Full transcript
0:00Believe it or not, you're not behind on
0:02AI, yet. Because the truth is, most
0:04people still don't understand how to use
0:07it well. By just watching this video,
0:09you'll know more about AI than 90% of
0:11the people out there. And if you make it
0:13all the way till the end, you'll be in
0:15the top 1% of all AI users. I've
0:17personally tested thousands of AI
0:19prompts across [music] every major
0:21platform, built dozen of custom AI tools
0:23for my own companies, and helped
0:25hundreds of founders turn AI into their
0:27most profitable team member. So, today,
0:29I'm going to show you how AI works, the
0:32fundamentals you need to use it, how to
0:34master it, and how to future-proof
0:36yourself so AI makes you a [music]
0:38weapon instead of replacing you.
0:39Starting with how it works. Essentially,
0:42you've got all the world's information.
0:44I'm talking videos, podcasts, documents,
0:47Reddit threads. And then AI analyze it,
0:49and then it did this mathematical
0:51equation. It broke all that text into
0:53[music] tokens, and then it tries to
0:55predict the most logical next token
0:57based on what you gave it. It's just
0:59like if you say, "Twinkle, twinkle,
1:01little," it's going to say, "Star."
1:03That's why most of the answers that come
1:05out of any chat AI have a lot of em
1:08dashes. It's not because AI likes em
1:10dashes, it's because humans like em
1:12dashes, and it's trying to replicate
1:14what it's seen us create. And this is
1:16why if you [music] give it really good
1:18information, the output is really good.
1:20Because it can use all that input to
1:23build context. [music]
1:24It's kind of like if somebody tells you
1:26all the details of a story and then ask
1:28you a specific [music]
1:28question, you're more likely going to
1:30get it right because you have all the
1:31details of the story. That's what AI
1:33needs to give you the best outputs, and
1:35that's where people fall short. The
1:37truth is, AI isn't magic. It's just
1:40pattern recognition with a fancy suit
1:42on. So, now that we've covered how AI
1:44works behind the scenes, we need to talk
1:46about the fundamentals and how you need
1:48to use it the right way. Starting with
1:50the right way to prompt. The more
1:51information you give it, aka the prompt,
1:54the better the output will look. Did you
1:56know you can give it like two, three
1:58books worth of information about your
2:00question before it even gives you an
2:02answer? Most people are giving it a text
2:05message. It's giving it nothing. And
2:07then you're expecting it to do magic. A
2:09few months ago I was working with a
2:10founder that built this really cool AI
2:12technology and he's working on the
2:13marketing. And he was having a really
2:15hard time getting the marketing to
2:17actually communicate what his product
2:19did. He spent like three weeks on it,
2:21iterating, going back and forth, trying
2:23to get the messaging right, trying to
2:25get the headline right, trying to get
2:26the copy right. All I did was sit down
2:28and I gave it the format I'm about to
2:30teach you and I hit enter. And within 12
2:33seconds it outputted the perfect plan,
2:36structure, and strategy for him to get
2:38his customers [music] to buy. It took
2:40less than 15 minutes to implement it on
2:41his website and all of a sudden he's
2:43converting that traffic into customers.
2:45What did I do different? Here's how it
2:47works. The first thing is the role. You
2:50got to tell it who to be like. It's kind
2:52of like act like a professional dancer,
2:54act like a killer musician. So all I did
2:57when I was helping my buddy to fix his
2:58prompt is said act like a world-class
3:01marketing strategist that focuses on
3:03conversion for software. What that does
3:06is it takes all the freaking information
3:09in the world that's all been cataloged
3:10and structured and it says, "Okay, out
3:12of all of this, only focus on this
3:14piece." And then with that it can then
3:17give you an answer that's very specific.
3:19If not, it just looks watered down. The
3:21second part to your prompt is the
3:22context. [music]
3:23I want you to give it all the background
3:25information. If you have marketing
3:28documents, you have transcripts from a
3:30call, you have any kind of standard
3:32specification, the more you can give it,
3:35it can use that in its little brain for
3:37that prompt to guide it in its output.
