961 - Gamifying Social Status by Combining Game Design and AI w/ Harvey Pratt (Simcluster)

Mat Sherman (00:00.89)
All right, we are recording here with Harvey Pratt, who is the founder and CEO of SimCluster. Welcome to the show. How's it going?

Harvey Pratt (00:08.352)
Yeah, yeah, it's awesome. I'm calling in from Kansas where it's, you know, currently snowing, very, very cold. Thank you for having me. It's been a long time since I've a podcast. So yeah, looking forward to chatting. Love talking about what I do.

Mat Sherman (00:21.584)
Yeah, excited to dive into it. Before we dive into it, I have to ask, what brought you to Kansas?

Harvey Pratt (00:27.694)
Yeah, so we just raised a financing round and this was the first time I'd raised, I'm an engineer first and foremost, this is the first time I'd sort of gone spanners down for just a month running around yapping, talking to people and just wanted an excuse to lock in and just work. Now this round was done. One of our co-founders, Salim, lives in Kansas with his wife and kid, wife and kids, I should say, young family.

and it's a lot easier for the other two of us to come and see him. Plus you can get a stonking Airbnb in Kansas for not very much at all. Compared to Bay Area prices, this is extremely cheap.

Mat Sherman (01:07.9)
I that. I'm based in Phoenix and I feel like Phoenix is cheap compared to the Bay Area but I feel like Kansas is cheap compared to Phoenix so you probably get a great deal there.

Harvey Pratt (01:17.068)
Yeah, yeah, it's amazing.

Mat Sherman (01:20.038)
So let's dive into what you're working on. So you're building some cluster. Can you kind of give us a high level overview of what that is?

Harvey Pratt (01:28.77)
Sure, so SimCluster is a video game, it's a social network simulation video game, sometimes we call it a parasocial network, and really it started with this idea of why sort of four years after the first Transformer models, or really the studio name is the Promenade Studios and SimCluster is the name of our latest game, really this insight of like...

We had these Transformer models four years ago, five years ago, and I remember looking across when they first came out and thinking, damn, someone's gonna build some really, really cool video games with those. And that's kind of been the consensus for the past four or five years, there's gonna be these really cool video games pop out eventually. But half a decade later, this is beginning to look like crypto video games, this is beginning to look like AR, VR, where you just don't see mainstream video games come out with this.

breakthrough technology. So SimCluster was really our sort of idea of like, how do we push the sort of nascent AI video game space beyond the current paradigm, which is effectively just like character AI plus plus and sort of text RPGs. So, you know, really like when we were setting out to design SimCluster, the first game we did was called Promenade Isekai and that was a text RPG. And we learned an awful lot.

I'd say the most important thing we learned was that users hate typing. So really, we were like, how do we build a fun AI video game where AI is a novel mechanic, as opposed to doing an RPG and plugging AI into it? How do we build a video game which is uniquely facilitated by AI, is impossible without AI? And that really shaped the whole SimCluster design process.

So I'd say the most notable things about SimCluster is looks a hell of a lot like Twitter, and I believe we've built the first AI video game that you actually play without typing, which was a huge, huge thing for us. I think I'll leave that there for a moment, because there's all sorts of ways I could take that from here.

Mat Sherman (03:33.03)
There's a lot of questions I want to ask. think I'll ask the natural next one that pops in my head is it's a game that you can play without typing. It looks a lot like Twitter. Let's say I start playing the game or I want to play the game. Just walk me through what that looks like. What do I need to start playing and how does someone play SimCluster?

Harvey Pratt (03:55.222)
Well, first thing you have to do is join the waitlist at simcluster.ai. But as and when we let you off, we really had this idea of like, once you begin exploring this space of what is a uniquely AI facilitated game, it's actually kind of a nebulous problem. You what's a game you couldn't build before AI? And we really landed on this idea that, well, actually, for love nor money, pre-Transformer, you couldn't build a video game about a social network.

right, where would the people come from? You had this sort of hellish cold start problem and you would just end up building a social media network. So that was kind of the first jumping off point for us. We knew whatever we had to do had to be some kind of social simulation, right, had to be some kind of like multi-agent sim as a prerequisite.

and then really like we started thinking like what is fun about social media i have a pretty weird lens on social media where i think twitter is probably the most successful video game on the planet maybe tiktok and i think you know there's been this process across all consumer apps for the past like 15 years of like product managers going to their bosses and saying can i gamify it more can i gamify it more can i gamify it more you know how long do you do that before something's implicitly a game so we sort of like the way we approached this was

quite interesting. was like, let's reverse engineer Twitter as if it was a game, right? Figure out, you know, what do people actually want to do on Twitter? And it turns out Twitter, more than about having banger posts or more than being about accumulating followers, is really about earning clout, right? We sort of surmise that there's this invisible score across the whole site for each person called clout.

