Learn to Build Software with Gemini 3 (In Under 10 Min)

In the next 10 minutes, I'm going to show you how to use Google's free tools, Notebook LM and Gemini 3 to build almost any kind of software. And the example I'm going to walk through today is this AI agent that uses Gemini 3 in the

background, but it's trained on your specific knowledge. These aren't thin

specific knowledge. These aren't thin demos, but real products that users can log in, sign up, and pay for. And once

you get the gist of this, you'll be able to create highquality working software with just [music] a single prompt.

There's also a simple trick I want to show you that'll help you continuously improve your software without getting trapped in an endless circle of bugs.

You can use this to launch your own startup or to make some significant revenue immediately by doing this for other companies. I ran a marketing

other companies. I ran a marketing agency for 10 years serving some of the largest software companies in the world and we spent a fortune building our own custom software. So, I'll share with you

custom software. So, I'll share with you what you might want to charge for projects like these as well. AI models

like Gemini 3 have just crossed an undeniable threshold. And it's moments

undeniable threshold. And it's moments like these that I feel confident in my decision to shut down my marketing agency and go allin on AI. And I hope by the end of this video, you feel that same confidence as well. So, here's what

we're going to get into today. We're

going to start by using Notebook LM to perform a gap analysis and turn that into a project requirements document to find opportunities that other people miss. Then we're going to gather up some

miss. Then we're going to gather up some design inspiration and generate a few different design ideas so you have complete control of what your software looks like. The third step is the

looks like. The third step is the foundation, and this is the counterintuitive part that I don't see hardly anybody talking about. And if you get this right, you're going to have a solid template that you can use over and over again. Then we'll build out the

over again. Then we'll build out the core features, and this is the fun part where your app really starts to come to life. But it's important to know how

life. But it's important to know how hard you can push the AI in one single prompt, and what might be best for a phase two. Then we'll test and deploy

phase two. Then we'll test and deploy your software using the simplest path to get your software out to the world. And

finally, we will look at how you can continually refine your software on an ongoing basis. The first step in this

ongoing basis. The first step in this process is the research step, and I've covered this extensively in other videos. I'm going to link to those in

videos. I'm going to link to those in the description. I'm also going to link

the description. I'm also going to link to this ebook in the description. I make

an ebook just like this for every single video that I create. It's going to include all of the prompts that we go through today and a whole bunch more actual clickby-click instructions on how you can get these things set up. This

ebook and dozens of others are instantly available to anybody who joins my Patreon. Again, check out that link if

Patreon. Again, check out that link if that is interesting to you. The first

prompt that we're going to use inside of Notebook LM is this prompt to search Reddit to find pain points around your customer and the uh problems that your software will solve. We're going to copy

that right into the discover feature of Notebook LM. Then we're going to use

Notebook LM. Then we're going to use this prompt to look for the competitors in your industry. We're going to drop that right into Gemini to find the established startups or market leaders

in your space. Then we're going to take each one of those and return any and all Reddit reviews for your major competitors inside of Notebook LM. From

there, we can now use this prompt to look at all of that source material and identify the top three gaps where the current solution are failing the users.

And this prompt will then go into taking those gaps and generating a product requirements document or a PRD. That's

going to be basically the core prompt of what we want the AI to build. Once you

have that document, you want to save it into a new file somewhere on your computer. The next step is the design

computer. The next step is the design step. So find a few different pieces of

step. So find a few different pieces of software that you like the look of. And

in Chrome, you can just go to file, save as, and save that HTML in that same folder as well as a screenshot of the software that you like. I would do this for the homepage as well as the core

dashboard page of the software that you'd like. Next, you're going to fire

you'd like. Next, you're going to fire up cursor and you're going to open that file where you've saved everything. Make

sure you've got Gemini 3 Pro selected here. Then we're going to copy and paste

here. Then we're going to copy and paste this prompt right out of the cheat sheet that asks the AI to create three different HTML versions. This allows you to look at a couple different ideas that the AI has generated and make sure you get that design exactly how you want it.

Next, we're going to build out the foundation using this prompt here. This

selects the design that we like the best and tells it to delete the other files and asks it to begin building the software using PHP and SQL light, which is a very simple tech stack that I've

had a lot of luck with. Some of the more complex tech stacks can lead you down a lot of different rabbit holes. For this

step, we want to make sure that we're focusing the AI and just building out the user authentication with Google as well as the Stripe integration for payments. And that sets us up in the

payments. And that sets us up in the great direction for building all of our core features on top of that. Doing the

reverse is way harder, trust me. So,

we'll copy this back here into cursor.