3:40The third is the command. You have to be
3:42very specific. Whatever you want it to
3:45do, tell it. Be very explicit. I always
3:48say make the implicit explicit when
3:50we're prompting. And the last the fourth
3:53is the format. Often times for me, I
3:55don't want these long, lengthy outputs.
3:58I want bulleted lists. I want it short
4:00and concise. I want it in a table
4:02format. I want it in a PDF. The format
4:05is how you take that output and then use
4:07it in other tools. You might want
4:08[music] to put it in a CSV so that you
4:10can import that into another tool, but
4:12if you don't tell it that, it's not
4:14going to know what format you want. If
4:16you do those things,
4:17>> [music]
4:17>> you will notice a complete difference in
4:20the output. I'll even go as nerdy in the
4:22format where I'll say, "Here's the
4:25format." Then give it a template [music]
4:27to follow, then hit enter, and it can't
4:30go outside of those parameters. The
4:32truth is is the quality of your output
4:34will never exceed the quality of the
4:36input to your [music] prompt. So now you
4:38know how to structure your prompt, but
4:40there's this other crucial step that if
4:42you don't know, you can't learn AI
4:44properly, which is you got to pick one
4:46tool and don't switch. The amount of
4:48people that come to me and they're like,
4:50"Hey, I'm using ChatGPT and Claude and
4:52Gemini and Grok." And I'm like, "Show me
4:54your prompts." And it's Crayola. It's
4:57simple. They haven't mastered the tool.
4:59Learning AI is like learning an
5:01instrument. If you tried to learn a
5:03little bit about the piano and the
5:04guitar and the drums all at once, you
5:05wouldn't be really good. But if you
5:07decided to pick one and you decided to
5:09be a master at it, the others become way
5:11easy after it because you understand
5:13music. You understand tone. You
5:15understand how to read sheet music. Same
5:17thing with AI tools. Pick one [music]
5:19and master it before you switch. Once
5:21you understand how one works, the others
5:23are way easier to pick up. And I will be
5:25honest, the people that keep switching
5:27tools all the time, I just call it
5:29procrastination. Masters go deep before
5:32they go wide. So know the tools, know
5:34which one's right for you, and go deep
5:35on it. So I've already mentioned the top
5:37three. Okay, we've got Claude, Gemini,
5:39and ChatGPT. I'm going to walk you
5:41through each one and tell you exactly
5:43which one does what best so that you can
5:45pick the right one for your current use
5:47case. So, for example, if you do a lot
5:49of writing, creative work, deep
5:51thinking, code, well, Claude's got
5:54everybody beat there. And it could
5:55change in a few months, but for right
5:57now, I'm telling you Claude's the one.
5:58The other one, if you do a lot of
5:59research, you need up-to-date
6:01information, you live inside of Google
6:03tools, you can't go wrong with Gemini.
6:05It is an incredible language model, and
6:07it's the one a lot of people that are
6:09like Google specific are using. GPT is a
6:11great one. It's the first one to get
6:13really popular, it's got a lot of
6:14integrations, can't go wrong with it.
6:16And before you start hammering me in the
6:18comments, I understand there's a bunch
6:20of others. I live my life on a site
6:22called Hugging Face, which is like the
6:23nerdiest AI community. So, don't tell me
6:27about Grok and all these Chinese models.
6:29I get that. I'm just telling it for most
6:31people that are like, "I'm ready to be
6:33[music] masterful at this so I can
6:34become a weapon." These are the three I
6:36would recommend. And by the way, if you
6:38choose to go towards ChatGPT, I've got a
6:40full freaking masterclass that I'll link
6:42at the end of this video that goes
6:44through everything to make you a pro.
6:46And if you're a business owner that's
6:48trying to break seven figures and you
6:49want to go even deeper, I've got an AI
6:51implementation playbook that is what I
6:53use in my companies to implement AI, and
6:56it shows you everything. Just find me on
6:58Instagram and DM me AI business, I'll
7:01send it over to you for free. So, here's
7:03the thing. If you made it to this point
7:05and you [music] feel like you've got
7:06everything covered so far, the cool
7:08part, you're in the top 10% of AI users.
