal with this stuff than the little guy at PPQ is. So we basically just. It's all on the provider and if the provider. And that's why you get all those crazy errors sometimes where it's like asking you or it's turning down a very tame query for no reason. It's just because it's hitting these very automated sensors that are not always accurate. Right, gotcha. So yeah, that's. We're not really dealing with it. We're passing all of that responsibility and management onto the end provider. Yeah.

Guy Swan (92:36)

Have you found any, have you, have.

Matt Alberg (92:39)

You pushed the limits?

Guy Swan (92:40)

Have you seen which ones will like gone? Because I've done that. I've done that a couple times. I'll make like a random account and just see because I think I got banned from. I was trying to make a dick butt with Mid Journey and I was trying to figure out how to trick it a couple of different ways. And I think, I think, I think that account got banned. I can't remember but I was, I was trying to get around it because I like to, I like to quote unquote, hack the, the LLMs every once in a while. But I'm curious if you've just seen which models behave differently because I know Grok is a little more open on its. Yeah, like it will curse and it will be. You can make it sound like a butthole. I'll use that specifically for that sort of thing.

Matt Alberg (93:30)

The different models do do different things. Like the Google models seem to be the most heavily censored Grok, maybe a little bit less. But what we found is I guess the Chinese models are the most appropriate for like, you know, having the sauciest prompts or whatever. But even they get, they get censored as well. I've done this before with, I think it's less true now because they kind of got caught. But like during the election time, I would, I would ask for Donald Trump images and Donald Trump doing different things. And like, it was actually so pathetic how, how much it was censored. Even just to generate a picture of the guy, you couldn't do anything.

Guy Swan (94:16)

Dude, politics is the worst, man. You gotta do a local model. If you're doing like a Trump or a Biden, it's like, that is inappropriate. This is propaganda.

Matt Alberg (94:26)

Yeah. And so I think they've actually kind of gotten a little better at that. And by better, I mean they're hiding their extreme ideology a little bit more. They still have that extreme ideology at these big companies where, you know, Trump bad and, and all that stuff. But yeah, that's, that's the beauty of being able to try different models because they absolutely behave differently. What, you can, what, what, what stops you. With one model, you can plug it into another model and, and oftentimes it will work. So that's the great thing about bringing all the models into one place. You can just keep experimenting.

Guy Swan (95:04)

Yeah, yeah. Heck yeah. Well, I'm curious. I haven't, it's funny, I've not actually made an account. I've just been using. I have.

Matt Alberg (95:14)

Lots of our users don't have accounts.

Guy Swan (95:17)

20 or 30 credit IDs. I mean, not, not that many. But like I just have a bunch of credit IDs that I will refill.

Matt Alberg (95:25)

Just to explain to the, to the, to the user or to the viewer here. The way that PPQ works is when you make that deposit the first time, it will basically put money or credits on a particular credit id, but we store that locally on your client side so that credit ID sits on your local browser. So if you come to us from, and you don't have an account and you make a deposit on one device and then you come to us from a different device, it's not going to have the original credit ID that you had with your first deposit. And so that does lead to our users kind of having different credit IDs for different devices. Or even in the worst case, if they have a preserving browser, they're deleting their credit IDs on a lot of the time. Yeah. And so we're looking into how to improve on that. We also do allow for account creation where we store the credit ID on our Server side we also store the user chats and stuff, but that's an option on our website depending on how much you want to trust us. But in the future what we would like to do is instead of. So we, we would still store the credit ID locally but we could encrypt the users. So currently all of your conversations are stored client side. And if you come from a different device or different. Or you delete your browser, your browser history, your conversations are going to also be deleted, which is a bad user experience. So what we could do in the future is we can make. You have to remember the credit id. That's unavoidable. Well, actually no, there are ways around that. But basically tie everything to a credit ID and then we encrypt everything client side and then still store all of your conversations server side but in such a way that we still could never get access to them. And that allows users to just plug in their credit ID on any device and they would be able to access all their chats and stuff but still protect their privacy. Another thing we could do is sign in with Passkey which then would allow you, you wouldn't even need to remember your credit id. Your operating system or whatever device you have is remembering your passkey for you.

Guy Swan (97:40)

I was about to ask a sign in with Passkey or a sign in with Nostr or something like that would be great. Just because I've already got the Albi it, you know, is already right there, you know, and then being able to create an account that way.

