#8: Mindstudio, the fastest way to build AI-powered apps

You've heard about large language models. Today is the one year birthday of ChatGPT. It's really crazy, but I'm seeing a bunch of little startups that are starting to move us toward an operating system, a way to do a lot more with AI than we could just talking to an LLM. And so I'm really happy to have our guest today. Who are you?

So my name is Dimitri Shapiro. I am CEO of a company called uai. You can find us at Y O U A I AI. We'll talk more about what we're building. But prior to this I was at Google for four years before that running product on three machine learning teams that were crunching all the data that Google has about people and how they use the Internet. Prior to that I was the chief technology officer of MySpace Music for those that are old enough to remember MySpace. Prior to that I built two other venture backed companies. One was a company called Veo Networks, which was a major competitor to YouTube. Prior to that I built a venture backed cyber security company called Econic Systems, Perimeter Security, Enterprise Software. And from 95 to 99 I built the web team at Fujitsu, big Japanese company. So I'm an old nerd. Yes, yes. And just Sean.

Yeah, and my name is Sean Thielen, Dimitri's co founder, CTO of this thing. My resume is much quicker than this. I've basically been building products with Dimitri since I graduated school and there was a period where I lived in his garage which was fun and stereotypical startup thing.

So on this Dimitri, what? On this one year birthday of CHAT GPT, what are you building? Tell us a little bit about your company, introduce what you're trying to do.

Yeah. We are as a society obviously really excited about this new era that we're experiencing together, which is our ability to interact directly with AI and do all kinds of sort of new things that we've never had the capability to do before. And that's extraordinarily exciting. And there are many wow moments where you can just go to ChatGPT and start typing with your thumbs. You are a blog post copywriter with 20 years of experience. Please write me a blog post about how AI will disrupt medicine and voila, out comes streaming a blog post. And we marvel at this thing. And we should. That is amazing. But once you get past the novelty of that, you start to realize what do I actually really do with something like that? Really do. I type like these long commands with my thumbs and I have to learn to be like this prompt engineer thing and that seems like it's more work to do that than just to write the damn thing. And by the way, when it writes it, whose style is it writing it in? It's not writing in my style. For it to write in my style, I have to teach it about my style and how do I do that? And so we realized this shortly after GPT launched and realized that the right way, we believe, to think about all of these models, whether they are language models or image diffusion models, video code, whatever, is not for humans, for end users to go and use them directly via chat interfaces, but to relegate them to being back end services that are abstracted from end users by what we call the application layer. And that's what we are building, it's the application layer of AI. Another way of putting it like the operating system of AI, which leverages these incredible capabilities in the intelligence layer, which by the way are evolving so rapidly that no human could possibly ever try to keep up and leverage that. And so they have to be somehow in the back end so that applications can leverage them, but not individual end users. And so anyway, so we built this very robust and broad platform that creates the application layer of AI, allows anyone to show up, watch a YouTube tutorial and learn to build AI powered applications to solve countless things that you might want to solve. We launched three months ago. Today there's over 9,000 of these AIs. We tend to refer to them as or AI powered applications that have been created. They're running in a bunch of enterprises and small medium sized businesses for consumers, for fun, whatever. And so democratizing building of AIs, leveraging all of the models that are relevant. We're model agnostic, so Today we support OpenAI's models, I think six of them now, Sean, five or six OpenAI models.

All of them.

All of them. Three anthropic models. We had support for Claude 2.1 15 minutes after it launched. Our users were able to use it. Meta Llama, Google, Palm, Open Source, Mistral, various others.

And there's a lot, a couple more this morning.

Yeah, yeah, a lot of models.

There's models coming out every day. Yeah, Tell us about what is the environment you're trying to build. How does it differ from Chat GPT, because you're hinting at it. But what with who is this aimed at? Who needs to use UA?