7:11But, if you stick around and implement
7:13what I'm about to teach you, you'll be
7:14in the top 1%. So, let's talk about
7:17mastering AI starting with pull versus
7:19push prompting. This is where we get a
7:21little advanced, but it's fun. Come with
7:24me for the ride. Don't run away. This is
7:26going to be good. Most people use AI the
7:29hard way. They do 80% of the work, and
7:32they let AI finish the last 20%. That's
7:34push prompting. You're telling AI how to
7:37do the task. So, you want to go from
7:39push to pull prompting. And the way that
7:42it's different is that you start with
7:43the outcome and then you let AI figure
7:46out the best way to get there. Instead
7:47of doing all the work yourself, which
7:49you're doing today,
7:50>> [music]
7:50>> you let AI do the heavy lifting. The way
7:52I like to think about it is push is kind
7:55of like when you used to get directions
7:56from a friend when they're like, "Come
7:58to my house. Come over." And then they
7:59tell you you got to turn left at the
8:01house and then right at the church and
8:02then you'll see a red mailbox. That's
8:05how most people prompt. Pull is like GPS
8:08where you let the AI direct you based on
8:11the destination you've given it. So,
8:12here's how you use pull prompting. It's
8:14super cool. First, you give it the role
8:16and the context like you normally would.
8:17It needs to know kind of who it should
8:19act like and what's the information
8:21about the scenario. Second is you give
8:22it the outcome-based objective, right?
8:24This is where you tell it what you need.
8:26So, you might say, "I need a sequence
8:27that converts cold leads into booked
8:29calls, right?" Very specific. It's like
8:31the destination. Third, you say this,
8:34"Ask me all the questions you need for
8:37you to create this for me and then give
8:39it back to me in you can say the format.
8:41You might say PDF or a table. Then just
8:43[music] answer all the questions. My
8:45little trick there is I use the voice
8:46mode where I can just hit that little
8:48microphone and then answer cuz they
8:49might give you seven questions and I
8:51don't want to type that much. But you
8:52give it all the answers and then let AI
8:54do the work. Watch it cook. Now, you
8:57might say, "Hey, I need to refine this.
8:59Ask me more questions to make it more
9:00relevant." And it'll keep refining. So,
9:02you keep essentially using, that's why
9:04it's called the pull method, you keep
9:06having it pull from you exactly what you
9:08need and let it go find the solutions.
9:11This is how software programmers have
9:12been doing it for a while, okay? Most
9:14coders are sitting there and we're
9:16coding through outcomes. It's called
9:17spec-driven development. That's now
9:19becoming the mainstream for how you can
9:22prompt. Now, pull prompting is crazy
9:24powerful. But if you really want to
9:26master AI, you need to make it personal
9:28to you. So, number five, master prompts
9:31for every role in your life. The way to
9:34think about a master prompt is kind of
9:36like a manual about a part of your life.
9:39So, if there was like a manual called
9:40Dan CEO, what would be in that manual?
9:43If there was one called Dan father, what
9:45would be in that manual? Well, I would
9:46assume in they would have everything
9:47about me, my life, the people in my
9:49life, how things are structured, what we
9:51sell. When you have these master prompts
9:54and you use it to give it to the AI cuz
9:56you might switch to different ones, then
9:58it'll give you better answers cuz right
10:01now if you don't, it's not personalized
10:03to you. It's just giving you generic
10:05auto completion instead of something
10:07that is laser focused to your exact
10:09situation. And remember earlier when I
10:11talked about giving it context, this is
10:13the context. And if you do it once, you
10:16never have to do it again and it makes
10:18prompting so much easier and faster.
