Loading summary
Transcript1 lines
- [00:00]
A
Today on the podcast, I'm giving you a walkthrough, a breakdown and an update on AI Box. Everything that my software company has accomplished over the last number of months as an update to our over 500 investors from our crowdfunding campaign for the $500,000 that we've raised. All the progress, everything we built, where we're at, all the drama around our recent cease and desist, we got from Vox.com, how the progress of all the development, developers, launch dates, all of that exciting stuff. This will be the episode for that, if you're interested. So the first thing I wanted to cover as far as a little bit of housekeeping is the recent drama that we had with Box.com, they recently sent us a number of months ago and I recorded a podcast episode where I kind of talked about this, but they sent us a cease and desist, essentially saying, you know, they were unhappy that if you search for Box AI, their AI product, we showed up in the Google search results because we're called AI Box. And so they're upset about this. The same we're infringing on the trademarks, et cetera, et cetera, because we had Box in our name. So, you know, I recorded an episode about this and talked a lot about, you know, why I thought this was kind of bogus, et cetera, et cetera. And I got a lot of amazing feedback from you, the listener. So I actually am very grateful for everyone that messaged me over on LinkedIn and commented and responded with good feedback. I got a ton of amazing recommendations of IP attorneys. And by the way, I will say that if anyone's watching this episode on Spotify, there's a video where I'll be showing screen shares and slides. Otherwise it's on YouTube and there's a link in the description to this podcast if you're on Apple or something where you can check out the YouTube, I'd recommend that because I'm going to show slides, show demos of our product and a lot of hands on stuff. I mean, I'll explain it all if you're listening, but I would recommend watching this episode if you're an investor and if you're interested in seeing a visual of what we've created and what we're working on. So in any case, with all the drama that went on with this cease and desist letter, we have found a great recommendation. An attorney that is representing us from Newman llp and essentially we sent a response to them. I won't give you like the full Document, which is quite extensive. But the gist of it is that essentially we told them that because the US Trademark Office has already allowed a bunch of other marks that include the term box, including ones that are in the computer software market. And because it is of its widespread use by third parties that have the term box in their mark, it's not distinct as its source. So because of that, Box is not entitled to claim its box mark as a quote unquote family. And it is also not able to claim that it has a monopoly on all marks, including the term box, especially when box is the, like the later part of the mark as an AI box. And there's great precedent for this in a lawsuit that happened between Box and Dropbox, which are literal competitors doing literally the exact same thing, and they ended up settling the lawsuit. So they're both able to use their name because neither of them wanted to lose the trademark in that lawsuit. So I think Box eventually just settled with Dropbox in any case, because of this precedent and a lot of other cases, and companies we believe are in pretty good standing. And so essentially we told them we'll be continuing to use the name and they're not entitled to this. So that's currently where this stands. And if more happens, I will definitely keep you up to date. There's a lot of interesting things in our legal tool belt, but yeah, that's, that's where it stands to date. And I hope that we can keep a great relationship with Boxes. It's a great, innovative company and, and I'm sure we could both coexist. Great. Okay, so with that all out of the way, definitely that took some time and resources to get this settled. And it was a bit of a speed bump for us in our, in our launch and our plans. Because the big concern here is right as we're about to do a launch, if we had to change our name, this was going to change a lot about how we had to do this launch. You could imagine that for over the last year, talking about this project on the podcast with this, this name and thousands of people on our waitlist, and all of a sudden, if we had to change the name, and all of a sudden we're messaging people and putting out market material and under the name of a brand new company, people wouldn't recognize us. Well, they wouldn't know who was emailing them. They probably deleted our market as spam. And so ensuring that we were able to launch and use this name, that we have essentially done all the marketing and built up all of our wait lists and customers with was pretty important to us. So once that is all out of the way, I want to give you a little bit of develop, a little bit of a progress update on our development. And the big thing that we've just essentially accomplished is building out a very robust backend that's very focused on security encryption and making this as secure as possible. So we have an incredible team that has helped with the back end, but this has been one of the more challenging aspects of our project thus far. So to be honest, we have an amazing architect for our backend who has worked extensively at AWS and in corporate security and back end security and does contracts for big companies in Silicon Valley. And the problem was finding people in house that could help fulfill this vision of the backend architecture. We hired multiple people in a row who would be around for about four weeks before we realized they just weren't able to keep up. And so eventually our team member who architected the backend decided that he would do the build out and built this all out for us. So a really big shout out to our team member, Nick Benyo, who is, who helped us architect and ultimately build out a lot of this backend. So our backend is now built and secure for our project. And in order for us to adequately test everything that needed to go into this, we also built out a admin panel for AI box. So one of the tricky parts with what we're doing