SPEAKER_00: The feedback I'm getting from the users themselves is they find this technology is just augmenting them. It's augmenting their creativity. It's making them go faster, further away. So they love it. And if you are, you know, have you seen that research saying if you are, the smarter you are, the better result you get out of ChatGPT. I think the same applies for images. The more creative you are, then you're going to get an even higher level of creativity. And yes, you can pretend being an artist, even though you're not one, you're not trained as a game artist. You can pretend being one with this or you can approach being one with this.

SPEAKER_01: But if you're a good game artist, you're going to be an even better game artist with AI.

SPEAKER_02: This Week in Startups is brought to you by Vanta. Compliance and security shouldn't be a deal breaker for startups to win new business. Vanta makes it easy for companies to get a SOC 2 report fast. Twist listeners can get $1,000 off for a limited time at vanta.com slash twist. House of Macadamias is the next big health trend. Get 20% off your first purchase and a free bottle of cold pressed macadamia oil at houseofmacadamias.com slash twist by using code twist20. And Crowdbotics. Great ideas can change the world. And Crowdbotics is the fastest way to turn those ideas into code. Get a free scoping session for your next big app idea at crowdbotics.com slash twist.

SPEAKER_03: All right, everybody.

SPEAKER_05: Welcome back to This Week in Startups. I've never been more excited about a technology, at least since the internet and the PC, than I am with artificial intelligence and the pace at which it's evolving.

Chamath Palihapitiya: And here at This Week in Startups, we are all over this trend. You can join our sub stack. You can subscribe to the podcast. Go to the YouTube channel, youtube.com slash this weekend. And you're going to see a constant stream every week. You're going to see five or ten of these amazing demos of what is being built in the AI space. Scenario is the startup today. They're using generative AI to create and produce creative assets for the video game industry. And the CEO and co-founder is Emmanuel Di Mastre. Welcome to the show, Emmanuel.

SPEAKER_08: Hey, Jason. Thanks. Glad to be here.

Chamath Palihapitiya: All right. So, um, we know the video game industry. We know the assets take a long time and are very expensive to produce. Tier one, where they call them tier one games or title one games. Uh, those are the ones that cost. Tripple games. What do they call them? Tripple. Tripple. Tripple. So, triple A games, those are hundreds of millions of dollars. That's Call of Duty. And then you have apps on your phone that might cost a million dollars to make or five million dollars to make. What are those Clash of Clans, Angry Birds games tend to cost? Do you know what those costs?

SPEAKER_10: Could be from like a few hundred K to dozens of millions.

SPEAKER_13: Actually, if you look at even like Angry Birds and so on your phone, it's pretty expensive. And it's not just the cost. It's, uh, it's, it takes a while and it's, it's risky. And those are double A games or what do they call those? Uh, double A games are typically PC games, console games, but not like Call of Duty, uh, level.

SPEAKER_10: God.

SPEAKER_17: Then you, mobile is a category, uh, part. And then you have all, all the new web three games wave. Ah, yeah. Which is new.

SPEAKER_18: So at a minimum, hundreds of thousands of dollars on average, millions of dollars, and sometimes even tens of millions to hundreds of millions of dollars.

David Friedberg: And a big part of that is creating these creative assets.

Chamath Palihapitiya: There are thousands of designers, you know, uh, working on some of these games. I understand, or at least hundreds of them working on a single title. Is that correct?

SPEAKER_24: Designers, um, game designers, product designers, art directors, technical directors, technical artists.

SPEAKER_13: It's a whole crew it's, uh, yeah, it's big.

Chamath Palihapitiya: It's big. Okay. And so you're going to use generative AI, which is to generate images, um, through typically text prompts. I'm guessing, but you tell me, and let's just do a demo here. I know you have an incredible demo, so let's see it.

SPEAKER_26: I definitely can switch to a demo. Let's, uh, share my screen.

Chamath Palihapitiya: Always great show.

SPEAKER_27: Don't tell. Okay. So here we go. Here we go. I see you're in an editor of some kind.

SPEAKER_00: I'm, I'm the web, uh, it's a web, uh, it's a web browser. It's Chrome, uh, and the website is app.scenario.com. Pretty easy. Um, what you see here is my, my homepage with some of the images I was making just before the, before the show, uh, definitely like, like gaming related. I'm like trying to, uh, you know, iterate around some concept around a treasure chest that I'm trying to produce and, um, look how consistent these, uh, images are, uh, in shapes and proportions and only the color is being, is being changed. And we are not in Photoshop at all. I'm prompting the color. I'm prompting the shape and I'm prompting the, um, the proportions of the image.

SPEAKER_30: Show us how that works. Yeah. So you're showing like, uh, what they think they call in the industry, a loot box or a treasure chest.

SPEAKER_05: You might find this treasure chest in a game. Uh, how did you prompt it? What did you type in to get that unique looking treasure test?

SPEAKER_00: I was, I'm gonna, I'm gonna show you, but before we do so, please, please understand these treasure chest came from what we call a generator, uh, which is a custom trained AI

SPEAKER_13: model with people's data. So instead of using a very horizontal model, such as mid journey, Dali, stable fusion, we do let people make their own custom model with their data. And that's what they love because when they mean when they say, so if I'm call of duty

SPEAKER_05: and I've, I'm on the fifth or sixth call of duty in the series where I'm angry birds making or plants versus zombies, I could upload all the artwork in every previous iteration of the game and be starting on third base.