Yeah, again, about 100 million people it seems have joined ChatGPT, maybe more joined, but they had 100 million monthly users by the way. And then traffic started declining rapidly for that and so if you want to go and play around with ChatGPT, or probably listeners here have already done that. ChatGPT is amazing again in this ability to type in commands and get responses back. If the average person out there, the average business person or the average consumer cannot be expected to join ChatGPT, learn how to use these things and type commands and really take advantage of it. So the opportunity that we're presenting to fellow nerds is to show up and build these again. AI powered applications, they're just web apps. You don't need to write a single line of code, you need to learn how to use mindstudio. Our platform lets you do it. You watch an 18 minute tutorial and then in an afternoon you become a master at being able to build AI powered applications for any kind of use case, enterprise, team, consumer, whatever. So then the rest of the world can simply use these AI powered apps that you've built to get their work done or entertain them or do whatever. And now, so chat OpenAI launched something a few weeks ago called GPTs, which some people that know what mindstudio does has. Oh, it's like Mind Studio. It is in a tiny little way. The GPTs allow you to within ChatGPT, within the walled garden of ChatGPT, package prompts, give them a name and so you can come back to them easily. There's like bookmarks is the way I describe it, bookmarks for ChatGPT. And they said in the future they will allow you to share those GPTs with other people. I would say that's like social bookmarks. Robert, you must remember delicious when bookmarks became social and that was amazing. That's what GPTs in its app store is, is GPT bookmarks and social bookmarks. Social AI bookmarks. But again, what we allow you to do is to create outward facing applications so regular people can use them without ever needing to understand anything about AI models, without ever needing to understand the nuance between is it Claude or OpenAI that I want to use or how to write a prompt, all of that gets abstracted. We nerds do that work and create end user products that end users can use and get their stuff done.

And so this, businesses that are, this is for businesses who can want to build functionality into websites or intranets and you're going to wrap everything up and make it, make the app work without having to know all this complexity or sign up for ChatGPT and pay your 20 bucks kind of thing.

Yeah. So today again, we're already seeing over 9,000 of these things that have been created there are AIs being created by just regular people that work inside of enterprises that are. One person that I actually interviewed on my video podcast, Michael, sits in a platform as a service cybersecurity company. His job is sales enablement. So his job is to make sure the salespeople, sales engineers and customer support people are well trained and well equipped. And so he's using Mind Studio to build AI trainers that salespeople engage with utilizing the data from their enterprise. And these trainers assess the salespeople, assess their deficiencies and then help them level up. Another one does that for sales engineering, another one does that for customers customer support. And then he's using it, as he says, like duct tape, like AI duct tape, to just solve all kinds of problems inside their enterprise. We're seeing that kind of bespoke use by somebody in an enterprise, brings it in, learns to use it, and becomes basically the AI superhero in the company. And we joked about it that now people come to him and say, oh wow, you're like this genius of AI that's able to do this or just.

You do this cross models, right? You can use a LLAMA model or an OpenAI model or some other model that comes along in the future, right? You're trying to build this kind of offering system across models, not be stuck with one model. And when I worked at Microsoft, we had a rallying cry, take no dependencies on another company, right? And using an offering system like what you're doing lets me be free of just one model. I could switch easily to another model to build my little app, right?

Yeah, or use multiple models.

I was going to say use, use multiple models within the same app, depending on what it is you're trying to do. You could imagine some kind of app that does some complex thinking and you want to fine tuned corporate specific or something model to do the initial data crunching. And then you want just a friendly chat model to do the chatting. And you want it to be faster and more performance. So it lets you leverage the most apt model for whatever the specific subtask you're trying to execute.

And I keep hearing about developers building architectures of AIs, clouds of AI, orchestras of AI. As one guy put it, he's building one for the Australian banking system, which he'll have dozens of models, some of which are checking each other's answers, right? Trying to reduce the confabulation or hallucination problem of these large language models, sometimes spit out some right. And give you an incorrect answer. You Have. If you're building something serious with consequences like a business app, you have to make sure you reduce as many of those hallucinations as possible, right?