Which, know, I know roughly how much clout you have, know roughly how much clout I have, and what's special about clout is that every time you post you burn clout. Right? Burning more or less depending on what you're posting about and how you're doing it. But then if you perform well you earn clout back right through engagement. And that's kind of how it works. We started with this idea of like, well, what if we just rebuild Twitter?

Harvey Pratt (06:01.262)
we make it so that you can't type and you just, you create content across all these different mediums, images, text, music, video, 3D, right? Just by clicking these things called fragments, we could, that's fragments are a whole thing we can talk about in the moment. And then you earn clout through engagement. So I'd say like, if I were to like break down the site, like in terms of its DNA, it's probably like 50 % cookie clicker, 50 % Twitter, 40 % cookie clicker, and then maybe like 10 %

doxels, something like that. And the doxels mechanic being this letting people interact without typing.

So yeah, you join the site, you generate content, and then you earn clout based on that content. And I think one of the unique mechanics we have is, I always stress this to people, where we sit in market is not as a Twitter competitor, not as a social media competitor. I think that's a really foolish space to compete in. I'm just not going to win competing on the quality of feed. And that's been the biggest issue with all of the AI sort of social media sites. It's just scrolling through the feed and you're seeing a bunch of content from fictional people.

who you don't know and it's interesting for as long as it's novel and then you just leave.

Right, I really wanted to engage head on with this idea of what distinguishes content from slot. Why do people care about some AI content and not others? And I think the thing we're really trying to do is ground this to X. a way, one of the core mechanics on the site is if I generate a piece of content and I then export that content in any way to X, if I earn 100 likes on X, I then earn 1000 clout in-game.

Harvey Pratt (07:42.928)
So we're really trying to sort of... One way to think about what we're doing is we're trying to build a video game for people who play X like a video game anyway.

Mat Sherman (07:54.342)
So I am big ex user. I haven't used it, you know, probably seriously, maybe for four or five years. Been on it for maybe a decade. I don't know. And I agree with you. I agree that that's how people use it. One challenge is that I don't really, you for better or for worse, I don't care about burning clout. So I just like post all, just like post, post, post, post, post. And then sometimes it hits and I'm like, great, people follow me. This is wonderful. But it definitely is a clout game. But.

Harvey Pratt (08:17.72)
Yeah.

Mat Sherman (08:24.134)
The only way to play the game on Twitter is to input posts, whether text or images or videos. So help me understand, if you can't type on SimCluster, do you talk into a little microphone? Hey, make this kind of content? Do you got neural links and you think content? How do we create content on SimCluster?

Harvey Pratt (08:27.928)
Yes.

Harvey Pratt (08:32.503)
Mm-hmm.

Harvey Pratt (08:36.758)
no.

Harvey Pratt (08:40.718)
So when I say the site is 10 % Dark Souls, I don't know if you know those games, but they had this very interesting mechanic where as you're walking around the world, you can interact, you can see messages from other players, know, the world is full of traps and enemies and sort of messages from friendly ghosts saying, hey, don't go over there, or hey, that chest is a trap. But From Software didn't want to deal with the moderation problem.

Right, they didn't want to have like freeform messaging because that's a whole can of worms, right? So what they did was they came up with this system where they give you like a hundred nouns, 50 verbs, right? So many adjectives and then just enough to kind of string together communicative sentences, but not enough to sort of go off piece, you know, write a book or whatever. And that was kind of the co-inspiration for us. And we came up with this idea of the fragment system, which my claim is it's the first

gamified prompt engineering system in the world. I think people are very interested in this idea of constructing the perfect prompt, but I also think people hate to type. I think that's something we sort of overestimate in tech over and over again is exactly like, or underestimate in tech is like exactly how much we as an industry spend time hands on keyboard compared to like normal people, right? So in SimCluster, if I wanted to make a post about, I know about San Francisco, I find the San Francisco fragment.