SPEAKER_00: You could upload all of it, or you could pick the, the, the, the certain portion of the games that you want to train a model on such as characters, backgrounds, vehicles, and so on.

SPEAKER_24: So in that case, we have chest, you know, what, what I can do before, uh, going into the chest one, I can show you another, even more relevant generator, uh, which, uh, which is something custom we've made for a demo, which is called bubble verse.

SPEAKER_00: So the bubble verse is a potential game made with that art. So look, look at the rendering, the, the colors, we're looking at a bunch of beautiful 3d

Chamath Palihapitiya: rendered, you know, looks almost like, uh, forgot the name of that, uh, famous director, Wes Anderson. These are Wes Anderson, beautiful colors, slightly Japanese inspired, perhaps, uh, beautiful, um, 3d models of bubbly type characters. And, uh, cars and vehicles and trees and homes and everything. So you only needed 25 images to upload into here to generate this. These might've been done by a human who made a collection of 25 bubbly images.

SPEAKER_24: Absolutely. And now let's start prompting, let's start creating from the same style.

SPEAKER_13: And so Jason, give me an idea of a, let's say a vehicle on it.

SPEAKER_00: Say maybe a, a, a background, a hamburger. Oh, let's do hamburger.

SPEAKER_40: So let's do hamburger. Yeah. Hamburger.

SPEAKER_00: Well, that's very easy. So you see, I just typed hamburger on the, on the left, uh, and now it generates. We're going to wait probably between 10 to 20 seconds for the images to, to generate its GPU based cloud-based, but why do we do so when I, what should the hamburger look like?

SPEAKER_24: It's pretty basic, right?

SPEAKER_43: Um, let's make the hamburger, um, let's make the hamburger be, uh, um, uh, on a bed of,

SPEAKER_05: uh, flowers and, um, make it the color, uh, green.

SPEAKER_00: Cool. So here is a green hamburger on the bed of flowers. So I'm just prompting generates, but before, yeah, before we, we wait for your green hamburger to show up, uh, let's look at your first, uh, session, your first, uh, so that's the first one, second one, third one. Hey people that, oh, by the way, this is like a character mixed with a hamburger, which, which happens sometimes.

SPEAKER_37: So for people who are listening, for people who are listening, what you're seeing is

SPEAKER_05: instantly something that I would say is game ready. Um, just in my professional opinion, if I was playing a casual game and it threw up a hamburger, any of these hamburgers would be acceptable to me as a legitimate item and

SPEAKER_50: it was created for free for now.

SPEAKER_00: Um, and look at the, the, the style of the hamburger is definitely, it's not photorealistic. It is not voxelized. It is in the style of your, of, of the person generator. And so I did not need to go into describing all the visual feature.

SPEAKER_52: Have you seen prompt engineering? Like people, like 50 words and it's super complex. Yes.

David Friedberg: You didn't have to say, do this in an Wes Anderson style with a color palette that includes these colors that is, uh, you know, uh, got depth and shadows or whatever.

Chamath Palihapitiya: And then we did green hamburger and here are the green hamburgers and they subtly made the bun green and they put some lettuce around it and all kinds of fun options that it just

SPEAKER_05: decided to make.

SPEAKER_13: Yeah. And so in your case, we do have your green hamburger.

SPEAKER_00: We do not have the bed of flowers though, which, which again could be expected. So on the, again, on the, on the left side of the screen for those who are on the, on the, on the audio, we have a bunch of settings and parameters to play with. Okay. And if I increase the guidance a bit and run another generation, it should emphasize, it should, um, increase the weights of the prompt. And so it's all about giving artists, mostly artists, game artists, giving artists the right tools, the right levers so they can control the AI and stop randomly prompting forever and ever on this code and, and, and some, uh, in some, some platforms. We, the, the game studio we work with, they want to get their image in the least amount of time. They want to have, they want it to be very efficient production has to be efficient. And so that's what scenario is about that, like stop, uh, exploring AI forever, train

SPEAKER_13: your own model that will drive consistency and then play with the settings. And then you, you nail your, uh, designs and so on. There we go.

SPEAKER_18: And what's great about this is you're really augmenting an artist's work.

Chamath Palihapitiya: So instead of an artist having to tediously make, let's call it 500 objects for this game,

SPEAKER_05: this bubble game, and then every week updated with, let's say they come up with 10 new characters, devices, or 50, that would be 50 times, 50 weeks, uh, 2,500 more. So you'd have 3000 objects in this game. Um, you could have an artist just make, um, 25 test images and then somebody else, a game designer can just with their judgment, do everything from there. So you really just need, uh, uh, somebody to start you out with a design style and then you're home free from there. Correct.

SPEAKER_00: Correct. Okay. Out of the 40,000 users that we have so far on the platform, the champions are artists, game artists. Um, and if you remember the controversy around artists and AI, I found that very interesting that actually the biggest champion for that type of technology end up being the artists themselves. All right.