10:20Every team member in all my companies
10:22has to have a master prompt. They have
10:24to show it to me. I got to make sure
10:26it's complete. I just can't imagine that
10:28they're efficient chatting with AI all
10:30day long and it giving great answers if
10:33it doesn't have that. So, now let's
10:34create one for yourself. The first thing
10:36is we're going to practice pull
10:37prompting. You can say, "I want to
10:39create a master prompt for my role as
10:42whatever your role is in life, you know,
10:43father, husband, VP of marketing and ask
10:46me all the questions you need to create
10:48it for me." See how we're going to have
10:50it pull from me? Now, it's going to give
10:51you all these questions so they can
10:53create this document and that's where
10:55you use the voice-to-text feature
10:56[music] and just answer each one
10:58conversationally. And third, once AI
11:01gives you your master prompt, go through
11:03it and just refine anything that doesn't
11:05feel right. And you can literally just
11:06talk to it and say like, "Update this.
11:08Add this. You're missing that. You
11:09didn't understand this properly." And
11:11it'll literally update the whole thing
11:13for you. The last part is to save it as
11:15a PDF and I like to just call the master
11:17prompt and then the role name so I have
11:18them on my hard drive and that way when
11:20I'm using different AI tools, I can
11:23always upload it so that it has perfect
11:25context on my situation in life. The
11:27truth is if AI sounds like a stranger
11:29and anybody you have got that output,
11:31it's because you haven't introduced
11:33yourself. The master prompt is how you
11:34say, "Hello AI, this is who I am." The
11:37cool part is like I said, you can use it
11:39in other tools. I personally spend a lot
11:41of time in a tool called Manifold, which
11:43lets you create a genetic workflows.
11:45[music]
11:46And when I first loaded up the tool, it
11:48didn't know anything about me. So, I use
11:49my master prompt to give it all the
11:51context so that it could do outputs that
11:53were very personalized. If I had to like
11:55play with it and tweak it and talk to
11:57it, it would have just took me a long
11:58time. This gets an AI from stranger to
12:01best friend. So, now you get the master
12:03prompts. You have a way to give the
12:04context to the AI so that you guys are
12:06best friends. But, what if you're like
12:07trying to get an output and you like the
12:10output and you want to like repeat it
12:12and keep it consistent? This is where
12:14you need the next tool to master AI.
12:16Number six, system prompts. Think of a
12:19system prompt like a recipe and it has
12:22all the structure and you have the
12:23ingredients, but the sequencing matters.
12:25How you do what you do in the order will
12:27dictate if you have a big pile of mush
12:29or you have a beautiful chocolate cake.
12:30So, if you're chatting with the AI and
12:32eventually it gets an output that you're
12:34like, "Oh my gosh, this is great. Why
12:36did it take me 17 minutes to get to that
12:38output?" This is how you solve that
12:40problem. Understanding how to create
12:42system prompts is the future. [music]
12:44So, here's how you build a system prompt
12:45using what we've learned so far. First,
12:47you got to start by saying,
12:48>> [music]
12:49>> "You're an expert AI engineer. I want
12:51you to create a system prompt that does
12:54X. Ask me all the questions you need to
12:56create." See how I'm using the pull
12:57structure? Next, once it gives you all
12:59the questions, answer them using the
13:01voice-to-text feature, then test the
13:04response you get back. The third, if
13:06something needs to be tweaked, just tell
13:08it to refine it. It's literally you code
13:10system prompts with words. Once it's
13:12dialed in, then you copy that system
13:14prompt into a custom GPT or a cloud
13:17project or a Gemini gem. And then you've
13:19got a system that works for you on
13:22repeat without ever having to give it
13:23instructions [music]
13:24ever again. I use this for strategic
13:27work, I use it for research work, I use
13:30it to create systems for my team to
13:32follow where I can give them the link
13:33and say, "Look, just give it the topic
13:35and it'll do something for us." That way
13:37you take the complexity of AI and just
13:39make it really simple for everybody else
13:41to use. You're basically building little
13:43machines that take care of work for you
13:45so that you can focus on the fun new
13:46stuff. Now, here's a pro tip. A while
13:49ago, there was a leak. All the complex
13:52system prompts from the top AI tools,
13:54I'm talking Perplexity, Notion, Lovable,
13:56and 20 plus others were actually
13:58collected and put together. So, if you
14:00want to go see what the world's best
14:03system prompts could look like, just
14:05search system prompts and models of AI
14:07tools on GitHub on Google and you will
14:09find the link and you can go look at
14:11each one of them to understand how
14:13powerful these can be created to do
14:16literally anything for you, like build
14:19mobile apps or reconcile financial
14:21details. So, now you know how to use AI
14:24better than most people ever will.