which purchased from the store. I then add that to the post. I could then add a tone, happy, sad, angry, these are all fragments, right? Sam Altman's a fragment, you're a fragment, I'm a fragment, my user is a fragment. You have styles as fragments, anime, know, cyberpunk, whatever. And each of these fragments has an associated cost in clout to use it. So for example, if I'm posting about the weather, using the weather fragment, I'm not really burning much clout, right? But if I'm posting

about Jeffrey Epstein or something, right? That should be like a 50-clout fragment just to even use it, right? Because it's sort of high risk, high reward making content about Epstein. And that's kind of the system we were trying to model. So every post you see on the site, you just give these fragments and then, on the backend, we have prompts associated with each of these fragments. We sort of do the magic part.

Harvey Pratt (10:54.094)
of just combining all of these different ideas you want to pull into this content into something coherent and throwing it out. And that's something that AI is very, very good at, obviously. And really, suppose, we were really, really focused on this idea of when I began engaging with the slot problem, I realized there are really three types of AI content. And there's only three types of AI content. If you take nothing else from this, I hope you find this interesting.

Right? So like when stable diffusion came out, a bunch of startups, including myself, all had a very, very similar idea, which is like, crap, well now I can just generate assets, you know, up the wazoo, so to speak. And I'm just going to be AI Disney, right? I've just solved IP. I can make IP forever and ever and ever. And that didn't actually work for anybody two years later, that hasn't worked for a single person, a single startup. doesn't matter how much funding you cannot name an AI character. That's not a foundation model.

like Claude, right? It's like, what, you know, why is that? Well, it turns out when you engage with it, it's like, well, Mickey Mouse isn't popular because he's a, you know, cute little black and white mouse in booty shorts or whatever, right? He's popular because he's a hundred years old. Like successive generations have grown up from him and you can't the same way. It's very hard to actually like convert cash into fame or profile, like in a durable way.

it's extremely hard to just like spend money or spawn concepts and then just get durable IP. In fact, I'm not sure it can be done. I think the key ingredient is literally just time and exposure. So when I talk about these like three categories, these three categories of content, what you have is you have category zero, say, there's four categories of content, Category zero is literally just slop, right? It's,

AI nonsense that you don't care about, people you don't care about, you know, whatever. Like the shit you just scroll past on your feed instantly. there's content category one is then content which is technically or artistically novel. Right? So think about when Suno first released and there were all these songs all over the feed or when Mid Journey comes out, just all the images on the feed, because no one's seen this before. It's crazy. The QR codes are an example of like

Harvey Pratt (13:14.922)
artistically novel but not technically novel. People knew you could generate images, they didn't know we could do these beautiful QR codes. That's not a great space to compete in, what's the issue with being there? It's actually really expensive to technically innovate and it's really hard to artistically innovate. So you don't see a lot of that stuff compared to category two, where category two is content which is about people or ideas people already know.

like a deep fake meme of Joe Biden and Donald Trump hanging out or whatever. do comparatively well. Images of Sam Altman or Jensen Huang, right? These do comparatively well. Right? And then category three is kind of the like Moby Dick of the space. It's what everyone's chasing in other words, I think. If not explicitly, they're sort of chasing implicitly. And it's really easy. It's just category one plus category two. Right? So it's like technically and or artistically novel and it's about someone you know.

and it turns out that's the shit people really, that's the good shit. That's what people really, really like and they're chasing. You what's an example of this? My belief is that the most popular piece of AI content ever is the video of Sam Bankman freed dancing at that concert. You saw this? know, Vigil had just come out, you so there's your category one. And then Sam Bankman freed is the last person who's going to be dancing in a concert and everyone knows him.

Mat Sherman (14:31.867)
Of

Harvey Pratt (14:41.964)
Right, that's your category two equals category three. Another example for people who maybe don't know that video is the Will Smith eating spaghetti thing from way back in the day. Right, and that took over the internet. And so sort of an interesting way to frame what I'm doing on SimCluster, or I should say what we're doing on SimCluster, is like the other thing you notice about category three content is normal people aren't producing it.

but who's making this, right? It's people who know how to call an API, who know what an HTTP request is, who probably know some JavaScript, whatever, right? It's not empowering actual creatives or artists. It's not empowering people who want to play creative video games, right? So really, I see our job with SimCluster is like your fragments are your category, your fragments are your category too.

