Today I'm going to be building an AI SAS

product right in front of you guys

pretty much live. I'm going to be

basically starting at the complete

beginning. I don't even have an idea

yet. I'm going to go through ideation,

planning, building, testing,

authentication, payment processing,

connecting it to a real domain. I'm

going to do all of that in front of you

guys so that after this video you

basically know exactly how you can find

some sort of idea, turn it into a

product, and then actually go look at

getting some customers for your product.

So, I don't want to waste any time.

Let's just get straight into the video.

So there's six major P's that we're

going to be focusing on today when we're

building this sort of like SAS sprint.

So the first P is pain. I'm basically

going to pretend that I am in your shoes

right now. You might be starting from

completely zero, a little bit of AI

knowledge, but you don't have an idea or

a product. So that's what I want to kind

of simulate today. So pain, we obviously

want to build something that people will

actually use. Now in a perfect world,

the way that I would actually go about

this is I would go get feedback. I would

say, "Hey, here's my business idea. How

much would you pay for this? Would you

pay for this?" and I would get a bunch

of validation before I actually build or

I'd build like a little P proof of

concept to just say, "Hey, here's a

demo. When I build this out for real,

would you build this? Would you guys

find this value?" And then I would go

build it. So, basically sell before you

build. So, we're going to be using AI

here to help us find some pain points

and find the business idea. From there,

we're going to step into promise. It has

to be super clear and basically one

sentence on the landing page. What does

this tool do? How does this tool promise

that it's going to answer the pain that

we already found? From there, we're

going to be actually doing the product.

We're going to be planning it out. We're

going to be building it. We're going to

be testing it. And then, you know,

that's basically the actual core product

that we're building today. The fourth

one is the plumbing. So, we're going to

have to wire up things like the payment

processing, things like the

authentication, managing the database.

Obviously, when you have a product like

this, the goal is that you one day can

scale it to, you know, potentially

thousands and thousands of users. So,

you need to make sure the plumbing is

there to actually handle that. The fifth

one we have is the packaging. So, making

sure we have stuff like the brand

guidelines, consistent colors, you know,

it just feels like a professional

product that someone could actually like

maybe go to the website and just buy or

subscribe to. And then the last one is

proof. So, basically all throughout this

whole thing, we're going to be doing so

much verification because at the end of

the day, I don't know how to read

Python. I don't know how to code in

Python. I don't know what other

languages we might use, but AI can do it

really well. But it's on us to be the

project manager, to be the one that's

saying, "Hey, verify this until you're

like confident that it's done and prove

to me that you're confident." So, it's

all about how can we actually utilize

the more intelligent and more capable

coding agents that we're going to be

able to use, but we still remain in

control of the judgment. And you know,

at the end of the day, if the product

fails, it's on us. We can't blame AI, we

blame ourselves. So, I'm sure you guys

are curious about what tech stack we're

going to be using today. We're going to

be using three main ones to sort of

drive everything, but I'm sure there's

going to be, you know, other APIs or

other payment processors. There's going

to be a lot of things that go into this

build, but the three core that I want to

call to your attention are we're going

to be using codec. So, I have the chat

GBT desktop app that I'm going to be

using. I'm going to be using codec

through this. I'm also going to be using

Claude, so Claude code inside of the

Claude desktop app. So, you can download

that as well. And then finally, we're

going to be using Glido to actually talk

to everything. It's basically going to

be the voice layer so that I'm not going

to be bottlenecked by typing. I'm just

going to be able to talk as fast as I

can. and do this as fast as I can. So,

if you want to sort of follow along with

this video, then I would definitely

recommend you get all three of these

things so that you can basically move

quick and you can have different AI

models and different coding harnesses

get different perspectives on your

build, which I think is really helpful.

If you prefer codecs and you just want

to use codecs or if you prefer claw and

you just want to use claw, that's

totally fine. But I found it to be very

helpful to be able to have different

coding harnesses and models take a peek

at what I'm doing and help me sort of

like play devil's advocate, stress test

it, things like that. Okay. Now, how do

I think about the difference between,

you know, our friend Claude and our

friend Chadigati or Codex or whatever

you want to call him? So, I really like

to think of Claude as my kind of like

thought partner. It feels more creative.

Especially if I'm able to come down here

and switch this on to Fable, it feels

like it has more of that wise old owl

sort of perspective. Whereas, when I

jump over here to codeex, this thing

feels like a Rottweiler. It is basically

going to take the task. It's going to

take the command I give it and it's just

going to execute and it's going to keep

executing and keep verifying until it's

good until it's done. And so I think

it's really great to be able to use them

together in tandem. And with that being

said, I'm going to start off our build

here in the pain section trying to find

an idea. I'm going to be using um Cloud

for this. And by the way, if you guys

are curious on where I'm actually

building this, what I did is inside of

my Herk 2 project, which is my AIOS, if

you guys have been following me for a

while, I've got a folder called Other

Worlds. And I made a new world in here

called AI SAS Sprint that's completely

empty. And this is the folder that both

Codeex and Cloud Code will be working

out of so that they can see the same

stuff. So anyways, let's get started

here. I'm going to open up Glido and I'm

just going to start yapping. All right.

So what I'm trying to do today is build

an AI SAS product. I'm trying to show my

audience how much they can actually do

with AI. And the first most important

part of building a SAS product is we

have to have an actual painoint. We have

to answer a problem or solve a problem

that the market actually has. So a lot

of my audience might not have you know a

big YouTube channel or a big community

to leverage data from and we have that

ability. So basically the idea is I want

you to use our primary data which could

be things like YouTube comments um

people talking in our school community

what are the problems they're having but

then also I want you to go on more of

the publicly accessible data like

secondaries forums of Reddit, LinkedIn

X, Instagram. I want you to look through

this type of stuff and help me identify

a painoint that we can actually solve or

a few pain points that we can solve.

What we're looking to do here is we're

looking to build some sort of AI SAS

product that realistically can be built

in one day. Something that is very

visual and the promise is very clear. So

it's not like a super super technical um

like CRM sort of product that they would

have to get onboarded to and that they

would have to get familiar with over the

you know a week or two get trained up.

I'm looking for something that people

can sign up for and get value

immediately. Like within five to 10

minutes, they can use it, they can

understand it, and then they'll get

value. So, what I want you to do here is

I want you to act as the orchestrator. I

want you to spin up different sub aents,

and I want you to do research, scrape

different methods, and then at the end,

come back, consolidate everything, and

just give me clear signals of a couple

ideas, and then from there, I will

choose. Let me know if there's anything

specifically that you need for me as far

as like API keys or if you don't have

access to certain data because I've

already set you up with my Herk 2

project which you can go in there and

you can use you know you might have more

reach into some things but just let me

know if there's anything that you can't

reach that you would really like to. So

obviously guys because this is sort of

more of like a challenge style video I'm

going to assume that you have a little

bit of previous like cloud code or

codeex knowledge. If you don't then I'd

recommend checking out this video right

up here. It's going to walk through a

lot of the basics, how you set up

different like environment variables,

API keys, and it will make more sense

when you then hop into a video like

this. So, if you want to check out that

full free course, check it out and then

hop back over here because you can see

here my cloud code was able to have a

YouTube API, school, community access,

X, firecrawl, perplexity. So, it's

spinning up five different research

agents in parallel in order to start

doing all this. So throughout this

challenge, I'm going to be thinking

about how can we do multiple things at

once just because we're trying to move

through this fast. And you know, it's

already 9:00, so we've already, you

know, used basically 15 minutes. And I

have a feeling that the majority of the

time spent in this challenge is going to

be the AI processing and us just waiting

for things like these background agents

to run rather than me actually building

or prompting. So that's why we're going

to try to stack as much as possible. But

I do want to wait until we know the idea

because then once we know the idea, I

think we can start having a different

terminal start to build out some of the

packaging like you know the title or not

the title, the name, the logo, the

colors, maybe even start building the

landing page once we have an idea of

what the promise is. But right now we're

just kind of bottlenecked. We're waiting

on the whole ideation and research

phase. So I'll check in with you guys

once I figured out what the product is

going to be. It's okay though because

while we wait I can have some coffee and

I'm working on a little Hermes agent

right here. So, you know, I'll just keep

monitoring but I can still be productive

while that's going on.

