The NEW Way Of Marketing in 2026

Our world is going through some crazy times right now when it comes to AI.

Many companies have a smaller headcount.

Sales and marketing teams are getting smaller and not because of budget cuts.

It's because one person with ambition, taste, and the right tools using AI can now leverage and outship entire teams. In this beginner's guide, you'll learn how to set up and build a full system to

leverage these tools without hiring anyone. I'll show you how to get set up,

anyone. I'll show you how to get set up, the fundamentals of what it takes to use these tools, and how to ultimately generate copy or content around what you want to do with your business, whether

it is to generate emails, images, or videos. And believe me, most people

videos. And believe me, most people still have no idea this is even possible, which brings a total unfair advantage. So, if you're a founder,

advantage. So, if you're a founder, freelancer, or just sick of hearing the same outdated advice, this is for you.

No BS, no gatekeeping, just pure knowledge drops. I'm Brandon, founder of

knowledge drops. I'm Brandon, founder of Top ofunnel, and I build enterprise AI and automation systems for a living.

First and foremost, one of the biggest things that I talk about all the time is you really need to get into a platform that just really leverages the true power of AI and that is none other than

an IDE called Cursor. Cursor is

obviously a fork of VS Code or it's an AI coder, which you can see here. So, it

does a number of different tasks, namely building code in which you can pretty much do anything. The reason I love cursor so much is because it's easy.

They ship fast. They ship all the time.

They're always updating and they really are helping a lot of AI coders build great things really quickly, such as myself. So, I encourage you to go to

myself. So, I encourage you to go to crusher.com. Go ahead and download for

crusher.com. Go ahead and download for Mac and Windows to be able to get started. So, we're at the Vibe Marketing

started. So, we're at the Vibe Marketing empty repository or what used to be an empty repository. I basically restarted

empty repository. I basically restarted my entire project. I have a lot of projects that I usually like to work on specific to GTM and sales and marketing as well. But one thing you'll find here

as well. But one thing you'll find here when you actually start in a code editor such as cursor, the thing that you need to start with is just an empty folder on your hard drive. I think one of the

biggest and challenging things for people that are not immediately coders is to conceptualize how an IDE, which is an integrated development environment

such as cursor actually works. One of

the key things here is all of these are nothing more than just files. And then

we have our agent right here where it has access to these files. And then this is nothing more than actually opening up the files to be able to interact with it. And I can reference these files into

it. And I can reference these files into the agent to actually understand context. It'll actually read it and we

context. It'll actually read it and we can manipulate and generate files and code and pretty much anything. Right now

you might ask me, well Brandon, why don't you just start using Claude? Why

don't you use chat GBT or Manis or whatever AI tool, right? The short

answer is I use these every single day.

Yes, I use Claude. I mean, I use Claude quite heavily, especially with MCP servers to be able to connect to different things. I also use Chat GPT.

different things. I also use Chat GPT.

My wife and kids use Chat GPT, although they don't use it for school cuz it would be really bad. All of them are really wonderful. So, what I would say

really wonderful. So, what I would say is start with Claude. Definitely have it on your computer. I mean, I use it every day for small task, but if you're really trying to stand up a good VI marketing

or sales intelligence or even GTM mastermind, you really should be building an overall repository or intelligence here around what we're looking at, right? And what I'm going to

do is I'm going to get you acquainted with cursor so that way you know how to poke around, you know what it means, you know kind of how to set it up. By the

way, this will be on my GitHub repository. So that way if you just want

repository. So that way if you just want to fork it or clone it if that speaks French to you, really all you need to do is just grab the URL to the GitHub repository, bring it up in cursor and

just say clone this repo and then put in the URL and it will pull it right away and you'll be good to go. Right? When it

comes to marketing, when it comes to data, when it comes to even working on a computer, we are doing four things.

That's it. Four things. I think this is always true. And when we use AI, even as

always true. And when we use AI, even as of this recording, if I were to recommend tools to you, it's probably going to be in a month that this is going to be outdated. So, like, what's the point, right? As of this recording,

I use Claude 4.5 Sonnet, which is a total workhorse, but in a month or two, who knows? We might see Claude 4.7. I

who knows? We might see Claude 4.7. I

would say the fundamentals still don't change. So, what are those fundamentals?

change. So, what are those fundamentals?