Matt Alberg (97:54)

Yeah, we like, we like Passkey more than Nostr because NOSTR is, you're still loot. You're, you're, you're. Now you're tying your identity, you're tying your queries to this endpub. Right. So it's a little bit deep anonymizing and so, and also passkeys just work on is a more accepted industry wide thing. So it would probably be passkeys.

Guy Swan (98:18)

Yeah, yeah, yeah, yeah. No, Noster still has key problems that and I think it's, it's the, it's the biggest. Are you a heavy Nostr user?

Matt Alberg (98:31)

I am now. As I said, like PPQ has more customers and more engagements.

Guy Swan (98:37)

That's right. Yeah, yeah, I would definitely push that harder. But I do agree that I really think it's been totally neglected because I think that's the biggest hurdle is how Noster treats keys. And I don't think a lot of the developers or user. I think, I think there's so many nerds there's so like, like heavy technical people in it. They're just like you just got to handle keys and they don't recognize how big of a barrier that is for a typical like a normal audience, so to speak and how just utterly intimidating it is. And I hope we're actually creating because we just have to because I don't want that user experience for Air Drive, but we're actually creating a system that maybe we'll even make as a nip or something. Like we'll, we'll suggest making it broader because it's essentially something that only paradrive will even recognize as, you know, these are my keys because nobody else is going to know. You know, like it's just going to look like a bunch of random keys to everybody else. But I think that's like one of the big things and it lends exactly to the problem you just we're talking about is that now I'm definitely tying it to my identity, you know, like to my no impub. It's like okay, well oh whoa, Guy Swan's here. Look, let's you know what's Guy Swan up to on the, on the platform? And I think having the option of just one time or, or sub keys that aren't actually attached to your identity and actually making it so that's not like a conflict using thing, you know, is a very, very difficult problem to solve.

Matt Alberg (100:33)

This probably ties to another problem we have which maybe your, your thing could, could address is like we have this guy who does part time community work for us. So I gave him our in the private key of. Because he has to use it to like make posts and stuff. But like if he were at some point no longer a part of that company, like can I rotate out my private key or am I just permanently effed forever?

Guy Swan (101:00)

Yeah.

Matt Alberg (101:00)

And I have to hope that he doesn't like take over the account at.

Guy Swan (101:03)

Some point, you know. And it's super simple too. I don't think the problem is as complicated as it sounds is it's just attestation. Like it's just attesting to the fact that this is a key. And so you're not actually giving people private keys or giving devices private keys. The device generates its own private key and it's on everything locally. And then you have your main key which is separate and safe and all you do is sign. All you do is say is this is also my key or this is the key to my producer and my producer can publish and then you can just sign a new one that says you can't use this one anymore. So if they leave, they leave and they don't take any critical information with them, you just revoke the key. I mean it's a simple key approval and revocation system that I think we overthink and or ignore the, that this is a major, major UX problem for, for Noster. Now we're a little bit away from the AI stuff, but I'm curious the reason what got me thinking about this is that like I wanted you to convince me to get a account. Like what do I, what do I get with an account other than like my saved history but I kind of have that with my, my credit thing and I don't know, I'm always saving little keys and little stupid things so I just have that available to me. Um, but what's the, what's the big reason that I would be benefited from an account?

Matt Alberg (102:36)

The main thing is that now that you have an account, your key so it's linked to like your, either Gmail or your GitHub account or whatever or just your regular email address. And then we are storing your credit ID for you and we are also storing your conversations for you. So the big thing is it's just more cross compatible. You can log into PPQ using your Gmail or whatever email you want across devices. You don't have to write down your credit ID or anything like that. That's the big draw. The thing that you're giving up is your chats are now stored server side. Technically at this moment we could go in there and read your chats if we wanted to. Actually, I think that's not true. I think they're encrypted, but even in transit I think that's where we could read your chats is on the way into the database. We could technically read them in the clear in the future. We could encrypt things client side.

Guy Swan (103:37)

I was about to say client side. So it's a tunnel. Yeah.