Like we just say, okay, there's all of these concepts you're already familiar with. You can create fragments, you can tag other people. Everything is guaranteed to be relevant to somebody. Otherwise it wouldn't be in the game. And then your category one comes from the fact that we just add new mediums to the game all the time. So it's like, okay, you want to do music? Great, we'll just add music. You want to do Vigil, right? You want to do lip sync, lyric videos, whatever. Great. We just add those. You don't have to worry about it. And then you just get to throw your fragments into these new mediums.

And then suddenly you're letting people create this quite expressive content, which they would have to be technical otherwise to create. And you're building a gameplay system where they are actually getting, you know, what's the core sort of in what's the meta game of SimCluster, right? We've actually already seen some players who by playing SimCluster well are earning profile back on Twitter, which is super, super cool to see, right? Because we're trying to sort of bond these sites quite closely.

Mat Sherman (16:29.658)
I feel like you need to have a very interesting brain to have come up with this idea. I feel like this is a very original concept. This is not like a, I saw this company and we're going to make an iteration of it. This is a completely original idea, at least from my perspective. So what were you doing before SimCluster? Like how were you able to...

Harvey Pratt (16:50.914)
Yes.

Mat Sherman (16:52.636)
come up with this or co-found this kind of concept, seemingly out of thin air, like walk me through the origin of the story.

Harvey Pratt (16:59.79)
Yeah, the shorter answer is I was doing everything wrong before I started doing a few things right. Right, that tends to be how things go. But I know I'm from the UK, always had a foot in the door in the US, my dad was a contract...

know, these rounds, is contract engineer with your military. And so spent and spent sort of two or three months in and around the US growing up every year. And so always kind of fascinated with the place came out here for college when I was 19, dropped out of college over COVID. And funny enough, I was doing a music degree at the time weirdly, dropped out of college over COVID.

Because I wanted to figure out I just saw DeFi blowing up in crypto and I was like I can make a couple million dollars like every other sort of 21 year old in tech that summer. Whatever, it didn't happen, but that's what I was thinking. But got very very lucky, you know Twitter changed my life where I got super burned out on that crypto project. I was living on my mum's couch, you know, just total burnout failure, you know, and I saw a tweet from Balaji Srinivasan.

He was like, hey, there's this note taking app called Roam Research. And he's like, oh, hey, somebody should build name density recognition into Roam Research. I was like, oh, great. So I did it. And I sent him the GitHub repo and went to bed, woke up the next day and he'd retweeted it. It had gone gigaviral. At least it felt gigaviral to me at the time. Had venture capitalists in my inbox, had CEOs of companies I liked in my inbox. Hey, this is pretty cool. Do want to talk?

And then one of the guys I had in my inbox was a co-founder of my first business, a guy called Igor Lenterman, who sort of came to me with the pitch of like, hey, I saw the same tweet, I tried the same thing, but like, you your code is better, but I have a lead on some capital and here's what I think we can do with this as a business. And so we, it was this sort of very funny fundraising cycle at the tail end of 2020.

Harvey Pratt (18:58.37)
where so far as could tell, a lot of VCs were sitting around bored like the rest of us, except they were all on this new app called Clubhouse. Right? And so like, they would do this thing where they'd pull like young people up on stage, have them do pitch competitions, then they'd just like give them money, and then they'd make you check in every evening, effectively as a form of like reality TV. So I went from like burnout in the UK, and I was a non Twitter account at the time too, right? To meeting a total stranger in a different country and having a million dollars in a bank account with him.

No computer science degree, was CTO at that business. About the luckiest start you can get in tech. God knows, we ran with it, but I'm very, very lucky. And the original pitch of the business was actually we were pitching a company called Atomic Search, which was meant to be a sort of AI search engine. The sort of example we used at the time was...

Wouldn't it be cool if you could ask like who Michelangelo grew up with and I that returned as a JSON or whatever and it just fits it out. And that was like crazy future tech at the time. We thought we were going to kill Google. Right. And we weren't those guys. Like we, we were not like the, you know, we were not the all star technical team who was going to take on open AI, but we sort of grinds on that for a year. And I think that's why I cut my chops on like a lot of early AI stuff. Lots of like symbolic approaches to not just the neural stuff.

And then really after about a year, started applying some of the crypto, some of AI stuff we were learning to crypto. And sort of progressively over successive pivots, founded an AI search business, ended up selling a consumer crypto business. And we ended up selling that to, sorry, is the recording? Pause. Oh, okay, sorry. did, yeah. Sorry, it froze out for a moment. Yeah, ended up selling that business to Gabe Layden.