SPEAKER_64: Founders. If you're a SaaS or services company that stores customer data in the cloud, you need to be sock to compliant from third parties in order to close big deals. And you need to use Vanta if you want to do this quickly and easily. Vanta makes it so easy for you to get and renew your sock to on average Vanta customers are sock to compliant in just two to four weeks. Compare that to three to five months without Vanta. Well, they partner with over two dozen audit firms and they have been trained to file those sock to reports directly within Vanta. It's a total no brainer for you. You're going to basically make it easier for your sales team to sell those big clients who

Chamath Palihapitiya: need that sock to compliance in order to engage with your product. Tons of my portfolio founders use and love Vanta. They all report back that they had an amazing experience. And that one, that one major customer, right? That lighthouse customer, that whale in your system. Listen, bottom line, you can't afford to lose sales for something silly like not having your sock to. Vanta also helps with GDPR compliance. And you know, that's a big one. You got to get your GDPR in order because they take it pretty seriously over there. So here's the best part. Vanta wants to give you a thousand. All right. That's ten hundies. A thousand dollars off at Vanta.com slash twist. That's Vanta.com slash twist for a thousand dollars off your sock too. Um, are any of your 40,000 users using this in production yet? And if they are using it in production, what is the distance between, uh, what has to happen

David Friedberg: between you making this bubbly green hamburger on a bed of lettuce and it getting into the game? How much more artistic work has to occur?

SPEAKER_13: So they definitely use it in production today. Not only they use it in production to make the game assets, but some games have this part

SPEAKER_00: of the game where the, the, the creator of the object is not just the artist. It's the player. The player gets to in game generates whatever reward weapons, skins, buildings. I mean, depending on the, of course, the, the, the theme of the game. And that's you it's AI powered UGC. Amazing.

Chamath Palihapitiya: So if you were to clear out all these prompts, we did, and we just said, do, um, a zombie,

SPEAKER_37: um, teenager, um, wearing, um, a football Jersey.

SPEAKER_05: Um, this could be, if I was playing a zombie game, I could say, okay, well, I'm a football player and I'm a teenager. I'd like to make my own character and it would make a bubbly designed zombie teenager with a football Jersey. Yeah.

SPEAKER_00: I'm doing it right now. So I first made a, a batch, uh, of your prompt, which was zombie teenager wearing a football Jersey. And because I know a bit about it, I, I prompted 3d rendered full body. And then I'm going to do apocalyptic, uh, just to give some more, uh, you know, color. Um, and I'm going to add like, maybe blood, um, and then run another. Yeah, we're not.

SPEAKER_75: And then, so the language model knows what a zombie is. It knows what a teenager is.

David Friedberg: It has some conception of these things from the real world. It matches it with the style, but even beyond a designer doing this, the game could, uh, a game player could decide, uh, to add some characters or make a level themselves. Or I suppose you could have a random generator that just looks at, uh, just looks at common

Chamath Palihapitiya: objects in the world and just makes random objects. And if people engage with them, it could make more of them. So you could actually have a game where it just randomly types in zombie hamburger football Jersey and you get whatever out of it. And this could create infinite universes with infinite characters.

SPEAKER_00: Yes. So here's the zombie coming for those who are listening. Uh, what we see is blurry pictures at the beginning, very noisy pictures. That's what diffusion is about. You add noise and you remove noise. That's the, that's where diffusion comes from in stability fusion, typically. So, um, definitely very, very consistent, uh, zombies coming up with a Jersey, uh, Jersey. With a number on it. Yeah. Number on it. So let's pick one of them, Jason, which one would you like? Uh, maybe the 18. Number 18 looks great. Number 18.

SPEAKER_81: Yeah. Number 18 looks great.

SPEAKER_00: 18. Um, we're going to do variants around 18. Uh, the only issue in that case, maybe, maybe let's pick the next one because we, yes. So let's do the, this one here. Yeah. Great. Because we love it. And now I'm going to not only prompt the same phrase, but I'm going to iterate around that image. Um, and I'm going to do image to image and with variants will be created, uh, around the same, uh, shape or proportions.

SPEAKER_13: And I'm not going to go into like the tiny details, but something super cool just came out, which is component composition control that lets you, uh, play again with the, the sort of input you want from that image. And we're going to have the results in a few, uh, in a few minutes. So in a few seconds, actually.

Chamath Palihapitiya: How much CPU and cloud is this taking? Um, we, we know these are resource intensive. So as you're doing this, you seem to be creating four images at a time. Um, these are really good looking images. So how much CPU is this using? If I, could I be doing this on my local Mac book or is it using 10 Mac books in the cloud?

SPEAKER_13: No, it's, uh, AWS, uh, based, uh, infrastructure. It's pretty, pretty, uh, GPU and server intensive. Uh, I think we've spent, uh, more than two 50 K just in GPUs, just for the beta, just in a few months.

Chamath Palihapitiya: And your company has 250,000 in these GPUs dedicated on servers. So, um, you have that as a fixed cost and there's a certain number of images that create, uh, could create at a certain standard. If you were divide those number of images into that cost, does it wind up being a dollar per image, 10 cents per image, a penny per image, or a fraction of a penny per image?

