E49: Paul Itoi on Stakwork, Sphinx Chat, Earning Bitcoin, and AI vs. Humans
Podcast
Video
Episode Clips
show Notes

Paul Itoi has built two of the most popular Lightning applications to date, Sphinx Chat and Stakwork. These platforms already enable people to chat and earn over the Lightning Network, and Paul’s vision for where these platforms will go next is incredible.

In our wide-ranging conversation, Paul and I discussed the concept of earning Bitcoin, how AIs and humans can work together to truly build a world computer, and how decentralized knowledge graphs can help people learn faster than ever.

→ Sphinx: https://sphinx.chat/

→ Stakwork: https://stakwork.ai/

Sponsors

→ Voltage: https://voltage.cloud?utm_source=kevinrooke&utm_medium=Youtube&utm_campaign=1mo

→ ZEBEDEE: https://zbd.gg/

At the end of every show, I answer any questions listeners send in over the Lightning Network.

To ask a question, send a message, or to support the show, download Fountain from the App Store and load your wallet with a few sats.

→ Fountain: https://www.fountain.fm/

→ More Episodes: https://play.fountain.fm/show/P6XXuSPg6f2rj4ECB0fT

→ Lightning Address: ⚡kerooke@fountain.fm

Links

→ Twitter: https://twitter.com/kerooke

→ Books: https://www.kevinrooke.com/book-recommendations

→ Blog: https://www.kevinrooke.com/blog

Timestamps

00:00 - Intro

02:26 - Paul Itoi Intro

05:29 - Why Build on Lightning?

10:36 - What is Stakwork?

14:53 - How Humans and AIs Can Work Together

21:10 - How Earning Bitcoin Can Change The Way We Work

32:52 - Stakwork’s Real World Impact Today

46:19 - Sphinx & Their Decentralized Knowledge Graph

51:47 - Knowledge Graph Use Cases

1:04:36 - What Else Will Be Built on Bitcoin?

1:08:49 - The Lightning Round

transcript

Paul Itoi:

When a new user signs up, they can do tasks and get paid and have that Bitcoin in 60 to 90 seconds. $10 a day or so is a major change in someone who is living in the Philippines or El Salvador or any country where your earning opportunities are limited. I can't think of anything else in Lightning that's more powerful than that. It beats spending all day long, so it really is a team effort between what an algorithm can do and what a human can do. What if that turns into the true world computer? Right now, if I go to Google and search trampoline routing, everyone should do that. See what Google tells you trampoline routing is or AMP or Taro or whatever it is. You're going to get garbage back. No one is going out there and building new knowledge first in a web page. That's the last place knowledge ends up. Our most limited resource is our time on this planet, we're trading that time for Tik Tok videos will be viewed as the most tragic waste of humanity. I don't want to keep training a central algorithm on what I like that does not have my interest in mind. Your algorithm is working for your education and your entertainment, not what YouTube thinks you should watch next to stay on their YouTube app.

Kevin Rooke:

Paul Itoi has built two of the most popular Lightning Network applications to date, Sphinx Chat and Stakwork. These platforms already allow people to chat and earn over the Lightning Network, but Paul's vision for where these platforms will go next is incredible. In our wide ranging conversation, Paul and I discussed the concept of earning Bitcoin, how AI and humans can work together to build a truly global computer, and how decentralized knowledge graphs can help people all over the world learn faster than ever. I've also added Paul to today's show Splits. So if you enjoy this episode and if you learn something new, the best way you can support the show is by sending in Sats and comments and questions over the Lightning Network. Just a quick shout out before we get into today's episode. Today's show is sponsored by Voltage. Voltage is the industry standard and next generation provider for Lightning Network infrastructure. Today's show is also sponsored by Zebedee. That's Zebedee, and Zebedee is your portal into the world of Bitcoin gaming. You'll have more from Voltage and Zebedee later in the show. Paul, welcome to the show. Thank you for joining me. I have a lot of questions to ask you about Stakwork and Sphinx and the work you've been doing on the Lightning Network. To give listeners a little bit of a background on your time in the Lightning Network and the work you were doing before. Can you explain exactly how you got involved in the Lightning Network and what was it that kind of drew you in?

Paul Itoi:

Sure. I had started working on a free nutrition app, actually to teach people how to do Ketosis. And so I brought that up a couple of times in the past, but we are having a ton of trouble getting work done for the app. We are receiving several hundred nutrition labels a day. And if you ever looked at food packaging, it's terrible for optical character recognition. So the idea and the dream is, okay, you just take a photo in the app, and then this food item magically appears so that you can log in and track your carbohydrates for ketosis. Well, we were paying people actually in Argentina 20 plus dollars an hour to do that data entry work. And it was a terrible experience for them because the work was really boring. And even at that dollar amount, we were actually having trouble keeping people because the work is so repetitive that they would quit. And there's enough to a nutrition label, you have to get it right. And there are so many different versions of the numbers. Is it milligrams? Is it grams? Labels in Europe include carbs and their total carbs. For the US, it's actually netted out into a separate line. So if you screw that up, someone can get very upset with you. So we actually had to train people and pay people over PayPal to do that work. And we were having trouble keeping people on. And so I read the Lightning white paper, put two and two together and said this could solve our biggest issue. And so we created a site where we broke the work down into tiny pieces. And this is multiple iterations to figure this out. And then with Lightning, we could actually pay people one cent, two cents to do just one line of that nutrition label. And so that was the magic. Unlock for us was a combination of Lightning that lets you do micro payments to anyone in the world. So now you have a workforce instead of who can I pay via PayPal it's, who can I pay in the entire world that would be able to access this job on the Internet? So if you have a phone and you have Internet, then you can get a Lightning wallet and get paid Bitcoin. And so that's really what we did. But then being able to break the jobs down into tiny tasks made it so that we could get the work done without spending the 2 hours training every person. Right. It's similar to the Henry Ford manufacturing process, where you do division of labor, and each person knows how to do one task. Instead of having to build the entire car and training one person to build each car, you have each person doing the tasks that they know how to do. So by lowering the training, increasing the geographic reach, we were able to get this work done at a fraction of the cost of what we were doing.

Kevin Rooke:

Before back in 2018. How did you know to build on Lightning and not to build on some other blockchain or create your own token. Because there were a lot of pitches back then where people did claim that you can do instant transactions and free transactions on other chains. How did you know to focus on Lightning then?