Matt Alberg (103:41)

So that's the trade off currently. But as I said, I think I haven't checked in a couple months. I really should be checking this more often. But it's still majority of the users on PPQ are not bothering with an account. I think they like the fact that they just put a little bit of money on a credit ID. If they lose 50 cents, they lose 50 cents, you know. But I think many users we have, they just keep track of their credit ID and they just plug it into various devices as well. It's not that hard to plug in your credit ID again to the website and you know, bring. Restore your balance.

Guy Swan (104:15)

So yeah, as I say, if you, if you kind of get that system from a, from just like a mental perspective and I'm so used to doing so many different stupid key and verification and OTP and all that garbage, as.

Matt Alberg (104:28)

You say, it doesn't translate to a normie who like, does not.

Guy Swan (104:31)

They're not used to that, you know. Yeah, I am.

Matt Alberg (104:33)

But honestly, like that's one of the most popular, maybe the most popular user question we get is like, hey, I have this credit ID thing, like where am I supposed to plug it in at? And then they plug it in and they get their balance back. But then they're like, hey, where are my chats? Where did my chats go? And I'm like, oh well no, your chats were always stored locally, so we don't have those chats. You'll have to start fresh new. Or you go to your other device and you can export your chats from your old device and then import them in the new one. But nobody wants to do that.

Guy Swan (105:05)

That's clunky for people. For Normie. Yeah, yeah, yeah. For a typical user. No, I think the idea of like encrypting and storing them either like client side or having a way. Hell, maybe even. No, yeah, technically, of course. Browser. No, no, yeah, it would, we could plug it into a browser back, back end with a browser. You could.

Matt Alberg (105:30)

We should talk.

Guy Swan (105:31)

We should talk. Paradrive's not ready yet for that type of use case, but that would actually be a really, really great one because one of its core things is how do you sync across many devices in the background without the user knowing that to just kind of have like this, this blob in the background that they can always access and then having that front end key be able to decrypt it. Yeah. So that, that might actually, that might.

Matt Alberg (105:54)

Actually be use case.

Guy Swan (105:55)

Well, okay, we'll talk. I'll, you know, go ahead and figure out your, your kind of simpler server client side. Just encrypt some stuff.

Matt Alberg (106:03)

But yeah, yeah, we're still in the idea phase of like how are we going to use these pass keys or. And we, we also thought about the noster thing. So yeah, it's, it is a bit, it's, it's, it's something that's coming to PPQ in the next six months, I'd say is we're going to change things around drastically to make it a lot more user friendly on the account side.

Guy Swan (106:22)

Yeah. What other things did you have other than the account stuff. What's, what's coming in the next year? You know, what's, what's your, what's your future plans? What do I, what can I expect to improve my PPQ experience?

Matt Alberg (106:39)

Yeah, I mean the biggest, the most user facing thing change will be like we're just going to continue adding the latest and greatest models. So anytime you read about an awesome new cool model, whether it's image, video, text, some AI tool or whatever, we're going to get it up on PBQ as soon as possible. So that way the user can almost expect that like, oh, I read about this thing on Twitter, by the time I click over to ppq, it'll already be served up there. We're there already with chat models where the second, the minute that a new model launches, we generally have them available on our website. We're getting there with image and video as well. But then the big, big stab that we're going to be making, the thing that could make PBQ go from kind of like a small community of users to like a gorilla website with millions of users, is this kind of creator monetization program where we allow outsiders to create some cool new image model or a new AI tool, serve it up on PPQ, and then also allow us to distribute their model on other websites that are ordered orders of magnitude larger than us. And then we do it all with lightning payments. So basically the creator can just put some AI tool behind an endpoint. They will then we're still thinking about this. But even asking an AI creator to spin up a lightning node is too much. I want AI experts, I don't want bitcoin experts. Right. So they don't have time to spin up a lightning node or create like a lightning address. But we, what if we on our website allow an AI creator who knows very little about bitcoin, they just click a few buttons in our website and they spin up their own lightning address. But that is still non custodial.

Guy Swan (108:34)

Right?