Absolutely. Go ask ChatGPT to do your work for you and voila. This is silly, right? Again, if you want real applications, it's not just please do my work that doesn't make sense.

I think also if you want real robust applications in enterprise and for business, it goes so far beyond just LLMs. Like, even if the LLM is the kind of underpinning or the core technology at the base of it, that's like saying you're going to build everything on MySQL or Postgres, you have to do the rest of the work too. Like you still need data storage, you still need front ends, you still need all the security and access and deployment and all of those pieces. So by being able to give that to everyone, I think you can actually build real meaningful applications instead of just some magic genie that's hopefully going to solve all your problems.

On the other side, since I like thinking of you as an operating system, on the other side of the operating system, Chat GPT can't talk to my phone, can't talk to the GPS in the phone, can't talk to the sensors, can't talk to the screen, you can't ask chatgpt, hey, brighten my screen. It doesn't know how to do that. It's not hooked up to any Android or iPhone screens. Right. It's not able to talk to the APIs in the device. Are you guys thinking of heading that direction when you're building an operating system to be able to talk deeper to the hardware underneath you.

So there are already people that are taking Mind studio and building AIs with it. And again, when you publish that AI, the AI is a web application. So it's got a URL, it's a web app. You can pass it parameters as any web app can, and they're taking those and they're wrapping them in native wrappers for iOS and Android, publishing them to the app stores. When you do that, then your wrapper has access to the operating system in various sensors and then you could hook it up to pass data to the AI that is wrapped and doing so. It's already possible. We don't have any plans to complete that loop ourselves here. That's not what we do.

But it's really the other thing that's potentially interesting there too is we have the thing that we call custom functions, which basically allows third party developers who do know how to write code, to write kind of these JavaScript functions that execute in a sandbox that can do all kinds of things. So we've seen people do that for like home automation stuff. Like anything with an API, you can build a function for it and make it available to the community.

Or your lights, these lights have meters, right? You can talk to them.

Yeah, yeah. I think the kind of people who. This is a sort of weird tangent on it, but like the kind of people who like nerding out over home assistant and connecting all these things, like they're the ones who actually want to build out the workflows and customize and fine tune things. Just ChatGPT is, is too magic and too obviated and hidden. It doesn't give you access to that lower level of parameters and customization that you actually want in order to get the thing to reliably do what you want to do.

Yeah, the other thing that.

Why didn't you guys start this company? What caused the foundation of this company? What, what's.

Sean and I are nerds. And when we see things going on in the world that we get excited about, we have this hunger to go build stuff. And so it was really as simple as that, where it was clear to us that these things are amazing. DOS, the AI version of DOS, which is ChatGPT, cannot be the thing. And that we like building platforms we actually don't like. Sean and I don't actually like building products. We like building platforms that other people can use to build products. And so this was obvious to us and we just built it.

Yeah. Tell me a little bit about competitive landscape. I know there's companies like Abacus that build. They have a large language model, they have a vector database, they have a whole system to make it easy to build apps. How do you see the competitive landscape for this kind of stuff happening? And how do you differentiate?

Sean and I have another, I think weirdness, perhaps that's unorthodox, is we can call it that we're not proud of it, but it's reality. We tend not to listen to other people's music, meaning we tend not to pay attention to what other people are doing. One, because we're busy doing our thing. And two, I think it just confuses people. I read Jimi Hendrix was like that. Like he didn't listen to. To music other than Zone. And unfortunately we're the same. So I'm not sure what these things are. We see the obvious ones that sort of come in the periphery, which is you've got lots of people that are building again, in the intelligence layer and they're building their own proprietary models and there's lots of interesting innovation there. And then some of them, like, again, OpenAI, have built like these interfaces to these models or Even like these GPTs that really confuse people or where are you playing? Are you in the application or are you in. In the intelligence layer? There's lots of companies, it seems, that have built entire companies on a use case, like Jasper or Writer, or like all of these companies that said, oh, there's now this capability. Let me complete the loop for a use case. And so again, we've seen a bunch of those kinds of folks. There are libraries at the development level like LangChain that we stumbled across where if you're a developer, this is a thin abstraction for accessing the APIs. Not sure, again, very different sort of layer at which you're engaging and for what you would use. So I think very different users. We actually haven't seen anyone cross our periphery that we could say, oh, that's similar to us. But we're certain there'll be plenty of people that will follow our lead and build stuff.