Paul Itoi:

Honest answer, we didn't. So I have a draft Google Doc. I'm going to dig up somewhere of a white paper that I was exploring on building a rewards token for using our app. And it was on an ERC token on Ethereum. And I was all excited telling our developers, oh, yeah, we're going to do this. One of the guys who worked on our app, actually, he was one of the first people who are helping us get our iOS app off the ground. Created one of the first ERC 20 tokens. I think it was numerous and so he was deep into that world. We had stopped working together that time, but that's how I heard about it, was tangentially through the work he was doing, which I think is actually still an interesting use case. But we looked at it in a Google sheet. You could do the math on how many users we had, how many rewards we needed to grant, and our little app alone would have saturated ten X what the Ethereum chain could do at the time. So that's the lesson. You can't physically store transactions in layer one. There has to be some sort of layer two. Enter the Lightning white paper. Oh, this is interesting. This actually will work. And then it was a waiting game to get Lightning out on mainnet before we could start using it. But as soon as it came out, I think we were one of the earliest use cases out there of just getting the job done. So I've always taken the approach of what problem do we need to solve? And let's do the minimum amount of coding and tech to solve that problem to see if it actually solves it versus hey, I really love Bitcoin or I really love Lightning. Let's see what we can make with it. It's more what problem do we have in real life now and how do we solve it? And so luckily Lightning just came along at the right moment and we were open enough to understand that this was a real possibility. Luckily, we never went down the altcoin route, seriously. But I did explore it and actually saved me a lot of distractions later because any time everyone brought it up, I said, hey, is there a layer two? No. Okay, not interested. It just doesn't make sense. And I also think Bitcoin has an advantage over everyone else and that the idea of layered technology. Layered money. I think I can't remember the guy who wrote the book, was it Nick or something like that. But the whole concept is you can't decide you're going to add a second storey to a building you've already built later on and say I'm going to tack on these other layers later. I really feel like having that as part of the Bitcoin approach from the beginning is super important.

Kevin Rooke:

And now in the last four years of building on the Lightning Network, how has your view on the Lightning Network's ability to solve your problems changed or has it changed?

Paul Itoi:

I feel like I'm a bit of-I’m very neophyte on Lightning. Every time I go on Twitter and I see these people who are true experts, I say Andre, Ryan Gentry, Fiat Joff, all these people who know this world so well. We're trying to use the minimum amount of Lightning features possible to solve our problem. Whereas I feel like there are people who know way more about the technical underpinnings than I do. And I'm relying on them to say, hey, can we keep Lightning scaling? Can we keep it peer to peer? These are all things that are important to us, but we really aren't on the cutting edge of Lightning. So I would say that we started off very cutting edge, and now I feel somewhat behind, but I feel like Lightning, even the version that we know and understand and use were built on Lightning Labs, LND. We've added support for C-Lightning on Sphinx. So we're very much back from the bleeding edge of what Lightning does. But also Lightning works for us. We can do what's available now and the payments go through. People are happy with earning on Lightning, on earning Bitcoin over Lightning getting paid. And now there are other companies we just started working and talking to a company called Pouch.ph, I think is the URL. So they're making it much easier to off ramp in the Philippines, lower fees, not having to go on chain. So we're just doing our little slow plotting thing, and then other companies are popping up that are making it easier to go take the fats that you earn and make them useful in your daily life. So those are all the parts. So I feel like it works the way it is. That's not very exciting from a technology standpoint, but I'm not waiting around for something. We're very happy with how Lightning works right now.

Kevin Rooke:

Awesome. Now for listeners who aren't familiar, can we do a little breakdown of exactly what Stakwork is and what Sphinx is, and we can get into the specifics, further into the episode?

Paul Itoi:

Sure. Yeah, that's a great point. We're a standard business, so we are a C Corp in the United States and our product is Stakwork. This is how we make money. So companies that need work done, and we start at the very low level of data annotation. So if you have an image that needs a box drawn around a stop sign, we certainly can do that. And then we get paid US dollars by these companies to get this work done. And so then we buy Bitcoin with those dollars, we open channels over Lightning and then we post these tasks on our website and then people log in and do those tasks and then withdraw the Bitcoin that they've earned over Lightning. So in a minute you can sign up. We're not right now full. So we're not taking new users, but when a new user signs up, they can do tasks and get paid and have that Bitcoin in 60 to 90 seconds. It's pretty amazing. And once people do that, they have this, AHA, moment of, wow, they're not entering an email address, they're not entering a phone number. They're not entering their Social Security number wherever they are. And this is open to people who are non US citizens, but they can basically earn money instantaneously and have that value. So I can't think of anything else in Lightning that's more powerful than that. It beats spending all day long. Sure, I can buy a sticker or something like that. That's cool. But in terms of a viral use case and a big moment in your life when you haven't had the ability to earn and suddenly you can earn with just your phone, it's a big moment.

Kevin Rooke:

Yeah, that's incredible. Just switching from spending or having to buy to being able to earn directly. I think it's a big moment that a lot of people still haven't clued into. Can you speak to some of the most popular jobs today? You mentioned a couple. What are people doing today and how are you kind of breaking the tasks down that companies send you and having individual workers work on little elements? How does that work?

Paul Itoi:

Yeah, I'll tell you what we're good at. We're good at images because it doesn't require language stuff. And we're just now getting into language processing, which we're not great at yet. We're really working hard on that. Actually, this morning right now we're going to do an all hands

where we have to actually help get the work done because we're behind on getting the customer deliverable then. So the task would be this is one of our longest running customers and has done millions of tasks through the system so far. They'll send a two to three minute video of a depth sensor. So this looks like LiDAR or Radar. It's a black screen with a ton of dots. And so we will pull frames from that video and then each person will work on one frame. So we'll actually take 30 frames a second, pull every fifth frame, and then have someone work on a frame and label just that frame, or we'll break a 30 second video into five second videos and then have one person look at that 1st, 5 seconds, the second person look at the second, and so on. So we always try to break the work down into human sized chunks where your brain is really good at doing 3 seconds of work versus watching a video for three minutes. So our approach is always take what the customer is asking us to do, break it down into these things we call logic legos in our system and you string them together just like a programmer would string together code. And then we assign those certain tasks to people and then we also have algorithms do some of the tasks. So it really is a team effort between what an algorithm can do and what a human can do. And then it makes what we can address, almost anything that a human can do, knowledge, work that can be digitized, can now be broken up in this way and then just the parts that need to be assigned to human or assigned to a human. But now instead of that human being in a desk somewhere in your office, they could be anyone with a mobile phone who values Bitcoin.