and interacting with so many different AI models and companies was making was essentially keeping track of tokens, usage, data storage on a platform basis, but also on a user by user basis. And another issue is that a lot of these AI models, for example, that we could get API access to, not all of them do essentially a bill as you go where you pay for some tokens, you use them as your accounts getting low, it's automatically buying them. Some of these platforms make you purchase your tokens ahead of time. It puts it into like a bank essentially of tokens. And then as it gets depleted, you're in charge of like refilling that manually. So in order to essentially facilitate for that and make sure that we weren't, you know, having like some sort of AI model run out of its tokens and all of a sudden it's not working on the platform. We had to build something that tracked and monitored the usage across all of that. So we can see on a user by user basis which AI models they're using, how many requests they're sending. We can't see any of their Personal data or anything that they're saying. But we can track all of, you know, kind of the usage metrics. We can also track error rates over time if there's any bugs or errors. All of that is being tracked on a platform and user level as well. As we've created, we're essentially tracking all of our users token usage. So month to month your tokens will expire and when you have a subscription, you get a certain amount of tokens. If you're on the annual plan, you can keep your to, you can keep your tokens for the full year. So if you're worried about tokens expiring, I just recommend getting on the annual plan and you got, you know, everything for a whole year. You can use it as fast as you want. If you use them all up, just go to higher tier, you get more tokens. So, but in any case this is an important metric for us to track as you know, knowing how much, how many tokens people are using. The, the, the, the platform becomes more profitable if tokens are burnt and not used. This is all just important for us to know for, for forecasting. And in addition, we're tracking all of our storage. So you could imagine if someone went and created, you know, a thousand images or was taking up, you know, a hundred gigs of storage, this would be very expensive for us to, to store all of their, all of their data, all of their assets that they create with their AI tools. So we have a, a panel where we can see how much everyone's using and it, it goes along with whatever subscription tier you're on. If you're on a really low subscription tier but you got a ton of storage, kind of like Google Drive. If you're using too much storage, you got to either delete your files or upgrade to a higher tier. So you know, that's just a cost thing that you know, unfortunately we pay for that. So we have to make sure that that's included in kind of our costs and what users are getting. So our launch timelines right now we have built a brand new product called the AI Box Playground. I'm really excited to give you a demo and explain to you all a little bit about what that is and why we are launching this. First, we're shooting for a November 24 launch next month of this platform and shortly after we'll be focusing on rolling out the full AI builder platform. And anyways, I'll give you a demo of what the Playground is and then I'll explain the differences between the two platforms. Why we chose to launch a playground first. What that helps us accomplish and all the reasoning behind that. So the next thing I wanted to talk about is the AI box playground, this new tool that we have built. So this is essentially a platform where you can go and access all of your favorite AI models all in one platform and you can chat with them all in one chat. We've built something called, called AI Box Default, which essentially uses AI to help determine, depending on your query, which AI model would be best for answering your question. So an example of this is if you said, you know, you know something like write me some code in HTML for a hello World website, right? So obviously something basic people can do a lot more complex stuff with code. But if you're doing something simple like this, the model would actually go and pick, would pick a model to, you know, it knows that it's generating code, it knows this is an image or isn't audio or something like that. And it's going to go and generate something that can essentially create the text or the code for this. It's gone ahead and it decided to pick 01 Preview to generate this code. And we also have built in code, essentially panels, code panels that you can see just the code that it generated. So you know something that kind of annoys me with chatgpt other models sometimes is like the code is mixed in with a, with their like explanation and how to use this code. And so you gotta like try to copy just the right place. We have a simple copy code element. So that's great. So in any case, maybe you're like, okay, it picked this. I didn't really love doing it with 01 preview. I wish it could do it with something better. You can go and manually click to rerun that exact same prompt with something like, something like sonnet 3.5 from anthropic. And you can also go and try to get it, to run it with Google Gemini Pro. And so you have essentially all of these different models writing this code and you have essentially in the chat like a tab where you can tab between all the different models. The other thing that you can do is open up a compare panel so you can look at the two models side by side or different, multiple models side by side to see the code that it generated. And this is super cool and useful with code. But this works for text documents, this works for images on different image models, this works for audio on different audio models. So this is fantastic. And you can also go and just choose exactly what model you want to use. So for Example, if I was like, okay, you know, this website's great, But I want 11 labs to say the words, you know, hello, world to embed on my website. I mean, more likely you're probably going to get it to do some sort of voiceover clip for a video or a podcast, but you could get 11 labs to generate that. And the real power is that you can then go rerun the same thing with