Okay, guys. So, it's almost 10:00 now.

So, that took quite a while. But this is

arguably the most important part because

the idea, the pain is the entire

product. That's what everything is built

around. So, let's take a look at what

this scraped. It looked through about

16,000 YouTube comments. It looked

through about 8,000 comments and, you

know, within posts inside of my plus

community. It looked at about 4,000

tweets and it also looked at Reddit. It

was still running at this moment, but I

asked how many it did. And it looked

through about 20,000 comments on Reddit.

So, this is the whole idea of being able

to outsource the thinking and the data

collection, but not outsourcing the

understanding. And that's exactly what

we're doing here. Because in order for

me to read this many comments and scrape

as much as this just did, it would have

taken me far longer than about an hour.

But now we have to look at this data and

understand what it's telling us and what

we should build. And it basically gave

us these final three recommendations. So

a client proposal and handoff pack. So

essentially what this is is you are

putting in basically the build or the

proposal that you are sending over to

the client and then it will come back

with a bunch of stuff like a document, a

diagram, a bunch of other things. So

kind of like helping you run an AI

agency. Then we have a social visual

generator. So carousels, infographics,

you know, to be able to post things on

different social medias. and then an ad

creative resizer. So, honestly, it

recommended this first one. I think that

fits more with what my audience is

actually looking for. It's a product

that I could fit right into an existing

product. So, it just makes more sense

and I think that it's very practical.

Now, one thing about this is there's

truly no moat that I see because someone

could use this product, they could then

um basically use it for a month,

understand how it works, understand what

they like about it, and then build their

own claw skills and their own systems

around it. And that is something that

you just have to understand when you're

building something out, especially right

now, the product game is kind of scary

because of how fast different tools are

evolving and how different things are

becoming open source. So really what I

want to recommend here is I was able to

leverage my own primary data and see

what my already paying customers want

and what their pain points are. And

that's exactly what you should be doing.

And if you don't have an existing

product or audience, you should really

be looking at your own subject matter

expertise. If you know a lot about

construction companies, then you should

be doing this sort of research and and

scraping for pain points but within the

niche that you actually understand.

Because otherwise, if you send off sort

of a generic prompt like this, and

there's nothing specific to you and your

knowledge, then you're going to probably

get the same AI recommendations that

everyone else is doing if they're

sending off some sort of product or some

sort of prompt like, hey, you know, help

me find a good business idea. So, the

idea is really important. And I also

want to show you guys something else

that it surfaced right here. So,

initially, remember how I said like,

hey, you know, maybe like 5 to 25 bucks

a month. It said based on the data that

price is too low and now it has direct

quotes. Reddit surfaced 835 price

sentences with a consistent band. So

like just a tool that does one boring

thing really well can charge 50 bucks a

month and people are happy to pay that.

You need 100 to 200 people paying 25 to

50 bucks a month for something that

saves them time every single week. So

let's say we're going to go off and

build this client proposal handoff pack

sort of product. Let's start figuring

out now how do we actually move forward

with this? We have the pain. We need to

start building out the promise, the

product, the plumbing, and the

packaging. And all the while making sure

that we're verifying and proving that

this all works as we're building. So

before we get going, let's just make

sure that our files are all set up so

that we're able to stay organized and

not get too messy. This is the only

thing that currently exists in this

project is one folder called research

with five different markdown files of

YouTube, school, blah blah blah,

competitive landscape. And these are

pretty in-depth, right? Like if I click

into this competitive landscape markdown

file, you can see that this thing is,

like I said, very in-depth. It goes over

pricing. It goes over tools with

different, you know, like actual

metrics. All of these are different

sources that I could click in and look

at. So, there's a lot of data here. I

mean, this thing ran for like an hour

just doing research. So, we're going to

be able to use all of this stuff later

because it's already in our project. And

as I'm using codeex later or clawed

later, it can read through these things.

Okay. So, what I want to do now is just

get set up with like our cloud. Mmd and

our agents so that the rest of the

project understands what we're doing.

All right. Great. So, we're going to go

ahead and move forward with sort of that

um client handoff product where they're

able to create an account. They're able

to send in sort of like, hey, here's the

workflow I'm going to build. Here's the

business that I'm building this for.

Kind of the portal that they're able to

manage some of these client projects

from. And they'll get the deliverables

like a branded PDF. They'll get

visualizations. They'll get diagrams.

And they will get other things to think

about to send over to the client like um

credential handoff information and

client expectations, things like that.

So, now that we know this is the

direction we're going to take this

product in, let's go ahead and set up a

claw.md file and just duplicate that

over to another one called agents.mmd

because I'm going to be utilizing both

codeex and cloud code inside of this

project. So, just help us get set up

with the project. Um, the idea once

again is that we're building out this

product in one day. Yeah, just go ahead

and get set up so that we can keep on

sort of like planning and building and

getting everything set up. Okay, so

we're getting some of this foundational

stuff set up and then we're going to

move into a planning phase. But before

we do that, what I want to do is I want

to get started on the packaging before

we actually start building this thing

out, just because we can do these things

at the same time. They're not really

dependent on each other. So the

packaging, so I'm going to be using

codeex for this because it has GPT image

2, which I think is the best AI image

generation model out there right now.

And we're just going to get started on

ideulating on the name of this product,

as well as what it's going to look like,

the color scheme, the feel. Okay, so

inside of our project, you should see a

research folder. We just did a bunch of

research on what product we actually

want to build inside of this project.

And we're ending up going with a client

sort of handover sort of like a AI

agency client management product where

people will log in, they will be able to

upload information about the project

that they're working on with a client

and it will create them, you know, a

proposal, a PDF, maybe even like the

audit. It'll give them the deliverables

based on their natural language and it's

going to be professionally branded as

well as there'll be images and diagrams

and whatnot. So, that's the idea for

this product. Ask me any questions if

you have any about it so that we can go

ahead and design, you know, like the

name of it and um you can give me a

bunch of different logo variations and

color schemes so that we can have a sort

of like a design system ready to go for

this entire product. And remember, these

things are working in the exact same

folder. You can see here the agents MD

and the cloud.MD just got created by our

other claude. And we have the research

folder with the five markdown files. So

these two projects are working on the

exact same thing. These two coding

agents are working on the same stuff. We

just have to make sure that they're not

like overwriting each other, which right

now isn't an issue because they're

working on very different things. Okay.