Kevin Rooke:

That's incredible. Are there any jobs? I'm just thinking of the possibilities here.

Paul Itoi:

Right.

Kevin Rooke:

Like you start with images, you move to language, you move to more complex tasks over time. Are there any jobs that cannot be broken down? Like who is immune from being disrupted by this?

Paul Itoi:

Yeah. So Jim one of our earliest investors, he nailed what we were trying to do within the first five minutes. He said you guys are going to be competitive or will be the first practical version of general AI. So the example here is chess. So there is a lot of great chess Bitcoiners out there. But from what I understand, the best chess computer will lose to the combination of a human with multiple chess algorithms running. So a very good chess player with two very good chess engines or chess simulators will beat just a computer or just a human. So in the machine learning world, that's called the Centaur model where you have a horse with a man's head on top. So this is a computer with a person's head on top. This combined version of man and computer and the way to interface between the two seems to be the winning model. So your question of what can't be disrupted when you combine what humans can do with the efficiency of what algorithms can do and you have software that can right size the task to those two choices, human or computer? As far as I can tell, there really isn't much that would be outside of that. We're just starting right now if you think of what Stakwork is as a computer that combines humans and algorithms. Right. And we have a very simple no code language that lets you configure this on a web admin panel for our customers to set these workflows. So you'll say, hey, I've got this video, I need to break it into these steps, then the next I need to have a human draw box, then I need to have a human check that box, then I have an algorithm crop that box. Now I generate a JPEG. So that's what you would normally hire someone in an office to sit there and do. We can, instead of hiring humans to do that, we can freeze that logic that you want to have happen in our system. And then humans can just do the parts that they need to do and then the computers get better and better as they shadow the human and then steps that humans do today will be automated tomorrow. So with this type of system, I don't think that just humans or just computers can beat a system like this.

Kevin Rooke:

Makes sense. Okay, yeah, 100%, yeah, it makes sense that there's going to be roles that computers are better at and roles that humans are better at. And working together gets kind of the best of both worlds. What do you think the biggest change to people's lives will be in a world where computers and humans can work together in the manner you described?

Paul Itoi:

Yeah, it really is. I really think about. I come back to Adam Back a lot on this question. And so the whole idea of the magic to me of proof of work, I was talking about this with some people last night. The whole idea of bridging this physical world. What Bitcoin does is let's say this is the onboarding process where proof of work links value. How does a computer know that if you're trying to solve a scarcity problem, how does the computer know that something was applied? It's an amazing tool for linking physical expenditure of electricity and energy into this chain that we all value. And then if you think about what we're doing, so that's the proof of work, onboarding value from the real world. And then if you think about what we're doing on the back end with Stakwork, we're taking those Sats that were frozen energy and now we're unfreezing it and trying to get a human to do a task. So it is the reverse of proof of work in a way when we transfer those US dollars Fiat to Bitcoin, Bitcoin into a channel over Sats. And then we've now digitized the human incentive. So someone who's sitting there in Nairobi is going to do a task and spend their time, their real work, their human cognition in return for a Sats that is a representation of that frozen electricity from before. So it's kind of a crazy system when you can link the real world to digital scarcity in the formation of Bitcoin and then take digital scarcity and then trade it for human scarcity, which is your time. And so that person I believe is making the purest transactions. I have what my brain can do. I have my moments on my life here on Earth that are finite. That's the one thing we can't get more of. And then I'm trading that time for this energy that's stored in Sats. So to me there's this beautiful symmetry of work in and work out and the medium there the medium of exchange is now truly Bitcoin that can be moved around the world for free and in a censorship resistant manner. That's the craziest thing. So anyway, I got off on a tangent there, but I feel like I don't know if that makes sense of just what this whole thing is trying to do is pretty wild. And the fact that that's part of a larger process that you're step four in a 20 step process, you're a human part in exchange for the Sats is step four and a 20 step process? The only question then becomes, well, what processes are best suited today versus later? So if software eating the world has been going on since that cliche was phrased, and now it really will be these logic trees or Stakwork eating the world when you have this ability to combine human cognition and computers together. So anyway, I don't know why I thought of that right then, but that was something I just felt like right before our call I was thinking about it. Is this kind of reverse process to the creation of Bitcoin?

Kevin Rooke:

Yes, that's really helpful. That's helpful to provide context for people. In a world, then where- one thing I'm trying to understand more is why hasn't everyone realized this mechanism yet? Why we have so few people who are earning Bitcoin today? Everything you said makes perfect sense. It seems like we have the tools to make it so that anyone can earn for any tasks they perform through Bitcoin and having a mobile phone through remote work. Why is there not more of this?

Paul Itoi:

I think there should be, really. But I think you're talking about the Venn diagram of people who know Lightning and then people who need task work done. Those are two very separate things. And there are other companies in the space now that are looking at Salona to pay people. I mean, I've hopped on the phone with a couple of these guys and they say, we really want to do what you're doing. And I said, well, make sure you choose Lightning because it's the most liquid asset and we're going to do Solana or something like that. I'm like, okay, that's fine. They are competitive in a way, but there's so much work out there that this process will be ten different companies. And I just hope that people invest in doing it over Lightning versus another asset that has less acceptance and liquidity out there. So I do think that it's the next phase. If you look at what happened with W-2 employees, that's what I graduated from College. I worked for my IBM for 30 years, and I retired. So that's long gone. And then you get into frequent job changes where I have a W-2 employment with the Corporation for four years on average, and I go through my life doing six different companies. Then you get into, okay, I'm not a W-2 employee or I am a W-2 employee. Now I'm remote, Covid did a ton of that. Now you see gig economy. I'm not actually an employee, I'm a contractor. And then the next step will be, okay, well, I'm not actually taking a long term contract. I'm doing work on a piecemeal basis. So we're doing that as a company to hire programmers. Right now we have a website where we're experimenting. People can go there. If they go to Sphinx Chat, they can click on the community and people page and developers can create a profile and then look at tickets and they can actually grab a ticket and earn Bitcoin for doing programming tasks. So to your question of why aren't more people doing this? Well, not everyone has this kind of micro task work that they need done. So we're starting there with Stakwork, data, annotation, simple processes, and then at the same time we're working down from the top and saying, okay, can we hire programmers this way. So if we can do ticket based hiring? So instead of, hey, here's a four page legal agreement with we own your code and what's your social? I'm going to send you a W-9. All that friction for working with someone. What if you could just post a task in a bounty denominated in Sats and solve a programming problem, ideally in real time, with someone who has spare time and is willing to help you out? So most programmers right now go to Stak Overflow and we'll try to look up and selfserve the answer. What if you had a real time Stak Overflow where people were hanging out and they get a notice saying, hey, someone needs help with Rust right now? And then you have 20 minutes. You jump into a buddy coding session, and then you try to solve that programming problem in real time. You don't know the person's name, you just know their pseudonym. You pay them some Sats. Hey, thanks a ton. That person could be a Google engineer, they could be working at your competitor, they could be in Ukraine, they could be anywhere. And so you're just getting that help. You're interfacing, no friction because the payment rails add the friction, and then everyone walks away happy. Problem solved. Satoshis paid, and then you move on to the next topic in your life. So that's pretty exciting to me.