Mat Sherman (20:41.135)
No, no, you're good.

Harvey Pratt (20:46.754)
business called limit break. you know, Gabe was an early investor in our business, but he's also just a legend in the games industry. He found a company called machine zone. Where if you ever saw an ad for like game of war fire age with like Kate Upton or Arnie Schwarzenegger, and like the 2010s, that was him. Right. Just really, really sort of invented the modern free to play games industry. And it was through like a lot of chats with him. You know, I was like, after the end of that, like,

We'd sold the business. was pretty, I was looking across at this. This is like, we're now in 2023. And like, I was, like I said, I like when we were founding and the first Transformers had come out, I'd looked across at the AI game space and I'd seen like AI dungeon and whatever. And I was like, Oh, that looks really cool. It's going to be some crazy shit to come out. And imagine my surprise, like I stick my head up three years later and kind of check into AI video games. Hey, what's going on? And it's still like AI dungeon.

We've had all of these technical advancements and there's just nothing new. I really, digging more and more into the space, I wanted to work in the space because I grew up loving video games. I know a lot of boys grow up loving video games, but I was always the weird one, think, who loved them a little bit more than everybody else. I really just had played everything.

very strong opinions on games and I, you know, like, unfortunately I'm a sort of joke, unfortunately I'm an artist, you know, and I like the idea of making beautiful things which other people enjoy. And, you know, once I got thinking about it, if you want to make a really impactful piece of art, I think it probably has to be a video game. You know, I think if there's like, I think like...

what pieces of art which are made in current year will be sort of enjoyed in a thousand years time. Same way there are some paintings, right, some writing we enjoy in a thousand years. Yeah, there might be paintings, there might be writing, though less and less each year that goes on. I think video games are probably the final form of art in some sense, insofar as like you can think of them as like an artistic superset, where like a video game can fully describe the richness of a painting. I could put Van Gogh's Starry Night.

Harvey Pratt (23:07.05)
in an arbitrarily beautiful rendering engine and walk around it in 3D, and at a certain point it would be functionally indistinguishable from the real thing. You could not take that same simulation and render it with 100 % fidelity by painting it, or by writing about it. That's what I mean by it being a superset. The fact that we call them video games is definitely a misnomer.

like the medium's 50 years old, right? And that more speaks to their position in market, I think, than anything else. And really the more formal technical term would probably be software mediated art, maybe just simulation, right? But I think if we called them that, we'd kind of understand how important they are that they're here to stay. And I think like, so that's kind of the background on me. And then where SimCluster came from was like, we joined the space and we started making RPG like everybody else, but like...

There were so many non-intuitive things in the space. Well, I think the biggest thing I learned is that Silicon Valley very rarely produces great video games. And you just have all of these breakthrough technologies. You have AR, VR, AI, GPS, crypto. Which of those, the only breakthrough game I can name with any of them is probably Pokemon Go, maybe Ingress. And it's like, why is that?

and kind of digging deeper and deeper and also why do consumers hate this stuff? Right, there's a lot like if you go on TikTok and you try and market your AI video game or you go on YouTube why AI sucks in video games people hate this stuff absolutely and why? It's not to put too fine a point on it but all the games coming out are kind of lazy and I just found myself in this weird situation where like

working in this industry, just felt constantly like I was the only person who actually kind of liked video games. Like there's this real sort of users are losers mentality in games in Silicon Valley, it's like, know, like maximally extractive revenue models, right? And like these actually work for proven formulas of games, which scale, but this is why...

Harvey Pratt (25:27.754)
This is why no one's really done interesting things with like, well, very few people have done interesting things specifically with AI games formulas, like at the minute. And it's also because the classic games industry does not want to touch these, right? Like games devs and like industry devs. think of these, I think of them as like two separate kingdoms almost. Well, like if you're a young person and you want to become a software engineer, there's almost this sort of hard fork you take in the road. Once you graduate where it's like, Hey,

Am I gonna go and become an industry dev and polish buttons for a fan company and collect 250 grand a year, you know, TC, but time on 25, right? Or am I gonna go and work for EA and have to do crunch and get paid 100 grand, right? But I get to work on shit that people know, right? And it's like totally different tool chains, totally different languages in some cases, different like games are not built like regular software products.

and that's kind of something which Silicon Valley is only really beginning to realize now. Right. And so I think like, I'm flattered that you would say this is like a novel idea and that you'd say nice things about me for having made it. I think like, there's just comparatively few people here and I would just plea if anyone is listening to this and has the slightest interest for like a novel AI video game.