So as you can see, the clawmd and the

agents MD has been set up. What I'm

going to do now is I'm going to use a

skill called session handoff, which is

basically just going to give us a

summary of what's been done, files that

are important so that I can clear the

session and paste it back in. And if you

guys want to access this skill

completely free, all you have to do is

go to my free school community. The link

for that's down in the description.

You'll go into my free school, you will

go to the classroom, and you will click

on all YouTube resources, and you can

download everything for free that I ever

give away for my YouTube videos. And so

now that it's giving me this full

handoff, as you can see like where it

started, what we talked about, key

files, all this kind of stuff. I can

basically just copy that message. I can

clear out this um session because you

can see we were at like half of our

context window. Paste it back in. And

now it's basically going to pick up

exactly where we were on a fresh

session. So I'm going to use this

particular session to help us sort of

with the planning of the whole product.

But before I actually start prompting

this to do the planning, let's jump back

over to Codeex because it has some

questions for us. I apologize for the

jumping around and I know it might feel

a little chaotic, but this is us trying

to move fast and this is pretty

realistic of the way that I typically

like to work. I am sort of juggling

different sessions and I'm moving

between them relatively quickly. I've

found out that I can do that in a way

where I'm not sort of like um decreasing

my quality because of the jumping around

and because of the context switching.

Sometimes obviously if you're in deep

flow you need to stay on one. But when

I'm setting up a lot of these

foundational things I like to jump

around. It just helps me feel like I'm

doing more with less time. And you know,

we're kind of up against the clock here.

Okay, so the strongest positioning it

says here is to turn messy project notes

and workflows into polished client ready

deliverables in minutes. I like that,

but I don't feel like that's strong

enough yet for our actual promise. And

in order to do that, we have to sort of

think about our avatar and the pain

point that they're feeling. But it is

starting to kind of create a vision for

um what this is going to look and feel

like. So it's asking me about some

questions here that I'm just going to go

ahead and turn on Glido and start

yapping about. So to question number

one, I would say the main moment here is

that after they have gone through like

discovery calls and sales calls and

they're ready to send over some

deliverables to to actually close the

sale, they need that to be branded. They

need to be professional and they need it

to be consistent. And I think that our

avatar that we're looking at here are

people that are just kind of getting

started in the AI agency game or they've

been running one and they have a few

clients and they're starting to get

overwhelmed trying to manage the

deliverables and the things that they

need to do. So, that's kind of the

avatar we're looking at. We're obviously

targeting AI agency owners and their

pain point is just feeling overwhelmed

and feeling like they don't have a

structure to what they deliver to their

clients. Now, for number two, I think

the naming style, it needs to feel

something. Yeah, I think literal is good

because we have one very specific

promise that we're trying to make and it

should align with the name so that it

feels very consistent. It's very clear

what this product does. How should the

brand feel? I think it should feel

like premium consultancy. Number four,

should it be visibly associated with

Nate Herk? It should be completely

independent of my my name, my brand. Is

an available.com required or are there

domains such asapp? That's a good

question. Um, don't worry about that

right now. We will figure out the actual

domain to host this on later. I just

want to prioritize a good name for the

actual product. And then are there any

colors, brands, logos, or visual styles

you strongly like or dislike? I would

like to stay away from the color red and

I want this to feel very modern and

clean, but I don't want any of this to

obviously feel AI vibe coded or feel

like AI slop. So, it should feel

professional. It should feel clean and

help us also create the like oneline

promise that's a little stronger than

the one that you gave us earlier. And

then from there, I want you to start

iterating on different um logos,

different concepts, different color

schemes, all that kind of stuff so that

we can lock in a final decision. Okay.

So, now that that's off and running, I'm

going to let this thing cook. I'm going

to hop back over to the Claude session,

and we're going to get going on the

planning. So, it's asking us some

questions here, and we're not going to

worry about that yet. I'm just going to

block off that, and I'm going to tell it

what I wanted to do here. So, here's my

thought process. There's a lot that's

going to go into the build, and we're

probably going to want to utilize

different sessions to be able to build

different things at the same time

without getting in each other's way. So,

that's what I want this session to help

me figure out. So, what I'm actually

going to do is I'm going to switch over

to Fable because now we're really going

to get into the whole planning and sort

of like looking down the line and trying

to see if we can look like look around

the corners, right? So, I'm on Fable

here and I'm going to have it help me

sort of plan out the structure. And what

I want to do is I want to have like a

plan document in our files. I want to

have something that all the agents can

look at and pick up and basically see

like a running task list of what's been

done and what needs to be done and

what's maybe in progress so that they

can see what everyone's working on. And

I think that having a plan doc like that

will just keep this project feeling more

organized. So let me start yapping into

Claude here. All right. So I actually

have a different codec session helping

right now figure out the name, the

oneline promise, the logo, the color

scheme, stuff like that. Your job in

this session, you're Fable 5. You're

super intelligent. You need to help me

figure out and plan out the actual

build. So what that means is your

deliverable right now is a plan

document, something that you can use

when you are delegating work to other

sub agents to actually build. You're not

the executor. You are just the manager.

You're the project manager here.

So like I said, figure out what this is

going to look like from a building

perspective. If you don't yet have a

clear vision of the actual product, then

ask me some questions and let's

brainstorm about that a little bit until

you feel confident that you know exactly

what the user experience and on the

technical side what needs to happen in

order to build this out.

So here's my vision. The PC version one

is we're going to turn basically a

discovery call transcript and a sales

call transcript, however many they want

to dump in. They will be able to dump

those in and that's going to kick off a

new project in their dashboard for

client A. Let's just call it for now.

The AI on the back end is going to

analyze those transcripts, analyze any

of the extra information that the user

has given us. And it's going to create

them a full deck on what was talked

about. And this is really important

because the goal of this deck is to

persuade the client to actually engage

in a partnership. So, it needs to at the

beginning highlight the pain, right?

like, hey, here is the pain that you're

feeling in your business. Here's where

you are constrained and here is the

solutions or here are the solutions that

we talked about in order to actually

resolve those constraints for you and

here are the expected outcomes that we

can receive. So, what's really important

when you're working with a client is

that you are showing them that their

investment in you, their investment in

this project is going to receive a 10x

on that. So, we need to show them, hey,

here's the pain and after the solution

is done, this is why you're going to

have a 10x on that investment. This is

basically

the justification for the price. And

then we need to show them the numbers

that are going to move because we're not

selling the the the journey. We're

selling the destination. So, hey, for

example, we talked about your pain of

feeling like um you don't have enough

business coming in. And what we realized

is that you actually need a better

system for follow-up and reactivation of

leads rather than getting more leads in.

So, we were able to reframe you there.