Kevin Rooke:

Yeah, it does sound exciting in the progression you described, though, I want to play Devil's advocate for a second because you described this progression of like starting out. You have a W-2 employee working at IBM for 30 years. Every progression along the way gets the work commitment or the level of commitment to the company is kind of broken down or turned into a smaller chunk or smaller time frame. Is there a downside to having no friction there, into enabling anyone to hop between whatever opportunity they want, whenever they want? Is there any downside to that?

Paul Itoi:

I believe there is, to people who want central planning. So if you think about employment, the transaction you just made is your Corporation through manager think office space, the movie. Here I am trading my security that I have some- and it's not real. Right, because I have an employee badge and I'm a W-2. Almost everything these days is at will in the United States at least. So you can get fired tomorrow. It doesn't really matter. It's actually, I think a bit of a scam. You're presenting this illusion of security and so that prevents the person from looking for the next job because they have the illusion that they're in a job. Right. And there's a commitment there, and I think there is some. But that illusion really is the person who's accepting the income, saying, I'm giving up my free will of things that I particularly want to work on. And I'm committing in return for this security or illusion thereof, that I have this job in this two week paycheck. And so I think it makes people less hungry and less productive to operate in this fantasy. Really. It's theater. It's work theater. That's why you go to a company, pick any company: Hertz rental car. How many people work there? Go to Oracle? How many people walk around the office? What do you people do? Why are there 25,000 people in this office? What on Earth could 25,000 people be doing? And you'll find incredible, unbelievable waste across the board because there's cross incentives and that's fine if you- not everyone wants to be a Hunter gatherer out there in the world. And I'm not saying this is for everyone, but if you really want to do creative work, learn and be on the cutting edge of your field, I would find a very tough argument saying that a W-2 employee in a large Corporation is the best pathway to that goal in your life. If you want even the illusion of security because you had a rough childhood or you're in a bad spot in life, of course, like go and have comfort and the free sushi lunches and stuff like that is not a bad thing. But for people who are prioritizing doing their most productive work and learning, then you want uncertainty, or you're willing to accept uncertainty to be able to work on just the things that are the right thing for you to work on at that moment.

Kevin Rooke:

Now, if in breaking up tasks and lowering that commitment to a company and hopping around to different jobs, if in that you're being more productive, do you think that the workers who choose to take on that are also going to start earning more and out earning their W-2 salary counterparts?

Paul Itoi:

Yeah, I hope so. It's almost like the podcasting world. I was listening to Joe Rogan the other day talk about what's his name, Howard Stern. Howard Stern would ridicule podcasting and just saying, my big time contract with my satellite radio company is going to beat your crappy little podcast any day of the week. And so that's similar to, well, the old model feels more secure, but the new models that arise could certainly outstrip the old model. To me, the curiosity is, would Adam Curry's value for value concept work for something like this where let's say that you have a machine learning programmer with several hundred thousand dollar-a-year salary is stuck for 4 hours if they don't know how to do something. But then someone else, this elf can hop into their computer and help them solve that problem to the Corporation to save those 4 hours, what value would that be? That might be thousands of dollars of value. It might only be 40 seconds of that person's time if you were to somehow find the right person and get them in that conversation. Anyone who's coded knows it's. Often in the moment you explain the problem that that act solves the problem. But if you're just sitting there banging your head against the wall, searching, Stak Overflow, it's a terrible way to try to solve problems. And so value for value would be I just spent a minute and I made $5,000. That would be an incredible moment of, I think, validation for this model that requires a lot of people involved, of course, in good matchmaking algorithms. So there's a chicken and egg situation here, but I think it's possible, I really do.

Kevin Rooke:

Now, in the mechanism you described in that case, that's not a voluntary payment, right? That's almost like a contractual thing. Like I have 40 seconds of my time to give you. You need this problem solved. You put up a bid or an offer and I'll perform the task for a set amount of money. Right? It's not an after the fact.

Paul Itoi:

It could be both ways. I could see it be both ways. I think people will start with an explicit amount to get people going because it's a trust issue. Hey, I want to spend 5 hours of working and there's no money guaranteed here. I might not do that, but after the 10th one where you realize you're under pricing the value you're giving and then so you could have fixed value, I'll give $10 for someone to try to solve this CSS problem. And then after it's solved, it's done so nicely, you might tip that person $20 again. That's how you can bring them value for value into the transaction. So you might have a minimum. It's a hybrid where I have a minimum where no matter what, for you to even pay attention to my problem, I'll pay you this much Bitcoin. But then if you solve it well and great, then I'm going to bonus you. And then having a way for the person doing the work to be rated and rewarded with some sort of badge or point, and then have the person doing the work reward the employer, the person posting the task with, hey, this person tipped me 300%. They're very cool to work with. That should go with your reputation as you go. And so this whole pseudonymous economy is something I think that Belaji Sreenivasan really popularized. He's very good at talking about it. He's more on the East Side, but the concepts are there. And then you get into identity. You've had all the key pieces here, Daniel, on distributed identity. And then you have John Carvalo working on synonym with Web of Trust and reputation management. So just picture all of these different pieces being built to support what I'm just describing right now.