Matt Alberg (108:35)

That's the dream we've all been hoping for for a decade now. Yeah, it's not the perfect solution, but it might just work. Is using the Spark Breeze SDK where you can come to ppq, you can literally in like three clicks create your own lightning address. That lightning address is still considered non custodial, where we cannot confiscate your funds or anything like that. There's a lot of gray area there which I think that a lot of folks will be, you know, diving into. But the bottom line is that if I'm an AI creator who's not bitcoin native. I need a way to spin up some sort of lightning address so that way I can accept payments for my tool. Right. So I'm hosting this tool on the Internet. Requests are going to be coming to me, I need to be able to charge for those. So we'll create some documentation for this. Where you host a tool on your own servers, you post up that lightning address Every time a user on PPQ uses your particular tool, we will send a lightning payment to you in real time. Also, anytime a user uses your tool on another website that PPQ has distributed on, PPQ will send you a lightning payment. And then we will collect money from that other website on like a monthly basis. Because they're all still fiat native. Right. Like for example, there's a website out there called poe.com they're kind of similar to PPQ. They offer all the big tools in one place, only they're on a credit card subscription. Of course they have this creator monetization program where people can come and monetize, they can create a tool and monetize it themselves. The problem with the way that PO does it is that they're only open to user to creators coming from 20 some Western countries and they all need to submit business documents to PO to get accepted in the program. And they need to have a stripe account account. Right. So somebody in their basement just who makes a cool tool, they don't have all that stuff. Right?

Guy Swan (110:38)

Yeah.

Matt Alberg (110:39)

And so most of the people in the world are not going to become creators on PO for that reason. But what if they could go to pbq, generate a lightning address in five seconds and just slap that as a paywall on their tool and then we are the ones who have the PO account. So we have the business docs, we have the stripe account, whatever PO pays us for distributing their tool and then we're just firing off real time lightning payments to the creators. We don't have to kyb this person. They can be anywhere in the world pretty much. And we can just send payments to this anonymous endpoint. And I think it's all perfectly legal and fine.

Guy Swan (111:18)

That's awesome.

Matt Alberg (111:22)

That's the stab that we're going to be making. We're still working like in the idea formation phase of this. It's a big job to, to, to make all this possible.

Guy Swan (111:31)

Sure.

Matt Alberg (111:32)

But yeah, imagine coming to PPQ a year from now or two years from now and instead of just seeing all these vanilla AI models, you also see all these other Tools. Tools that allow you to create a podcast from a bunch of documents. Tools that allow you to do, like, some sort of image iteration where you're, like, making slight changes to an image. Tools where perhaps it goes into your Excel spreadsheet and AI. A PPQ API key is you're using some sort of Excel tool that somebody created. Yeah, that's the idea of where this could go in the long term, where it just becomes this omni tool place where anything you need to get done, whether it's on a productivity level or a creativity level, you can go to PBQ and there's going to be a tool for you to use very easily. Yep.

Guy Swan (112:28)

That's awesome. That's awesome, man. I tell you, the. That kind of, like, this is something that I'm shocked. And maybe it kind of makes sense at the stage we're in, is that people are still just kind of like, generating stuff at random or for fun or for kind of like stock footage stuff. Like, it's just generic things. But the one thing that I'm still just shocked doesn't exist yet in AI because. And it's probably just because the tooling is really difficult for this, but this is the use of AI is modifying a video, like, in kind of an itemized fashion. You know, like the. Like being able to generate an AI explosion from an image of something is awesome. Like, because, you know, it's vfx, you know, like, that's the. That's the ultimate question is, can we use AI to do this? But I can't really take a video of me or someone else in front of something, keep the video, and then have the thing blow up in the background, you know, like, I. I can't change just the background. Do the mask. Like, I have to break out Da Vinci, which I do do that. And, you know, sometimes I can just kind of like, do that myself. But then, you know, if I'm masking out myself, then now I have to do fake lighting. I have to figure out how to modify the lighting on myself to make it look like the thing's actually behind me and jump up the contrast really fast and put a red outline around me. And you're like, I have to. I have to edit the shit out of that thing to make it even look remotely good. And at that point, why don't I just drop in a VFX of a explosion from one of the. One of the, you know, things that I've already got. But, like, when somebody figures that out, the closest thing is probably firefly with Adobe. But when somebody really figures that out, like how to piecemeal AI into like, this is the background, change this thing, change the foreground to fit that video, but keep the thing that was recorded, that's where like, this is going to explode and it's going to F with Hollywood, like real bad. Because anybody can tell a story with that. And right now it still requires an enormous amount of skill and legwork. As someone who has a background in editing, editing and film production, it's still. It ain't a cheat code yet. It's only a cheat code if you just kind of want generic stuff and you don't want a consistent character. And you, you know, but, but I know it's that. That's exciting. And I think that that may be the type of environment in which something like that actually increases. Because what. Or it comes about because what you end up doing is, you know, thinking of like editing software. And of course, you can apply the Same logic to LLMs and image generation and anything, anything else that you're doing, but just because of the context of video editing, you know, DaVinci is just a collection of, you know, 10,000 different tools of one little thing you might want to do with video put into an interface. And so what you need is like that for AI, right? Is it like, okay, well, where's the tool for adding, adding an effect into the background? Or where's the tool for changing my character clothes? Where's the tool for, you know, like that sort of thing?