SPEAKER_00: It's more of a, it's more of a penny to a fraction of a penny though, uh, depending on the settings, uh, you might end up with images that are pretty expensive. If you go into like more steps and more, um, uh, you know, advanced, uh, parameters and higher resolution images, you can go into, uh, uh, you know, maybe up to like a fraction of a dollar per image.

SPEAKER_05: Oh, wow. You could be spending 10 cents or 25 cents an image. This is an extraordinary thing because we've lived in such a deflationary Moore's law. We didn't know what we would do with all these CPUs. We thought CPUs and GPUs and storage was getting so plentiful that we had no use for it because, Hey, people can't even, you know, tell 4k from 10 ADP from 4 ADP. Like if I showed the average user, they probably wouldn't know the difference. Whereas here, because we're doing generative AI and it has to do so much GPU work, this is actually using a lot of the GPUs in the world. And it's going to be constrained for some time, or is this, um, something you think software is going to become more refined and we're going to just have these costs plummet 90% a year.

SPEAKER_24: So costs will plummet. Maybe not 90% a year right now, definitely they will plummet.

SPEAKER_00: We have seen some demos of stability and running on a phone, maybe not the, not the type of advanced, uh, calculations we do, but some, some, some day, yes, diffusion models will run on phones, whether it's for text, uh, like TPT, you know, like, uh, models or whether it's for images.

SPEAKER_13: And one day, because it's not, it's not quite there yet, full, like 3d models, it's coming, um, especially when you make a 3d models right from the image you just generated in that case, we have the initial zombie, uh, with the Jersey.

SPEAKER_00: And then I made a few variants, uh, without even touching the prompts.

SPEAKER_13: And from these variants, it's reasonable to think, uh, 3d models will be made just in a click. And that's, that's extreme, extremely, uh, deflationary as you just, uh, as you just said.

SPEAKER_18: And you could take random prompts, um, where you could ask, you could train chat GPT for, or just a language model.

David Friedberg: You could say, go look at, I don't know, Reddit or Twitter, find different characters or find inspiration from these locations.

Chamath Palihapitiya: It could be a newspaper. It could be, uh, YouTube videos. Let's assume you have permission.

SPEAKER_05: You could say, look at the top trending characters or people or scenarios on this website. It could be an open source one. It could be like, um, 500 pixels, right?

SPEAKER_88: You say go on 500 pixels and just start making me images or characters, uh, and just start building up crazy amounts of libraries. It's extraordinary, uh, where we're at right now, um, in terms of building stuff.

SPEAKER_24: Um, and develop localized versions of games, develop seasons per, you know, holidays, Halloween, Christmas, news, memes, you know, uh, like this could be highly personalized.

SPEAKER_00: Any, anything that goes into the prompt box could be, uh, you know, plugged into other, uh, pipelines.

Chamath Palihapitiya: So this will drop the cost of making, this drops the cost of making a season of, let's say they're doing the Christmas season, the holiday season of plants versus zombies. What did you guess a game like that would spend on that refresh in terms of time and money that, and then how much would you reduce it by?

SPEAKER_24: I, I got the first numbers from the first customers we have, um, feedback is, um, processes are 10 X faster on average. Um, so they work 10 X, 10 X, you know, faster.

SPEAKER_00: And for a, you know, small studio of 20 people, which is rather small, it's not riot or blizzard, but for a small studio, they would typically save at least 200 K a year using scenario. Just on concept arts, not even in game assets, just concepting will save them 200 K a year. Then you, then go, you know, in game assets and then come marketing assets, which is a big part of the game industry as well.

SPEAKER_24: How to automate the marketing.

SPEAKER_05: Got it. So they like to make YouTube ads, Tik TOK ads, Twitter ads, whatever, all of those ads falls on the creatives. Now they could be making them without having creatives, having to anoint every single one.

Chamath Palihapitiya: So this is going to knock 20%, 30% off the cost of running one of these studios this year. You think?

SPEAKER_100: Uh, pro it's, it's probably right. Yes, um, probably right.

SPEAKER_50: Yes. Yeah. Yeah.

SPEAKER_05: Uh, which means they can either save that money or they could do 30% more work and delight their customers 30% more, which is probably what they'll do.

SPEAKER_88: Um, this is extraordinary. Um, where will you be in one year?

SPEAKER_00: We go to other media, uh, beyond 2d. Uh, we, we are expanding into 3d, making these, uh, not just 2d images, but 3d models. Um, we integrate as we are doing it right now into the game engines, which is a good, good. Go to market for us because that's where most of the game developers are. They're on unity. They're on unreal. They might want to do there's just this on unity and unreal. Um, obviously we, uh, improve everything around asset management, collaboration, you know, bring the whole studio together. Um, and, um, there's a lot to do on the roadmap, um, better, better trainings, by the way, better AI models.

SPEAKER_24: Train faster, uh, more efficient fidelity will go way up.

Chamath Palihapitiya: You'll be able to do 3d models, uh, which means like a fortnight character could be made in here, maybe, um, and then actually just dropped into the game. Um, and so that's going to save an massive amount of money and time.