Kevin Rooke:

Yeah, I want to dive into identity in a moment, but before we do that, I want to help listeners get a grasp of exactly what Stakwork is doing right now. Can you share some of the stories of the real world impact that working on some of these tasks is having on people's lives today?

Paul Itoi:

Yes, times are tough right now. So we are in a comfort stimmy check world over here in the US, in developed countries are, I believe, in a false sense of wealth right now. If you read that book When Money Dies, it's creepily similar feeling to what's happening right now. But there are places where our site is popular, where on average, let's say it's six dollars to ten dollars a day worth of Bitcoin because we limit people, we cap out the jobs that people can do. And so we don't want to see people spending more than 2 hours of their time hunched over a phone. And so we limit every job by the number of tasks that can be done. And the ideal target is 2 hours or less. And so in our target range of pay, at the very lowest level, if you're just doing, say, a number entry, it might be around two dollars to three dollars an hour. But you pretty much graduate that just once you make it past there where you're not a bot or just entering garbage and trying to cheat the system, then you're quickly into the four to $8 range on income on US dollar terms. And so $10 a day or so is a major change in someone who's living in the Philippines, El Salvador, any country where you're earning opportunities are limited. So we get stories. I mean, some of them are some are super tough right now. If your quality drops, we just get a support email an hour ago from someone who says, hey, I've been working on your site for a year, and I just lost access to this very high paying task that they like to do because your accuracy score drops. And I mean, the stories are gut wrenching, but it's literally, hey, I'm feeding my family off of this. This is paying my rent. It's a major portion of the income. And so for them to lose access to that, whether it's our fault or their lack of attentiveness is rough to see, but it speaks to what income means to people. And so to me, it's a very important thing for us to recognize. And that's why we don't want it to be the full time income for anyone, because we are a new company and we're figuring this out as we go. We never want to rug pull anyone on anything. Even if people start to rely on income, you don't want to go away the next day and move someone's entire income. And so people are using it obviously for life. And it's a very critical piece for many people's lives. And I think that the fact that we're speaking in Bitcoin to people makes Bitcoin a central asset to them. Hopefully the association of Bitcoin is what's enabling me to pay for this food or this rent. So Stakwork is the brand name, but the medium we all work in is Satoshis. And I think this is the positive part for Bitcoin that our company can have is on the Gini coefficient, which is the spread of the asset to as many people as possible. So yeah, we typically work just one on one with people, but we have people in a women's shelter in Costa Rica doing work. We have youth coming out of a church in El Salvador doing the work. So we have small pockets of groups that we have as well. But 90% of our users are just single anonymous users working on the system. And then the experiment we're doing now is we're starting our own Stak phone brand. And so we're actually providing smartphones to people who can't afford it or they have one that's not great. If you go down to the outskirts of Manila, people are often sharing a phone. The phones are not great, the batteries don't last. And so if we can buy these at wholesale and then provide that to someone who doesn't have a great phone, then they're more effective worker on our platform and we can buy these at a discount so that they're not getting paid by us. And then going paying retail for phone service and a phone. We would like to start the Stak phone brand starting in El Salvador, Central America, Philippines, and really focus on providing phones and service to people in return for their spare time. And they will never pay a penny of cash or Bitcoin to pay for a phone. It's just spare time. We're hoping a billion people will make that transaction and then our workforce just grows from there.

Kevin Rooke:

And what does that mean then to have so much in a stage where you get a billion people working on Stakwork, that's a lot of tasks that are now being outsourced to a truly global community. Right. How do you think about the changing dynamics between work today is not may be perfectly siloed because there is remote work and there is the capacity to do things in different countries, but it's challenging a lot of the work that is required in Canada is happening in Canada, and a lot of the work required in the US is happening in the US. What does that difference look like when everyone has access to this liquid system for exchanging their time for money?

Paul Itoi:

Yeah, it is cliche. But what if that turns into the true world computer where it is a combination of these thousands of algorithms that are good at very specific items? What type of pet is this? Is there a cloud in this satellite image of the types of customer requests we get? And algorithms get very good at that. So you have thousands of algorithms that are good at it, and then you have millions of people who are doing the things that the algorithms aren't good at it. And so you could ask any question to this global brain and get an accurate answer out of it. Right now, the fastest our global computer can process a question. I mean, the fastest I've seen is say a 20 second turnaround time. That's an incredibly slow computer. But it's okay because this is the first version of this computer. So we have a first version of a programming language that lets people ask questions to this computer and get the computer to do tasks. We have fairly low CPU, power, and Ram. That's the number of people and algorithms in there. It's fairly small, so it's slow and small like any computer would start with. And the instruction set is fairly limited. But this doesn't have the same limitations of other computers. Whereas soon as your code gets stuck in normal entirely computing environment where it's all automated, this has a cheat code, which is all right, I'm stuck. Ask a human or ask for humans, you get out of jail free card, then the thing gets to skip that problem step and move on to the rest. Now, if you can make that fast enough, then I don't know why you would use a regular computer program or why you would care. So if I were Allstate insurance company and this is very pie in the sky, but this is just what you have to have the goal to try to get to. If I'm Allstate as an app developer, I would just say get damage quote and I would pass in an image of my car's fender that's dented, and then the app in 4 seconds gets back a quote that the company is standing behind. Now, was that software? Was it some magical AI algorithm? Well, before you send that $2000 check, you're probably not just going to send the algorithms output and live by it. We're not there today. But if you have the right number of humans checking all the parts and you do cut that a $2000 check in 4 seconds, that's a ten X experience over what any company has access to today.

Kevin Rooke:

Yeah. And you can pass that on to the consumer then.

Paul Itoi:

Exactly. That to me is an incredibly powerful computer to build. And how do you increase the Hertz rate of this computer and return answers faster? Some of the things we knew in the beginning is that our website is slow. There are certain mechanics of loading a web page and asking questions that is inherently slow. So one of the ideas in the very beginning was to build a connection to games. So you know how in games you often have to answer a survey or pay to continue playing the game. It's the business model for these games. What if we could just present tasks in the game? You have young people, incredibly adept with their hands already in an environment not wanting to pay Fiat or Bitcoin to move forward in the game. What if they could earn credits in the game or skin or whatever by completing tasks really quickly? So all the overhead of loading a web page, hitting submit, doing all these things. If you just have the ability to surface these tasks in a game to move forward in place of paying cash, then you could open up a number of tasks to so many more people who are already on their phone playing the game. So it removes a lot of the turnaround time in the process. So we'll get to those things as we go here, but that will help solve some of the turnaround time and make this a faster computer.

