we built an ai cheating app in 2 days @ 18 – YouTube Dictation Transcript & Vocabulary

1.roll on i know shorty and she doesn't want no slow so had a man last year and life goes on haven't let the thing loose girl in so long you been inside know you like to lay low i've been peeping what you bringing to the table working hard girl everything my name is Gabriel i'm Benji and we're two college students in New York City building an app studio this week we are launching our first web app product which is called Spy GPT which is simply an AI detector website and an AI humanizer website that basically helps you bypass AI checkers like turn it in check whether your text uh was written by AI and humanize any AI written text so that it can actually bypass these checkers just like ours and our competitors so let's get into the video so there has been a trend online recently of people building different study tools the reason why people are building study tools or just web apps or mobile apps that are targeted towards students is that there's a huge demand and huge market for it so for example our friend actually built Corsology which is a AI study tool that allows transcriptions also uh pretty sure like some really cool features basically just any tools that would enable students to work in a more efficient manner in school those are doing really well and that's actually one of the reasons why we're doing spy GPT in the first place because a lot of students they might want to you know use this tool to their advant advantage to excel at school things like that and there are other products for example stealth GPT for uh you know undetectable AI that are doing the exact same thing that are generating millions of dollars in revenue so that is the route that we're going with and we believe that the market is uh big enough and the demand is there to the point where it's not going to be too saturated when you know we're launching it now basically spy GPT has two main features the AI humanizer and the AI detector and for a product like this there are two different ways that you can go about building it there's a really expensive time-consuming way which is training your own model to actually recognize AI's patterns to and then actually you know train it to humanize AI text and also the other side again recognize AI patterns and determine from there whether the text is AI generated or not however there is a different way which is the approach we took and that is necessary if you want to build stuff in a really short amount of times and for very very cheap and that's the GPT wrapper approach now you might think that a product like this it is impossible to make accurate or actually make a valuable product uh by making it a rapper however uh there's a very simple way to do this and this is because it is all about understanding how the AI works and then just going backwards from there so basically we understand how AI generates text and then from there we just go through the generation process step by step and train the AI to recognize those patterns in the text that reveal uh the AI generation process if that makes sense so I give it I give it a text and I try to recognize patterns that are common when it comes to uh DI text generation process the different things that we have it look at is you know vocabulary diversity the use of personal voice consistency in the language whether it's too consistent whether it uses a overly formal tone and whether there are uh very very strict patterns that are easy to notice and also patterns that are not the same type of patterns that person's writing style would have you know if you think of a person writing their writing first of all is never going to be perfectly consistent throughout whereas in AI it will be now I can't get into the very strict technicalities but it is simply about recognizing uh the patterns that occur in AI generated text and this is through very long and elaborate prompts the reason we were able to make this GPT wrapper very very accurate is basically by sending out almost 10 to 15 prompts per each uh request so if I request a humanized text the AI will work uh on the same prompt 15 times before the user gets an answer back and using different models now since the user's text goes to through 10 to 15 prompts every time they're either trying to humanize text or detect whether it's AI generated means our uh our product is actually very very accurate and that obviously means it's going to be expensive on our end in terms of API costs but we're also charging users a monthly membership to use the product so we're still able to uh make a profit in the end these are things that you only learn through research and actually a the hard part of building products like these is actually just the backend research to figure out how you can actually make this first of all the product work and work better than other products so something that was really important for this one for example is to make sure the speed and getting the answer was there but the accuracy was even better than uh the competitors that were out there so in terms of marketing we're thinking having student creators since this is mostly a tool targeting a I would say young uh young adults or you know students having students creators increases authenticity for for the product people related to it more and for the videos you know we're just going to create videos on Tik Tok and Instagram have these creators um basically post themselves using this app or you know making some memes out of it i think that not only uh that is relatable but at the same time uh the use case of the app is clearly stated one of the mistakes that we made for pillar which is our first product and it's a self-improvement app is that a lot of the memes that we created are indirectly selling on the app meaning that the videos that we're making are indirectly promoting the app in a way that is not that obvious so the conversion rate on those ads where they you know our our videos got 40 50 million views however the conversion rate is really low on those so we decided to take a different route where the videos itself that we're making uh first of all it's going to be creator-based meaning that there's authenticity but second of all it's going to be entertaining but at the same time it's going to directly