SPEAKER_04: we're getting very close to going from somebody's mind and doodling to a finished product. Is that what's happening here?
SPEAKER_01: That's 100% what we are trying to make. I mean, we were trying to make that happen.
SPEAKER_04: And so filmmakers are playing with it. Other folks are playing with the people who have had their their minds blown by this. You've got hundreds of people paying 30 bucks a month for this dozens. Where are you at as a company? So we have a waiting list of more than 200,000 people right
SPEAKER_01: now. Wow.
SPEAKER_02: This Week in Startups is brought to you by Squarespace. Turn your idea into a new website. Go to squarespace.com slash twist for a free trial. When you're ready to launch, use offer code twist to save 10% off your first purchase of a website or domain. Masterworks is the first company allowing investors exposure into the blue chip artwork asset class. Twist listeners can skip the waitlist by going to masterworks.com slash twist and fitbod. Tired of doing the same workouts at the gym? fitbod will build you personalized workouts that help you progress with every set. Get 25% off your subscription, or try out the app for free when you sign up now at fitbod.me slash twist.
SPEAKER_04: All right, everybody, welcome back to This Week in Startups. There's a ton of AI powered creative tools that are coming to market, you may have seen Canva, open AI, Adobe mid journey, obviously stable diffusion runway, so many different products and tools to help creatives make more interesting output, faster and better and even allow maybe people who aren't that creative to get creative. All these platforms run in a similar way, enter a prompt, and then the model goes to work, you wait a couple of seconds, sometimes a little bit longer. And then it spits out an image, or maybe a collection of images, and then you refine those images further with more prompts, and thus restarting the cycle. It's not ideal, and it doesn't actually work that well, unless you have a lot of creative skill, and you can refine those images, it's very rare that the images come out fully baked and ready to go. But a new startup called crea.ai. That's with a K K r e a dot AI is working on a real time AI powered creative toolset that works in the browser, it went viral on x this week. So we decided to have the CEO on x is the website, formerly known as Twitter. Victor Perez is here today, and he's going to show us how it works. Victor, I don't think you've done any interviews about Korea yet. So this is a this week and startups exclusive, I think welcome to the show.
SPEAKER_01: Yeah, thank you so much, Jason. Super happy to be here.
SPEAKER_04: So let's get right to it. Show me what you've built and why people are losing their minds over it. Let's go. What we have here is a is a tool by the way, this was
SPEAKER_01: built in one week, right? So there was this new technology that got released. We were here at right when I'm recording in a party, where we decided to make like a fun interaction with this new technology. So everybody that was in front of the webcam would be turned into something else in real time. That's how these tools started. And after the day of the party, we realized like, Holy shit, this is actually very, very interesting. Should we build an interactive tool on top of this? And in a few days, we already had the whole thing together. And very like we divided the team. Some of us were working on the infrastructure on making an infrastructure that can scale because this is kind of crazy to have 1000s of people being generating images in real time. And some others we were working on the design. All right, so we're in a web browser, essentially, it looks
SPEAKER_04: like. Yeah. And on the left hand side, there is a pink circle that is layered on top of a blue rectangle. And on the right, we see what looks like a mushroom with a pink frog on it. That mimics the very simple circle and square. So explain to us what the prompt is here and how this all works. So the prompt here, it just blue mushroom in top of a pink rock.
SPEAKER_01: Yeah, like that's exactly like what what the AI is doing is is getting this initial image that we have on the left side, and turning it into something that looks very much realistic on the right. This is not nothing new, right? Like we've been able to do this with models like stable diffusion before. This technique is called image to image. So this is nothing new. What is new is that if I start moving this pink circle, you will see that the frog starts moving around in real time. Right? So this is right now giving me full control is giving me a whole new dimension for prompting, which is like right now I can prompt these AI model visually. And so what I could do right now is, for example, let's say that I want a, I don't know, instead of a pink frog, I want a blue bird on top of this blue mushroom. So what I could do is get the same shape, change the color, turn it into blue. And what we should be able to see is a blue bird on top of a blue mushroom that if I start and that's the thing like only through prompting right now with the stable diffusion, I have gotten something like this that doesn't look at all like a blue mushroom in top of a bluebird. Actually, that's kind of a hard one. Let's do a bluebird on top of a blue mushroom, that will be an easier one. So if I change the color
SPEAKER_01: here, like as you are changing the color palette, and then AI
SPEAKER_04: is automatically reacting to that. So by using the color palette, it's sending a prompt, I guess, saying, Hey, use this color to the AI in real time. Exactly. Yeah, it's just like, you can think of it as a, as a
SPEAKER_01: visual co pilot, where like, we are used to co piloting in text that it kind of auto completes what you want here, like the input images on the left side, and it's kind of auto completing it with this final result that looks very realistic.
SPEAKER_04: And so what's the back end here? What are you using? What's the language model or the image model?
SPEAKER_01: The image model is a version of a stable diffusion that is distilled with a new technology that is called consistency. And with these models, we're able to run stable diffusion. Yeah, we're able to run on single image in 40 milliseconds, which is insane.
SPEAKER_04: So it will do a single image in 40 milliseconds. And so it's almost to the point where it is when you drag and drop this and move it around. So if you were to move the bluebird to the other side of the mushroom, it kind of does it, you know, whatever, five frames a second, four frames a second something. Yeah, that's mainly because of the all the network delays.