Kevin Rooke:

I hope you're enjoying the show so far. I just want to give a quick shout out to our sponsor, Voltage. Voltage is the industry standard for Lightning Network infrastructure, creating layer two applications and services on top of Bitcoin starts with Voltage, where you can spin up nodes, get access to liquidity, optimize your node and much more. Voltage is leading the way as the next generation provider of Lightning Network infrastructure. And if you want to get a free trial and start using Voltage today, you can do so at Voltage Cloud. That's amazing. I like that world computer analogy at a mature state, do you think? Like I think about the different ways in which Stakwork can offer a better solution. One it could be just a higher quality solution, one it could be faster and one it could be cheaper. And so when you think about combining a network of humans and different AI engines together, building a solution, what do you think that main selling point is? Is it primarily are you going to win on speed or on quality or on, you know, or maybe it's a combination. Is it going to be cheaper? What's the kind of main selling point, do you think?

Paul Itoi:

Well, I think the task that we're targeting right now are relatively low on the complexity scale. So to me right now one of the biggest factors out there and motivators for people to change is in the back of their mind, the people are outsourcing this work to it's called the BPO market business process outsourcing market. So replace that with cube farms. And so you have overseas cube farms run by corporations in that country who have contracts with corporations in this country. And there's many layers to that. The problem is that if you have 5200 people in a cube farm, the likelihood that that individual worker has any say over their working conditions or their hours or their pay. There's an incredible imbalance of the power dynamic there. And so would anyone be surprised if the person actually doing the task to train yourself driving car are being treated as well as they could be? It's a very low likelihood. There's a bunch of articles out there. This is the new digital sweatshop and stuff like that. So our model, by being this Bitcoin approach is we can't force anyone to do the work. We have to offer a task at a price and the individual is making the decision, is my spare time worth this price? That's it. And so we can't fire someone, we can take the task away. But I have no leverage and our company has no leverage over the individual user. And so that should be the first motivation is just when you are putting a contract out to an overseas company and keeping it at arm's length and hoping that you don't hear about whatever is going on to get that work done, there are abuses happening in that side and it's not something that a bunch of people talk. No one's incentivized to really talk about it that much. So we just take the approach of can't leverage instead of won't leverage people. And I feel like that's very much the Bitcoin and meritocracy way to go about it.

Kevin Rooke:

Right.

Paul Itoi:

So I think that's one part of it. But sorry, I was getting I tend to get off track on your question. No, you got to bring me back any time.

Kevin Rooke:

Actually, I think people have a good understanding now of how Stakwork works and some of the components there. I want to get into another project that we were talking about last week. Sure. You're building a knowledge graph and you're building this knowledge graph for right now, I believe Bitcoin podcasts. Maybe we can start with just giving listeners an understanding of what a knowledge graph is. What your vision with this is? I think it was a really cool conversation we had last week, so I want to give everyone access to that.

Paul Itoi:

So I'll cover Sphinx a bit, which is also probably your first question. You said was Stakwork in Sphinx, and I skip the Sphinx part. So Sphinx is an open source project where we use the Lightning Network not just to send payments, but we are sending payments, tiny three set payments over the Lightning Network using the key send feature of Lightning that Lightning Labs came out with first a long time ago. And so we attach text to that Lightning payment and lets you do lots of things we can just talk back and forth. That text could be an invitation to an online meeting like this. So you could initiate a call, and then the call itself goes over the standard Https Internet, but the initiation of a call can take place over Lightning. So Lightning has that built in spam protection so that we can talk to each other, and we use it for our entire internal company process. So if teams out there want to try this, we've replaced. Originally, we use Skype and Slack and all the tools that everyone uses, and now we've replaced the entire thing, and we just use Sphinx for all of our work. We hook it into GitHub, we do all that stuff. So that's what Sphinx is to get to your question, which is the knowledge graph. So one of the things that we were part of in the early days is this whole podcasting thing. So you can consume podcast content through Sphinx and then stream Sats payments to your favorite podcasters. And there's tons of great apps that do that. Fountain, Breeze, and there's a whole list on the Podcasting 2.0 app site that you can go to to listen all stuff. Many of those have far outstripped the features that we have on Podcast. So why we started early, why did we pull back from developing that? So if I'm doing content consumption, I don't want just podcasts. I do not want to train the YouTube algorithm on what videos I watch, just like I don't want to train any central system on what podcast I'm listening to. So we would on Sphinx, spent the last six months supporting substantial medium YouTube, any video site, plus podcasts. And so we wanted to cover all of the different mediums and create a unified way to deal with all of these mediums. If we over index on podcasts alone, then the protocols and standards that we need to do, like Podcasting 2.0 is amazing. But the language and structure of Podcasting 2.0 might not fit with, say, Substack. And so if you want to apply value for value and Lightning to everything that you consume, we have to run the experiment of consuming other mediums, and you'll learn all kinds of crazy stuff. Like, instead of paying per minute of podcasts, you might want to pay per paragraph of Substack. And so all of these things have to be generalized into a standard protocol or a protocol that would allow you to pay for the content you want, voluntarily or required. And so there's lsats involved, and there's a bunch of pieces involved in this process that need to be flushed out before it can be 10X better than the existing experience out there. So you need all medium. That's number one. Number two, you need to have a standard way to compensate the person so that's Sats that's all the value for value payment stuff that works today in podcasting, but needs to work in other mediums it should work on this call we're having now. We should have video calls. You should have Twitch. All of those things can be done using the same model and needs to be integrated into this technology. Into Lightning, we need to Lightning file all of it. Okay. But now that you're consuming this content through Sphinx or fountain or wherever. So by removing yourself from a central algorithm, you also lose the recommendations that you need from that you're used to getting. So now when I open Sphinx with no recommendation engine, it's a crappy experience. I'm just having to go and find my favorite podcast. Look it up, Add. It feels like when you're used to things being fed to you, it feels very lame to suddenly have to pedal this bike yourself. So instead of training these central algorithms at YouTube, at Apple Podcasts, at Spotify, you actually want to train your own content engine on what your preferences are. So that's why we built this knowledge graph, and we're building it right now. That is a separate way to feed your own personal algorithm. It's a true bring your own algorithm model. And whether that runs on your own Lightning node or on your phone, that's okay. It doesn't really matter where that stuff runs. The important thing is you're training your own assistant and not someone else's.