Well, I put it in this beginner's guide because what it actually means is exactly these. There's four ways in

exactly these. There's four ways in which we work with data. We either get data, which is the acquisition of it. We

either consume data, which is the analysis and reasoning of it. Right? So,

when we get information, we now have to figure out what we're going to do with it. There could be a number of tasks

it. There could be a number of tasks like scoring leads, analyzing things, reading content, reviewing customer feedback or consuming, right? A large

part of this world consumes data and there's a lot of time and time wasted on consuming. The third one is generating

consuming. The third one is generating data, which is the creation of it. The

fourth thing is sending data, distribution, right? Could be sending a

distribution, right? Could be sending a cold email, could be posting an image on Instagram, could be posting a a tweet or an ex post. you're feeding the machine,

right? There's always going to be a get,

right? There's always going to be a get, a consume, a generate, and a post. The

biggest thing, at what point do those four things really mean something as far as getting it, and doing all those to to actually automate, right? For example,

we might have a client that will bring in thousands and almost a million rows of a CSV, right? We could have large amounts of data where a clay table,

frankly, no one wants to read that stuff. When it comes to maybe dduping,

stuff. When it comes to maybe dduping, making sure you're enriching, really qualifying your leads, getting rid of junk. It's a mess, right? This is the

junk. It's a mess, right? This is the current state of a lot of CRM systems. So, you can use something like cursor to be able to analyze this, have some true marketing intelligence, and then make a

decision from there. That is, in my opinion, best use of AI. your use case might be different but I think that's where it comes into and then the next thing here is where you use AI where

should you apply a human in the loop because you want to start small you want to iterate you want to verify you want to make sure that whatever that task was done that it was executed with high

accuracy so if you expect any system to be built out it's really important when you start you start small start simple don't try to overengineer something and then just let it rip right so there's a

reason why we started from the horses.

Then we went to a simple automobile and then you get into a car with that. You

know, you turn on your car, right? And

you hear like a weird noise and then you get to check engine light. It's like,

well, like what the hell? I got to take it to a mechanic or I got to learn how to fix it, right? It's why I own testies now is because I'm tired of oil changes.

So, enough about that. But this is the beginner's guide. And I would just say

beginner's guide. And I would just say look through it to know what am I doing here to work with data and then at what point can I separate what is needed for

a human to be in the loop or what is simply something where AI can do the heavy lifting for me and then provide me the upper level higher leverage tasks or

decisions that need to be done. And then

of course with generating data, generating you can write cold emails, we come up with copy, coming up with social media posts, videos, things like that.

That is something that obviously depending on your business, you can do a whole lot. But we're talking writing

whole lot. But we're talking writing cold emails, creating social media posts, designing images, producing videos, writing blog posts, things like that. So with that said, just look

that. So with that said, just look through this and then automation exploration here. I would say, you know,

exploration here. I would say, you know, there's cursor and then there's common other platforms that I personally like to use in this day and age right now, which is N810. Those are the core ones for us that work really well. There's a

lot of them out there. And then the other piece here is revolving around cursor and context engineering. Cursor

is great because you can actually provide a lot of context no matter what you're doing and it allows you to keep

this contained. If I were to go into

this contained. If I were to go into Claude and sure I could go to settings and you can create a basic role of what you want to do. There's always like a

little bit of a role to actually start define like who you are and things like that. But with cursor I can just create

that. But with cursor I can just create a folder and then I could define roles and provide context and guard rails around exactly what you're doing or what the agent should be doing. So there's a

doc cursor and then there's rules. Cloud

code has the same thing. They all have them. You're defining roles around what

them. You're defining roles around what the AI agent is allowed to do and what not to do and what tools to actually use for the job. These other, you know, Chad

GBT and claw and all these things, they have a very slight thing here. And so

best practices to be able to leverage these types of things. And so, uh, context engineering, I like to talk about the big three, which I believe are

frameworks, which is RTO. You define the role, you define the task, you define the parameters around the task, and then you define the output. Allow me to

explain. The role is the system prompt

explain. The role is the system prompt or the persona. So if I were to say in chat GPT, why is the sky blue? We don't

know what it's going to give me and it's going to probably go on and on and we have no idea. Okay, cool. All right,

very nice. Okay. So then if I said as a role, you are a kindergarten teacher and you are teaching a fifth grade class on

why the sky is blue. I need you to answer why the sky is blue. And so I defined the role as the kindergarten teacher. And you can see the output is

