Build Epic Startups: NotebookLM + Gemini 2.5 Workflow
By Blazing Zebra
Summary
## Key takeaways - **AI can replicate a startup team**: With NotebookLM and Gemini 2.5, individuals can perform the work of an entire startup team, accelerating idea development into a functional product. [00:04], [00:06] - **User pain points are key to startup success**: Founders should prioritize user pain points backed by valid data over ideas that are merely a founder's personal great idea, to avoid project failure. [02:39], [02:44] - **NotebookLM's Discover feature streamlines research**: The 'Discover sources' feature in NotebookLM eliminates the need to manually search for information on platforms like Perplexity, directly integrating research into the notebook. [02:02], [02:15] - **Iterative AI-driven development with Gemini and Firebase**: Gemini 2.5 can build an initial MVP, which can then be enhanced and scaled into production-ready software using Firebase, streamlining the transition from prototype to a full web app. [11:07], [11:15] - **AI for comprehensive marketing strategy**: AI can generate a full content marketing strategy, including hub pieces, blog posts, and even visually compelling infographics and interactive content, accelerating promotional efforts. [17:20], [17:56]
Topics Covered
- Can two free AI tools replace an entire startup team?
- How to find true customer pain points with AI.
- Build production-ready apps from MVP with AI in hours.
- Generate an entire content strategy and interactive content with AI.
Full Transcript
Do you realize that anybody who knows
how to use Notebook LM along with Gemini
2.5 can now do the work of an entire
startup team? Today, I want to show you
how to use Google's massive updates to
these free tools in order to research
build, and promote your ideas. I've
spent many hours experimenting with
Gemini's new superior coding
capabilities and 1 million token context
window to come up with this step-by-step
process for getting your ideas into the
world as fast as possible. You see
while I was running my marketing agency
we worked exclusively with software
companies from unicorn startups all the
way to the largest SAS company in the
world. And I've been shocked at how well
you can combine these two tools to rival
or even exceed the work of some of these
top tier teams. Did you know that Sam
Alman and his cronies are betting on
who's going to create the first single
person $1 billion company? After
watching this, I think you'll agree that
that moment is a lot closer than most
people think. So, here's the gist of
what we're going to get into today.
We're going to use Notebook LM
especially its new discover feature to
look into the pain points of your
particular customer. Then, we're going
to use it to look into trends in your
specific industry. Then, we're going to
use it to look into competitors in your
space. We're going to take all that
information and do a smart analysis on
it. Then we're going to pass it to
Gemini 2.5 to actually build out the
software to build out the marketing
website and to build out the marketing
content. And we're also going to look
into Firebase as a way that you can
really move from an MVP or a prototype
that Gemini builds directly into a
full-fledged web app that you can sell
to, you know, a bunch of people. So, it
all starts here with our trusty friend
Notebook LM. We're going to fire up a
fresh new notebook. I always get excited
when uh I fire up a fresh new notebook.
And I get even more excited now that
this feature is here. This discover
sources feature is truly a gamecher. I
know that word gets thrown around quite
a lot, but many of my videos relied on
going to Perplexity to find sources and
then dumping those into Notebook LM. You
no longer have to do that. This is a
very, very powerful search feature here.
And you got to know how to prompt it
though, because like anything, the
prompt matters. And for that prompt, I'm
jumping right into the cheat sheet. I
make a cheat sheet for every single
video that I create. These are all
immediately accessible to anybody who
joins my Patreon. Um, so yeah, check
that out. There's a link in the
description. Starting with the user's
painoint is one way to ensure that your
idea is successful. I've seen a lot of
software projects go off the rails when
it's just something that a founder
thought was a great idea but didn't have
any valid data to back it up. So, we're
going to grab this prompt and we're
going to dump this right into the
discover feature of Notebook LM, asking
it to go out and look at where these
people are actually voicing their
opinions. We're not looking for
editorials of what some person thinks
about what some other person thinks
about. We're looking for the voice of
the customer here. And the project I'm
working on today is a uh healthc care
app that I've been uh thinking about.