This is the solution. And now the metric

that we're looking for is you're

currently having about five no-shows per

week. This AI system is going to move

that no-show rate from five to two

within 30 days. And that's basically

like the metric that we're looking at

moving because that metric in the

business owner's mind has a very very

clear correlation to the bottom line to

the actual revenue of the business. So,

that's just an example of the way we

should be framing this. And then it's

going to get into sort of like the

technical details of what this looks

like, visualizing it, because business

owners and people that aren't technical

really like to look at things that are

visual. So, this deck will be very

professional, very visual, very outcome

driven. And then we also need to have

the ability for clients or sorry for the

users of this system to upload their own

case studies to upload their own

specialties so that every deck is

created in a personalized way because

we need these decks to feel personalized

from the actual agency owner themselves

with, hey, by the way, we did similar

builds for these clients and here were

their results. We did this, this, and

this. That's why we're confident we can

deliver this, this, and this. So, there

needs to be that aspect of

personalization as well. They need to be

able to add their own logo and their own

brand guidelines so that all of their

PDFs are coming out branded and

professional as well. And that's sort of

the general base idea that I have for

kind of like version one of this

product. So my idea with the to-do list

that you're building right now or sort

of this planning doc is that we're able

to see all of the different steps that

you imagine we need to build out so that

we can start to delegate work out to

different sessions. We can have

different agents working on different

work trees so they're not overwriting

each other's work and we can move as

much in parallel as possible to get this

done as soon as possible. Okay, so that

was a ton of information that I just

said in there, but like I said, the

majority of this build is actually just

me using judgment, me being the project

manager of agents. as you can see the

way that I've been talking to codeex and

cloud code and it's going to get even

more intense once we really start

building but it's a it's just basically

a lot of strategically prompting and

then working in verification so that you

aren't the one factecking every single

line but you are being able to basically

be given something from an agent that

says hey I'm confident that this works

and here's why here's what I tested all

right so hopping back into codeex let's

see how it's coming along over here with

the branding side so the collision check

eliminated several obvious names close

kit is already being used by nearly

identical AI proposal product. While

client ready is already being used for

an AI report service, the three cleanest

literal territories are client pack,

scope pack, and proposal ready. I'm

treating client pack as the leading

working name and we'll compare it

against the other two visually before

recommending a lock. The first three

studies confirming the name blah blah

blah. I'm doing two more. So, it's

starting to create some branding for us

as you can see. So, we have client pack.

Um, I actually really like this. It's

like there's a C and there's a P and

there's this color scheme which is

pretty professional. We have scope pack.

I don't quite like this one as much, but

it's really not too bad. We've got

proposal ready. I like this, but this

feels a little bit more AI. I don't know

exactly why. It just feels a little more

AI. It feels a little bit more playful

than kind of like premium. And then

we've got this client pack as well.

Okay. So, I mean, so far I really like

this one. I think that this feels

consulting grade. Um, let me close out

of here and see what it's telling us,

what it's yapping about over here. It's

okay. So, it's calling this a editorial

fold. Modern, approachable, and balanced

between SAS and consultancy. I really

like that one. These definitely feel

this feels like a I don't know, like a

financial wealth adviser or something.

But I think this one felt very SAS. This

one felt AI SAS, whereas this one just

felt trusted. So, I'm going to go with

this one. I really like this one. All

right. So, I want to go with a I like

the client pack editorial fold version.

Let's go ahead and lock that in. And now

we can also see the actual messaging.

Turn discovery calls into client ready

proposals. The oneline promise is turn

discovery calls into polished branded

deliverables that help close the client.

I like that. And the short brand line is

discovery in client ready out. Cool. So

before we lock in these three kind of

like pieces of messaging, let's see what

other AI agents think. So what I want

you to do now on the messaging, I like

those three, you know, the headline, the

oneline promise, and the short brand

line, but I want to make sure it's

extremely clear and it answers pain

points. So, what I want you to do is

spin up a few different agent personas,

business owners, agency owners, um,

small business owners, CEOs, and I want

you to have them read the messaging,

look at the logo and the name, and tell

us what they think about the branding

and if they trust it, if they think it's

strong enough, or if it feels very

generic. And once we have sort of like

this collection of other opinions, it'll

help me make a better decision. So, just

spin up those agents and tell me what

they think. You don't have to have them

debate. You don't have to have them give

you a deep analysis. just tell me what

their first impression of the brand is.

And this is something I do a lot and

it's really really helpful. Having teams

of agents, having different, you know,

sub agents or even debates, even war

rooms, having them discuss and argue

things and then give you the consensus.

It really helps you make decisions when

sometimes, you know, you have three

options that you like, but you just have

to get a little bit of persuasion in one

way or you have to hear, you know, the

holes in certain plans or the problems

with certain plans. That really helps me

out a lot when I am planning things or

when I'm building things out. Okay, so

it went through this kind of like sweep

of different personas from business

owner, agency owner, small business

owner, CEO, and it made a few tweaks. So

now the headline is turn discovery calls

into client ready proposal packs. Here's

the oneline promise, which is more

specific. Polished branded proposal,

scope, ROI breakdown, and project plan

without rebuilding every document from

scratch. So it hits on that pain point.

And then the short brand line is

discovery in client ready out. Now,

here's one thing I do want to call out.

Obviously, this in itself is not enough

to be a SAS product. I don't think that

it's enough of a pain point to really

sell, especially for high LTV because

people, like I said, will take it and

we'll do something else with it. But

what this is going to do is this is a

good V1 for us to build on top of once

we set up sort of like this hub. The

vision of this could be this is

basically the operating system that

people are going to be able to use

inside of all of their client

management. Maybe it's a portal now

where people can say, "Hey, log into

this portal, Mr. or Mrs. Client, and

this is where you can see our progress.

This is where you can see, you know,

submitting tickets, things like that."

this could really start to evolve based

on other feedback and other signals that

we're hearing from the audience or from

maybe our beta users. And so maybe this

is something that you get people on for

free to start. You start collecting

feedback and then you're able to charge

later because I know the hard part that

a lot of you guys are going to bring up

is like, okay, well the hardest part is

to get clients. And that is true. And

that's why it's very tough to start off,

you know, kind of like building a

business as a SAS product. And that's

why a lot of us are going through the

service-based route first because it's

easier to be sort of like hight touch.

it's easier to get some data on what

works, what doesn't, what are the

problems, and then sort of start to use

all that knowledge to build something

like a product. Okay, so two other

things I want to kick off now. We're

going to go back to Codeex, and what

we're going to do now is we're going to

spin up a one-page landing page. That's

basically a waiting list, and I'm going

to just let this thing go at it. So,

what I'm going to do is a /goal, which

means that I'm able to set a goal and

it's going to keep pursuing. And what

I'm going to do here is, hey, so now

that we know this information when it

comes to branding, we know the name, we

know the logo, the color scheme, I want

you to create a really simple,

professional onepage landing page. What

this means is I want you to create a

waiting list for people to be able to

sign up for access once this product is

ready. So we can start going around

getting feedback and validating if it's

something that people are actually

interested enough to get on the waiting

list for. So we should have obviously

the logo up top. We should have the

headline and the promise up top as well

so that they understand what they'd be

signing up for and then a little bit of

information down below as far as what

the product will actually do. So if you

have any specific questions about what

it's going to do and feel free to ask.

You can also investigate inside this

project. You can look at the planning

doc. You can let other conversations

that I've been having about it. But

basically just create us a super simple

onepage landing page that we can start

to either run ads to or you know hit to

our email list or drop in different

communities things like that. somewhere

where we can just send people once you

have actually finished building that.

Don't stop until you've validated it.

So, open it up, take screenshots, make

sure the waiting list works. Make sure

that when they sign up for um you know,

when they put their email in there, make

sure it actually captures somewhere.

Like, we need to capture that somewhere.

Whether that be a Google sheet or um it

sends me an email, whatever you want to

do, just make sure that it's actually

all completely functional before we

actually push that anywhere. Okay? So

that was me working in verification

because if I just told it without saying

that it would probably give me a landing

page that wouldn't work. It would just

be, you know, an HTML file. People could

put in an email and hit submit and then

nothing would happen. So hopefully

because I put in that verification loop,

it's able to consistently keep building,

testing, building and testing and then

give us back something that actually

works. And that's all part of sort of

like managing these agents better with

better prompts and things like that.