teacher. And you can see the output is clearly different. We're talking about

clearly different. We're talking about let's imagine a sky with a big ocean made of air and this and that and this and that. It's all sunshine and

and that. It's all sunshine and rainbows, right? You are a all- knowing

rainbows, right? You are a all- knowing college professor from Harvard University and you are about to present

to potential graduates an assignment on why the sky is blue. I need you to answer to them why you believe the sky

is blue and to explain in your role to make them believe the actual reason why the sky is blue. So you see how I defined the role in a very different

manner compared to kindergarten teacher and it is basically seeh it's going into the the weeds here or the sky but uh why not violet so what

does that mean in the real world well it could be you are a digital marketer you are a expert email copywriter professional sales development representative I'm going to challenge

you with this you only know what you know we only know what we know LLMs large language models We don't know what we don't know. These are the smartest artificial brains that have ever hit the

planet and it's only getting better, bigger, I don't know. It's only going to be what it is today until something else comes along. So, it makes you really

comes along. So, it makes you really wonder what is possible with prompting and knowing the words you use actually dictate the output that you get. So that

means every single word that I use, if I say you're an expert sales development representative or you're a mastermind, what does the AI actually think now?

Like hm I am expert or mastermind. Next

thing here is the task. So the task is what the AI models are going to do. And

the thing is is the LLM will do that task. It may not do it right. It might

task. It may not do it right. It might

do it right, but it will accomplish something for you. What I would say is the better you think the more specific

and articulate you are for models in your tasks, the better your output would be. If I were to say your task is to

be. If I were to say your task is to research the company and see what they do for a living. So the agent scrapes the website and says, "Yeah, I found this and I found that and rocket ships

and everything's cool and blah blah blah." Well, the thing is is you don't

blah." Well, the thing is is you don't actually know if that's true because it might hallucinate. It may have saw a

might hallucinate. It may have saw a word that is like, "Oh yeah, this is really good." And then again, it

really good." And then again, it actually might give you a bunch of fluff where you're like, "What are you thinking?" Like, "This is the worst

thinking?" Like, "This is the worst thing ever." Right? Well, if you define

thing ever." Right? Well, if you define the task and you articulate the steps in the task such as you need to go to this

website and step by step go through each and every page, including about us, I need you to extract something meaningful. And what I mean by

meaningful. And what I mean by meaningful is that a human actually sees something and could probably feel emotion by it. And then defining good

and bad examples. That is huge because you need to be able to give the eye model. This is how you color a piece of

model. This is how you color a piece of paper. This is how you don't color a

paper. This is how you don't color a piece of paper. Right? Marine Corps. I'm

Air Force so I can call this out. But my

buddies in Marine Corps be like, "Oh yeah, that's it." By the way, simplify if you're Marine. It's all good. And

step-by-step execution. break down the process, define the logical flow, specify decision points, right? Think

step by step. So what I would say is look at the websites, really try to look at exactly what the AI models or the like anthropic and open AAI what they

actually say, how to use the models the best because they trained it. So

therefore, they are their mommies and daddies and you should always go to the mommies and daddies when you're trying to learn about their children, aka their large language models. So that's that.

Moving on, the other and final point around context engineering that I want to call out is the output. The output is really a point where you want to think

strategically on what kind of output you want. You have yes and no. You have

want. You have yes and no. You have

qualification score. You have word limits, paragraphs, essay, like output JSON, output markdown. This is pretty much you just define the output and making sure that you're getting into the

weeds of that, right? Then you have why markdown's popular. So I would just go

markdown's popular. So I would just go into that. But beyond the basics, some

into that. But beyond the basics, some things that I talk about as well is the psychological aspect of what is a deep lexicon and what is surface lexicon. If

you think about it, Steve Jobs for example or Simon SK, if you ask them a question, there's actually a pause before they actually answer. there needs

to be someone who is sort of the um keeper and reiterator of the vision >> because they're using a deep lexicon.

And what that means is they're actually they're thinking deeper on words that they could carefully use in order to articulate what they're trying to say in a more meaningful and elegant manner

versus just saying fluff. And when using it with AI models, you're a lot more descriptive and you're a lot more articulate, which could actually improve your outputs from there. And so I would

encourage you to look into your deep lexicon and your surface lexicon. It's

basically deep thinking or shallow thinking in a nutshell. And then of course with context engineering, there is the other piece which is being able to pull in MCP servers. So I encourage

you to check out Enthropic. I won't bore you with a page, but go to the makers, learn how to really leverage exactly what they're talking about on why it's important to really be specific on the

particular task and to go into the details of how to actually execute that task and then define the output. Going

into the fun stuff, we have cursor.