What's the power of what you've built? What are some of the fun things that develop? First of all, you don't have to be a developer to use you. It probably helps in some understanding of what's out on the Internet and what kinds of APIs are out there, what kinds of things you might do. But tell me, what are the coolest things that you've seen your customers do? Maybe as a good question.

Yeah, so maybe I'll answer two questions. One is, what's the power of this? And I think the power is this sort of accessibility democratization of creation, that any person, without needing to write any code in an afternoon, can become a master of a new skill. Again, AI duct tape. The ability to solve their own problems or problems in their enterprises, their teams, whatever, by learning to use this new tool. That's extremely empowering. And so I think that's the biggest thing is this accessibility. And also the time required to build these things is in minutes, not days or weeks or anything like that. So you shrink the time, you allow anyone to be able to do it. And so that's obviously the magic sauce. And then all of this, like, abstraction of models and model agnosticism and all that stuff that's like, interesting and powerful, but the output of it is like anyone can build stuff and build it extraordinarily rapidly and diverse set of Things. So that's that. What interesting things we're seeing people build. First of all, we have a showcase on our website and you can go to mindstudio AI or Y O U A I AI. Both of them are so. And you can look at the showcase and you can see again, thousands of things people have built. So I think sort of beauty is in the eye of the beholder. All of those things in many ways are like fascinating. There's a parenting copilot. So there's code copilots, and that's amazing. And code copilots have transformed development, making junior developers able to perform as senior, making senior developers, whatever, ten times more productive. That's insane. It's amazing. Awesome. Why should developers be the only people that have copilots? That doesn't make any sense. Parenting is a big job. I have five kids under 10. Okay, yeah, that's crazy. You're on mute, Robert. I think. But the parenting copilot is an amazing tool that allows me to. When you set it up, it asks you what are the first names of your kids and their year of birth. So now it knows how old my kids are. And then from then on I can just give it situations. And because it's read all the books on parenting and all the studies on parenting and all of that, all of a sudden it's able to basically give me step by step instructions for each kid. Here's what you say to Una, here's what you say to Diego, here's how Noah's feeling. That's incredible. No need to ever read parenting books. And nobody got time for that. You get these things or you get now books that are being taken because they're all digitized anyway. And so people are. Authors are creating companions to their books. Well, they're publishers now. They're building entire new businesses they haven't launched yet, but will soon online studio that are taking over 1500 of their textbooks and creating companions for each one. So when you buy the textbook, you also can subscribe to the companion. The companion can paraphrase, can answer questions, act as a master practitioner. There are again all of these enterprise use cases I told you about the training of sales agents and such. I'm not going to mention what government, but there's an IRS type of a thing of a foreign government that's using this now internally like random thing. So we're seeing like a real giant diversity of use cases.

We watch an 18 minute video to, to learn how to use the system. Give me a taste of what kinds of things I would need to learn like a chat G. I have chat GPT, I know how to talk to that. But now I want to build an app for my business or for my life. What do I need to know to move from chat GPT to using what you're doing? Mind Studio.