SPEAKER_112: Um, if you love snacking, like I do, you got to find the perfect snack. It's not easy. You want one. That's going to remove cravings that fit your dietary goals, regulates your blood glucose levels. And gotta be something you look forward to eating. And you know what? I love. I love macadamia nuts. What a treat. They got so many health benefits. I'll give you the highlights right here. They are the only nut that is rich in omega sevens that's linked to natural collagen production, reducing inflammation, stabilizing your glucose levels and healthy fat metabolism. And the folks, my friends at house of macadamias are obsessed with making the highest quality macadamia products available. Chocolate dipped macadamia nuts are my favorite. But I also, sometimes I get that coconut craving. You ever get that? They got a chocolate coconut macadamia bar. That is amazing. And it hits the spot. It's not going to spike your glucose. It's something healthy, but it's a treat. It's a treat. And they just launched an amazing cold press macadamia oil with a buttery flavor and a high smoking point. Perfect for drizzling over meals or cooking. Use the code twist 20, get 20% off. And for a limited time, they're giving a free bottle of premium cold pressed, extra virgin macadamia oil with any purchase. Go to house of macadamias.com slash twist and use the code twist 20 house of macadamias.com slash twist and use that code twist 20. So they know J Cal sent you.

SPEAKER_118: There was one artist on Reddit. We read it on a all in previously and on this week in startups.

Chamath Palihapitiya: Um, this artist was lamenting how, um, you know, like this is taking a little bit of the art out of the game. And, uh, he was kind of depressed that somebody who was an average, uh, artist could now do his work.

SPEAKER_05: And now you have the ability for a prompt to do the work and eliminate their need to do the next 500 images. So they just have to do the 25, what's your response to those people who maybe are bummed out about this technology?

SPEAKER_00: The, the feedback I'm getting from the users themselves is they, they, they find this technology is, uh, uh, just augmenting them. They are, it's augmenting their creativity. It's making them go faster, further away. So they, they, they love it. And if you are, you know, have you seen that, that research thing? If you are, the smarter you are, the better result you get out of chat GPT. I think the same applies for images. The more creative you are, then you're going to get an even higher level of creativity. And yes, you can pretend being an artist, even though you're not one, you're not trained as a game artist. You can pretend being one with this, or you can approach being one with this, but if you're a good game artist, you're going to be an even better game artist with AI. And it's not just push a button, you know, it's like fronting, but like selecting the right models, the right training parameters, the right, everything takes a while. It's a new art. It's a new, it's just a new Photoshop.

SPEAKER_10: Um, yeah, and it, we, it will take years until people go to the end of this, this technology.

SPEAKER_75: It's not, we had architects who were very upset about CAD drawings and people using computers instead of using pencils and rulers and drafting desks.

Chamath Palihapitiya: And, you know, the, the fact is we're able to design, uh, you know, more beautiful homes, more unique homes, more unique spaces with CAD and share them around the world. Um, how many people in your company, uh, to get to this point, how many developers did it take to get here and how much time.

SPEAKER_00: We have 25, but we started as a team of six, uh, back in October, uh, went pretty fast to put the, the, the beta out in December. And that's when we started growing after we got some, the first, uh, seed money in, uh,

SPEAKER_13: in December, uh, growing from six to 25 in just three months, basically, um, mostly engineers so far.

SPEAKER_05: Got it. Any other applications for this? We hear a lot about, Hey, video games and Disney plus in the Mandalorian and the real

Chamath Palihapitiya: engine, and things are kind of converging. Do you see this at some point, um, taking over and doing South park episodes or, you

SPEAKER_05: know, uh, you know, clone wars episodes, bad batch episodes. You could start to create characters and then place them into what would eventually become TV shows. Cause you're doing backgrounds. You're doing art. It doesn't take much to have people move around the backgrounds. It doesn't take much to add dialogue. And if you can make the characters, well, certainly you can make voices and there's other people who can do voices. So I'm curious, where do you, is there a new place to go with this right now? You're helping game studios, but this feels like a new art form could emerge from your tool itself.

SPEAKER_129: Yes, we have some, we have some edge cases, uh, being, being, uh, being, uh, used right

SPEAKER_00: now by a few, uh, a few of the users, definitely movie, even though some other companies are much better, are doing movies with Gen.AI such as Runway, such as wonder studio. Um, they, they, it's, it's a slightly different sort of feature sets to develop. Uh, but like marketing, you know, um, all kinds of like any industry that needs some images will

SPEAKER_13: be based on Gen.AI or is going to be based on Gen.AI significantly. How do you charge for all this?

SPEAKER_00: It's just, we charge a subscription, uh, per accounts, which goes from 1k to 40k a year, depending on the size of the studio members, quotas, images. Uh, and we, we go into some enterprise custom pricing as well. Um, in addition to the SAS model, we have an API based model where people just go pay per API call,

SPEAKER_13: you know, depending on how much they use the API. Um, yeah, we just, we just started this week to monetize actually.

SPEAKER_79: So, Oh, really?