Cursor has the ability to switch different large language models. So,

different LLMs have the ability to a lot of them have pros and cons to everything, right? So Claude 4.5 is

everything, right? So Claude 4.5 is currently my powerhouse when it comes to all this stuff. Although GPT5, we have Cheetah, which still is unknown. They

say it's Grof Fast or Gemini 3, you know, all the things, right? We have all these LLMs that we can hop and you could be on auto or you can pick your model

and go from there. So that's super powerful. And would you use Chad GBT or

powerful. And would you use Chad GBT or Claude or any of that stuff, you're in their house and you can't switch around, right? And then we have MCP servers. So,

right? And then we have MCP servers. So,

if you simply go to settings, you go to MCP servers. MCP servers are what I like

MCP servers. MCP servers are what I like to call toolboxes to a bunch of tools.

Without boring you on this graphic, this is just for pretty purposes. It could be I'm going to explain it metaphorically in a couple ways. They started off with

like a USBC port for AI applications.

USBC drives are everywhere. Well, USBC

ports, I should say. It is a universal peripheral that is standardized in order for me to plug in any device through one

port. That's pretty much what it is. MCP

port. That's pretty much what it is. MCP

allows AI models to use services like ClickUp, HubSpot, any external source that you want in order to be able to connect to them and interact with it.

Right? So, I have a number of MCP servers that are all connected to my agent here and I could do a number of different things. I have fireall for

different things. I have fireall for example. So I can have it scrape

example. So I can have it scrape websites. This right here is a

websites. This right here is a description of how to use this tool.

Another example is I have GitHub. GitHub

has 51 tools. So it can do a pull request. It could add an issue. It can

request. It could add an issue. It can

create an issue. It can do a whole lot of things here. We also have think, believe it or not, like think is an MCP and it and it works. Basically tells

you, hey, be quiet and pause. Don't do

anything. We also have browser MCP. So,

if I were to go go to google.com and type in this is using browser MCP and then smile. Make sure to use browser MCP

then smile. Make sure to use browser MCP for this. It's going to browse the

for this. It's going to browse the internet and it will actually use a mouse. So, you can see right here, it'll

mouse. So, you can see right here, it'll click in, it'll type in. And this makes it super easy for me to to automate things. You can see a mouse there and it

things. You can see a mouse there and it will click and it will type it in. And I

can do all sorts of things. The thing

that's cool about browser MCP is it will also look under the hood. It'll see

console logs. It'll know exactly what I need to do in order to troubleshoot.

Look, it has the screenshot and everything. So, it could take

everything. So, it could take screenshots. It can know exactly what I

screenshots. It can know exactly what I want to do. And this really just circles back the feedback loop in order to know like, hey, I'm trying to do this. Like,

let's move on. Right? So, there you go.

There's browser MCP. Uh, context 7 is really cool for keeping up toate documents. So sometimes like an LLM has

documents. So sometimes like an LLM has a cut off date. It's not going to know like oh newest Python just came out or something like that, right? Then we have really cool ones like X like I could

browse my X profile and generate posts and then post them. I also have Sora and Nano Banana. So what's cool about Sora

Nano Banana. So what's cool about Sora and this is again looping into Vive Marketing is Sora right here. I just

straight up created this video and then thing is is it it created it and then store it to my hard drive and >> engagement up 12% week over week. Nice.

Let's double down on that format. I'll

cue it in the calendar.

>> I could literally say using Sora MCP because I have tools right here. Seven

tools. Create video, remix video, get video stats, and it will literally save to my hard drive and then I could do whatever I want with it. The same goes for Nano Banana. Now Nano Banana, I just

straight up said create something and it just created Vive Marketing. like cool.

It spelled it right. It's all good. It

saves to my hard drive. And there it is.

And then here comes Reddit. So Reddit is one of my favorites because I could say using Reddit MCP tell me what the latest

in cold email is for trends. I have it on YOLO mode. So there we go. YOLO mode

is basically autopilot. And it will uh continue to do that, right? And then you can see right here 45% context window.