I've recently struggled with some health
issues and I've had a lot of trouble
understanding my blood lab work, sharing
that blood lab work out. I've gotten it
in it exists in multiple places. You
know, there's a a portal here and a
portal there. I want to build a software
product that compiles all of the data
for my blood work in one place. I think
there's a lot of people that could
benefit from this, especially if you
tack on some AI capabilities of
interpreting that and and letting you
know, you know, what these tests mean
and potentially uh you know, next steps
there. So, I'm just saying find sources
on sites such as Reddit, Kora, and niche
forums to find user pain points
regarding understanding, accessing, and
sharing blood work, labs, and similar
healthcare data. Let's just submit that.
And the better you prompt this, the more
you can just select all of what is in
here. Cool. Some of these are really
awesome. I'm going to leave that first
one out. I don't know that that's
exactly what we need, but we're going to
import all of those different sources.
Now that all those are in there, I want
to label these. So, I'm going to rename
the source. And for all of these, I'm
just going to put user in all caps. Copy
and paste that. Save. This will allow us
to easily find our user data from our
industry data from our competitor data.
All right. So, there is all of our user
painpoint data. We're going to go back
to this discover and back into the cheat
sheet here. And I'm going to say gather
studies from leading consulting firms
and other credible sources on you know
exactly what you're looking to do. In
this case it's consumer healthcare apps
looking for trends. It exclude anything
that was published before 2024. So we're
just going to drop this in to find
credible resources on industry trends.
Going to import those. And these ones
I'm going to rename. just put this label
all caps trends in front of them so we
can easily see at a glance our trends
from our user info. Trust me, this is
going to come in helpful later on when
we have a bunch of sources in there.
Cool. So, we've got all of our trends.
We got all of our user. This one didn't
come through, so we just remove that. No
big deal. And now, finally, we want to
look for some competitors. And
importantly, we don't want to find uh
reviews of these competitors. We're
really just looking for their homepages
because that's what we really want most
from them. So, I'm grabbing this prompt
here, dropping that right into this
discover. Again, looking for the most
successful. Again, you can put in
whatever you want here. For me, it's
consumer healthcare apps and tracking
software. Please return only the
homepage or the app page of fast growing
startups or established market leaders.
Ideally, we want a mix of those of the
new uh hot products as well as the ones
that are, you know, really dominating
the market here. And we're looking for
their main links to their main
homepages. Does an awesome job of this.
So, really think about how you steer
that prompt. We got My Fitness Pal
Headspace, an ovulation tracker.
Perfect. That's an interesting one. Not
exactly what I was looking for, but that
could be helpful as well. That Y
Combinator. Let's import these. And I'm
going to rename these. I'm going to add
the competitor label to these. All
right. And just like that, we have a
powerful asset here for our startup
idea. We've got a bunch of information
about our user pain points, trends in
the industry, and our competitors. So
let's get into how we're going to use
this next. First, I want to retitle this
Blood Works, which is sort of the
placeholder name for this app that I'm
working on. Okay, so now we're moving on
to this step two, analysis and strategy.
And this first part of it, we're getting
back to those pain points. A very simple
prompt that just says, please analyze
the pain points listed in these sources.
We're going to copy and paste that in.
But the key here is make sure only these
user sources are selected. So we're
going to select all of our user sources.
Going to leave everything else
unselected and ask for a painpoint
summary. And this is exactly what we're
looking for. These health care systems
electronic records are very frustrating
experience miscommunications between
medical professionals and patients. So
now we're going to save this note here.
Then you got to click into the note and
convert this to a source. So now we've
added that as a source down here and
we're going to rename this to keep
everything straight. We're going to
rename this source user painpoints. So
that one jumps right out to us there. So
the next step here now that we have
those pain points analyzed is to cross
reference those against the industry
trends using this prompt. So we're going
to just grab this and in that same exact
chat string, we're going to copy and
paste this in right here into the chat
string. We're going to turn all of the
user sources off and these trend sources
on. So now we are cross-referencing
these trends with the pain points using
this prompt that basically says please
return an aggregate of the user trends
based on you know XYZ what we're looking
into as it relates to the conversation
above. Let that run and here is where
those pain points align with the trends
here. beautiful analysis that we can
just save as a note. And then once that
note is saved, we want to open up that
note and convert this to a source so
that we can reference it here in the
sources. And we'll save that and rename
that as our trends summary. So we have
the painoint summary, we have the trends
as it relates to our pain points summary
in here. And now is when things get
exciting. We can start mapping the
features of our prototype of this
software. So now I'm going to refresh
the chat string. I'm going to select our
user pain points and our trend summary.