Please just do it. Now is the best time in history to make your like crazy stupid video game. It's not crazy. It's not stupid. Come and talk to me want to help you right But yeah, like and then sim cluster really was just thinking about Why hasn't this been working? You know how and then really engaging with the game design and really starting with this idea? They're like what is a game that you cannot build without AI because if I want to take on a multi-billion dollar studio

Right, it's like the median pitch in games seems to be something like, I'm gonna do the Sims with AI, or I'm gonna do Skyrim with AI. And I think what everyone in Silicon Valley forgets about doing Skyrim with AI is that you do have to be able to build Skyrim 2. Right, it's like, and there is a quality bar which the median consumer expects, right, and they're not playing with your game because it's an experimental technology. They don't give a fuck, right? Right, they just want it like...

Harvey Pratt (27:52.556)
That's just not a selling point, and I think that's where people go wrong in the space. There's this sort of Silicon Valley arrogance, especially with AI, of like, well, AI was the hard part, and games are the easy part, so we don't have to worry about game design. Right, and I think games, take so long to ship as well, that it's very, very easy to kid yourself for a very long time that things are working when they're not.

which is how you end up with an industry that's kind of like, or like a space that's kind of like a graveyard even five years in. I could go on and on, but I hope that's given you some sort of background or insight there.

Mat Sherman (28:30.52)
absolutely. That was was appreciate you sharing all of that. Going back into the gameplay, so I want to outline what my understanding of how it works and fill in the gaps. So if I would spend money or spend clout on fragments, that means I get clout in the first. Do you get clout simply for joining on day one or how do you accumulate clout from on day one that you have to purchase fragments?

Harvey Pratt (28:57.059)
Yeah, are a few ways to accumulate clout in the game. The number one most obvious one is through engagement. So every time someone likes your, I think about, know, what do you do when your feed is mostly sort of AI content, right? And like people aren't necessarily scrolling your feed for engagement. This is where the cookie clicker thing comes in. Scrolling, like liking a comment, liking something on SimCluster is like clicking the cookie in cookie clicker, right? So every time someone likes your content, you earn some clout.

You can then spend that clout on various upgrades and perks. So if I buy the self-love perk, now when I like my own posts, I get two clout. So fun stuff like that, right? And so you can really like, we took a lot of inspiration from Bellatro to, to big popular sort of poker rogue-like video game this year, where the idea is like, your, skill expression in the game really comes from, well, can I buy these perks and can I post at a certain time? Can I use certain fragments so that when someone likes my post,

actually I'm earning 10 clout per like or I'm earning 25 clout per like right just for how I've set this up. You earn clout for completing challenges in the game so you know like make 10 posts, like 10 posts, things like this those are really sort of that's to get you your first sort of clout wallet to start spending. You earn clout when you're tagged in things so like you know I talking about content cool things we can do on sim cluster

in terms of content, still think we're one of the very few sites on the internet, I can't think of any others actually, where like if you joined, I could grab your fragment and my fragment and I could throw them in a post and I get a photo of us both, right? And then I can make that into an anime style or something like this. And because I've tagged you, you're earning 10 clout every time you're getting tagged, right? Then, you know, music, I think we're the first.

Mat Sherman (30:41.03)
Well, that's cool because it kind of resembles the real world. mean, if you're buddies with, yeah, it's social network, right? Yeah, it makes a lot of sense.

Harvey Pratt (30:44.686)
Well, that's the goal.

Harvey Pratt (30:48.494)
Well, we've also done something cool on the music for I could talk about AI music for a long time, but with the first with the first game in the world, I think to hit these two criteria, which is we are the first game in the world with a fully user created and contributed soundtrack in the all of our soundtrack to the entire game is made at runtime by players and goes into our soundtrack, which is awesome. And also the first game in the world, obviously, with a fully AI generated soundtrack. And how it works is, you know, like

there's always music playing and it's always made by other players through fragments. But then if I listen to a song all the way through, they earn royalties in clout. If I like a song, they earn royalties and they can put all sorts of bonuses and perks on that. If I save a song, actually buy it off them. Right? And if I want to add a song to my playlist, I spend two clout, they then get that two clout. Right? So it's really just building this economy where like, you know, thinking like

If I go on SimColustre and I go on the leaderboard and I see that somebody has 20,000 clout, what is that meant to represent? Well, you know, if that number was visible on Twitter, what would it represent? Right? It means you're either very, very good at gaming the algorithm, right? You figured out some like MLG strategy that's working for you, totally valid. It means you're high profile and people are just engaging with you a lot anyway, right? And it's basically only one of those two things. And that's really what we're trying to model.

with every gameplay mechanic we build.