Okay, so now I'm going to go back over

to our cloud session where we had Fable

give us this plan. So I'm going to open

up this doc and we're going to take a

look at it real quick. We have our build

plan for the pre-sale deck generator. We

have some decisions that were locked. We

have the user experience. So they sign

up. They have a one-time setup. So they

can put in their logos. They can put in

their case studies that they had. They

can put in anything else. They can start

up new projects. So they'll put in the

client name, the client business. They

can dump transcripts. It'll get

analyzed. It'll have a review screen on

everything that's been sort of discussed

before it actually creates the PDF. It

generates the deck and then they have a

dashboard list with all of their past

projects and other things that they can

do, edit them, regenerate them,

whatever. So, here's the structure of

the deck. We have the cover, where you

are today, the real constraint, the

plan, what moves, the math, how it

works, why us, scope, investment, and

next steps. And then we have the actual

architecture of the app. So, it's going

to be a next.js app router. We're going

to have Superbase on the back end. We're

going to have a Stripe checkout. Okay,

perfect. It's going to have two AI calls

per project, which will be to analyze

things and to create the deck. So, this

is pretty in-depth, right? This is

basically the build plan. We have phase

zero, the foundation. We have phase one,

and then we have phase 2. And there's a

lot going on here, right? Like there's

basically A, B, C, D, E, F, G, H,

different work streams in here. And the

idea is maybe we can have these all

being built in parallel. We'll see how

that's possible or not, but we're

obviously going to be using Fable to

delegate all of that around because it

knows how to do this better than I do.

So, here's what we're going to do now.

I'm going to do a /gal prompt in cloud

code as well. This is not just codec

specific. It's also available in cloud

code. So, I'm going to be able to set

this condition here. Now, I'm going to

prompt this thing in a very very

particular way where I'm once again

telling Fable 5 that it is the project

manager. It should not be executing

anything because that would waste a lot

of our Fable credits. It should just be

delegating work around. And real quick,

guys, one thing I wanted to throw out

there before we really get into the

weeds of this video is that building and

scaling a SAS product successfully is

not easy. And I hope that because of

this video, you don't think that that's

the point I'm trying to make. You know,

I'm I'm seeing what's going on behind

the scenes with Glido and how complex it

is and how much the team is working on

the build and constantly iterating. It's

not easy, but what it is is 100 times

easier than what it used to be before

AI. And I think that a lot of us have so

many ideas that we want to build, but we

just never do it. So really, the point

of this video is I want you guys to see

in one day what I'm able to achieve just

by not being technical at all and just

by being curious and prompting and

managing agents and having them verify

things. The way that I'm doing this and

sort of the the mindset that I have when

I'm building all of this out today is

really what I want you guys to actually

take away from the video. So now that

I've got that out the way, let's just

get right back into it. Okay, so you are

the project manager of this entire

project. You need to see that it from

end to end all gets implemented right

and all gets built right because you at

the end of the day are responsible for

any of the mistakes here. So what I want

you to do is you're going to work off of

that build plan. You're going to work

through these phases by delegating work

to different sub aents using opus or

sonnet whatever you see fit for the

complexity of the task. So what I want

you to do is you can use dynamic

workflows, you can use just sub agents,

you can use agent teams, whatever you

see fit because you know how to do this

best. You know how to manage the project

best. Your job is to do this as fast as

possible. Meaning anything that can be

done in parallel, do them in parallel.

There's no need to wait and do this as

an assembly line. So figure out based on

your build plan what can be done

together and what has to you know what

has dependencies and just execute this

as fast as you can. Um if you need to

use different work trees, feel free. If

you want to just spin up a bunch of

different sub agents, feel free. But

what's really important is that you're

having the agents write progress back to

either the build plan or each agent

independently is creating a new folder

and is saying, "Hey, here's what I've

done. Here's what I'm still working on."

So that you can always check in on them

to see where they are to make sure that

they're not overwriting each other and

to make sure that you're constantly

verifying what they're doing and giving

them feedback. So your end deliverable

to me is basically a finished V1, a

finished P that we can get in there and

we can test and see if we're ready to

kind of like move on to the next phase

of this thing. A few things I imagine

potentially getting in your way would be

setting up like a superbase connection

because you might not have that API key.

Setting up the Stripe integration if

there's anything like that. Build out

the supporting infrastructure around it.

Build out the plumbing and then you can

basically flag me at the end when you

finished and say, "Hey, I need this API

key and I need this API key." But that's

the only thing I should have to do. You

should be able to then take the API keys

and set up everything else that has to

do with the front end and the back end.

You should be thinking about how people

pay and what it looks like when they get

onboarded, where their accounts get

made, how that happens in the back end,

and how we as admins are going to manage

the infrastructure in the back end of

our users and our databases and things

like that as well. This has to be idiot

proof on the customer journey side as

well as on the admin side. So that's

your goal. Go ahead and do whatever you

need to do to accomplish these tasks.

Okay, so that was obviously a very big

prompt, a huge slashgoal. So, this is

probably going to take a few hours, and

that doesn't surprise me. So, what I'm

going to do, though, is I'm able to use

this session on my phone. As you can

see, we have remote control, and we can

turn this on. As you can see, that just

got turned on, which means I can go into

my phone and I can actually manage this.

So, right here, I'm not sure if you can

see, but this is the prompt that I just

sent off, right? Um, I'm able to manage

this now from my phone, which means I'm

gonna be able to maybe go make a quick

bit of lunch, maybe go on a walk, maybe

go down to the gym, and I can just keep

checking in on this so that I'm not

wasting time because this is, like I

said, going to run for a little bit

because that's really what we're doing

now. We're kind of waiting. We have um

Claude over here working on the actual

build. We have Codeex over here working

on the packaging. And I guess the only

other things that we could do maybe is

we could go ahead and give them a

Superbase API key and a Stripe API key

or something so that they could start

building in the actual integrations. But

for now, I'm not going to mess with

them. I'm just going to go ahead and

wait and then we will come back when

these are done and we'll see what else

we need to do. But this is why this

stuff is so powerful. I mean, I was able

to send off one request and it's been

building things, viewing them,

iterating, viewing again, verifying,

iterating, using the sites integration

to see if it's working. It's, you know,

creating this type of stuff, and now

it's just going to keep going until it

has proven to me that it's all working,

which is just really, really cool. And

then on the other side, with what

Fable's doing right now, it basically

was able to work off that plan doc, and

now it knows exactly what to do, and

it's running a bunch of different agents

in parallel. As you can see, it's

spinning up tons of different workers to

do one very specific thing, which is

really, really cool because it's really

tough to have one agent do everything.

It gets confused. The context fills up.

You get the rot. You just get all of

this type of bloat. But if you're able

to have one main like fable manager that

delegates off tasks, spins up new

agents, and has them do one very

specific thing very well, the quality of

everything is just so much better. As

you can see, it's already finished all

these agents. right now. I can view the

transcript and I can see exactly what

they did and I can keep looking into

what they're doing in real time. And so

we're still, like I said, very much in

control, but we're not having to be the

middleman between everything. So don't

underestimate how powerful this stuff is

when you're able to kind of clearly have

a vision for something and then have

something like Fable or Codeex or

whatever it is help you sort of

orchestrate in the right way using a

bunch of different agents to help you

achieve that vision. And so really the

main point I'm trying to make in this

video is that there's a lot of things

that you might not know. Like maybe the

idea of, oh, how do I even integrate

Stripe into my own app? How do how do I

even do that? You don't have to know.