Totally great, great question. So what you Learned by Using ChatGPT is hopefully is that depending on what you type, depending on what your prompt is, you'll get a response to anything you type. But if you create your prompts in a certain way, you can get the model to be more precise, to think from a different perspective, things like that. One of the things you can take is leverage that prompt engineering training that you got by playing with ChatGPT and that's part of building an AI powered app using MindStudio. We too have this prompt things, that's one. The second thing you get is you get to choose what model you use and you can instantly switch between models. So it's just trivial, it's a dropdown. And so learning a little bit about the differences between what all of these models are, you can read about it or you can just experiment with them really easily and start to understand that different models should be spoken to, prompted in a different way. The other thing you need to learn is how to create multi step automations that can ask the user of your AI powered app questions, present interfaces to it without talking to the LLM first, but just get user input. Let's be specific. Let's say I'm building a blog post generator and using Mine Studio behind the scenes. The blog post generator is going to use GPT 3.5, which is a really turbo, which is a really fast model, much faster than GPT for example. Right. So that's what I'm going to use. And so I choose the model and I create the prompt that explains to this thing that it is a blog post generator and how we want the blog posts formatted and various sort of constraints that it should have and all of that. And then I need to really think about like the user that's going to be using it, the different kinds of users and what data might the LLM need to be able to generate the types of blog posts that would be compelling to the users.

If you're building blog posts for orthodontics, it's different than building a blog post for a lawyer, right?

That's right. And so your app could be actually a generic app. And one of the questions it could say is who are you? What's your profession, what do you do another one could ask another of these questions. Could say, please provide to me a bunch of samples of your writing. So just upload a bunch of samples of your writing so I can learn how you write. Another thing it may ask so is like personalization questions. Your profession isn't changing all the time and your writing style isn't changing all the time. So you can create these onboarding wizards for your AI. Just like mobile apps have an onboarding wizard that you engage with first, these AIs can do that too. So you can collect a bunch of info from the user in onboarding and then for each use that they use to generate a new blog post, you might ask them a whole bunch of other questions. What is the topic of the blog post? Who is the intended audience of the blog post? Are there any other things that you'd like me to consider when writing the blog post? Do you want me to scare the audience? Do you want me to inspire the audience? And so these are the things that you as this developer of this AI need to think about and learn how to create these front end interfaces. In Mind Studio, it's trivially easy. Again, 18 minute tutorial will teach you all of that. So those things then you need to learn again, you'll learn it in this 18 minute tutorial is how to leverage your own data and how to give your AI more data, not just use its knowledge of the world, but upload hundreds of files, for example of different types that it can then use to reference in creating responses to the end user. Some people may have heard this term rag retrieval, augmented generation or you know, these things that have to do with vector databases. We trivialize all of that for non technical users. And so you get to do rag without knowing how to spell rag.

This where I'm starting to learn a lot more about the Internet again because ChatGPT, when I asked it for the hours of my haircut place, it went off to three separate websites like the Chamber of commerce website, the Supercuts, which is the retailer I was asking about website and some other website and grab data and then analyze that data before showing me the hours to make sure it had the latest hours. Can you build apps that do that kind of thing? Go out to YouTube and find a video that has a certain kind of bio in it or I think you showed me some of these demo earlier when I was meeting with you, right?

Yeah.

What kinds of Internet apps can you build and talk to the Internet and have it gather data and analyze websites or tell me a little Bit about your philosophy there?

Totally, yes. Again, part of the things that you learn in Building AIs is to use these multi step automations. And in these multi step automations you can do a number of different types of things. You can ask the end user for input. Again we were talking about what's the blog post about and things like that. You can then make calls to APIs, whether they are these different language model APIs or other types of models or any API on the Internet. And we have a system as part of mindstudio called Custom Functions that allows developers that know how to write JavaScript to create new modules to extend the capabilities of our platform so that non technical people can simply use them in their automations. And so there are already a bunch, maybe a couple of dozen now of these modules that allow you to call Zapier and allow you to call Google and allow you to like interact with all kinds of third party APIs to get data in, process it, modify it, send it to this LLM for doing something, bring data back, send it to another LLM yesterday, publish the video, just.

Add a quick clarification that it's more than just APIs, you can also just scrape websites. That's one of the amazing things with LLMs now is that kind of you don't need a structured API anymore. It can parse the website and get the data you want back. So you can do both the structured API calls and also just scrape a website and magically turn it into an API and get data from it.