OpenAI's voice model as well. And then you can also pull up the Compare tab and see both of these audios side by side to see which one you prefer. You can listen to them both. The other cool thing that we've integrated is settings, AI model settings into all of this. So right now I just ran a test where I had OpenAI's model do the fable voice, but they have like five or six different voices. So maybe I'm like, okay, well, I want to try nova's voice to, you know, also essentially run the same prompt. And so you can go and have it run the, you know, do a voiceover for the same text as before, but now with the Nova voice, and you can open up the same side panel to compare, you know, Fable's voice versus Nova's voice, and you can listen to the two side by side to determine which one you like better. This is incredibly powerful for comparing different AI models, different outputs, deciding what you like or getting it to rewrite it, or even doing something. Like, you get ChatGPT's O1 preview to write the first draft of a document, but you're like, I want the tone to be better, so then you get Sonnet to, you know, rewrite it and just focus on the tone. But maybe you wanted 01 preview because it did a little bit better of job with adding bullet points and facts, and then you move it over to another model. So really great with model collaboration as well. In addition to all of that, we also have a media uploader. So you can upload an image, a PDF, an audio file, any type of media, and you can ask it, you know, what is the text in this image? And it's able to do image inference, see kind of what is written down in that. In that actual model. And something interesting is that, well, you can upload this. You can upload so many different types of media. So this is kind of what you'd expect. I think in my demo I accidentally asked it while it was open to OpenAI's audio model. So this again is a great reason why I would recommend using our default is because on the default model, it's able to determine what, what the best model is for this. But in any case, you can do the image inference. You can also go, okay, the other feature that I absolutely love with this because, yeah, so you also got all the settings. You can change your, your temperature, your top p max tokens, context length limits in all of, you know, for all the AI models that you're using. So this is phenomenal. I'm looking at the settings for sonnet 3.5. The other thing that I absolutely love about this platform is we created something called Media Storage. Now, this solves a problem that I personally have had a ton, which is I'll be creating an image or an audio or usually like an image or something in a chat on ChatGPT. And I was like, okay, I made this image like two weeks ago. I don't remember what chat, because I do like 10 or 20 chats a day. This thing's two weeks ago. Like, there's no way I'm going to scroll back and find it. We have a media storage that every piece of media, whether that's audio file or images that you've ever created, shows up in here. You can go click on the image or whatever asset you created. You can see the prompt you used to create that. You can see the model that was used to create that, how many tokens it took. You can either delete it or download it, or click to view the chat. So you go directly back to the chat where this asset was actually created. You can see kind of the conversation that you're having before and after and everything else. You can see that in this chat I was doing, I, I uploaded an image and, you know, said, what is this YouTube video about? And it's, you know, kind of talking about what's inside of the YouTube video based off of this thumbnail I uploaded. So that's the image inference I was referring to before the code, the audio. All of this is really easy to find. And especially with this, especially with the media storage, it makes this a phenomenal option. So this is essentially the new product that we'll be launching, the AI box playground that I'm super excited to be rolling out shortly, getting all of your AI models in one platform and it's all for the same price as Chat GPT, but you get everything there. So all of your favorite models. So this is. And we're, we're adding new models all the time. So this is something we're super excited about. I wanted to talk a little bit about what some of the benefits of essentially launching this, this playground before launching the full app builder. So the first thing that this allows us to do is focusing on testing. We have, you know, thousands of companies and professionals on our wait list right now and to roll out, you know, everything we've built essentially to everyone. Rolling out this playground first allows us to test five key areas. Number one is account management. So our whole user account creation workflow, make sure that's all optimized, all the authentication, AWS incognito integrations for secure user authentication, all of our payment processing, our stripe, how we handle all of our transactions, all of our back end infrastructure. So we're using Amplify lambda functions and GraphQL API testing, token management via Lambda and DynamoDB interactions, all of the AI service integrations through our ECS. So we're really able to run our back end infrastructure. Everything that we have built for this playground is built for our full platform. So this is kind of allowing us to test it on a, on a smaller scale before we unroll iterations, which I'll talk a little bit about later. But the fifth thing I want to mention is the data storage specifically. So we have S3 to S3 buckets on AWS to store images and audio and all of our media, but making sure that we have that rolled out correctly. Our MongoDB storage solution verification with all of our media storage. So I wanted to talk about some of the strategic benefits of specifically testing this. The first is that we're able to have a really focused development, so we're concentrating on just the core functionalities before we start scaling to some of the more complex features in the builder. But we're able to essentially test the core, the core development of the platform, while getting essentially will be essentially testing all of our core things to make sure it's strong enough for it for the next phase. So we're also able to do