SPEAKER_75: So do you have a sales team doing it or you just took these 40,000 people who are using the free

SPEAKER_05: product and just said, Hey, you hit the upper limit consumption based. You need to sign in and pay, put a credit card in, we, we do, we have a small sales team for,

SPEAKER_00: uh, for the tier one, tier two customers, uh, like game studio of a seven size, uh, plus myself, one in Europe, one in the U S and we have a, we, we do have a customer success team as well. It's not just about, you know, signing. It's making sure that they are doing a good job and they, they are adopting it at scale, but it's, um, yeah, it's going well is I'm super, super bullish. I've made a first company before, which I sold was good, but this one is going to be, you know, a hundred X bigger.

SPEAKER_31: What was the pre you previously you were in lidar.

SPEAKER_00: If I understand from my notes here, it was, uh, aerial capture with drones, lidar and photogrammetry.

Chamath Palihapitiya: Oh, so you were sending drones up with lidar on it to do what?

SPEAKER_00: Uh, um, all kinds of inspections, you know, mapping, uh, of buildings and, uh, infrastructure construction sites and so on. We were not flying the drones. We were doing the software to process 2d and 3d imagery at scale with AI in the cloud. So it's kind of a natural evolution, but instead of like scanning, we prompt.

SPEAKER_140: And that was a niche market, right? Like sending those drones up to do construction sites.

SPEAKER_10: It wasn't supposed to be niche. When we started according to, uh, you know, BCG and McKinsey and so on, but the drone market didn't end up being that big.

Chamath Palihapitiya: Um, what do you, why do you think the drone market didn't become that big? Didn't provide enough value to users?

SPEAKER_00: Oh, value was, value was enormous. It's more like regulation and it's not that easy to fly a drone anywhere, anytime, you know, with the, with the rights, uh, uh, uh, material resolution and so on.

SPEAKER_144: Um, so the pilot limitation and the regulations killed it.

SPEAKER_00: Yeah. You can only wait, you can only carry so much in a drone. You're never going to see, you know, you know, 500 pounds drones flying tomorrow to, to, to capture you on your next construction site.

SPEAKER_26: Um, yeah. And let alone, uh, let alone, uh, the, the transportation drones.

SPEAKER_18: You know, it's, it's an amazing technology, but physics is physics.

Chamath Palihapitiya: Regulations are regulations and things in the air when they fall from the air can do damage. And I think that is, I think one of the lessons people learned was the utility of drones, um, is based on weight and the utility is based on being in cities or populated areas in some cases like delivery. And, uh, that's challenging, uh, because. People do have, uh, the regulator seem to be very cautious, uh, in the, in the drone space.

SPEAKER_88: And here, you don't have to deal with regulators. Um, you just have to deal with software and how much you can delay your customers.

SPEAKER_147: For now, for now, for now. Yeah. What do you think? Yeah. Regulation AI. I think if you, I think it's there.

SPEAKER_149: Yeah. Good.

Chamath Palihapitiya: Well, I was going to say, um, if you, if there was a, uh, a place to start thinking

SPEAKER_05: about regulation, where would you, as somebody who's an insider start to think about regulation? Uh, at what point in the stack at one point in the code, the hardware, the application, the geography, where would you even start with regulation?

SPEAKER_24: I think it starts with the training data.

SPEAKER_00: How do you, how did you get it? Where did you get it from? Sure. How did you select, uh, you curate the training data? Um, that, that is, uh, that is the, the debate today around, around chat GPT and like, what data has been used and how, you know, how do they control what comes out and what gets, uh, you know, um, um, uh, eliminated or, uh, or, um, uh, um, just like what gets, what gets out and what gets blocked. Uh, the same will go for images. Stability fusion was trained on a very wide array of 5 billion images scrapped from the internet. Uh, people loved it, but a lot of legal teams find challenges around like, do we have the right to generate commercial, uh, images?

David Friedberg: The answer is clearly no, you cannot take Getty's images and make derivative works from it, especially if that infringes upon their ability to do the same business.

SPEAKER_152: At least here in the United States, that's how the law works. It's clear as day.

SPEAKER_00: There'll be tons of settlements, which is why people use their own models. And which is why Adobe released firefly, which is ethically trained on, on images. They have the rights to use. And I believe in a few months, every single, you know, gen AI image generator will be trained on, on, uh, on data that people will have the right to use at some, in some, in some, you know, at some point.

Chamath Palihapitiya: There's plenty of open source data out there. And if you wanted to train on Getty's, you should just ask them and pay them a fee and they might charge you a million dollars or they might charge you, you know, there might be a photographer out there or a group of photographers who own these who just say, yeah, you can train your image generator for $1 per photo. And if you need a million images, great, that's a million dollars, uh, per year.

SPEAKER_31: There's some reasonable amount of money to be charged. So it could be a great income stream in the way streaming is for musicians.

SPEAKER_24: And income stream could be renting or leasing your generator instead of selling the final

SPEAKER_00: image, such as the final picture, the final photograph, just rent a, the use of your model as an artist. And it's like a much more like a much passive, uh, income that could be interesting for them.

SPEAKER_24: It has to be, um, that's fascinating, but yeah, it's coming.