Yesterday and we'll link to it, we'll provide a link, you can link to it. Yesterday we published a short Little tutorial on YouTube. Three minute tutorial that does the following. It presents the user with just a text box and says enter a URL, you can put in any URL and then it takes that URL and it makes a call out to it scrapes the URL and gets the data back from the URL and then passes that data to Claude to Anthropic's model and says given this content here, create a spec for a game about this content and create a game spec. And then it returns that and then it passes that to a code model and says take this spec and write a Python script to implement this game that was just created by this other model. And again you can watch a three minute demo that explains to you how to do all of that. You never have to write a single line of code.

Now you can build calculators, games, all sorts of even the supercuts example I gave you, ChatGPT answered me about the hours, but it didn't hook up to the reservation system that they have. That's an API that it's not hooked up to. So I could use your system to hook up to that API and make myself a little reservation app for Supercuts. Go and embed that into the Supercuts website.

That's right. That's the other thing we didn't talk about is that these things are web applications. They're embeddable, they're on hundreds. Sure. Over a thousand now websites that. And in fact today we launched something which is a visual customization so you can change the colors and styles of it so it looks really nice inside your website, embedded in it.

Yeah. So there's going to be a whole bunch of people who create these kinds of gluey apps. Right. They're not really writing an app from scratch. They might be integrating what has already been built into a new web page. Or have a copilot for haircuts.

Right.

If I was working on Super Cuts, that's how I'd be thinking. It's like how do I put a AI on my web page that will help customers go through the process of signing up for but a reservation. Right.

Yeah. And take it really zany to the next level. The right thing to do then is to have millions and millions of those and then orchestrate some kind of meta agent that is able to interact with those. So when you do go to ChatGPT and you say hey, book me a haircut, what it's actually doing is it's using that other agent that was created to do that to then get the result back.

And that's where you start really acting like an operating system to me. What's an operating system? And a, a mouse move call to an API call to a lot of DLLs underneath windows. Right. That was the old operating system. This new operating system is different, but similar.

Similar. That's right. Yeah. Like this intelligence layer where we used to have to write like all the exact connections down to the to silicon to be able to do it in an operating system. The intelligence layer says I got you. Don't worry about any of that stuff. Yeah, you just write the logic, tell me what you want me to do. Write the logic, write the user interfaces for the end users and the rest is taken care of by this. And again, with these multi step workflows you're in control. You can collect information from users, you can log it, you can pass it to any Third party system that you're integrating with in five years, you guys.

Are going to build Jarvis, right? So I'm just going to talk to the system and it's going to have thousands of different things it can do and it's just, hey, build me a website, right? I might be talking to my, might still be talking to Siri or ChatGPT and I might be going off and talking to thousands of different things to do different tasks. Right?

Yeah, yeah. Meaning we're not going to build it ourselves, but we will enable the building of all of these AIs that do specific tasks. So that then to Sean's point, your proxy in this case, as you said, Siri can say I need, Robert needs to do this, hit the right endpoints to be able to do all of this stuff. And so Woodford's stating the easy sort of enumeration and creation of all of these made just for purpose AI's today.

Yeah, I think the operating system analogy is, I think useful in understanding kind of this horizon as it exists in the industry. Because as Dimitri and I see ChatGPT and all of those today, that is, that's great. You've got the command line and you can write shell scripts and these prompt guides are basically like you're getting the magazines and you're typing out the code to make your programs. Like we're in that era of comput right now for this new world and what we're doing is taking it the next level to whatever Windows 95 applications, things like that. And so what you're imagining is I think a little further down the road, but this is, I think the trajectory that the world is moving toward.

I dream of a day where I have a pair of glasses on. I'm just talking to AI all the time, right. And it's doing stuff, hey, build me a website for my business or hey, answer that guy's email. But we're starting to get tastes of that. It's just some of it fails. And so I'm looking at companies like yours with a high interest.