a lot more efficient debugging. We're able to optimize our resource allocation. So how much money we're spending to develop this. The nice thing launching the playground is that we'll be able to immediately start making money. All of the risk mitigation, all of the iterative improvements, we'll be able to start focusing on that. So I want to talk a little bit about the development focus distribution, which essentially just means where we're at, what, what sort of we focused on for each project. So in this current release of this playground, about 70% of the work that went into this from a development standpoint was on the back end, the back end was, is incredibly robust, incredibly secure, focusing on encryption and security and privacy and all of that and building something very, very powerful that can scale to the entire platform. So about 30% of it was front end work and a solid 70% was just the backend and making this work. Now once we switch, now that this backend is all built out and ready to launch this and we're testing it once we switch to the next phase, which is the builder platform, which is of course a much more complex platform, the backend is already done and tested and so our, our builder launch is just going to shift to focusing essentially on a front end emphasis, making sure all the features and everything's polished up on that. And I'll talk a little bit about the status of the builder where we're at on this, but essentially this backend focused phase right now is really to optimize the performance and reliability of some of our key components before we expand to some of the user facing features. So I want to talk about a couple of things that we're still, that we're just finishing up before this November launch. So you know why we're launching in November. We do have people coming on immediately. Like we've already had some people coming on right now doing beta testing. We'll be rolling that out to more people. If you're on the wait list, which I recommend, AI box, AI you can get on the waitlist, I will be, you know, first come, first serve and we have a long waitlist but we'll be rolling this out and we would love to give you beta access. If you're on that waitlist, look out for an email. The things we're focusing on right now is our Stripe and Lambda web hooks. So that's essentially implementing all of the payment processing, our Toca, Lambda and service. So essentially our centralized token management, every single AI API and model has a different cost associated with it and so we have to convert those into our own tokens on our platform. And managing all of that is, you know, an exciting project to say the least. The last thing is our cloudfront hosting. So this is really just preparing AWS cloudfront for hosting and launching the application. So those are three things we're currently working on. The other thing I wanted to talk a little bit about is some of our progress updates. So the builder designer, output view, homepage components are all 90% complete and I'll talk about each of those. So the builder is essentially where you link together different AI models. This is our drag and Drop platform and that we're calling the builder the next phase. Once you've kind of linked together all your AI models and figured out what you want to build, we have what we call the designer, which is a phase where you essentially design what a user that's going to use your AI tool, what they see. You can change the colors, you can change the backgrounds. We have a lot of interesting things. Adding photo backgrounds, adding your logos, a lot of different stuff is all determined in the designer. After that, we have the output viewer and that is what your actual asset looks like. So if you're creating a document, how the document is formatted is the output viewer. You can actually go manipulate that so that a user gets exactly what they're looking for. And then you of course have to publish that to the marketplace. So that's going to be the homepage component. So all of those four components, we're about 90% complete on the overall builder platform. One of the bigger holdups was finishing all of the back end and making sure everything was secure and all the security and privacy was all a big enough focus and was all done right. So now that that is completed for this, for this playground that we're launching, this is about 90% complete and we'll be working on launching this shortly after. The remaining tasks we have for this are some UI element finalization, some functionality debugging. We have to transition our code to a consistent snake case convention because some of the tools in our backend need that. So that's just, you know, housekeeping, essentially. By prioritizing this playground launch right now with a really focused approach, we're setting a really strong foundation for future development and making sure that a more stable and reliable system is available. As we start to scale this, we're also essentially putting ourselves on a really good path for a smooth integration of a lot more complex features, boxes and other things that will be launched in subsequent phases. So this is essentially what we are, what we're currently working on and the reason why we're focusing on this playground launch and testing before we do our full builder rollout. So this is something that I have been working on for a very long time. I'm incredibly passionate about this. Obviously, if you're interested in getting on the waitlist to use this product and you're not already, go to AI Box AI this will be rolling out next month. So I'm super excited for anyone that was an investor in the AI Box Republic crowdfunding campaign, depending on the level of investor and where you invested at we have different perks and we'll be sending emails out to people that are whitelisted for the beta and essentially having people come on and test and get early access to it as well. So super excited to give the update. If you have any questions, feel free to reach out to me over on LinkedIn. And again, if you are interested in watching this and some of the demos, if you are just listening, there's a link to YouTube in the description where you can go check out all of that. Thanks so much for tuning in and I hope that you all have an incredible rest of your day.