SPEAKER_05: Well, you could do it if you had somebody who put a collection in here, like this bubble collection, you could have an artist who builds that on spec, who's in college. And they say, here's my bubble collection. Anybody who wants to generate on top of it, um, it's going to be, you know, $1 per exported image. And, um, then they would know who paid that $1 per export image. And they create those 25 images. And if 10,000 images are exported per year, that's $10,000 you as the platform take 30%, they take 70%, you made three, they made 7,000. That's a pretty fantastic business for a kid in art school who made 25 images that took, I don't

Chamath Palihapitiya: know, 10 hours each. So, and that's money they would make forever.

SPEAKER_118: So there's a reasonable trade of services here. Yep.

SPEAKER_23: And then, you know, this happened in stock photography. Stock photographers were, um, were villainized at some point because they were like, hey,

Chamath Palihapitiya: you should send a photographer out to take a picture of the golden gate bridge for $1,000. Let them develop the film and they should pay charge $1,000 or 2000 for that image. You shouldn't let people buy a golden gate photo for $5 or $10. So the world advances, more people have access to this incredible content. That's better for society. Sorry, if people are, you know, sorry that it's deflationary, but when things are deflationary, people can do more creative things, right?

SPEAKER_26: Yeah. And free time, um, to, um, to, to do more.

SPEAKER_10: And, um, you want me to drop another, you know, buzzword, it could be also blockchain could be involved to, uh, track these, uh, flows of images across different owners.

SPEAKER_00: Uh, which has not been done yet, but it's a, it's a good, I think it's a good use case of blockchain, uh, that combined, you know, GNI. Explain how that might work.

SPEAKER_159: Yeah. Explain how that might work.

SPEAKER_00: You know, well, a generator might be, uh, tied to some sort of smart contract. So any image that's, that is being made from that generator could be tied to the generator and then to the artist via some revenue share stream, um, you know, uh, written, coded into the contract. Um, or I mean, like look at GNI, it's a, it's the abundance, right? It's like millions and millions of images, uh, being made and shared and content overall, you know, how to be able to track it via some sort of like, uh, trusted ledger could be, could be interesting. I'm not saying we should all do it on like right now, and we are not even doing it, but with the web three people, the web three games that use scenario, they're all about it.

SPEAKER_17: Incredible. Yeah. You know, yeah.

SPEAKER_163: Well, this has been amazing continued success. Where's the company based? San Francisco.

Chamath Palihapitiya: You know, this is like your, maybe the fourth AI company to come on who is basing in San Francisco. And it does seem like there's a little, uh, you know, revitalization, uh, in San Francisco of people putting their AI companies in the city itself, not in the surrounding area. So, hey, listen, continued success. And, uh, we'll see you all next time on this week in startups.

David Friedberg: Bye bye.

SPEAKER_64: Probably the most common challenge I hear from founders is related to building.

Chamath Palihapitiya: Either they are in technical and are searching for a technical co-founder or they can code, but they're just spread too thin. This is one of the first major obstacles you're going to face. And I know how discouraging it can be, but there is a solution. Do you have a great idea, but you don't have a technical co-founder? Well, Crowdbotics can be your CTO as a service. Boom. Just like that. This means you can focus on building an awesome product and delighting your customers. Rather than wasting your time on infrastructure, planning, architecture, compliance, and all that boring stuff. Crowdbotics also offers professional scoping to help you flesh out your project at the MVP stage and beyond. So cut out the hassle and get back to building that perfect product for your delighted customers. And when you think Crowdbotics, I want you to think getting your time back to focus on product. Product drives everything in a startup. So let the folks at Crowdbotics show you how it works. Schedule a free scoping session and get your detailed build plan at crowdbotics.com slash twist.

SPEAKER_114: That's crowd B-O-T-I-C-S dot com slash twist.