You have to outsource that thinking and

that research to something like Claude.

And then it will say, hey, in order to

do this, I need you to set up this

connection. I need you to give me this

key. And then I need you to enable that

in the actual UI. And then you can just

go ahead and do that. And it will take

care of the rest. I'm not trying to say,

"Hey, in this video I'm going to make a

million dollars with my AI app." I'm

showing you what it looks like to

actually build the thought process of

outsourcing the thinking but still

keeping yourself in the loop in the

right areas and hopefully motivating you

guys to realize that you can give AI a

lot more ambitious goals than you

actually thought that you could and you

can really just start to leverage this

even more to do even more yourself. I

mean, just take a look at all these

other agents we have. We have currently

nine that are running anywhere from

intaking and reviewing the UI, the data

layer, the API routes, viewing the PDFs,

um looking at the extraction spikes and

fixing them, even checking on Puppeteer

to make sure that it can go ahead and

click around on things. It's super super

cool. And because Fable isn't doing

this, because it's Opus or Sonnet doing

all this, our limit is going to last a

lot longer because we have a weekly

Fable limit, which we've only used 3%

of, even though we've done so much work

already with Fable.

All right, guys. So, take a look at

this. The client pack landing page is

done. So, we can go ahead and open it

up. And it looks like this. Honestly, I

think it looks really good. Like, you

can tell it was made by AI because it

has similar formatting. You know, the

big hero text, the kind of the the hero

image over here, but it's not bad at

all. And when you think about these

little images that it had to create, it

definitely has a nice kind of like

professional vibe. Even the depth in

this little logo here. I really like

this from a first pass. We could

obviously iterate more if we want to.

I'd say this little pulsing icon thing

is very AI, but this looks good. The

next question I had for it then was

like, okay, so how are the emails being

stored? Because obviously that's the

whole point of this is to collect emails

and it did all these verification

checks, right? It said that the emails

are being stored in the site's database.

So that's not something that we really

had the ability to access. So then I

came back and said, "Okay, so how do I

access these?" It says that you could

put it on a different password protected

weightless page. And I said, "Yeah, go

ahead and set that up." And so after it

did that, now we have a different admin

page that I can go ahead and sign into.

And when I sign in, I now get this

second page. I can sign out. This is,

you know, the email that I signed in

with. I could refresh this. I could

download this as a CSV, and I'm able to

see the most recent sign up, the total

signups, and I get all this data. So,

let's see if this actually works. I'll

go into here. I'll put in a fake email,

and then we'll go ahead and join the

wait list. See what happens. It says,

"You are on the list." We'll go back

into the admin. Let's see if we can give

this a refresh. There we go. We now have

a new entry signed in right here. And

now we have basically an email list

growing of people that are interested in

this product. So the next step now,

while Claude is still building out all

this stuff, obviously, like I said, this

is going to run for quite a bit. So

we're going to keep letting this thing

cook. What I want to do now on this side

is we want to actually be able to expose

this to an actual domain because right

now you guys can't see. Let me just copy

this. This is the domain that's being

hosted on right now.

Clientpack-weightlist.nateherk.t

tragic.

Which is cool because that is a URL that

anyone could access. But what we want to

do now is we want to be able to get this

to something that we actually own and it

looks a little more professional because

if you sent people to this URL, it just

doesn't have the right vibe, right?

That's clearly a chatbt website. So what

I want to do is push this onto a GitHub

repository so we can keep that codebase

and we're going to push that to

something called Verscell which lets us

basically host that URL. So,

okay, now that we have the site done, I

want you to create a new GitHub

repository for this. So, you should be

authenticated into my GitHub already. Go

ahead and create a new repository for

this. You can just call it the client

pack landing page. And then once you've

got that pushed, I'm going to connect

that to so we can actually deploy this

on our own site. Now, if you've never

used GitHub before, it's very, very

simple. Just type in github.com. It's

free to create an account and this will

actually let you sync your Versell

account with GitHub. So go to GitHub,

make an account and then you're going to

have to authenticate in to Codeex or

Cloud Code. Just basically say, "Hey,

here is my GitHub account. Um, help me

authenticate in and it will basically

pull up like an OOTH signin." You'll

sign in. It's super super easy. And then

it will have access to help you with

your repositories or basically just a

codebase. And then you're going to also

go to Verscell. And this is where you're

going to be able to manage other things

that you want to put onto a real live

URL. And this is where you can actually

buy your own custom URLs to put on an

app. So for example, right here I've got

nateherk.com.

And this will deploy by default on

natek-.vercell.app.

So it's doing the same thing that chatbt

does when it has like the chatbt.

But then you can also connect your own

custom domain which is super super easy.

So what you want to do that it referred

to earlier is you want to go see if your

domain is actually available. So you can

go in here to versel. You can go to

domains and you can either buy one from

here. So for example for client pack I

could say clientpack.com.

That's obviously unavailable. Um what if

I do like getclientpack.com? This one's

available for $11.25

as you can see. And we could buy this

straight here from Verscell. And then we

could just say okay cool. This is the

project that we have set up for client

pack. use this domain rather than using

a different one. Right? So that's how

you would do it if you wanted to buy

your own custom domain for your landing

page, for your website, for your app,

whatever it is. Okay, so that obviously

worked super quick here. It pushed this

to a GitHub account, to my GitHub

account, and it called it client pack

landing page. So I can click on this

link. This is what the GitHub looks

like. It has basically all of the files

and folders that we need in order to

host this site. And what this means is

we can have version control here. So if

later we update the color or we update

some of the copy, we could push that

back to GitHub. So now we have a working

version and we can go back and see

different versions if we changed

anything or anything that we added. And

that helps us be able to protect our

codebase. So now you're going to go back

into Versell. You're going to click on

add new project. And then all you have

to do is you're going to import the Git

repository. So right here it's going to

prompt you to connect your GitHub

account if you haven't done that. So go

ahead and connect the GitHub account.

And then you can literally just click on

this is the GitHub account or sorry the

GitHub repo. I can literally just click

on import. And then I can deploy this.

And all this is going to do is now it's

just going to take this landing page and

our admin site and it's going to put it

on our own URL that we can own like I

said. Okay. But there is an issue here

which sometimes happens with deployment.

So I'm kind of glad this happened. I can

show you what to do. So you can see it

says 404 not found. I'm going to go to

the dashboard. We're going to click on

the actual URL that it gave us and it

says 404 not found. So what I'm going to

do is I'm going to take a screenshot of

this because codeex will know how to

actually solve this. So I'm going to

copy this screenshot. I'm going to paste

it in here and say okay so I connected

my GitHub account. I connected the

GitHub repo that you just made client

pack landing page to Verscell and when I

tried to deploy it this is what

happened. I'm not sure if because it

tried to deploy this with Vite or

whatever happened but if you need more

info from Verscell let me know but this

is what we're currently seeing and you

need to fix this. And just like that,

you guys can see it's still working, but

it's already fixed the issue. I go back

into my versel, I refreshed it, and you

can see we now have this showing up as

the preview. And when I click into the

actual domain, we now have the actual

site that we were looking at ready to

go. So, for the most part, our packaging

has been done even though the product

isn't yet done. And I think it's really

important to realize that you can do the

whole like selling before you build idea

to actually validate that what you're

about to do, the time you're about to

put in to this product is going to pay

off. And then we even had codeex without

asking go ahead and verify that all of

that is working. So we are good on that

front. And Fable is still cooking up

over here. So I'm just going to keep

letting it do its thing. And because we

got the remote control on my phone, I'm

actually going to grab some lunch, run

down to the gym, and just keep checking

in. and then whenever it's done, I'll

come back and give you guys an update.