SPEAKER_04: And so how much compute power is this using? Or have the stable diffusion models gotten so good that they don't require as much compute, you can run this model really on like 3090 or like some GPU that
SPEAKER_01: you could have on your computer right now to make it work at these speeds, we are using a 100.
SPEAKER_04: Got it. And so if you have an a 100, which costs, I think 20 grand, still something in that range. How many people could be doing this in real time on one of those units? Right now, I think that we should be able to support from
SPEAKER_01: four to 10. So in other words, you need about, let's call it $3,000 in
SPEAKER_04: compute to be doing this in real time, if you were divide the GPU by the number of people using it. So that's what desktop computers cost today. It's not that big of a deal. Where is this going? What else can it do?
SPEAKER_01: Oh, yeah, I think that we should think of this demo as the worst you will ever be. This is the dumbest like the AI will ever be. This is about to get way more efficient and it's like the quality is about to get way better. So we can expect that in the who knows when because like breakthroughs are very hard to predict. But I can I can think that in a few months, we'll be able to see these demos running on your own computer, maybe on your m1 m2. I wouldn't be surprised if that happens. And we're kind of like setting up everything towards that.
SPEAKER_04: And Apple just launched the M3. And so right there, chipset is obviously getting powerful enough to do this. And maybe next year, you'll be able to do this on a desktop computer or a laptop using Yeah,
SPEAKER_01: yeah, like I don't really know when it's gonna happen. But it's 100% gonna happen. Yeah. If your landing page looks terrible, I'm out. We all know
SPEAKER_04: that you see an ugly website, you skedaddle, you leave, you're done. So you need to stop settling for okay or good and start using Squarespace. So you can be excellent and extraordinary. It's an out of the box business solution to build beautiful websites, engage your audience and sell anything you want. You know, Squarespace is amazing features, or just templates that are always optimized for mobile drag and drop web design with their fluid engine, advanced analytics, marketing analysis, sales data and more. And with Squarespace, you can create an online store or start a blog at the click of a button, create a subscription business for members only content and so much more. And you can do this all simultaneously. It's the simplest, most effective and best looking way to start a business online. So here's your call to action Squarespace comm slash twist for a free trial. And when you're ready to launch, go to Squarespace comm slash twist for 10% off your first purchase of a website or domain. You guys are in a hacker house there. I see like three or four people writing code behind you. Where the heck are you? Is this your startup here? Yeah, yeah, we all live and work in here.
SPEAKER_01: Are you in the Bay Area? Where are you? Yeah, yeah, we're in San Francisco. This is more like the the Korea house, right? Like we have founders founding engineer is also living here in a room. There are like some friends in the city that we just like set up in here so they can sleep. How long do you guys how late you guys stay up coding? What's the bias?
SPEAKER_04: This few this few weeks has been a madness with all this growth
SPEAKER_01: and like trying to scale up all the GPUs and onboarding users. We stayed here until very late. Normally, Diego, my co-founders stays here until three 4am. Normally, I normally go to sleep at one. So people are grinding it out. You guys are super motivated.
SPEAKER_04: And so is this a company now that you've built? Have you raised money for it? And then who are the customers that you're trying to get on board? And what are they using it for? Totally, yeah, this is a company we raised our seat
SPEAKER_01: ground last year. And the customers are mainly creatives. We are right now focusing in kind of a consumer product. So the spectrum of people that get interested in this is pretty wide. Like we have from professional film directors that they are creating shots for a mood board that they want to show to their art directors. For example, we have a film director for example, we have graphic designers that are using this tool to make letters like kind of typographies or to make all sorts of illustrations. We have 3D artists also making like kind of brainstorming or making like very low or yellow like images that they can show to their clients to confirm if they can proceed to go and to the professional tool and actually like spend time doing it high resolution. You charge 30 bucks a month for this tool, which allows you and
SPEAKER_04: people don't know this, but this is how like Ridley Scott works Ridley Scott, the film director, aliens, Blade Runner, gladiator, etc. He makes these Ridley Scott like tiles, basically little drawings on the set, and then they work with the cinematographers and everybody make it happen. I'll share my screen here because I have a couple of the demos that were shared on Twitter. Yeah. So here is a demo somebody did just drawing the moon and we see the trees behind it. I'm not I can't see the prompt there. It's a little bit too small, but they put in some prompt obviously to make a spooky nighttime foggy place I'm assuming but they start drawing and when they're drawing, you know, very rough sketches of waves or whatever, it just makes this incredible evocative image. And so this feels like we're getting very close to going from somebody's mind and doodling to a finished product. Is that what's happening here? That's 100% what we are trying to make. I mean, we were trying
SPEAKER_01: to make that happen. Like getting a great interaction, great getting a great communication with AI. So you can use it end to end from pre idea until a product that it's 4k resolution, extremely detailed and that you can use as a final result or for whatever client you have. Yeah.
SPEAKER_04: And so filmmakers are playing with it. Other folks are playing with the people who have had their their minds blown by this. You've got hundreds of people paying 30 bucks a month for this dozens. Where are you at as a company?
SPEAKER_01: So we have a waiting list of more than 200,000 people right now. We have people paying but right now that's actually an issue because we are still working on scaling this up. We are doing it gradually. We started rolling out invites during these past days and we are doing it very slow while we monitor how well the GPU cluster handles all the requests. And yeah, we had issues because people were paying thinking that they would get access right away. So yeah, only yesterday we got like more than 7000 or $8,000 of people paying and yeah, so all that people will get access of course, like either today or tomorrow. But yeah, like we're still not not letting them pay because we want to monitor all the cluster and we want to make sure that everybody can have a good experience.