Matt Alberg (116:00)

And I could see creators on the platform. The way that I see this first happening is instead of this omni model that just does any particular thing that you want, they're going to create thousands of little, extremely specific models that put whatever you want done. You'll have to kind of find the tool that, okay, it only changes the background and there's going to be a single person in the foreground and in the background. It's going to have this like insane comfy UI workflow happening right where the creator developed this very specific path of like, start with an image, then only focus, then then do like out painting on the background, then convert it to a video, then upscale a video, blah, blah, blah. It's going to be specific, like number of steps. All of that will be abstracted away from the user, but the user will just click and it'll say, like, this tool accepts an image of a person in the foreground and nature in the background, and it changes only the background and it doesn't mess with the foreground. At all and it applies X, Y, Z effects to it, blah, blah, blah. And it'll be a very specific tool that you'll pick from among many different tools. And I could see a single creator just creating that extremely specific workflow where they're doing all the DaVinci stuff in an automated way behind the scenes and the user just gets the final output and they don't have to understand any of that stuff. Yeah, yeah.

Guy Swan (117:32)

That's honestly the leap that I think as far as like, because that's, that's what I found myself doing is that the thing that I use AI for is code has become this, it's just become this accessible tool. You know, it's like when you, when you have like a little tool that you can just like, I don't know, convert something or whatever, I don't know, you just like put it into your workflow. Right. Well, somehow AI has been, been this way to like, even, even the coolest thing is that like I can modify open source code that's just sitting on my computer. I've modified apps that I use that I downloaded off of GitHub and now I can't update them because it will, it will take out the update that I made, you know, or like, maybe I need to go into like try to push a commit, be like, hey guys, you, you guys want this thing that I made for your app? But it's wild that like I can like, I'll just be like, I just don't like the way this looks. I just want this in a different color or something. And also I'll just like go in and be like, can you change the JavaScript for this or the CSS or whatever it is and make this different. It's all just, all the things on like all the tools on my computer are no longer static.

Matt Alberg (118:41)

You know, I thought that it was fascinating when you talked about, you drag and drop. You have this video audio, video transcription icon now where you just take a file and you drag and drop it into something in your dock. And it's basically this open source code that you've kind of customized to do something for you. Yeah, that's really cool and novel. And I read a tweet the other day about like when you, in the future, maybe if you hire a new employee, right. You're, you're not just hiring an employee, but you're, you're hiring the employee. Plus all of they're going to. Every employee in the world is going to have like a personal bag of tools that they bring with them to the job on any given day. And it's like, I don't know, it's. Yeah, it's harkens back to like when you hire a new carpenter, he brings his own tools, right. He's got a big tool chest. It's the same thing when you hire a developer, when you hire a content creator, whatever. They're all, they're all going to have these like custom built AI tools that you know, are a part of their daily workflow. That's very interesting.

Guy Swan (119:42)

Yeah, yeah. No, I'm, I'm a, I'm a super sucker for drag and drop stuff, man. You wouldn't even believe it. Every single tool that I make, in fact I've literally made an app that, that I'll probably publish soon because it's, it's taken a long time but I've kind of made it stable and I've worked out some really clunky UX stuff with how do I connect the tools to the scripts. But I was doing so many drag and drop icon things because I just, I love that workflow. I just, I dropped this into my archiver. It zips it, verifies it sends it off to where it's supposed to go, clears it off my hard drive, you know, and that's, that's it. It's just, it's just a five step script that just runs and I drag stuff to it. I could drag.