When I was at Google for a moment, I think for a quarter, I got thrown into product managing. Google Glass. Yeah, and I know Robert, you remember Google Glass.

I do.

And there have been pieces that have been around of this thing. What was missing is the intelligence layer and these connectors. And once this again gets a little bit more mature than a lot of old ideas that didn't succeed can be revisited and could potentially now succeed.

Tell me a little bit about the company. How did you get it funded. Are you looking for funding? Tell me about the business side of this house.

Yeah, so the company is actually seven years old. So Sean and I started it in September of 2016. And so this is another platform from the company that we started in 2016. Prior to this, we built another platform called K O J I. Koji is a creator economy platform that is used by over 700,000 creators to engage their audiences on various social media services, monetize them, et cetera. And so we built that. We've raised 36 million over that time and actually can't say much about Koji at this moment, but you might see it in the news soon. And yeah, so we're focused on this Mind Studio product now for almost a year, nine months really.

Very cool. Before get you to give me a demos. But yeah. Anything else that I should be asking? Because it's a new world. I'm trying to figure out the right questions to ask about this.

No, look, I think your questions are spot on. I think again, the really interesting things about this is once you give this capability to everyone, all kinds of things get developed that you wouldn't have ever thought to develop. Right. This is the same thing as we've seen on the web. Once you allowed everyone to publish, you had all of this content being created that would have never been created. But because it's crowdsourced and because the barrier is so small that it can get created. But in fact, creating a website is harder and takes a tremendous amount longer than building an AI to launch, which on average takes less than an hour. And so I think again, that's where the real excitement is going to happen is in the diversity of use cases where we'd look and we're like, oh wow, I wouldn't have thought of that. Like the parenting copilot, on one hand, once you grok it, you're like, duh. On the other hand, like, why would you have thought about that?

And so academic consultants may be an interesting example. Dmitri, what's that? The Academic Insight people might be an interesting example.

Yeah, there. There are two people, Jessica and Kimberly, who are consultants to higher education, universities, colleges, etc. To faculty, supervisors, staff, PhD students. Two months ago, they were consultants that sold their time and their Expertise. They discovered MindStudio and on their own again learned to use it and have now created over 80 AI applications. You can find them on academicinsightslab.com and they for academics to use. So a suite of AIs for academics to use. They sell access to that suite for $59 a month, recurring in the first month. This is public information that they shared with us on a video that I did with them. In the first month, they had over 60 subscribers, primarily to this $59 tier. They have another, smaller tier, three universities asking them for enterprise pricing. So here again, you have these two women who have never written a line of code, claim not to be technical, but yet have now built over 80 AIs and have built a SaaS business in a month. And so, like, these kinds of dynamics, we think, are the fascinating part of all of this.

Yeah, and I would also add to that that I think there's a lot of noise and hype and frankly, confusion around AI right now. And I think one of the most powerful things that you can do is just play with it and make things. And it really, I think, demythologizes a lot of it. I think it makes it a lot more real and concrete, and I think it makes kind of the broader social discourse around AI a lot more concrete and practical. Once you've actually gone in and played with these things and used them and know what they do and where they fail and where they're strong and where they're weak, like, is it, I think, just engenders such a different relationship with this technology than a lot of people who are confused by trying to read and keep up with things.

Totally agree. That's what attracted to me is I can switch models real fast and try things and learn how it works. Right. And get demystified and understand what's going on. Thank you for what you're all doing there. Where do we all learn? Where do we learn more about you? And try this out?

Mindstudio AI. You just go to mindstudio AI, sign up for free. You can create all of these things completely free of charge. You can create them, you can deploy them, people can use them. We do have paid tiers that if you want to take your AI and embed it in your own website, you have to be on a paid tier. If you want to charge users, end users, for using your AI on a monthly basis, you have to be on a paid tier. If you want to get rid of some of our AI UAI branding, then you need to be on a paid tier.