SPEAKER_04: Yeah, and so I could see this working for graphic designers, people building web pages, people doing illustrative work. And now you could have, you know, if you think about just journalism, vector in journalism, there was an art department that would do illustrations. Now you can have the journalists themselves say, Hey, I'm doing this, you know, editorial about I don't know, US China relations, I want to do an image of the US as an American Eagle and China as a dragon and, you know, just start riffing and, you know, show the Pacific Ocean, whatever, put Taiwan in the middle and you know, I'm just riffing here like a journalist might and it could make an incredible image that would normally take an illustrator, you know, a couple of days and cost what would probably be low thousands of dollars and it could just be done by the journalist in 10 minutes. It seems like you're pretty close to that, huh?
SPEAKER_01: Yeah, we have we are almost there. And for people who spend a little bit of time on learning how to use these tools, I would say that we are already there. Like people are already making things that they can publish in like a blog post or like in in the news, etc. Where you can already get very nice quality.
SPEAKER_04: Do you have any news outlets using this for illustrations or cartoons or that kind of stuff? Because I was just thinking like the Dilbert guy, you know, he does everything digitally, Scott Adams, but now he could just talk to prompts and you could make a Dilbert AI and you know, with chat GPT forward, it could be making jokes to or at least giving you ideas for jokes, you could make a verticalized version of this and everybody could make Dilbert paneled cartoons for their organization or to do marketing. I mean, I guess if you allowed it and licensed it, it would be crazy.
SPEAKER_01: Yeah, yeah, totally. Like right now, we've seen it in a couple of places, I should start with some guys that they were working on a project. It was not exactly journalism, but they were doing like kind of a documentary of people that were tortured by the police in Barcelona, which it was kind of a little bit crazy. And they were telling me how this tool was pretty interesting, because they want images, right of the of these people that were tortured, they want they wanted them to create the exact places where that happened. I think that the main difference of what we are doing is that you have full control, like way more control over the final image, like you can really decide this object needs to be here, this person needs to be there, it needs to be in this position. And you can make this with very, very simple doodles, then refine it with the prompt and the AI will 100% understand what you mean. And you will be able to yeah, to get it like with way higher quality and with without much this is a composite documentary technique. If you
SPEAKER_04: can't get, you know, to recreate what happened, they'll do illustrations, moving illustrations. And there was a really great film, the kid stays in the picture about Robert Evans, I don't know, have you ever seen it? kids days? No, you should watch it. Just an incredible documentary. But in this documentary, they will show you know, moments in time, but they take pictures and they kind of layer animation behind them and kind of bring things back to life that, you know, they don't have documentary footage of and so this is a classic technique, but you it's expensive to create animation, right? You can even if you're outsourcing it to China or Korea has a big animation outsourcing business. Here, you could just make incredible illustrations yourself and yeah, fill in the blanks for people when they're trying to get a visual and that would typically be on a documentary budget. That could be like half the budget, a third of the budget could be making those maybe a third of the budget could be making those illustrations. If you're making a three or $4 million documentary, might spend a half million dollars a million dollars making that art and that art direction now, who knows, I think could be done for close to zero. Yeah, totally. And the
SPEAKER_01: same the same idea is going to be applied everywhere from product photography to yeah, like advertisement gaming, I think that we will start seeing AI everywhere because it just makes sense. You can you're going to be able to use this technology end to end and get the same if not better results than the ones that you're getting right now.
SPEAKER_04: Okay, this is doing static images. Is there an ability to do moving images yet or loop videos?
SPEAKER_01: Yeah, so we've seen people recording their screen while they while they move the shapes, which was very, very interesting. So we've seen like people doing these kinds of kind of even psychedelic animations, because you see like all these things changing all the time. But there is one model that got open source a few weeks ago, it's called Animate Diff. And we are already thinking on ways how we could apply this technique to this video model. And once that happens, we should be able to make a pipeline that works from video to video. And you're able to get extremely consistent frames out of your your like very simple shapes. So yeah, that should be possible very soon.