Mat Sherman (32:18.108)
When you were giving me the origin of the story and what led up to all of this, you kind of took, I don't know if you'll call this, but you had some critiques maybe on Silicon Valley on how they think about games and maybe disrespecting the process of building games and what goes into that. What was it like raising money from those people? Were you able to share your experience on was it easy, was it hard, was it painful, was it easy? I don't know.

Harvey Pratt (32:44.366)
The short answer is, like everybody, will share cautiously, so to say, because I'm going to have to raise money at some point in the future. Venture scale investment specifically into video games is a relatively new phenomenon in Silicon Valley. The predominant financing model for video games has traditionally been the publisher model, where you get an advance, then you ship your game, and then they help you market it.

Mat Sherman (32:55.398)
course.

Harvey Pratt (33:14.144)
and then they take some cut of lifetime revenue. And that's worked very, well for the duration of the industry. I think the VC model is quite different. Well, obviously you're expecting a venture scale return. You're expecting billions and billions and billions of dollars out of the game. And it's quite interesting if you chart all of these cult classic games people love, Deus Ex, Elden Ring, things like this.

they would actually be terrible VC investments, right? In terms of their annualized revenue since launch, these would be considered failures in the average portfolio. So there's already kind of a disconnect there. so already it's a certain type of game that's incentivized to get produced, and that's like live service, anything with a subscription plan, like MMOs, anything with high-frequency in-game payments. This is why they all got so excited about crypto video games too.

I would say what I'm seeing, the process of raising money has been, it's like it always is, it's very very hard until it's easy. You know, and really I think my thinking around this has mainly changed from...

I think a big unlock for me was when I just started turning up to people, identifying as an artist, not trying to do the business plan pitch and just saying, I'm looking for champions. I'm looking for people who love this space and love what I'm selling, so to speak, and love the game. With SimColustro, it was awesome because we ran a failed fundraise earlier this year because that was one of the shocking things of like...

getting into AI video games where everybody agreed the potential of the space was huge. And then I got into it sort of four years late and there's this whole swathe of like dead, invisible dead companies before me where they had their five, $6 million seed round in like 2022 or something like that. And then, you know, running around for nearly a year and just hearing over and over again, like you're a smart chap, like you're well-considered about the space. We like the stuff you build. We wish to God you were working on something else.

Harvey Pratt (35:29.198)
Right? Like over and over for a year. think VCs have rightly learned to be very, very cautious of anything in the AI game space. And I think this is broadly true of AI consumer as well. Like the chat app founders are finding this as well now, right? It's not as easy as just spinning up a character AI clone. It's not as easy as just making Skyrim with AI. So I think like really the unlock for me has been

So yeah, just turning up and saying I'm focused on building a video game first and foremost, like it's it's beautiful. Like there's a UI making sure it demos extremely well, like turning up and demoing first thing, making the site as as legible as possible and really explaining my thinking. But yeah, I'll have to get back to you on that one as I keep doing it. It sucked until it didn't. We were extremely lucky early on to meet A16C Speedrun. Those guys were great.

We were extremely lucky to meet Siam Banister, Dan Romero, the guys from Social Graph, all absolutely awesome people. Ralph Reichert, people who've sort of taken bets more on the process of my thinking in this space than sort of the baggage that has come out of this space historically.

Mat Sherman (36:55.036)
Cool, I realize we're at time. Do you have time for one more question before we call it a day? My question is, feel like, I'm curious on your long game vision here. from what you're saying...

Harvey Pratt (36:58.68)
Yeah. Yes, I think.

Mat Sherman (37:08.924)
It's like cool, it's like a social network. There's a lot of different components. You kind of get gamified cloud in a way. You implement AI everywhere. You've thought through a lot of really creative elements. But is it gonna look like this in a decade? Or in other words, where are you rowing to here? And what's the final vision here that you're trying to build?