relate to the product so that people when they view the uh video they understand that you know oh the tool is there and then they know where to get them so that the conversion rate would be much higher falling down ain't nobody here to play around push it to the edge i won't back down cuz it's time to go hard and go home one way up no way out i give it all all for the family the reason why packaging is so important is because we have to deliver and explain the value of our product in a very clear and direct way and when users land on the website usually in this in cases like this one where they're just looking for a tool to humanize their uh AI written text or they're just trying to make sure that their essay doesn't look like it's it was written with AI they're looking for solutions really really fast right they want to get it done in 5 to 10 minutes they don't want to be spending an hour signing up and stuff like that so packaging is really just figuring out what the fastest onboarding flow can be and the best onboarding flow that can then get lead to actually paying users right so through the onboarding flow we have to it has to be quick but we have to show them the value and give them a reason to actually pay for this product and to keep coming back uh you know day after day and that's uh what the packaging is really about and that's why so many people can build there are so many developers out there but most of them are you know employed because uh not that being employed is a bad thing right but it's just because the actual hard part of scaling products is the marketing and the packaging of it not necessarily the building of it unless you're going into you know spaces where uh you're like innovating with a completely new technology but when we're talking about SAS products uh you know mobile apps things like that it's really about the idea the concept and then how you're actually packaging it and marketing it and especially for products like this the packaging is a really big part of the marketing of it and the reason we also chose dark mode is because it's more intriguing if we're to make content for Tik Tok and Instagram which we're going to do where you see creators dark mode generally looks better and we took inspiration from products like Turboarn for example cuz we really liked uh their products aesthetic so if you actually look at the pricing we price it at weekly $4.99 uh monthly $9.99 and yearly $49.99 and that is just to make the monthly option appear more attractive and more appealing to the users because um if you just do the math right if you want to go with the weekly option your monthly cost would be around $12 and yearly also saves a lot of money as well but I think uh monthly is what we're trying to go with and especially you know once we get that reoccurring revenue going on it's going to be a really good source of passive income so if you guys saw the last video we were building five web apps in one week so we actually had to outsource uh a lot of the work for these web applications and this means we had to look for a lot of developers and obviously the last video helped us get a lot of applications of people who wanted to help uh but also we were able we had to look for a few ourselves because we need really skilled people that are not just really good at what they do but can also do it really really fast under a small you know a small amount of time and we learned our lesson uh working with a few not so good developers but uh still learning process and the really good thing is that once we find a good developer it's not like they hit a do a project and they're gone we but we actually have them as part of the team long term so anything else we want to build we know we have a reliable person we can reach out to that will deliver uh really high quality work to us and the thing is we didn't actually uh outsource everything to the developers we'd uh outsource like the building of the core functionalities of these products which is really the hard part and then we take we're taking care of like the packaging so the part that's like a little bit more related to marketing and the part that's really our strength like we're really good at marketing and scaling apps and there's no reason why we should spend time you know those hours and hours coding uh the really hard and complicated part when uh really our strength is packaging it in a way that actually gets sales and makes money so that's what we did this time and it really turned out great for uh spy GBT there are also some things that we can't have developers do like setting up payment integration with Stripe purely because you know for the safety of our bank account we obviously don't want to leak that information but other than that developers can usually handle most of the backend work for marketing we're also thinking about doing Google search ads because this is a web application we cannot do Apple search ads so but also I'm really curious as to how it would play out if we have that little boost on the browser search functionality maybe some students would search you know how to bypass has turned it in or things like that and then we just hit them with spy GBT you know that's a that's a perfect target audience that's a perfect way of marketing as well but um usually from what I heard from my friends is that Google search ads are expensive the cost of acquisition for one customer is much higher than as than you know just creatorled marketing in general so excited to see what happens and ideally that's going to work out if not you know we're just going to spend a bit of money on that you know test it out if it doesn't work it does not work at least we tried we hope you guys enjoyed the video if you're interested in joining the marketing team development team design team or just our team in general feel free to email us at [email protected] the easiest way to reach out is Instagram cuz we do check our DMs every day hope you guys are excited about uh Spy GPT cuz we are really excited but yeah we'll see you in the next video say my name three times like I'm Tony Diamond V as I put all my roll i ain't never believe nothing that you told me these [ __ ] keep sneaking and they turn me to my homie before I had these mess I was all so lonely ready with [Music] my phon you'll be getting best Harley yeah