SPEAKER_04: So how did you get into AI? I know you went to Cornell University, I know you published the paper on animation. But how did you get into this? And how long have you been doing it? So everything started around 2017. So six years ago, and my
SPEAKER_01: story is before going to university, I was very interested in artistic things, especially on music, I see that you have a guitar in there. I've been playing since I was eight years old. And I had a music band through all my teenage years. And I was just like, yeah, recording music, making photos, making like all sorts of creative things, painting graffiti, all of this. At some point, I decided to study computer science in university. And in my third year, I got introduced to image processing. And with image processing, I started to learn a little bit more about neural networks, and artificial intelligence for image processing. Very shortly, I discovered that you can not just process images and detect objects, but you can also generate. So I got introduced to things like DCGAN or like StyleGAN, that these are these early generative AI models for generating images very realistically. And at that moment, something clicked in my head. And I went crazy to understand everything about how they work. I started to read a lot of papers, make a lot of implementations and use this technology creatively. And I guess what I saw at that moment, it was a new kind of creative medium. Like if I wanted to record a, I don't know, like a hip hop beat, I needed to learn how to use this program. If I wanted to like a 3D shape for a video, I had to use like two, I had to learn how to use Cinema 4D. And I was like in this loop over and over and over. And I think that with AI, what I saw is something that can execute your ideas for you, which I think that is the most important thing on any artistic or, you know, unlike the artistic process, I think that the really important thing is like the idea that is behind the process. And the execution is just like something that is in between and that enables you to get these ideas out there. So I think that since I saw since I started messing around with AI, I really saw this. And that's what I've been was done inside of Nvidia, I think. And it was in video. I
SPEAKER_04: remember there was a famous Uber engineer who did this person does not exist, which was a website that would make an a photo of a person that doesn't exist. And that kind of blew people's minds, because then all of a sudden, this concept of stock photography, where you have stock photography models, do things like, I don't know, pour a bowl of cereal, but you have to find a person and bring them to a studio and have them, you know, spend a day doing inane things like pouring a bowl of cereal or, you know, eating cereal from a bowl. Now the whole concept of stock photos is just, you don't have to do that you can just make one. And style again, it's still going Yeah. Or is it stable diffusion is just leapfrogged it so fast. Yeah, I
SPEAKER_01: think guns are not a thing anymore. I think that's it will diffusion jets, it was better like the main difference is that it's able to do everything whereas guns are only good at doing a single thing like you can get guns making faces or making cars, for example, but they will only be able to do that. And there's some new research that tries to fix that, but it's not working too well. So stable diffusion is an open source product. There's a
SPEAKER_04: company called stability AI, that is the company by some of the people who worked on the project or created the project, and they've raised billions of dollars, and they're going for the gold there. But anybody can fork stable diffusion and do what they want with it. Correct? Yes, totally. Yes. It's open source. Yeah. So there are a ton of startups coming out of that space, how many people are working on that project and explain to the audience with who are unfamiliar with open source, what the activity is within stable diffusion, and then how code gets committed and, and what that's like right now, because this is a very unique moment in time where there's a gold rush. There's so much creativity. There's so many ideas. What's happening to the open source project of stable diffusion? Because I don't know if they've ever had a project that's had so many people who want to participate. You tell me,
SPEAKER_01: yeah, I think that what is crazy about stable diffusion is that they are making accessible this thing that costs so many millions to create, right? Like, stability AI goes ahead and shares the weights of this model that they spend millions of dollars on GPUs in training, and now people can use it freely to do whatever they want. Once that happens, what you see is like a crazy amount of improvements like a year and a half, I guess it was August last year is when it got the first version got released. And I remember like at that moment, the best thing that you could try it was Dali. And we were all mind blown by Dali. And when stable diffusion appeared, it was not really that good. And it just took the open source community a few weeks to get this model to a point where the results were even better than Dali. And now of course, it is like so much better than at least the previous version of Dali. So what I think that is interesting about open source is like, it shows you the power of people of the collaborative work for making something work like people really, it's so crazy how we before like all the all the breakthroughs that came in the AI space, they were all coming from research institutes, or they were all coming from companies like Nvidia. And now it's crazy how a lot of the techniques that we are using in Korea to make the quality of the images better are coming from random people that who knows where, where they are, they are sometimes they just learn how to code because of this technology. But they are so passionate, they're so driven to make this thing work, that they go ahead and make it and they make it public to the community. So everybody has access. So this compounding of Yeah, like like this community effort, it's it's what it is what it is. contributing meaningfully to the code base at this point, do you
SPEAKER_04: think?
SPEAKER_01: Well, there are many code bases in the end, like you have on the one side, hugging face that I don't really know how many 10s of 1000s of people probably 10s of 1000s 10s of 1000s. Yeah.
SPEAKER_04: So explain to folks who are not familiar with what weights are and why, who understands and knows the weights is important. Because we do hear about open AI, which some people refer to as closed AI now, like who has the weights for the their models is important. Microsoft has access to them, nobody else does, is I think the public positioning here. So why is it so important that the weights are shared? And then what are weights? How can people think about those if they don't understand that concept?
SPEAKER_01: Sure. So in the end, these AI systems that are so successful, like stable diffusion, they are at their core neural networks. And the way how neural networks work is they get an input like an image, they pass this image to a set of weights or to a set of neurons. Like it's the same idea, right? Like we call weights what, because in the end, they are numbers from like they are like float numbers that they weigh the features that need to be generated, for example, or comprehended, and then you get the final result. So these weights in the end is like when you train these AI systems, what you do is you for example, in the case of image recognition, you have an image as an input, you pass it through all these weights, and these weights give you an output. And if there's a car in the image and the weights end up deciding that there's a, I don't know, a dog in the image, you will be able to update all these weights. So the next time they don't predict a dog and they are able to effectively predict a car. But essentially, this is everything that you're doing when you're training these kinds of models. You're just like making sure that you find that combination of weights that have knowledge and that can understand the input that you have and that they can give you the right output. So when someone release the weights, you can use all these all these knowledge that it was extracted with the with the neural network and you can, yeah, you can generate images or generate text, do text, understanding, etc.
