This one prompt makes GPT-5 INSANE

Chad GPT5 is the best AI model ever released. The issue is if you don't prompt it correctly, your results are going to be horrible. After using GPT5 non-stop since it came out, I've come up with a prompting system that will make your results 10 times better. I'm telling you, if you steal the prompt I'm about to show you, you'll be oneshotting things with AI you never thought possible. Let's get into it. So, I came up with this master GPT5 prompt. I'm going to show you in this video. We'll go through all five stages of this prompt. So, there's five parts of it. You'll be able to copy and paste it and use it however you want, but I'm going to walk you through each one of these five stages in the prompt. I'll unlur for you step by step. Last week, after a lot of complaining about GPT5, Open AI released an entire cookbook of how to use it. This is what that cookbook looked like. And it's this huge guide on exactly how GPT5 works and how you should be talking to it. So, I basically went through all of this and I went in and I created a master prompt based on the recommendations that OpenAI gave. This basically takes hours of advice OpenAI came up with and puts it into one simple prompt. You can copy and paste, fill in the template, and you're going to start getting incredible results. It's really critical you stick through this and learn the entire prompt because this is going to be the only way you get really good results out of GPT5. Unfortunately, GPT5 isn't that good if you're not prompting it correctly, which I understand people don't like. But if you do use it correctly and you use this prompt I'm about to give you, the results you're going to get are way better than any other model you can use.

So, let's get straight into it. Let's get into the master prompt. The first thing you want to give GPT5 when prompting it is the role it's taking. Right? Just straight up the role. Very simple, one sentence. What role is GBT5? The reason why roll is so important in the prompt is GPT5 only does exactly what you tell it to do. Nothing more, nothing less. So, if you want it to do something really well, you need to give it a very specific role. So, the example I'm going to be showing you throughout this video is we're going to be building a master prompt for planning and building out an application. This is going to be an application for YouTube creators. It helps them plan videos and make scripts. And so I'm going to build out an entire prompt for but first giving the role you were a product full stack app planner for indie creators. So for you depending on what you're doing right you'd make your role you know you're a product full stack app planner for video games or whatever you want or you're a master researcher for trending news items whatever that task is you want to do you have to think about the perfect role for the AI. So the first line in our master prompt is going to be that role, right? We want to make sure right away, right off the rip, before GPT5 goes into anything else, it knows what its role is. So here's where things

start to get interesting going into the second part of the master prompt, we have a control panel. And so if you read through OpenAI's cookbook for GBT5, one of the biggest parts here is the amount of options they give you. They talk about making control the verbosity of the model. They talk about controlling the amount of thinking the model does. There are a whole bunch of factors that on the negative side GPT5 isn't fantastic at figuring out on its own. Out on the positive side, it's fantastic if you control those parameters and tell it how much to think and how verbose it should be. So, going back into the master prompt here, you want to put a control panel with all your first main prompts, right? You want to control reasoning, verboseity, the tools it uses, for it to self-reflect, for it to metaphix, and I'll go into what those two mean in a second. But you want to have when you build your kind of first master prompt, when you're taking on a major task, to have this control panel where you control all of these variables. So, going back into our prompt we're building here, I put in the control panel. For this use case, I'm going to have it do ultra think, so think as hard as possible for this prompt and planning out this app. I'm going to have the verbosity be medium, so it's not a ton of extra information. I want to have it use most of its tools, so web, code, PDF, and images. I want it to have self-reflection on, and I wanted to have metapixing on. So, real quick, you can probably figure out what these three mean. And for a full list of tools, I'll include that down below as well for what tools GPT5 can use. But what does self-reflect and metaphix mean? These are two very important features of GPT5 that actually makes it very powerful. Self-reflection is basically GPT5's ability to when it's about to execute on a prompt, it'll actually reflect on what it's doing in that prompt and improve the prompt before it executes. Right? So, if you give it a prompt, it's about to execute on it, it will actually self-reflect on that prompt and figure out ways to improve it before it actually does the tasks in the prompt. And this is a really powerful feature in GPT5 you need to be taking advantage of. This is how you get really good results. And then there's also metaphix. This is another really powerful feature. And again, these are all things talked about in here in the GPT5 cookbook. Metaphix is actually its ability to reflect after it executes. So self-reflect reflects before it executes. Metaphix reflects after it executes. And so after it executes on your prompt, it'll actually go back, look at the results, and improve the results if the results didn't match exactly what you were looking for, right? And so these two you want to have on most of the time. As long as you have time to spare waiting for the answer, you want to have these on. If you're in a rush and timeliness and quickness of the model is important, you'll go ahead and you can turn these off. And I'll go at towards the end of the video into a lot more detail on how these two work because there's more to this prompt that actually ties into these two options. But basically what you want to do here is with every one of your first prompts you do, right? You want to have this control panel and you want to control these different variables in GPT5, right? So for reasoning, you can do think, think hard or ultra think. For verbosi, you can do low, medium, and high. And for tools, here are your different options. If you don't want to choose specific tools, you can just use auto. But getting into the granularity of controlling each one of these variables is going to give you way way better results with GPT V. I'm not kidding. If you do these things, your results are going to be amazing. Equally as important as this control panel is what we are going to go into next in this master prompt. So the next