This shows how much of the context window I used. And the power here is once this gets to like 95 to 100%, it's actually going to compress it. It's

going to summarize the context window of my current and then in the same thread, I can actually continue it and really steer the AI direction. I'm not losing context. And if I want to restart it or

context. And if I want to restart it or whatever, I can just pull in additional files and it will be just fine. Right

from there, we have instantly, good old instantly. So, I created this MCP

instantly. So, I created this MCP server. It's due to release here any day

server. It's due to release here any day now. We are going through the final

now. We are going through the final stages of testing, but it works really well and I use it in cursor. For

example, I'm going to say like in instantly create a well, I might actually have context around a client of ours. So each and every client that we

ours. So each and every client that we serve, we have our own cursor instance and then I will push it to GitHub. So

that way myself and my employees will have the same synced repository around what we're working with. And then I could just simply, you know, scrape the web. I could pull in a CSV. And then

web. I could pull in a CSV. And then

this will also know the persona and everything around what that client actually does. And then I could say,

actually does. And then I could say, hey, you know, create a campaign. This

is kind of what I'm thinking and make it 100 words or less, etc., etc. And then from there, we will actually be able to create copy. And I say, cool. Send it to

create copy. And I say, cool. Send it to instantly create the campaign. And then

let's hit play. And it'll do all these actions. I do have more endpoints that

actions. I do have more endpoints that I'm baking into it now, but this is at the final stages for getting released here very shortly. Uh, it's definitely a powerhouse. It's a lot of fun. Now,

powerhouse. It's a lot of fun. Now,

going into more details here. So, we're

all set up in cursor. We set up the role. We talk about what it means to

role. We talk about what it means to have RTO, right? The next thing I would say that's really powerful is using agent skills or Claude skills, right?

This just came out October 16th as of this recording. It was just a few days

this recording. It was just a few days ago. and Claude now has skills to

ago. and Claude now has skills to improve how it performs specific tasks.

So there's a lot of different ways and this is also in Claude, but we can do a whole lot of stuff when being able to create skills which again this is the

nature of leverage and being able to do Vibe marketing or Vive selling in such a way where you don't necessarily have to train an employee, right? The biggest

thing here is being able to leverage this. And then what I would honestly do

this. And then what I would honestly do is just copy this URL and say, "Hey, let's create a skill for um digital marketer, right? And it's going to

marketer, right? And it's going to create what you call a skill.md file.

Let's say there's a specific color schema. Let's say there's very specific

schema. Let's say there's very specific branding in your company that you want to do. Well, you can actually say that

to do. Well, you can actually say that and it's going to create the skill for you very, very articulately. So, we have brand guidelines. you have canvas

brand guidelines. you have canvas design, skill creator, you know, all these things. So, you could generate a

these things. So, you could generate a skill that is a digital marketer or cold emailer, whatever that is. Make it a skill in this project and then you could

use MCP servers and all that to pull in and even create content like what I showed you with the image here to create something that is of that brand and then

therefore post it on X or any other platform that you wish to do. So this is the overarching context when it comes to that stuff, right? So I would say all that to say that is in my opinion the

best way to leverage tools when it comes to MCP the foundational elements of you know RTO which is RO task output really

ask yourself when it comes to the four ways that we migrate or use data which is obviously acquiring it consuming it creating it and then distributing it and

what are those four ways that we can apply AI and what are those four ways inside of those that we need to have a human in the loop because it's just too critical. There's a lot at stake here

critical. There's a lot at stake here when it comes to business. And then that way you could really start to really build a true intelligence here where this right here, I mean, I also have

augment, you know, and you can bring up the cursor, you can bring up claw code, you know, you can bring up Gemini. So,

we have so many things right here that I mean, I could use three agents right here and just start hacking away. And I

haven't even gone into the background agents here, right? So, background

agents are a whole different game. being

able to use them on your phone. It's uh

it's definitely a lot of fun. So, that's

the end of the guide. I truly hope you got something from this video in order to inspire you to go out and do great things because at the end of the day, solving problems and providing true value is what matters most. If you like

this video and you're interested in vibe marketing, click here to check out the next video. In that one, we actually do

next video. In that one, we actually do a challenge to start and sell a service-based business in just 10 hours and use a lot of Vibe marketing principles here to get the business started, get some understanding of the

customers, whip up some copy, and start selling. So, if you want to apply what

selling. So, if you want to apply what you just learned and see how it's done in real life, I'll see you over