Harvey Pratt (37:25.346)
No. The big, there's two big trends I've identified which sort of govern three, let's say, which govern how this is built. And I'll try and get through them quickly. The first thing you see is if you look at the most popular games in the world right now, which I would say is indisputably Minecraft, Fortnite and Roblox, all of them are doing the same thing, which is they are basically spawning like proto social networks inside of themselves.

So you look at Generation Alpha, it's extremely common that Fortnite is effectively used as a social network. Where they log into lobby, they just message, and that rings true to me because I did the same thing with World of Warcraft when I was 10 or whatever. So you have this process where the games people want social networks anyway, which makes a lot of sense when you have social networks of these hugely retentive apps and as a game you live or die by retention, so if you can roll it in, great.

Right. You also have this process. So like in terms of the SimCluster build, it's like, well, actually, like, why don't we just make the game the social network? That seems like a pretty safe bet to me. Right. Right. Rather than like just build Fortnite, but drop the shooter. Right. And like if you'll end product, if you'll end, if you're metastasizing into a social network, we'll just start there. Right. And build that. So that's kind of the product. And then if you look at IP, one of the things you notice in IP is like,

Everyone's converging on the same model for IP, which is the multiverse model in some sense So like Marvel just spent over a decade and billions of dollars educating audiences on what a multiverse is and it is like that and that's kind of like fine like We all get what the concept is, but you go back 10 years. That's pretty high concept sci-fi stuff Right it and it really was once you look at it as a public education effort. Why did they do that? Why would you do that? Yeah, and the answer is pretty simple. It's because you ship

Deadpool vs Wolverine or Deadpool and Wolverine and then suddenly you get to have your Gen X original X-Men fans or your millennial original X-Men fans sat side by side with your like Gen Alpha Fortnite players who know Deadpool from Fortnite and you can cross pollinate those audiences and you can sell IP right and I think like the same you'll probably see a similar sort of phenomenon in social media but we'll end up with something like a social media

Harvey Pratt (39:50.252)
multiverse, think AI kills platform lock-in, IP across social media. And then the final part I see is like, know, payments, how's this make money, whatever. Something I think about a lot is that creators sort of got one, two, three combo by tech, then crypto, then AI. So if you're like a musician, right. If you're working musician, Spotify killed you, right. There's just nothing you can do. And then crypto said it was going to fix it. And then they fleeced you.

and an AI stole from you. Like you're not particularly happy. But you know, what did they get in return? Well, they got access to viral distributional upside. So something I think about a lot is if like I'm a 19 year old in Berlin and I go viral in my bedroom with a song I wrote, I'm seen by more people than the Beatles at peak in like an afternoon. And that's, that's insane. Like the numbers are really difficult to comprehend. But what you notice is that people don't quite make money. Like the ways people make money off of that is very second order.

You know, it's like maybe maybe they get lucky and a record label exec slides into their DMs. Maybe they're smart or thrifty. They set up a store or merchandising or something. I think like if you're building a game about morality, you know, my my insight on the crypto side and how you build a crypto game is the exact same as how you build the AI game. It's you have to build a game that is uniquely facilitated by crypto, right? It's like AI Farmville won't crypto Farmville won't cut it.

crypto magic gathering won't cut it right these things work fine without crypto but what I do see is I think there is similar consumer demand for viral financial upside that there is as there is in like viral distributional upside I think meme coins are sort of a beginning of people addressing this and so I think in the limit for us I think one way you could think about sim cluster is what I'd like it to be is you could think about it as if steam was an MMO

at the same time, so you're playing the games as you're making them and you're distributing them, and then you're monetizing natively on the platform by going viral. And I think if you can nail that, you've killed Spotify, you've killed Steam, you've probably killed Coinbase. So that's the opportunity I see. I think all media is basically ending up in the same place in video games.

Mat Sherman (42:14.192)
That's fascinating. I feel I could ask 100 questions about this, but I'll save it for podcast number two in a year or two. Where can someone find a SIM cluster and how can they get on the wait list?

Harvey Pratt (42:24.802)
Yeah, go to simcluster.ai, sign up to the waitlist. You can sign up with your Google account, your Discord account, your Twitter account, or your email. Yeah, come as you are.

Mat Sherman (42:37.308)
Cool. Well, thanks so much for coming out of the podcast and I wish you best of luck building this out.

Harvey Pratt (42:41.592)
Thank very much for having me. Cheers, man.