part of the master prompt is a really simple task. So usually what people do when prompting GPT5 is they actually do only step three and none of the other things. They usually just give it a task. So this step three is probably the one you're most familiar with, which is giving the model a really simple one-s sentence task. The reason why you don't need to go into super detail here is the rest of your prompt is going into the important detail. This is just simply what you want the model to do. So for this example, when it comes to building and planning the app, we're going to say plan and scaffold a minimal app called YouTube topic scout that finds trending ideas, scores them, and generates a script outline. And so basically we're building like a YouTube planner and scriptor for us. And so the task very simple just planned and scaffold out this minimal app and we gave a short description what we want this app to do. We don't need to go into a ton of detail here because the rest of our master prompt goes into all the other details we need. So again just the task just exactly what we want the model to do and nothing else. So next, the fourth and

second to last part of this master prompt that you really need to learn to get incredible results from GPT5 is going to be inputs. Right? So this is probably the only optional part of the prompt. Right? If you haven't done much planning or thinking about what you want, you don't need to put this inputs, but if you have context you want to put in, this is very important. Basically, what inputs are is important context you want to give GPT5. This is basically going to be a list of notes, of links, of other data, of other thoughts, of other ideas, of comparisons, things like that. I'll show you a couple examples in a second here. But this is important context you want to give to the model before it does its execution for the task you're asking for. So, for example, for building out this app that we're planning here, I'm putting in the user. So, who I think's going to be using this app, the core loop, what I want the experience of the app to be, different information about how I want them to score the inputs of the app and score what the topics are that they're planning in the app, non-negotiables, tech preferences, tone. So, I'm just giving as much context of things I already thought about for this app that I wanted to build. Again, this is completely optional, but you want to be brain dumping in as much context as possible. If you're building out or planning an app, you just want to put in all the different ideas you have for the app already. And this is going to be really important context. Maybe if you're having the AI build you content, what you would put in here is other comparable content or other comparable content creators, you want the content to sound like, right? This is going to be all the context that's important for this prompt we're building for GPT5. As you can see, our master prompt here is looking pretty good. We're giving it a lot of information so far. At this point, it's already probably giving you way better results than any other prompting you've been doing with GPT5. But let's go to the final part or the final part of the master prompt for GPT5. So, the fifth part I want to show