Matt Alberg (120:23)

Are you using 30 files? I can drag one for any of this. A what? N8N. You've heard of that one? It's kind of an automation.

Guy Swan (120:30)

That one sounds kind of similar to Zapier Familiar. But no, I'm literally using like Bash scripts and Python scripts in little packages. That's it. Okay. And I'll be using paradrive on the back end when I get. Just. Because that's part of one of the things I want to use it for because a lot of the things I have are clunky or unreliable for transferring stuff between devices. But like I said, I have the whisper, the transcription, drag and drop. And I started building so many of these little things that now I've got a app called Drop Click that I've built for myself which is literally just a little drop zone. And then you drop stuff in it and it comes up with this list of like my most used little scripts. And now it inside.

Matt Alberg (121:12)

What do you want to do to this thing that you dropped? Just dropped in.

Guy Swan (121:15)

I execute scripts or whatever and so, and, and, and then I made a template system so that I Can easily build scripts because one of my biggest problems was like, okay, how do I package this so that I can make it a drag and drop thing? It's like, okay, well what if I made a drag and drop environment like a, like, just like a little thing and then I made it so you can run scripts inside of it and so you don't have to figure out those two or three packaging and getting it to work on your system steps. And. But, but it's actually really, it's a great little app. I use it all the time. In fact, after this is done, I'm gonna, I'll have a mov file that's like a pain in my butt and it's like huge. And so I drag and drop and I converted to MP4 Web Optimize. Yeah. And it's in my, in my drop click app. So hopefully that'll actually be useful to somebody else out there. Um, but yeah, no, this, the people. I'm shocked by how many people who don't realize how useful this is and, and people who think their jobs are going to go away because of it. And I'm like, dude, I've, I've been hiring developers. Like, like when, when I've literally started working with developers, I get them all a cloud subscription, you know, because like, I'm just like, I want you guys to be faster, you know, Like, I would even hire somebody who was just really intelligent and kind of understood architecture and design of these things, but couldn't code. I would literally just hire them and get them a clog code subscription if I just reliably knew that they could beat their face against problems and find solutions to it and knew when things were getting too bloated, you know, if they could understand that concept, then just, just let. Call Claude code it. Um, it's like, no, this is a freaking tool. This is like an unbelievably valuable tool and you can do a hundred times as much with it if you just use it. If you, if you, if you utilize it properly.

Matt Alberg (123:05)

I don't know.

Guy Swan (123:06)

People are afraid of it. It makes no sense.

Matt Alberg (123:07)

Yeah, explore. I was talking to this VC guy a few weeks ago and he had kind of like a somewhat, I think his background is mechanical engineering and he had, he had, you know, learned Python a little bit, but never used it on the job type thing. And, and he kept saying to me, I was telling him about PBQ and I was telling him, like, you know, AI is incredible. Like it's, it's. I use it every day in Many different ways. And he's like, yeah, I need to get into that. And like, no matter how much I said like, dude, just like do it. Just put your face in front of the screen and just start talking to AI. Like literally you can just say to AI, hello, AI, I'm a huge idiot. I don't know anything about this, but I just want to solve this one problem. Can you help me out? And then it'll start talking back with you. And, and, but yeah, there was just this barrier where he's, he's like putting it on a pedestal of like, it's like a to do thing.

Guy Swan (124:01)

I gotta figure it out.

Matt Alberg (124:02)

Yeah, it's like there's nothing you figure out. Literally just put it in front of your face and start talking to it and you're gonna get where you need to go.

Guy Swan (124:10)

You know, that's the biggest thing is that I use AI to figure out what I need to figure out. Like that, that's the, that's the thing that I think people don't get is they think it's intimidating because they like, how do I even start? Go, Literally go and just say, how do I start? Like, AI is the interface that you need to start down that rabbit hole. That's what I do with everything now. It's like a general, it's a incredible generalizer. If I start anything, I usually kind of get AI to explain the basic framework. If I'm ever like, somebody's like, just change this setting to do this. And I'm like, what is that setting? Like, I don't even know what this means in this interface. I just go ask, how's guys? Like, what is this? What is this thing? Explain how this relates to the other pieces I'm doing. Because I don't ever touch this setting because I'm scared of it and it doesn't make any sense to me. But now that's great. That's great. And PBQ is like a perfect example of something like that. Because that, that's like a. Those are one off things. Those are one off things. Not something that I'm going to need a chat GPT subscription form. I just want to do one little thing and God knows how many different things I might need in any. I don't know, Like, I think it's a great, it's a great partner. It is a great partner right now that I have found with having like my two main AI subscriptions, you know, knocking it down from five to two or whatever. Um, and then for all the things the little One offs and stuff that I do need from time to time. Not having to spend 150 on them, you know, just being able to spend the five or the six bucks that I need and it's, it's really beneficial and I got to give you a shout out for it just because I can use lightning and I can, I don't have to have an account, like I can just throw it up on my stuff. Um, because, I mean that's why I was like, dude, we should just do a show. Because I like the tool. I like the tool a lot. And kudos to what you guys built. Thanks.