14:26That's the truth. But, here's the thing.
14:28AI is evolving fast. I'm talking every
14:31day. And if you want to stay ahead, you
14:32need to make sure AI will never replace
14:35you. Which takes us to our last point,
14:37but the most important one, which is
14:39future-proofing yourself. Which comes
14:41down to taste, vision, and care. There
14:44are three things that when I sat back
14:46and I asked myself like, "What are the
14:47areas of human existence, of life, that
14:49I feel is going to be really hard to
14:51disrupt?" I feel like things like our
14:53taste for what is great and our vision
14:56for things that don't exist that should,
14:59or caring about other people. What a
15:01crazy idea. I really don't think that AI
15:02is going to be able to do that well.
15:04Simulate it, maybe, but I'm talking
15:06heartfelt, empathetic conversations.
15:08That takes you. Machines can optimize
15:11what already exists, but humans imagine
15:14what doesn't. First thing is taste. You
15:16got to immerse yourself in excellence.
15:17[music] I think of it like standing
15:19underneath a waterfall and allowing
15:21amazing, world-class things to be poured
15:24over me so that I might have the
15:25possibility of picking stuff up. So, if
15:28you're like massively interested in
15:30music, go subscribe and use your social
15:33media to learn music. If it's code,
15:35software, I want you to use your feed to
15:38feed your mind. Go and follow all the
15:40top people. Go subscribe to the
15:41newsletters. The more you consume the
15:43best in the world and you get to
15:45understand what greatness looks like,
15:47the better you build your taste. The
15:48second is vision. Schedule thinking
15:51blocks so that you can sit down and
15:52think about the future. Visualize
15:54concepts cuz your brain doesn't think in
15:56words. I don't know if you know this,
15:57but like if I say pink elephant, you
15:59just saw it. Okay? You didn't have the
16:00words written down in your mind. So,
16:02even just doodling, planning, exploring,
16:05reading books that'll expand your
16:07horizon, that gives you a sense of
16:09what's possible or maybe how the world
16:11should be shaped around you. And that is
16:13your job to shape the AI to help you
16:15with the vision. And the last one is
16:16care. And this is why I'm so bullish
16:19around AI doing all the boring and dirty
16:22and dangerous work for us [music] is
16:23then you get to celebrate your clients.
16:25You get to celebrate your team. You get
16:27to be with your family. You get to use
16:29it to create more cuz you were born to
16:32be a creator. Like that's what you're
16:33here to do. Then you have the time to
16:35collaborate with people authentically.
16:37Start thinking about ways you can get
16:39out in the world and show that you care
16:42and contribute and just connect with
16:44people. I really think that AI has
16:46enabled us to have this deep desire to
16:48be around other people. you see events
16:50having record number of people showing
16:52up and sold out in no time cuz we have a
16:54deep need for connection. Do these three
16:57things and AI will become your weapon
16:59instead of your replacement. Now,
17:01believe it or not, you're still early.
17:03So, stop panicking if that's where
17:05you're at. Most small businesses haven't
17:06even started using AI in a meaningful
17:09way. So, just by watching this video,
17:11you're already ahead of the game. So,
17:13here's what I need you to do. Just pick
17:15one daily task [music] and have AI do
17:17it. I don't care if it's move
17:19information from one system to another.
17:21Just 10 minutes [music] of action beats
17:2310 hours of tutorials. So, just leave me
17:25a comment and leave that commitment
17:27below. I want to know, what are you
17:28going to commit to to really upgrade
17:30your AI skills? And remember,
17:32>> [music]
17:32>> DM me AI business on Instagram if you're
17:35looking to grow your business and I'll
17:36send you my AI implementation playbook
17:38so you can scale [music] past seven
17:39figures. Like I mentioned, if you want
17:41the full ChatGPT masterclass to get
17:44ahead of 99% of the people, just click
17:46here and I'll see you on the other side.