And I'm going to grab this prompt. Copy
that right in. This just says using the
selected sources, please outline the key
features of an app that addresses the
identified concerns here. Awesome. And
this does a really detailed job. So what
we want to do is focus it on something
that is more of an MVP or a minimum
viable product or a prototype that we
can start with and build upon. So that's
where I use this prompt. That'll
simplify this down a little bit to
something we can wrap our heads around
initially. Awesome. This has now
simplified that down into some key
features for that MVP. And now I'm just
going to follow up with this prompt
which is sort of a meta prompt. It's a
prompt asking for a prompt that we can
then put into Gemini to start building
this app out. I'm asking it to use
JavaScript specifically because that's
the language that I'm trying to focus
on. And I have a whole another video on
my path of learning how to code with AI.
I'm going to link to that video now
because I think there's some helpful
things in there if you really want to
build production ready software. But
let's see what this comes up with. All
right, we've got a pretty robust prompt
here that we can now go over to Gemini
with. So I'm going to copy this out of
notebookm. We're hopping over to
gemini.google.com. We're going to drop
this right in. Got to clean up some of
this beginning here. And I'm going to
title this blood works. We got to turn
this canvas feature on. We're going to
let that run. And there it goes.
Building our JavaScript app. And there
we have it. This is a working app here.
This is the prompt that we entered in.
It ran for a while. And I want to show
you a trick if you want to get even more
serious about this. Go to this code.
We're going to select all. And moving
out of this MVP prototype into a real
legit piece of software, we go to
Firebase, another Google product. Get
started. Log in here. And I'm dropping
in everything that we created from that
working prototype. Right in here, I'm
going to click run. And then I'm going
to follow it up with this prompt, which
just says, here's some code from an app
I created. We can use this code as is
because otherwise it kind of assumes
that you're grabbing it from somebody
else and it wants to reinvent the wheel.
So got to tell it know hey I built this
so we can take it from here and then
progress it just like that ask it say
feel free to make any enhancements you
want to improve on it and I like to say
use this shad CN which is a anytime you
add that in it's going to make it look
even better I think that is a tailwind
CSS thing that that makes these web apps
look really good. So, we'll copy and
paste that in and we'll follow that up
changes I want to make and it's going to
go through and start actually improving
this and enhancing it in various ways
and actually building it from basically
one file into a whole set of files. So
it's kind of hard to see here. I'll zoom
in, but it's now taking really just that
what was a onepage app and building out
all these different Typescript files.
really turning that into a true code
base that then you can edit and start to
improve upon just like any other, you
know, software tool out there. And this
Firebase Studio is really awesome from
moving from a prototype into a legit
production ready piece of software. This
rivals Cursor and others. Um, so
something to check out as it all stays
within that Google ecosystem. Okay, so
we've built the app and now we're going
to design this homepage for this app and
our notebook LM notebook is going to
come super handy in there. Remember all
that competitor uh information that we
pulled. We're going to use that now.
We're going to grab this prompt. We're
going to copy and paste that in. We're
going to make sure all the different
competitors are turned on here. We want
to give it access to these competitor
resources. Now, we're also going to give
it access to this MVP prompt so it knows
the tool that we're building and it can
cross reference that with all these
different competitors homepages and just
saying, "Hey, craft a prompt for a large
language model to generate a visually
compelling and effective landing page or
homepage for this product." So again
that is a meta prompt, a prompt that
we're using to create a prompt that
we're going to put into Gemini to create
the marketing web page because, you
know, when you go to uh a new piece of
software, you don't go right into the
software. You go into a homepage that
you can then click get started and you
know, access the app that way. So that's
what we're going to use to build out
next. All right, so that has run and
created this landing page prompt.
Remember to save it as a note and then
save that note as a source. And this is
the prompt that we're going to use that
summarizes the tool and all of the
language from these other uh competitors
into a prompt that will build our
homepage for this thing. Grabbing that
copying and pasting that back into
Gemini here again. We want to turn
canvas on and let's start building the
marketing website for this product. And
we're running here. All right. And there
we have our little app here. I want to
change the name here to make sure it
says Blood Works. There we go. And it
rocked those icons. I've had trouble
with those
icons before, but this time it did it.