Okay, so I have not looked at this at

all yet. It says that it's done. It says

version one PC has been finished. And we

can go ahead and open it up. I'm

prompting it right now to open it up in

a different local host for me. So, it's

spinning that up right now. But here's

what got verified, not just built. So,

it said the full customer journey with

real model calls, meaning once a

customer comes in here and puts in

information, it will send off that

information to an AI model. Now, it

probably chose an enthropic one because

this is clawed, but that's part of

what's baked in here. And so,

essentially, that's what a user is

paying for when they subscribe to this

on, you know, a monthly subscription. We

could have maybe done like a pay as you

go billing, and you have a choice there

as far as the model, the business model

that you want to run. But what it chose

to do here was you have a subscription,

you probably get like 10 generations a

week or something like that based on how

much you're paying. And then we're

technically eating that API cost. Now,

as you start to scale that up, there's

going to be different types of problems

and other things to think through, but

that's the way that it's currently set

up. From there, it said the deck is

actually good. I screenshotted every

generated PDF and looked five layout

defects found and fixed, then reproven

on a fresh generation. A geometry fit

check script now guards all 10 slides.

So, one pack is going to cost about 16

to 20 cents. Every AI call logs, tokens,

and dollars to a specific event or to a

specific place in the admin panel, and

that is what got checked. Now,

obviously, this isn't a finished thing.

This is V1. We still have to set up a

few API keys. We have to get a couple

things sorted that are open items like

superbase forcell anthropic keys, things

like that. But we have something now

that we can sort of look at and give

some feedback on. So the demo is live.

We can go ahead and open up our local

host. So I'll click on this. And that

just opened up this page right here,

which is a local host. So no one else

could actually touch this. But this is

kind of the UI that we're looking at.

This is client pack. We have projects.

We have brand and case studies. So this

is where you as a user could come in

here and say here's my agency name,

here's my color, here's what we

specialize in, here's our logo, and then

we can also put in some case studies. So

I think the idea here is that every deck

is generated in your in your style. So

your logos, your colors, and your notes.

And then if there's a relevant case

study, you can put them in here so that

they can be automatically added into

your decks and into your, you know,

potential sales material. We also have

billing here. So right now we're on a

free plan. Zero out of one decks. The

free account covers one deck in total

and then we can go ahead and switch over

to a paid membership. So, it's a super

clean, super easy to use UI. Now, let's

take a look at here. We have two client

decks. One of them is a draft and one of

them is ready. Let's just go ahead and

click on the ready one real quick. So,

radiant skin med. If I open this up, it

basically opens it up in a little

project portal right here. We can see

the deck is ready. We can see the pain

points, the investment, the value per

month. We can view deck, edit, findings,

regenerate. And here are the transcripts

that we put in to this specific um

project. So, we put in a discovery call

on August 2nd and a follow-up call just

as a text on August 6. So, it looks like

when you add a transcript, you can do

text files, doc files, VTT, as many

files as you want. Or you can paste a

transcript in right there. So, that's

pretty cool. We also have project

details. So, I can open this up. We have

client name, what they do, anything else

we should know. Okay, perfect. Now,

let's go ahead and actually view this

deck. Okay, so it opens it up right

here, and we have 10 slides that we can

flip through here, or we could also

download this as a PDF. But here's what

we have. So, this is our agency name.

This is where our logo would appear. And

you'll notice that these colors are

basically being presented because this

is the color that we entered for our

agency. So, Radiant Skin Medspa, that's

the client. Fill the chairs you already

booked without raising ad spend. So,

prepared by us. Here's the date. And

we'll go ahead and go to the next slide.

So, it starts off with where you are

today. Five appointment slots a week go

empty. Blah blah blah. Inquiries you

already paid for, wait days for a reply.

And then Botox clients need a touch-up

at 12 weeks and nobody asks them. So,

what this did is it pulled specific

quotes where the client themselves told

us the pain. And this makes it a little

bit stronger cuz it's like, oh wow, I

did say that this is definitely

something that we have, you know, a pain

point in our business. And there's three

of them backed by real quotes from our

discovery call. So that already gives

them a sense of like, okay, I trust

these guys a little more. They're

clearly not just like putting me through

some AI generated like prompt. They're

pulling specific things that I said in

our call together. We keep going here

consistent with our branding, consistent

with our color screen. This is the real

constraint. They came in basically

saying, "Hey, we need more leads." But

what actually is true is you need 30

minutes after an inquiry and 12 weeks

after a visit. So this idea of the

reframe is positioning us as the

consultants, the long-term thought

partners who are really thinking about

the business and not just saying, "Okay,

oh, you want this? Here you go." We're

not just taking orders. We are actually

helping think about the business. So

here is the plan that we talked about.

We have three steps right there. We move

on to the actual metrics that we're

trying to move. So you can see five per

week goes down to two per week. The

first reply about two days goes to about

under five minutes. And then the

repeatable injectable bookings per month

is about four per month. And we want to

get that up to 10 per month. So we're

really showing them what it will look

like when you actually work with us.

Meridian Automation. Here we get into

the math. So this is basically what you

invest and this is what it returns in

year one. Obviously this has to be

backed by real data, real things that

they said, real numbers of what the

automation or what the solution is worth

to them is worth to their business and

value. But this once again is just going

through the framework of we have a

system as consultants. This is how we

work with businesses. This is how we

show value. This is how we show the ROI

and having this sort of consistent way

that you're presenting these decks is

going to help you come across more

professional. It's like you have your

own framework. Now, here we can see we

sort of dig into the actual processes.

So, here's the manual process and here's

what it will look like after launch as

well as a little bit of a visualization.

Now, this could be reworked a little

bit. I think this could be better. But

for a first step, this helps us be able

to really really visually explain, okay,

this is the trigger. This is a step that

happens. Here's maybe where a decision

happens. Here's where AI is involved.

Here's where your team is still

involved. So, it once again starts to

create a better picture in the client's

mind. We move on to why us. So, this is

where you can show specific examples

from your case studies and you can edit

the information. Here we have the actual

scope, what's included, what's not

included. And then we have the

investment and next steps. So, once

again, this is not like the formal

contract that they would be sending to

the client to get them to sign. This is

more so like, hey, we had three calls.

Here's what we learned. Here's what we

want it to look like when we work

together. are you guys in or not? So

that is what the PDF actually comes back

as. Now once again, I could go back to

this project. I could edit the findings

if I wanted. There's a lot of different

things here. We can change the severity.

