YouTube Transcript:
This one prompt makes GPT-5 INSANE

Skip watching entire videos - get the full transcript, search for keywords, and copy with one click.

AutoDub

Understand YouTube Foreign Videos

Immersive YouTube Voice Translation

Break language barriers, embrace global quality content

Solve Foreign Video Barriers Instantly

Video Transcript

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 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 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 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.

This is how you 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. But here's the thing. 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 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

Click on any text or timestamp to jump to that moment in the video

Most transcripts ready in under 5 seconds

One-Click Copy125+ LanguagesSearch ContentJump to Timestamps

Paste YouTube URL

Enter any YouTube video link to get the full transcript

Most transcripts ready in under 5 seconds

Get Our Chrome Extension

Get transcripts instantly without leaving YouTube. Install our Chrome extension for one-click access to any video's transcript directly on the watch page.

Add to Chrome — Free

Works with YouTube, Coursera, Udemy and more educational platforms

Get Instant Transcripts: Just Edit the Domain in Your Address Bar!

YouTube

←

→

↻

https://www.youtube.com/watch?v=UF8uR6Z6KLc

YoutubeToText

←

→

↻

https://youtubetotext.net/watch?v=UF8uR6Z6KLc

YouTube TranscriptPreparing your results…