SPEAKER_170: Hey everyone, I'm Presh, and today we're going to walk through how to build NoCode apps. So first, we're going to talk about why build NoCode apps in the first place. We'll cover things like who is it for, and then we'll jump into some of the pros and cons. Next, we'll talk about some of the best tools to use when building in NoCode. And these are going to vary depending on what you want to build. So hopefully after watching this presentation, you'll have a better understanding of what NoCode is capable of, and then you can get started on third base picking your platform. And lastly, we'll wrap on some of my favorite resources in the NoCode community to help you further your education. So let's get into it. So why build NoCode apps in the first place? It's going to come down to two things. So one, you're non-technical or non-technical yet, and two, you want to move quickly. So for example, you have an idea and you want to put that idea in the marketplace as quickly as possible. Test and see if it's worth pursuing. Over the next couple of slides, I'm going to show you how before even building your product, you can utilize some cool tools to get to the third base of building your product. So using NoCode tools to start this process of building your product might look like this. Step one, you get your idea. Step two, you might want to share that idea or validate it a bit. And this can be done by tweeting it, writing a blog post, or sharing it in the subreddit. Step three, you might want to start making a mockup of your product with tools like Figma or Canva or Sketch. And I include design tools in NoCode because they've really designed them so anyone can learn them quickly. And they're just delightful and fun to use. Step four, okay, now your product mockup is done. So it's time to build a simple landing page. And remember, the goal of this is to put something out there and see if there's interest of the product. So it's not mandatory to have a overly designed website. So I use tools like Webflow or maybe even Notion, Squarespace, Card. Step five, okay. So now the website is up and we want to start capturing people who visit the site that might be interested in the product. So we can do this with a simple email capture tool. And I like to use Mailchimp in this case. Step six, and lastly, with the emails, we can now start funneling customers or potential customers into a community with products like Slack or Discord or even Twitter. Keep the conversation going and keep your potential customers engaged. Now we're going to talk about the pros of no-code development. Now I've narrowed it down to three things. Agile development, low to no maintenance, and cost efficiency. So in terms of being agile, you can put out an MVP as soon as you have the idea and you just need to do it with one person, which is yourself. You don't need a whole team. Next, there's no maintenance. Since you're building on top of platforms and ideally big ones with big teams, we're piggybacking our product on them. And so they manage the maintenance for you and updating everything. And lastly, it's cost effective to a point because you don't need to hire developers, which are expensive. The other thing is you're dependent on the platform you build on. So if one day they wake up and decide to remove a feature that you use every day, you might be at a disadvantage there. And lastly, it can get expensive. Again, depending on if you're using multiple tools, you want to be thoughtful with the kinds of tools that you're using. Now we're going to talk about the tools you might want to use when building no-code apps and what's actually possible. Let's get into it. So to start, the first decision we need to make is what type of product do we want to build? Do we want to build a simple website with a gated community slash paywalled content? Or do we want to build a marketplace type of business similar to Airbnb, where it's two-sided, you got a supply side, a demand side, and you want transactions to occur? Or do we want to build a mobile application for maybe the App Store or the Play Store? So there are tools for all of these examples, and we'll dive in deeper right now. Let's start with a simple subscription website or subscription business. These are three tools that will get you started, and you can build an entire business off the back end of this. The first is Webflow, which is the visual no-code website designer that I mentioned earlier. And this is popular now because you can customize fully, and they've got beautiful templates. The next is a tool called Memberstack. And what this allows you to do is allow your website to have a signup portal or dashboard, and allow you to charge customers either one-time payment or a subscription model. And so really turning your website more into like an app or a community. And lastly is Zapier, which is an automation slash API tool that allow you to connect and talk to different applications. So you can have your Webflow website, talk to Memberstack, and you can do that through Zapier. So here I'm just on Webflow's template website, and you can see they've got beautiful designs here. I'm in one example called London, and I can buy this if I want to use it. You can see it looks like a very professional website here. It's got nice animations, nice colors. And this looks like, you know, it would cost tens of thousands of dollars to build, but you can replicate this with just a click of a button. Okay, now here we are on Memberstack. Again, Memberstack allows you to create like a signup portal and create a login for your users on your Webflow site. And so here you'll see this login portal. That's kind of an example of what it will look like. And then you've also got a pricing page. So that's, you know, if you want to charge or put a paywall for some of your content, you can do that as well. Here's an example with one of our portfolio companies, Sol Savvy. And they initially started with just completely no code tools. So you'll see their website, you know, you can go in, you can sign up for the waitlist, or you can skip the waitlist and put your information and sign up. All of this can be done within Webflow and Memberstack. Lastly, on the subscription website stack, we have Zapier, which is the king of automation and connecting different apps to talk to each other. So you'll see in the screenshot, they've got thousands of integrations into all the main apps like Google Sheets, Mailchimp, Calendar, Slack, Typeform, Stripe, and so on. So the main purpose for Zapier is for apps to interact with each other. Moving on to a marketplace type product, these tools will get you started right out the box. So the first one is software.io and the second one is bubble.io. And both of these will get you on third base because you can use some of their marketplace templates. And the best examples I'll show you here are Airbnb clones on both platforms. All right, so here I am on software.io, and they've also got a great template library. I'm going to go to one of their examples here, code.live. And this is a marketplace example built right on the platform. And so you can see you've got some houses here that you can click into. You've got a search feature, you've got explore spaces, and you can even go as far as booking a spot on the platform and that'll take you to like a landing page or stripe, a stripe powered page to complete the transaction. This is another example using the bubble platform. So again, another marketplace type product. You can see you've got different locations here. You can even create your own listing. And so again, designed beautifully, works perfectly, and you're building this from scratch, but you can also use a template and get you again on that third base. Lastly, moving on to mobile app development in the NoCode ecosystem, there are two main tools that I've used. And the first one is Bravo Studio, and that's bravostudio.app. And the other one is glideapps.com. The first one, which is Bravo Studio, is more of a visual designer. So you can use Figma or Adobe XD to design your app. And so what Bravo Studio does is it connects those elements together to make into actions. So you're basically leveling up your prototype, be a functional product. The other app I've used is called Glide. And Glide allows you to create web-based but mobile-friendly apps to do powerful things like create management tracking tools or a conference app or a workout app. And I've even seen a template for an Instagram clone. Now let's wrap up with some resources you can use to learn more about building NoCode products. So the main resources that I like to use are makerpad.co and nocode.tech. And both of these sites have extensive libraries of content that you can get started from beginner all the way to advanced. And you can see on this screenshot, you can start as, you know, as basic as just creating websites or you can go to building mobile apps. And again, this is all NoCode, so they've got you covered. And that's going to wrap today's episode of how to build NoCode apps. Thanks for listening.