SPEAKER_04: Listen, public markets can be volatile, don't I know it. And if you're looking for a unique asset class to diversify with, let me tell you about blue chip art. Blue chip art has historically been uncorrelated with the stock market. And Bloomberg reported that as equities dipped in 2022, blue chip art had its best year on record last year, the big three auction houses posted record high revenues of a combined $17.7 billion. But here's the problem. Blue chip art has always been an exclusive asset class until masterworks with masterworks. Anyone can invest in fine art without needing millions of dollars. This is because masterwork securitizes blue chip pieces, then sells the shares to investors and masterworks provides liquidity. To date, masterworks has sold over $45 million worth of art and net proceeds have been paid out to everyday investors, not billionaires. masterworks has more than 840,000 users and north of 800 million in assets under management, aum and twist listeners get special access to skip the waitlist. Just go to masterworks comm slash twist that's masterworks comm slash twist to skip the waitlist past performance doesn't guarantee future results. See important disclosures at masterworks comm slash CD. So one of the big debates has been on copyrighted images or just the training data. What is the state of the training data inside of stable diffusion today? It was trained on what and then how is it being trained as it grows? And then how do copyrights if at all play into that? So stable diffusion was
SPEAKER_01: trained with this data set called lion five p. This is a data set that comes from pairs of images and image caption that it was a script from the internet. And it's like these kind of crazy scrapers that they go all through the internet, and they get they get these, these pairs. So there's definitely copyrighted images in the data sets that it was used to train stable diffusion. And right now the situation is well, there are like several cases going on with mid journey, stability AI, etc. There's still not a verdict of where this is going to go. But yeah, my sense is that the really important thing is how you use this technology, not how you train it. And it's equivalent with what happened a few years ago with Google Images, that they were trying to sue Google because of showing like the copyrighted images when you search for something on Google Images. And in the end, like you need to answer the question of is this a transformative technology? Or is this a derivative one? If it's a derivative, like you are making a copyright infringement, you're making money in the same way that the creator of that asset is making money and you cannot do that. But if you realise that you can do this, that without this technology, it's impossible from a text generating an image. In this case, you could say that is something transformative. It was not possible before. So it's not making any copyright infringement and it should be allowed to exist. So we'll we'll see how that case works out. You know, for the in
SPEAKER_04: the case of the stock images, if I had a stock image library and it was used to train it, I would say, well, I should be able to train my own model. And I should have that opportunity to then make stock images based on the library of stock images that I spent decades acquiring. There's some sympathy to that stock image library holder. Yes. In the community, or is the community just like, hey, we can do it. It's too late. We can't unpack this. What is it? What do people in the community think? Yeah, I think that the power that this technology gained out
SPEAKER_01: of being trained off of all this data is superior than, than like all these stock photography companies that they had all all these images and copyright. For me, the really important thing is on the people actually spending a lot of time on creating certain things. Like for example, imagine that you are an artist like people, the famous, like, we all know, know people. Right now the AI got so good, thanks to having seen so many images of people. And that right now everybody should say well to say this thing, but in the style of people or this other thing in the style of people, and they are able to make money just out of that. I don't think that that's something positive. But I think that there's something very, very special and being able to say, okay, do this image in the style of people that mix it with the style of, I don't know, like Michelangelo. And suddenly you have something new that you were not able to have before. And that it can be 100% considered something artistic and something new. For me, it's like way more important to protect like the copyright of artists, then to protect like, Oh, yeah. So if you you if you evoked people, or
SPEAKER_04: you use people in the training set, it could just say, Hey, you know, you have to use the people version of this, just like, you know, there might be other versions made eventually Star Wars characters, etc. If you want to make Star Wars characters, and you want to evoke Star Wars and Darth Vader, you just pay the license. And so the industry has got to figure that out at some point. And I think it's going to be a negotiation because what a mess like, what would happen if it had to be retrained? Is that even possible? Or you'd be starting over if it had to take out all the people's if it had to take out getting images or whatever it was scraped on with that lion five, what would just that's holding back?
SPEAKER_01: No, that's kind of what Firefly is doing. For example, like other way. It's training their model using other stock, which is everything. It's non copyrighted images. And you know, like it's gonna have its use case, like if you just need like a regular image of a woman drinking a coffee next to the beach, like very basic stock image for that is going to be it's going to give you a great result. But if you are an artist and you want to give you like certain styles and you want to play with things that come from copyrighted images, you will realize that there's a very, very boring model, and that most artists are not happy with it.
SPEAKER_04: Fascinating. Well, listen, continued success on this is super exciting. I want to let the dev house get back to work. I can see everybody's out there having a great time. This is the spirit that built Silicon Valley and you know, amazing for America, but you're from Spain. Yeah. Right. Yeah. Barcelona. You're from Barcelona. Oh, one of my favorite cities, my Lord. Any city where you eat dinner at 10 or 11pm and call it supper. That's my kind of city. I miss Barcelona terribly. How's the startup seen in Barcelona? There are a lot of entrepreneurs there. I know that they they've had a little bit of a challenge with the economy and yeah, it's hard. It's hard. Like you have a
SPEAKER_01: lot of handicaps from the government itself. Like it's hard to hire, it's hard to fire. It's like legally it's hard to create a startup. It's not like in the US that in a few days, you can have your startup set up. But on the other side, there's a ton of talent. I think that in the US people are better at selling themselves. Like you find a lot of people that they are not actually that like crazy good. But when you talk with them, you will think that they are like the best programmer that you've made in your life. And in Barcelona, it's more like this person thinks like they are not that good. But when you talk with them, they are actually the best programmers that you've met in your life. So yeah, that's kind of what I've seen.
SPEAKER_04: I know there's a lot of red tape there to start companies and taxes and firing people is hard hiring people. Everything's just crazy. It's like France. And also, and if you want to raise money right now in in
SPEAKER_01: Spain, it's extremely hard. I was talking with friends that they are already doing 250k of annual recurring revenue, and they were getting valuations of 6 million.