Kevin Rooke:

Right. What do you think some of the applications will be that people use this knowledge graph structure for?

Paul Itoi:

It will be to start. We did it for our own purposes. So I was falling behind, as I mentioned before on Lightning, I couldn't even tell you what some of the stuff is. I mean, I know what trampoline routing does. If you ask me to explain what it is, I get it. Sort of. I'm going to mess up the details. So I want to be able to go to the knowledge graph and say search trampoline routing. And then right now, if I go to Google and search trampoline routing, everyone should do that. See what Google tells you trampoline routing is or AMP or Taro or whatever it is, you're going to get garbage back. Because we've left the job of content indexing, we feel like that's being done by Google. But Google just does one way linking of web pages, and no one is going out there and building new knowledge first in a web page. That's the last place knowledge ends up. It's being created on podcasts like this. It's Stefan Losera talking about multisig, or having guests on and talking about these topics. Guy Swan. I mean, all of these people are doing the work of pulling knowledge out of people's heads and then communicating it to their audiences over audio, which is way more efficient than reading about it in an article on a web page. So new knowledge is happening off of web pages, but it's not being indexed by the company that we thought was indexing everything and making it available to us. So what we're doing is actually taking, we've taken 1500 podcast episodes so far on Bitcoin alone, and we're adding three to 400 a week right now. So our rate is accelerating and we're going backwards in time. And so if I go to our knowledge graph, which the topics are almost like the constellations of the universe, they're all just floating around, linked to each other. I can search for the word trampoline routing, and then I can get six clips back that are all from different shows in different episodes on that topic. And then a virtual podcast is structured and I can hit play and listen to those six clips and get the latest information on trampoline routing from the audio content where it's being created in the first place.

Kevin Rooke:

Fascinating.

Paul Itoi:

So human knowledge is not happening on web pages. That's what's being indexed. And it's not searchable. So the new information where the good stuff is happening is actually the least searchable content out there. And if you take the Podcasting 2.0 chapter Tags, that's like the birth of this whole thing. And then you apply that to the YouTube timestamps. That's another version of this is where it gets back to protocols and standards and having these things searchable and machine readable. So having the topic in a Substack trampoline routing, should pull that for you and auto read it to you if you happen to be in the car. And then it should grab the fourth to 6th minute of a YouTube video that's time stamped on trampoline routing, and it should grab three podcast episodes that have talked about it just a couple of minutes that cover that topic. And then I listen to those six topics for 15 minutes, and now I'm up to speed on trampoline routing.

Kevin Rooke:

This sounds a lot like a new version of Google. Is that the right way to frame this?

Paul Itoi:

I think it's just doing the job in a modernized way that no one else is doing it, and then we're doing it again. Our pattern is do it for our own purposes. Now I'm listening to stuff through Sphinx. I consume most of my content through Sphinx now and then. I'm now training a model that is learning what I'm interested in, and it's working for me. It's going to suggest to me, oh, you just understood trampoline route. You should look at AMP. And I go, yeah, what's AMP? I should look at that. And I listen to those ten minutes. And that's also content discovery. I'm discovering your podcast or Stefan's podcast because I'm hearing that three minute snippet. And I go, you know, this is really good. I'm going to just stay on this one. And so what a better way to experience the topics you're interested in and your algorithms working for your education and your entertainment? Not what YouTube thinks you should watch next to stay on their YouTube app, right? That's like an incredible waste of time when we fall for the YouTube tricks and they're really good at it.

Kevin Rooke:

They are very good. Is there a way to tie in Lightning here and payments in the job of classifying information? Because that seems like it's one of those things again where maybe an AI engine can do that if it's a YouTube timestamp. But what happens if it's like unstructured or the data doesn't quite line up to the description?

Paul Itoi:

So we are hiring people right now over our best sources. Stacker News. So Keyan's site there where we got the initial people to do it. You posted stuff on Reddit and everyone you're banned as a spammer. If you say, hey, I'm going to pay you Bitcoin actually Stacker News, we post it there. We have great people who come on board. And so we're using the Stakwork model of pay Bitcoin for people to see this initial database. So it's already using our own tool to build this graph and then using Lightning to pay people to build the graph. But that model of an entity like us paying to build something doesn't work. So what we're working on, we would love input from people is gamifying the creation of this graph. So we just started, we've got the Bitcoin ones pretty huge right now and we'll hopefully move backwards in time. And our goal would be, is this the best way to learn and understand Bitcoin related topics? I don't think there's a great place to understand this right now to have people go, oh, you want to understand multisig will go to these sites and listen to these 15 episodes, listen to minutes, think about how you share information with people. It's incredibly inefficient. So we will need to gamify the creation of these graphs. We just started one on Austrian economics. So there is no the Nakamoto Institute is amazing for this, but there's no one capturing all the different podcasts and videos talking about Austrian economics. So we're moving to a cousin of Bitcoin into that topic. We also had an off site where teaching people to do fly fishing. So Sam, the amazing project manager on this, built this for fly fishing too. So if you've never been fly fishing before, we ingested hundreds of videos and podcasts on the topic of fly fishing and you can just search for Fly Fishing 101 and watch 20 minutes and have a reasonable start to the whole thing. So there is no tool I'm aware of today that does that for you.

Kevin Rooke:

And this could expand to everything over time, right? Is there a limitation on what kind of knowledge it could serve?

Paul Itoi:

No. And then each graph could be hosted by Lightning node and you could earn Bitcoin Sats for storing and serving this graph. And then to make it financially interesting to those nodes, to the creators into the people building the graph, you could even have. And this is where it gets very fuzzy. But a futures market for these ideas. So let's say you are against big blocks and you vote no on that and you don't believe it will be approved. You could stake some Bitcoin in the DLC or in Lightning and then actually earn money for being right. And then that money, the house, the graph takes 10% of the winning, just like any betting market out there. And then that incentivizes people to create the graph. And so this is a way to incentivize people to use the wisdom of the crowds to predict something. And then that prediction can be taken as information for someone who's trying to make up their own mind about a topic. CTV is a great one. Is CTV good or bad? Well, I don't even know what it is. How do I know what it is? Okay, now this market for betting, whether it'll be merged or not, could actually fund content creators who go, Whoa, five Bitcoin staked on this question. It's a hot topic. 10% of that is going to the people who create the content, who help educate the people placing bets on this. Whoa, okay, I'm going to make content for that. I'm going to tag things in the graph that are related to this topic. So again, creating the market dynamics that incentivize people to build it and use Lightning as the incentive engine to do it. Lots of people have been trying these things. There's probably six different projects I track. But what's the first thing they do? Oh, we're making our own token to incentivize people to build this graph. They pre-mine the token, they give some percentage of it to the people out there. They try to get the price to go up and then incentivize people. That's a house of cards. Even though the idea, I think, is great, we wouldn't be working on it.