We can move it up or down as a different

painoint. And we can really customize

this as much as we want. As you can see,

there's so much here that we can do

because obviously we want to make sure

that this is legit, right? So we have

all of these metrics we can play with

and then we can hit regenerate deck and

we get a new deck. I'm curious what this

one looks like in draft mode. So, I

think the only reason this is a draft is

because there's not yet a transcript in

there. And once we had a transcript in

there, we would go ahead and hit

generate deck or analyze deck and then

we would get something back. So, for a

first PC with the main functionality

here being taking transcripts and

turning it into decks that you can play

with and, you know, iterate on with your

own branding and case studies, I'm

pretty happy with this. Like, this is

V1. This is the first time I opened this

up. And now what we have to do is make

sure that everything is working because

there's probably some bugs in here that

we haven't yet found. But we also

haven't gone through like the onboarding

phase. So if I go ahead and click log

out in the bottom left, this doesn't

actually do anything. You know, maybe

it's cuz it's blocked, but I can't hit

sign out. Like yeah, nothing's happening

when I hit sign out. So I think the next

step that we have to go through is kind

of like stress testing this piece, but

then also getting that onboarding flow

set up. So I'm going to go back into

Claude and we're going to just chat a

little more. But one thing I want you

guys to pay attention to is all of that.

We could have done that all with Fable

5, right? And I'm not sure how much

better it would have been, but look at

this. We only used 1% of Fable since we

last checked in. So all of this

basically coding and decision-m Fable

was routing to different agents and we

weren't actually using Fable's real

compute to build out that code. So

that's really cool and it's a great

trick to use when you're sort of like

managing these agents. So, I'm going to

go ahead and once again do a session

handoff because we are kind of hitting

that context rot territory and I just

want to clear the context to throw off

another task while still keeping us like

on the right path. Okay, so I have this

handoff message. I'm going to clear this

chat, paste that back in, and I'm just

going to say here is what just happened.

Don't do anything yet until I tell you

what to do. So, I just wanted to read

that and get familiarized with what what

what's going on. Now, I'm also going to

open up Codeex and I'm going to open up

a new chat inside of our AI SAS Sprint

project and paste in that same session

handoff because what I want to do here

is I really like to use Codeex as my

verifier when it comes to an actual app

and opening it up and playing with it.

Codeex is much better with computer use.

And what I want it to do is I want it to

basically come in here, spin up a new

project, generate transcripts, paste in

different types of documents, paste in

different things. And I want Codeex to

try and break this app. I want it to

find bugs and find holes in it so that

we can go back and fix them. So that's

exactly what I'm going to do. So you can

see inside of Codeex, it says, "Okay,

cool. Here's the context. It's running

here." Blah, blah, blah. Okay, Codex,

here's what I want you to do. This is V1

of client pack. You can see it open in a

local host. What I want you to do is I

want you to use your computer use and I

want you to click every button. I want

you to play around with this. I want you

to try to break this. I want you to try

to find bugs. Your goal is to tell us

what needs to be fixed so that we can

fix this before any customers actually

come in here. I want you to try to

upload different formats. I want you to

try to generate transcripts or sorry,

generate decks. I want you to play with

different buttons and find where there

might be either like a visual bug or an

actual technical bug, a limitation in

the functionality of this app. And once

you've tested this um you know through a

bunch of different workflows and tried

to find edge cases and you either have

found a bunch or you found none, once

you feel confident that you've tried

enough stuff to confidently say that

this version is ready, then go ahead and

let me know what we need to change.

Okay, cool. So this is going to open up

the browser. It's going to start to play

around with it a little bit. And that's

great because it's not going to change

anything. So what that means is over

here back in claude I can have it start

building out some other stuff without

interfering with what Codex is doing. So

I'm going to come in here and say okay

cool. So right now I have codeex testing

everything the version you just gave me

in the local host. It's going to start

testing things generating new decks

playing around with different buttons

and seeing if there's any bugs. I will

come back to you with what codeex finds.

Now what I'm curious about from you is

once again you're Fable 5. You're the

orchestrator. You don't execute

anything. You're basically just um

you're basically just orchestrating a

team of sub agents, opus workers, sonnet

workers to help you build this out and

you are the project manager who's

responsible for everything. I'm now

curious about the onboarding flow. We

have to figure out what does it look

like from the beginning into the actual

dashboard when a client comes in and

they pay, you know, they subscribe and

then they onboard, they get logged in.

Where do we hold all that data in the

back end? And what does that journey

look like for the customer? Because it

has to has to be quick. It has to be

easy and seamless. Otherwise, we're

going to lose paying customers in that

onboarding process. So, that's what I

want to build out here. And I want to

make sure that I'm giving you everything

you need as far as like Superbase

integration, Stripe, whatever you need

here. Let's start building out that

whole onboarding and the payment

processing flow. Okay, cool. So, we have

these two other prompts coming back.

This one will probably take a little

bit, you know, of testing. And then this

one over here will probably be going

back and forth a little bit to answer

some questions and to set up some things

like superbase and whatnot. But real

quick in codeex, if I see over here the

browser, we can actually see it trying

things out. So right here you can see it

tried to enter an accent color like it

tried to type in red, but we obviously

need to use a hex color like 275 blah

blah blah. So now it put that in. And so

this mouse right here is codeex. This is

codeex using a browser. You can see it

clicking around. You can see it moving

around. And this is what's so cool is

it's just going to test a bunch of stuff

and then it will come back and say,

"Okay, I tested these a thousand

possibilities which would have taken you

like 12 hours to do." So, so using

codecs here as a computer use verifier

really really fun and also really really

helpful. Oh, I just realized that these

lights weren't turned on. There we go.

Now we got the the good vibe in here as

there's a bunch of clouds coming over

Lake Michigan. Hopefully it doesn't

start storming on us. And while it's

running through this, let me show you

something real quick, which is a slash

by the way, which means you can ask

Claude a question without interrupting

this main kind of like chain of thought.

So, what I'm curious about is the text

stack that it wants to use because I

remember it mentioned Superbase and

Stripe, but I just want to make sure

before I go off and like create a new

table or create a new account. So, while

you're thinking about this, I'm curious

the text stack that we're going to use

on the back end for authentication. I'm

not sure if you were planning on using

Superbase or Stripe or what all are the

things that you need, but if you could

let me know what we were planning so

that I can go real quick and spin up the

accounts and get the API keys ready and

stuff like that. So, as you can see,

it's a little side chat, right? It's

just it's super super cool feature when

you have a little one-off question like

this and you don't want to interrupt it.

And as you can see down here, the answer

we got is that Superbase is going to

handle the authentication. So, it's

basically going to store our users.

It'll be our customer database and then

Stripe is going to be just the payment.

So, it's never going to touch off Oth.

So, that's super clear. Now, we know

what accounts we'll have to go set up.

You can see that we're going to have to

get some API keys and things like that.

Superbase. And then from Stripe, we'll

have to get a secret key, probably a web

hook secret as well. It's already got

our anthropic API key. It's already got

this other stuff. So, cool. I'm going to

go ahead and close out of that chat. And

now you can see the main session has

some questions for us. So, what does a

new user get before paying? This decides

whether checkout comes before or after

the first deck. I would say yeah, free

deck, but it's watermarked. That's

perfect. Final price point config

currently says 39 bucks a month and 390

for the year. Research band is saying 29

to 49. Let's just go ahead and go with

the recommendation since it did go ahead

and do a bunch of research. So, I'll

click on the recommendation and now we

will see what comes next. So, the

journey that it wants is there's a

pricing page and you can sign up and

then you check out and then there's a

three-step onboarding which gets you

into your first deck. So, we're