SPEAKER_04: Well, if you think about that, that's if it was 300, that'd be 20 times revenue. Yeah, I mean, you can get 6 million probably coming out of an accelerator right now with a prototype. So or 10k a month in revenue. So yeah, it's it's that would be on the lower side. But if they just move the company, or they domicile it here and work from there and have that lifestyle, they could just become a Delaware seek work, which is I think what a lot of people are doing now.
SPEAKER_01: Yeah, yeah, that's what we recommended them.
SPEAKER_04: Yeah, absolutely. All right, listen, everybody go check out Victor's company. They're doing amazing stuff. It's k r e a.ai Korea.ai continued success and back to work there and order some pizzas or whatever you guys do whatever your job is. We're doing the Brian Johnson diet here.
SPEAKER_04: Oh, you guys on that Brian Johnson vampire everything you guys are just eating like a pound. What what's like a pound of what's in that? Potatoes a day? What do you eat?
SPEAKER_01: Yeah, so it's a the first plate is like a broccoli, cauliflower, mushrooms and lentils and a little bit more of virgin oil and something else. Then he's like a pudding with like a chocolate pudding with a lot of seeds and walnuts and and like very, very heavy. Yeah. And finally to dinner is a salad salad with mandarins. But yeah, he's like mainly it is like a 15 blueprint diet. And so how long have you been doing it? And
SPEAKER_04: does it make you feel like a superhero or something? What does it make you feel like?
SPEAKER_01: I mean, we're growing. Like, I don't I don't know what happened. But since we started with this diet, like suddenly the startup went up.
SPEAKER_04: Oh, you're saying that the performance of your startup went up. So is a direct correlation between the blueprint diet and the productivity at your company? Totally. Yeah. I go. All right. Well, there you go. Forget about your lifespan, your health span, lowering your biomarkers. Also, just get on the blueprint for performance. You write more lines of code.
SPEAKER_01: I just think like you we're not gonna die out of out of this this diet, right? So that's that's our mode. We don't die.
SPEAKER_04: Yeah, I mean, that is essentially what the goal is in Silicon Valley. We're doing all this work to make all this money to plow it into companies and experiments so we can live forever. Right now the entire world is absolutely appalled and at all of us at the same time in Silicon Valley that the central goal is to generate wealth, to create technology to eventually live forever. I mean, just I mean, it's kind of a joke. But if you think about what AI is doing, you're going to eventually have your blood and your markers and your body scan and AI is going to start looking at this and it's going to figure stuff out. Oh, yeah, it's gonna figure stuff out about life extension is gonna be crazy. Yeah, yeah, I cannot wait for that. I already have my order
SPEAKER_01: ring. This is very, very helpful to try your day today. But yeah, I cannot wait until we have like really powerful AI is to try your health. Yeah, it's coming. It's coming. It's also going to
SPEAKER_04: help them if you had to have surgery, God forbid, because you had something in your body, it's going to really make an incredible 3d scan and know exactly what to do. And the surgeons are going to be even more precise and eventually they'll be robotic. And that all that technology is in the process of being built, and then being distributed to everybody on the planet. So life expectancy could go way up. All right, listen, great job, Victor, and we'll see you all next time on this week. It's our bye bye. All right. You know, I've been on a health kick over the past year. And you know, I care about data driven solutions. And if you listen to this podcast, I bet you do too. So let me tell you about Fitbod. This is a data driven workout app that blends machine learning with exercise science. Fitbod creates custom dynamic workouts programs based on your fitness goals, your experience. And most interestingly to me, the available equipment, let's say you got a bunch of kettlebells, or let's say you're at some, you know, sparse gym at a hotel, or you're on vacation, you got nothing. Well, Fitbod will maximize your fitness gains by varying the intensity and the volume between your sessions and leverage the equipment you have or don't have. As the case may be, you can customize the length of your workout, what muscles you want to target and so much more. So let's say you want to get a 30 minute workout in and I want to do chest triceps and abs, but I'm staying at an Airbnb, there's no equipment, Fitbod can create a perfectly optimized workout for me based on these parameters, and it will do it for you to check it out. It's amazing. The design of this app is extraordinary. I was able to invest in it. That's how impressed I was with it. Fitbod takes the guesswork out of fitness, just open the app and start making progress. You deserve it get 25% off the pod subscription or try out the app for free. When you sign up now at fitbod.me slash twist. That's FIT BOD.ME slash TWIST for 25%
SPEAKER_03: off.