Kevin Rooke:

So you're building a just so I have this clearly a self hosted graph of knowledge that can also then be reflected as a market for people to understand topics better. They can actually benefit from the creation of content and the decision making of whether they think something will happen or not.

Paul Itoi:

Absolutely. And then you can earn by being early and right. Earns you this amount of money early. Right. And contrarian gets you more money. And just like we have social media managers today, the job that no one thought of podcasters are a job that no one thought of. There might be people who are literally the graph maintainers, who make a living for maintaining and being correct on topics. And so being able to prove everyone says they knew the 2008 crash happened, very few people put their money where their mouth is. This allows you to be right and proves that you were right early by the bets that you place on being right. And so that to me, is a likely way to incentivize this. And you turn it into an ecosystem where people are building this, it's Wikipedia with the correct incentives.

Kevin Rooke:

And without that, and all of this can be built on Bitcoin and Lightning.

Paul Itoi:

Absolutely. I believe so.

Kevin Rooke:

It's incredible.

Paul Itoi:

We're doing it right now, but we've only done one or two topics. How many topics are there in the world? Billions. But if you can get it to work for one, 2, 5, 10 and it pays for itself through these incentive models, then it has the potential to be huge. But why are we doing this? Because I don't want to keep training a central algorithm on what I like that does not have my interests in mind. And so, again, we go back to that thing where our most limited resources, our time on this planet, we're trading that time for Tik Tok videos will be viewed as the most tragic waste of humanity. And it's okay to do that. I watch stupid stuff all the time, and that's the transaction I'm willing to do to rest and entertain myself for 20 minutes. Thank you, YouTube. I appreciate it. But do I want to sit there and then get suckered or tricked? As soon as I let my guard down, I have to use willpower to extract myself. So I would much rather have an engine that I trust and content that I pay for that's in my interest. And so pop in English Spanish lesson up in my feed. Pop up something that's on sale at the archery shop that is local, that the archery shop paid my fee. One dollars to show me a bow on some sale. Thank you. I'll definitely take that ad, and I appreciate that. And I'll give you the dollar back that you paid to have your ad show up on my feed. So those are the kinds of reversals of advertising that I think are possible with Lightning and where you're controlling what you see, good things happen. When you're training a third party, bad things happen. And that's as simple as that. You have to move off to your own algorithm so you get to control your identity, your algorithm, and your money under this model, true self sovereignty.

Kevin Rooke:

Right.

Paul Itoi:

And it starts with how you spend your time.

Kevin Rooke:

Right now, I know we're running out of time here, but I want to finish off with one last kind of question on what can be built on Bitcoin, because we've gone through a few cases here that I think listeners will recognize as things that are not widely talked about in the Bitcoin ecosystem. I think there's still this notion in the broader crypto space that you can't build on Bitcoin, that you can't do things on it, that it's just a store of value. But in the Bitcoin space, we have a saying I think I heard from Elise Colleen. First, it was, if it can be built on Bitcoin, it will be built on Bitcoin. So I want to ask you to finish off what's something interesting that's kind of in the back of your mind. Maybe it's not ready yet. Maybe it's a few years away. What do you think else can be built on Bitcoin that people haven't quite clued into yet?

Paul Itoi:

Mobile phone platforms. So we would love to play. That's our goal is, do you really believe in a hyper Bitcoin eyes world that your main computer device will be governed by one of two companies? It's not going to happen. We shouldn't let it happen.

Kevin Rooke:

What's your vision for making that build on Bitcoin?

Paul Itoi:

So first, it's a multi iterative process, but you use the Bitcoin payment system to gather the best people on OS and hardware in the world. And so you have a flash mob of capabilities that may be coming from within the big platforms today, but you won't know because they're pseudonymous and they're providing their experience and expertise to create a third option. And most people are going to stay in the environments for our foreseeable future in Apple and the Google Play Store. But do we really think that we're going to be keeping on paying 30% to those two companies for everything that happens? And you're begging for permission to add this feature, and you have to explain what Lightning is. We spent unbelievable amount of time and resources doing that, trying to get our app unstuck on the App Store. So no, we're not going to tolerate that. It's ridiculous. It's an impossibility. So Bitcoiners will build a true freedom device. It may be calyx, it may be generic hardware. It won't happen overnight. It'll be an incremental step to get there. And we want our Stak phone that we provide to people to be that freedom oriented device. We want the 15 year old to grow up without surveillance capitalism. And it starts with the phone.

Kevin Rooke:

So I love that. That's a great prediction, and I think it's going to open people's minds to what can be built on Bitcoin and the ways that we can use Bitcoin to incentivize and create new opportunities where none previously existed. Thank you so much for the time. This was a great hour to discuss what you're building with Stakwork and Sphinx. And I have a lot more questions, so I hope we can do this again sometime soon. But before you go, where can people go to learn more about you? The work you're doing, Stakwork, and Sphinx.

Paul Itoi:

Yeah, come on, try. Come to Sphinx.chat. We really limit the number of hosted nodes that we want to offer, so we make them available. But run your own node, grab a Raspberry pipeline, start nine umbrella any of those nodes, and run a node if you have a team. We were looking for two to five person teams that want to use Sphinx and put in place of Slack or Discord. So get off those centralized systems. Definitely get off Telegram. We have a Telegram group, but ditch that stuff. Stakwork.ai Stakwork.ai is our new website, and so come check that out. We would love to hear from anyone who works in a company that has these tasks that they think we would be good at. So reach out. We would love to hear from people who are Bitcoiners who also work in companies who need this type of work done. Those are the ideal people we love to talk to.

Kevin Rooke:

Awesome. Thanks again for the time and hope we can do it again soon.

Paul Itoi:

Great questions. I appreciate you talking to me.

Privacy Policy
Terms and Conditions