you in this mass prompt, this is actually not the last part. I have one more thing I want to show you after this that's important to include in the prompt. So, stick around for that. But the fifth part that you actually want to customize in this prompt is the deliverable. So, this is a list of precisely what you want in return from GPT5. what you want it to actually output to you after it does all its self-reflection. So here's what it looks like for us in our example of what we're building here is when we're planning out this app. We wanted to give us a PRD, so a product requirements document. We want to give it a competitor scan. So we wanted to go research all our competitors for this app. We wanted to give us an architecture, an API spec, what the UI looks like, starter code. So this is actually a really complex list of deliverables. And if we didn't include these deliverables, GBT5 probably wouldn't output this. But here's the thing. GPT5 is so powerful that it can actually handle doing all of these things in one shot. Because we're using this master prompt to get the output from GPT5, it'll actually be able to get us all of these deliverables, right? Because we also included what tools we wanted to use, the amount of reasoning we want to use. because we included those options, it'll be able to oneshot a lot of really complex things. And so what you want after your inputs, after your context, is the deliverables of what you actually want out of this, the results you want to get from the model. And because we're doing all these, again, fine-tuning using specific variables, we're going to be able to get not only all these deliverables, but also all these deliverables done really, really well. But these deliverables are only high quality if you include this last part in your prompt, which is going to be this little private ops part here. So, this is down below as well. I'll have this entire example prompt down below, but private ops is basically describing to GPT5 how the self-reflect and metaphix works. In the cookbook that Open AI put out, they had things like how to score itself when it's self-reflecting, right? had talked about how you should be using a self-scoring rubric. And basically what that means is when GPT5 reflects on the prompt and reflects on its output, it should score its output from like 1 to 7 and then improve the output based on that rubric scoring. Right? This is how the GBT5 model was built. And now what we're doing is we're basically going in and we're defining that for the model. We're saying this is how you self-reflect. metapix. This is how you build the rubric for yourself. Self-reflect and metaphix so that you get even better results. Now, I know what you're thinking. I got this comment on my last GPT5 video, which is why the heck do we need to do all these things? If this was a good model, we shouldn't have to control the verbosity and the reasoning and all this. And you know what? I agree with you. I agree. You shouldn't have to do these things. And I think in the future, you're not going to have to do these things. I think in the future as the model evolves and we get GPT6, you're not going to have to control all of these things, it's going to be able to kind of figure it out on its own. But at the moment, you have to control these things. And in my opinion, it is very much worth it because here, I'll show you. I'll hit enter on this and we'll start running this prompt. At the moment, GPT5 from a raw power perspective, just raw power of what it can accomplish is the best. It is the best model out there from a raw power perspective. The outputs you get, the things it can accomplish in one shot. It is completely unmatched. So, yes, do you have to do a lot of work to get good results? Yes, you do. Would I rather it so you don't have to do all this work? Right. I would rather not have to do a lot of work. You're 100% correct about that. If you put in the work, you're going to get exponentially better results, which I'll show you in a second here as it thinks through this, your results are going to be way better. So, it's up to you. Do you want quick results? That's fine. You can use GBT5 like you were before. But if you want great results, you want to be controlling all these different variables. And here's the thing, you don't need to do this entire prompt every single time you prompt GPT5. That'd be ridiculous. If you had to do that, I wouldn't be using GPT5, right? But if you're starting a new conversation, you're starting to go down on a very important path, like you're building out a new app like we're doing here, right? This first prompt, you want to have all these variables and options. The first prompt, you have all this control that puts it's almost like the planning mode in clawed code in a way where it's like the plan controls everything else and everything else can you can do really quick. Once you got this really good strong first prompt in, you get your first results. From there, you can iterate very quickly. You don't need to use this master prompt every single time. But if you're going down a path where you're doing something really complex, like building out an app, that first prompt you want to use, you want to use this master prompt. And you're going to see why in a second once it outputs all this information. Okay, so

it's all done. It took 2 minutes and 21 seconds. Let's see what GPT5 came up here. It first it has the initial code. Okay, so it wrote all the code for the app. So we'll be able to take that code and we'll have V1 of the app. It also has C data in there. So you can run this and it'll have C data. But let's see what else we got here. We have the PRD markdown. Oh, and it had a metaphix applied. So apparently it built the product requirements doc and then went back and fixed it to make it better because it didn't match the scoring on the rubric. So here's the product requirements doc. It has a goal. Primary users, jobs to be done, MVP, what should be in the MVP, nice to have, success metrics, non-negotiables. Okay, so it has a competitor scan. So, it searched the web and found four competitors. I actually use Vid IQ quite a bit. It's a pretty good app. Uh, so this is real. This is true. These are true competitors architecture. So, it has the entire architecture. It has the schema for the database. It has the API spec, so how you would plug in the API to get data for the app. It even has wireframes. So, let's take a look at the wireframes. We'll open this up. Here's the wireframe. So, you can see what the kind of homepage, a little mockup of what the homepage would look like. Let's just check out what the uh the other ones look like here. I'll download this. It'll show you some other result cards for the different topics it comes up with. So, it makes the PNGs. So, we're making PDFs, PNGs. We're writing code. And then it even has an entire zip file for all the code that we can scaffold the app out with and an explanation of what's inside that zip. So, did what we asked for. That's pretty incredible. All in one shot. Now, from here on out, I can prompt it just as you normally would any other model. We don't need to use that huge big prompt every single time. Now, it already has enough context where we can just go back and forth now and tinker with what we have here. Using this master prompt is going to get you so much better results with GPT5. I promise you, if you use that prompt I just showed you, you're going to get the best results you ever got out of an AI model. If you learned anything at all, make sure to hit subscribe. Make sure to join the free vibe coding community I built, link down below. Only free vibe coding community on the internet. I promise you'll learn a ton. You'll love it. I'm in there all day.