SPEAKER_04: Everybody, welcome back to this week in startups. As many of you know, if you're building products out there quality assurance QA testing is a massive market in software, but nobody talks about it. It's something that developers have to do. You might call them chores or a best practice. So test rigger is a startup that's building a tool that streamlines the process of software validation, aka quality assurance testing, and they help companies to use non technical users instead of QA engineers for testing, which cuts the cost dramatically and the CEO of that company is Artem Golubev. Welcome to the program. Tell us a little bit about test rigger and maybe educate the audience on what software testing is and how AI is going to is going to change that. Yes, absolutely. Well, first of all, let me describe the problem
SPEAKER_00: as you probably know when you're building a software eventually you would need to test it as soon as you have paying customers to make sure you actually don't break the functionality. Imagine that you're running amazon.com and your customers can't purchase products. That's a disaster. And that happened before. Some companies are losing hundreds of millions of dollars, literally. So oftentimes you can't afford it. So you do test now imagine you're testing if you do it manually, it takes two weeks oftentimes for a smaller companies or two months on the larger ones, to be able to retest all the functionality even if you employ tens or hundreds of people who will date that your system works correctly. That is extremely slow. People do want to speed it up to be able to move faster and automate testing. Moreover, you can't even have testing done manually in 2023. Because imagine where is the security vulnerability that we need to almost immediately fix in production, you have to be able to do the release ASAP. If your testing takes two months, you can't do that. It's a must to have test automation today for all companies. However, 70% of all functionality today in 2023 is tested manually. How is that possible? Isn't it a contradiction somewhere here? The problem with test automation how people are writing tests today is that we are hardcoding how engineers wrote the product yesterday in minute details, as opposed to how it should function from an user's perspective. And of course, it changes on a daily basis. And all that automated testing fails, instead of telling being able to validate if it doesn't work or not.
SPEAKER_04: Got it. So in an example, like Amazon, I might do a search, find a product, read the product reviews, add it to my cart, then check out, pick my delivery options, pick my billing, and then, you know, order my whatever USB cables from Amazon after going through that process. In the software, the engineers might have written little tests for how that might work. However, if things change, the test may not change. And the test is written from a software perspective, as opposed to from the user's perspective. So there's two ways to test this. One is to actually have a user go order the USB cables and go through the checkout process. The other is your solution, correct? Yes. So basically, the issues you might get with automation,
SPEAKER_00: for example, search button now code differently, it's changed color, it moved to a different location, the input for the search is no longer called search, it's called find the product, and so on support, right. And that basically, imagine you have 10,000 end-to-end automated tests, that starts from finding that input, and they can't find it anymore. So everything breaks and there is an engineering overhead to fix it again. So this is exactly what test trigger is fixing. It allows you to explain how your system should function in the normal language. And the sticker would execute various instructions, emulating exactly how you would do it as a human. And that brings the not only unprecedented test stability that you can run it and as soon as your specification is still correct, it will still work. But also, it allows non-technical people, all of those tens of hundreds of manual testers that companies already have today to be able to build this automation, mind you, about 10 to 20 times faster than even engineers could because you don't need to write any code, you don't need to go into technical details whatsoever, you just explain how it should work. And bam, it just works. Got it. So you describe or a human describes how this human
SPEAKER_04: process should work in your tool, you charge companies for doing that, and then they run your system against their product, correct?
SPEAKER_00: Yes, usually there are so called test environments, where people deploy and test were released before moving it to be available to everyone. And this is where we usually run those automated tests. How is AI going to change all of this over time? Are we going to
SPEAKER_04: be able to just ask an AI agent, hey, here's a website, please pretend you're a user who and test every function here and report back what's broken and give me suggestions for how to make it better? Is that kind of the ultimate future of this where some AI has been given the role of a quality assurance tester, and they just know what to do? Well, AI as of today can emulate what humans do. Basically, the
SPEAKER_00: starting point is to replacing human in human work. And this is what what we do right now, right? Instead of executing various tests manually, AI can basically execute those automatically. So much so that our customers can copy paste their manual test cases directly into our system and our system will execute those.
SPEAKER_04: So you'll just describe it in plain English, hey, go to this website, do some searches, put some things in the basket, click checkout, and then your software runs, I guess, in a virtual desktop where it loads different browsers and tests. Hey, what does it work like in Firefox, or Chrome, or Microsoft Edge, whatever it is, and you can do like multiple browser tests and different speeds of computers and bandwidth? Is that part of the testing process to to do that matrix of here are all the possible platforms, browsers and and speeds that you could be interacting with on the on the product?
SPEAKER_00: There is such an option. Yes. As you can imagine, if you want to run on more infrastructure, it's becoming that much more expensive. Second browser will double your costs of third little triple and so on and so forth. But yes, of course, you can do that. That's the whole point of testing is making sure that your customers can use your system not only from Chrome, but also from Safari, including Safari on iOS, and from Firefox, Edge, Internet Explorer, sometimes, and so on, so forth. Fantastic. And so in terms of getting this product into
SPEAKER_04: customers hands, and being a startup, I know you went through alchemy accelerator, Y Combinator, how do you get new customers for this product? And is the customer base ready for this sort of paradigm shift in testing? Well, we have multiple channels and the onboarding new channels.
SPEAKER_00: So of course, we started as I guess everyone else, we have some outbound, we were able to sell to people this way. And they figured out okay, so now let's do some marketing and marketing works to a point where today with zero investment in marketing, we're getting more inbound business than we have from outbound. But now we also onboarding onto new channels as well. For example, in four is our customer, it's top five largest ERP systems on the market similar to SAP Salesforce and such. And we have 90,000 enterprise size customers. So we're working with them on partnering to help their customers to test their implementations of our ERP system. And that would become an example of one of the channels we are onboarding.
SPEAKER_04: Now is Microsoft also competing in the space? I know Microsoft's big on AI, they obviously have the co pilots, Azure, the relationship with open AI, are they kind of building in this kind of AI testing yet? And how do you look at the competitive landscape and why people should use you versus using maybe something from Microsoft?