So now to publish this app and website
to the web, I'm going to use Replet.
It's what I'm most familiar with. You
can again also use that Firebase option
but I'm just going to grab this prompt
here that uh will allow us to take what
we've already created and drop it into
Replet and get it online. So when you
open Replet, you're going to be met with
its agent here. And we're we're just
asking it, hey, take the document
attached and use this exact code. No
need to change anything. I'm simply
looking to create a new replet with this
uh and host it on Replet. So we're
adding that prompt in and then we're
grabbing the code from Gemini. So here
is the what we created in Gemini. So
grab this code, select all, copy, paste
that in, start building, and it builds
it. So, it's going to ask you to look
through a plan. This is all there. And I
I think you can just approve the plan
and start. And ideally, it won't try to
reinvent the wheel, but it'll just take
that file and help you host it right
here. And again, this is a place just
like Firebase where you can then start
to improve it and add features and build
it more into a full working codebase.
There we go. Looks awesome. Now, all we
got to do is click deploy. Uh, and
there's a few different ways to deal
with this. Autoscale usually works well.
Click, click, click. Keep clicking
forward. And there you go. Now we have
both the homepage here. This is live.
Anybody can visit it. And we have it
connected. When you you got to work
through this in replet and just fix the
link. When you click get started, it
opens the fully hosted app here. So
we've got the homepage. We got the app
all online. All just a few hours of work
maximum. Now we got to promote it. All
right. So now step five, we're going to
move into creating a content strategy
for our new startup idea. And this is
really what my background is in. I ran a
content marketing agency for over 10
years. Worked with all sorts of software
companies before I really decided to
pivot into this new world of AI. And I
started this Blazing Zebra channel.
Blazing. We got to move fast. and zebra.
We've got to, you know, identify our own
strengths, our own passions, and bring
them to this new world of AI. So, I hope
you subscribe and follow along with
these videos as you begin to pivot and
transition into your new AI career with
some intentionality, and ideally, it's a
more fulfilling career than any that
you've had in the past. So, for the
content strategy, we're going to use
this prompt here that just says
"Develop a content marketing strategy
outlining four hub pieces of content
each supported by four blog posts that
link to those different traffic sources
there." And we're going to copy and
paste this right back into our notebook
LM notebook. For this one, you can
pretty much have all of the sources on
and see what it comes up with. Awesome.
So, it's returned already a pretty
killer content marketing campaign. And
if you've done anything with AI, you
probably know how to create a blog post
or even a long form piece of content.
But I bet you don't know how to create
infographics and interactive pieces of
content. So that's what I want to show
you how to do next. So I'm grabbing this
prompt. This is another meta prompt. A
prompt that creates a prompt for
designing a visually compelling
infographic related to one of the hub
page ideas that it came up with. Very
cool. That is pretty big. We're going to
save that as a note as we always do. And
save that. convert that to a source. So
we have that. Going to copy and paste
all this right into Gemini. Turn on that
canvas. I'm also going to supply the
code from our main landing page here so
that it can follow the brand um you
know, look and feel that we've already
been working on. I'm adding that there.
You can also upload it here as a
separate file though. So, let's let that
go and see if we can create that. So
here's the result of that infographic
here. You can see it follows all of the
styles of our previous creations. We can
get that online using Replet the same
way we got the app and the homepage
online. And we can start to build out
our content portal that way. But there
is so much you can do from here. Not
only infographics, but you can create
cluster visualizations. Here's just a
peek at what this can do when it comes
to creating interactive content. I used
it to just create an interactive mind
map from about all the notebook LM
features limitations
etc. So you can create content like this
and get it online copying the code in
using that same process I showed you on
Replet or inside of Firebase. This cheat
sheet is absolutely packed with a ton of
good information. It goes through
everything that we went through here
today, but then dives even deeper into
each step of this process. So if
building out software and businesses
like these is something that you're
interested in, definitely check out my
Patreon. There's a link in the
description. You can support this
channel. Grab this cheat sheet and over
125 others instantly. There's some
coaching options in there as well. I've
got a video that dives a lot deeper into
the strategy portion of what we went
over today. So, if you're serious about
this, check that video out next.
Loading video analysis...