Matt Alberg (126:08)

It does seem to be coming together now. A lot of the things that I kind of theorized about why this product would be useful are actually playing out. The fact that you can use any model, the fact that it's like such a low barrier of entry. You don't have to commit to a subscription, just throw a few bucks at it, test it out. If you like it, keep using it. If you don't, then you only lost a couple of bucks. All of those things are happening now and I hear them daily from users and so yeah, it making me very bullish on this long term. And so yeah, we just need to get out there more. We need to get in front of more faces and make people understand that the mental frameworks that they're stuck in, whether it's a subscription thing or whether it's just not using AI in general, is a part of their normal daily flow. We have to somehow breadcrumb them to the idea of go to pbq, throw a few dollars at it, solve whatever problem you have. And if we can do that at scale to many people, then we're going to be in really good shape.

Guy Swan (127:18)

Heck yeah. Heck yeah. Well, dude, thanks for joining me. Thanks. Good combo. I'm stoked about all these tools and like I said, I'm a big fan of the, the product that you guys made and hopefully the bitcoin audible crew will go explore. Check it out. Throw, throw some sats up there and generate some, generate some images or explore with Sora too so they can post some random crap on Twitter. Be like AI is 10 new crazy videos. You've never seen anything else you want to direct people to or last minute thoughts or how to follow you. Find you on Noster, Twitter, all that good stuff.

Matt Alberg (128:06)

Yeah, just the website alone. Ppq, AI and then we're big on nostr. I'm sure you can find. I think we're at like paper Qostr Lol and then on Twitter you can find us as well. Those are our big three. We have a telegram group as well. You can access that from our homepage on pbq. So just give us a try. Reach out, have a little customer service widget. All of those messages go right to me. So yeah, if you have any questions about the service, I'm usually pretty good at responding within either immediately or within a few hours. Okay.

Guy Swan (128:43)

Hell yeah, man. All right, let's wrap it up. Thank you for that. I'll have links and details all in the show. Notes and we are out. Appreciate it, man.

Matt Alberg (128:53)

All right, thanks guy.

Guy Swan (128:57)

And that'll wrap us up. Shout out to the audio Notch. Shout out to Matt All Borg and ppq. In fact, I'm gonna head right back to Vibe coding right now to work on a few things. And I'll be using PPQ on my Linux machine. And don't forget to check out the links. Don't forget to check out our amazing sponsors. Now you can follow him on Twitter. And our sponsors, Leden IO they do bitcoin backed loans. Big fan of their service and pretty longtime user of it. Now then, Pubkey and Synonym. This is P U B K Y pubkey app. They've built a protocol, a suite of tools for re decentralizing all the services on the web. And it's so cool actually it's a perfect thing to Vibe code with. I'm literally doing that right now. Then the HRF and their financial freedom report. And then lastly get Chroma getting your lighthealth right, getting your your hormones and circadian rhythm right by getting your blue and green light blocking glasses, you name it. Check them out. Some really awesome holiday discounts.

Matt Alberg (130:04)

So.

Guy Swan (130:04)

Oh and don't forget to use my link by the way when you do that because I don't think my discount is good as their holiday discounts. But if you use my link, that at least helps them know that I sent you there. All right guys, thank you so much for listening. I will catch you on the next episode of Bitcoin Audible. Subscribe. Share it out. Let everybody you know, everybody you know except the people I might be buying Christmas presents for. Let them all know about this show and I will catch you on the next episode of Bitcoin Audible. Until then guys, take it easy.

Matt Alberg (130:48)

La.