SPEAKER_00: No, Microsoft is building some tools that we either use right now or will most probably use in the future, for example, to be able to better automate working with Microsoft products such as Word and Excel and so on so forth. Specifically, however, they do not have that system which overall can work with any UI whatsoever like ours and execute those, this kind of stuff in plain from plain English, very high level. How is AI just generally speaking, if we open up our
SPEAKER_04: discussion here, how is AI impacting how software is being made today? We hear about co pilots, we hear about, hey, I'm going to put my entire environment for my company into a verticalized AI and co pilot. So my code base is part of the language model helping, you know, whatever the 10th developer 11th developer on my team get onboarded. And sometimes I need to explain, maybe to the new person how the code base works. And AI seems to be doing a pretty good job of that. How is AI changing, just developers and dev teams, operations today? And then how do you think it will change it in the future? Current state is you probably have seen the presentation by
SPEAKER_00: GitHub, like literally a couple days ago, they already make engineers 55% more efficient, which is mind boggling. However, I believe the future is AI agents that would do stuff instead of needing humans to to engineer things. And test rigger is the first example of such an AI agent where we do not generate the code, we do not need engineers, you just write in English how it should function. And test rigger will execute it for you. I can show you a very quick demo if you'd like. Sure, give us a quick demo. I'll do it on best by that calm. Well, on the best by website. I see that. Yeah. And this is the
SPEAKER_04: test rigger suite. We're creating a test suite. It's a
SPEAKER_00: set of suite. Usually, you should be what operating system
SPEAKER_04: what browser you're giving it credentials to log in. Yes, yes,
SPEAKER_00: yes. browser. Yeah, we can generate the test. This is what you were talking about. Hey, how do you do it autonomously? Where you go? What is the description of the test? Yes, we provided
SPEAKER_00: the description of the system saying that it is ecommerce websites and electronics and it came up with suggested we selected Hey, give us so you said hey, this is an electronics
SPEAKER_04: website. Give us some tests the test that came up with the first one is browser electronics category and verify product listing. Second test at Kindle to cart and validate cart contents. Third test proceed to check out and confirm purchase success message. So these are tests that it's generating the AI is suggesting these How did it get those? Is that just a language model? Are you trained a language model on what tests are typically done on a website? Or specifically ID cameras website? So this is coming from LLM. Yes, it is suggesting some
SPEAKER_03:
SPEAKER_00: stuff like this is pretty, pretty high level. This part didn't didn't need to be trained. It's just our top 24 directly. How did the language model know to suggest those
SPEAKER_04: three things? And how accurate are those three things as tests? Well, you can modify those, right? Yeah, if you find
SPEAKER_00: out that they are not exactly what you would expect. However, it had been trained on the full internet and internet has a wide variety and library of everything around it. So you can know basic things based on the description kind of common sense where interesting, where the sticker provides the largest value and I'll show you let's add another case let's say test editing record. So test adding to cart Okay, yes. So I'm the
SPEAKER_00: say things like find and select a Kindle. And it's to shopping card. So I'm trying to come up with something that you would see typically in a typical test case. Alright, so you told that
SPEAKER_04: Hey, find and select a Kindle to buy on Best Buy, then add it to the shopping cart, proceed to the cart, obviously clicking on the cart button, and then check the page contains Kindle. So this is all written in plain English, you don't need to have a developer do this. And then it I guess it's looking at these commands and saying, Hey, I don't understand these commands. So let me let you map those commands. Is that what it's doing there? Yes, that's not be selected. Hey, use AI to execute those
SPEAKER_00: commands. Do not do anything else or just use AI directly. Whereas no specifications are specifically and what system is doing is will kick off the new server with oh yes, that we have selected, we'll start the browser that we have selected. And that it will go step by step through the screen.
SPEAKER_04: And is it making a little video there for you or just taking screenshots along the way of each step? What is it doing there?
SPEAKER_00: It can do both. We did not select the recording a video. So it's just taking screenshots. But it can of course, create a video. As you can see here, like it decided to enter Kindle into search, and click on the search button.
SPEAKER_04: Yeah, so you get the actual evidence of it working. Yes. So and those are commands that it came up with based on
SPEAKER_00: this prompt, or two commands in this case, enter Kindle and search. Second, it basically said it's done with that prompt. So proceed to the next prompt, which is at the shopping cart here, just click on that to cart. And then before proceed to the cart is clicked on go to cart. So it got to the shopping cart out there. And drumroll, you would need to confirm that it contains Kindle in this cart.
SPEAKER_04: Fantastic. So this is pretty groundbreaking. Now you can have a non developer write these use cases, test them and get the report back. Absolutely amazing. Well done. And at every aspect of what we're doing in building products is being done by AI now or helped in some way. So it looks like this is going to save people. I don't know, hundreds of hours a month, thousands of hours a year at an average ecommerce website or startup, correct?
SPEAKER_00: Yes, because it basically kind of common sense, right? What's important in software testing is the main knowledge, right? So yes, AI can come up with common sense suggestions for test scenarios, and so on and so forth, and figure out based on how your system works, certain things, but only you, as a domain expert would know what is truly most important, what needs to be tested and how, but you might not necessarily be an engineer, you might be an expert in the product, but not necessarily an engineer. And you shouldn't be you should be able to just explain how the system should work. And it should execute it for you. That's the kind of division of labor between the main experts that are humans and just machines that do stuff for you.
SPEAKER_04: Amazing artem great job, everybody check out test rigor calm t s t r i g o r and you can follow them on x formerly known as Twitter and you can follow our team on Twitter a r t e m t w t r well done and we'll see you all next time on this week and startups bye bye