And for this one, we want to make sure we're in planning mode because we want to get a good plan in place before we start building this stuff. Awesome. It's

come back with some questions. You want

to answer these to the best of your ability. And if there are things here

ability. And if there are things here that you're unsure of, just ask the AI what it recommends. Once it returns the plan here, you can give it a quick look or basically just click build. The final

part of the foundation stage is to simply ask the AI to run the software locally and walk you through how to set up that Google authentication. And

there's just a couple hoops you got to jump through. There's a bunch of details

jump through. There's a bunch of details about that in the ebook. But once you get that working, make sure to save a copy of that code base somewhere because that can be your template that you use over and over again. You don't have to

repeat that step each time. And now the fun part begins where you can use this prompt to tell the AI to plan out the rest of the core features based on the project requirements doc. And most

importantly saying, is there anything that should be left for a phase two?

This gives the AI an out and lets it prioritize a few things that it can handle all at once without trying to do too much. There's always going to be

too much. There's always going to be some testing involved. So any errors that you run into, just feed those back into the AI until you get a stable version. Once you have that stable

version. Once you have that stable version, go to GitHub and create an account if you don't have one already and then create a new repository. Copy

that link and feed that right back into the AI saying, "I now want to push this to GitHub." This is how you save that

to GitHub." This is how you save that working version so you can continue to refine it without worrying that you're going to break something that is already working because if you run into trouble, you can always revert back to this

working version. This is also a critical

working version. This is also a critical step for getting it deployed online so other people can access it by connecting GitHub to Railway. You can deploy this very easily. I'm not sponsored by

very easily. I'm not sponsored by Railway, but it's what I'm using to deploy apps like this. You just want to click deploy a new project and it'll look and find anything in your GitHub repository. You just click that and

repository. You just click that and it'll start to deploy it here. Once it's

up and running in Railway, there's a couple areas you're going to want to know about. These variables are where

know about. These variables are where you're going to add your API keys and other passwords that you set up. And

also in the settings tab, there is a networking area which is where you connect it to a custom domain or you can use just a basic domain that Railway generates. There's a ton of details

generates. There's a ton of details about all of this in the ebook that walks you through step by step exactly how to do this. That ebook and a ton of others are available right here in

Patreon. But if you're really serious

Patreon. But if you're really serious about this stuff, I have this software builder call. This includes two calls a

builder call. This includes two calls a month where we go in detail about the best use cases for AI and how to build software around them. And if you join this, I'm going to give you the exact

codebase for this agent so you can just immediately deploy it and start tweaking it without rebuilding all these different steps. So once you've got it

different steps. So once you've got it up on Railway, there's always some testing that needs to be done to get it working right. There's a few different

working right. There's a few different places you might want to look at. There

are logs here. There are logs here as well when you run into trouble. That's

where you want to copy and paste things out of there into the AI to get things working again. And it can also be

working again. And it can also be helpful to feed in this link, docs.railway.com,

docs.railway.com, when you run into issues, so the AI knows exactly how railway works right now. You just copy that link right in

now. You just copy that link right in here, like so when you run into trouble.

Once you've got that up and running, the sky is really the limit on how you decide to enhance this software using this same process of perhaps asking the AI to start working through phase two.

Push it to GitHub. It'll automatically

be deployed on your uh railway uh server and you can test it there. And if

anything goes haywire, you can always revert back to the most recently saved version in GitHub. If you want to try out the AI assistant I built here, just go to agent2.blazingebra.ai

AI and you can log in and just begin interacting with this. Remember, it's

powered by Gemini 3 in the background.

It's got these different personas here and I think you could easily charge 5 to $10,000 for something like this, especially considering many people are charging upwards of 20,000 for just basic websites. All right, there's a ton

basic websites. All right, there's a ton more in this ebook, including some security best practices to make sure that your app doesn't get hacked. I'm

also going to include the project requirements document that I used in this uh tutorial. That way you can jump in and try to recreate it directly from that project requirements document. And

remember, if you join this software builder group, I'm going to give you this entire codebase so you can just plug that in and get it up and running online immediately. That also includes

online immediately. That also includes calls with me where if you run into any problems, we can help work through them and really make sure that you're gaining

confidence in this new world of AI. And

once you've got a piece of software up and running, you're going to want to start marketing it. So, I want you to check out this video next all about how to create blogs that update themselves as the core part of your marketing. I'll

see you over there next. Make [music]

your dreams come true.