What is happening everyone? It’s Luke McGuire here, and we are live with today’s session. A huge, huge, huge welcome, and I am so glad that everyone has made it into this series. This is going to be a lot of fun over the next three core days and then two bonus training days next week. We are going to be going from zero to a complete build of any software app—whether it’s a Chrome extension, setting up any type of automation for things that you do every week manually that we can now automate forever so that you never need to do those tasks again in your workspace, or even just at home in your personal life if there are things that you do all the time on social media, Facebook, or in your email clients. Whatever it is, we can automate everything, and we’re going to get into it pretty quickly today.
Now, this little intro here, I’ve had to record again outside of the welcome webinar because of GoToWebinar having some annoying issues when we started today’s session. So rather than have you watch a very frustrating 10 minutes while I had to change computers, change back, and tell a bunch of jokes, I thought I would do this little intro here to get started.
The first thing that we covered on the session today was the prize pool that we’re going to be having coming up shortly. We’re going to be doing prizes for this masterclass. All you need to do to be a part of it is submit your work when you finish with it. If you follow this every single day that we teach it and implement it on the same day, by the next session, you’ll have everything of this three-part series done by the end of the third webinar.
To give you an example, there was one automation that I share later in this session that I set up about two weeks ago. It was taking around 10 to 15 hours of my week up for me manually having to do the same repetitive tasks over and over and over again, where I didn’t think it was possible that I could have automated it or had tasks and work being done while I was out. Then, little did I realize, when I could do that, it saved me 10 to 15 hours a week, which adds up to over one month of the year (and that’s in 24-hour time). Say I slept 12 hours of the day—which I would love to be true, however it’s not—but just say I did, that means that would be two months of the year that I have now saved instead of doing this one incredibly monotonous, boring job task. So again, whether it’s a product, a service, an online offer, a software product, or private mentoring—you name it, you can create it with AI. We’re going to get from zero to your complete build in the coming week. Give me one moment here as I change this over.
All right, so before we jump right into the rest of the core of the webinar, I just wanted to go over what you first land on when you log in here. When you log in at the link agent72-toolbox.netlify.app (the link will be around this), you enter your email and the password, which is VIP1234! in capitals, I believe it is. I’ll put the password around this video here somewhere, but you can log in with that, and you’ve got the email and the password as well. Once you’re in here, it comes to this first breakdown where it shows you the first 30 minutes of what you need to do as soon as you get inside Agent 72. First being, install Claude for desktop. If I click on this right here, it’s going to bring me over to the desktop build. Hit download, and you are good to go.
The reason that we use Claude on the desktop is because Claude has all its features that it provides on the desktop app, which you will not get if you’re using the browser version on Google Chrome, Safari, or anything like that. You won’t get a lot of the features like Claude Code specifically. You won’t get any of the features where Claude can control files on your desktop, like go into your documents, into your different Claude builds, and access things on and around, and even use the different products and services that you have purchased in the past as actual tools. Claude can work with any of these, essentially as if it’s a personal assistant to you. This really ties in shortly when it comes to… actually, you know what? I’m just going to let you guys see it firsthand. So go ahead and install it for the desktop. Inside that desktop app, you have Claude Chat, Claude Copilot, and Claude Code, and from there, you are going to be good to go.
The next tool I suggest downloading here is a tool called Linear. It’s a free tool; download it. The reason behind it is it’s going to be a huge help when it comes to building out individual job tasks in a large project build. Again, you don’t build a single one of these tasks; it’s all done via AI. This will essentially mean if you have a big project that you need to build out as a software tool, so there are a hundred different mini-projects that need to be done and set up, instead of sending one agent to go and do it, you can send all 25 to knock out all of the key parts of the build in parallel in just a matter of minutes instead of hours or even days. That’s a really neat extra feature that’s inside of Linear, which you’ll see more of shortly.
Outside of that, download ChatGPT. If you don’t have it already, it’s worth having a copy, even if it’s just a free copy on your desktop. If you want to take things to the next level, there is a tool called Codex. Codex is essentially like ChatGPT, and it’s run by your ChatGPT subscription, but it allows you to create code at a much higher level than ChatGPT does itself.
So let’s jump back over to the main training. I just wanted to get that intro tied in nicely here. Stay to the end because we build a lot in this session. If you follow it exactly, you’re going to get some pretty insane results that you’ll be able to share inside of the VIP community (the link is below). We created it on this webinar; join that too, and I’ll speak to you soon.
Let’s get into the main chunk of it. You’ve got Claude Code installed, you’ve got Linear and ChatGPT downloaded. Where do we go from here? The first, most important thing you’re going to be doing after this session is building your business brain out. This is going to give your AI an unfair advantage compared to most people when it comes to anyone doing work on any AI tool out of the gate, because they never even think to do this. I’m going to copy this prompt, and I’m going to go ahead and pull up a Google Drive link. Let me see if I can share this window.
What I’m going to do is grab this copy-and-paste text here. I’m going to run through it really quick. This is going to be the first thing you post into Claude when you download it. If you’ve done this with me before, don’t worry, we’re going to be moving forward through this pretty quickly. But if you haven’t, this is the most important thing to do straight after this session. You’re going to write in this prompt: “You are a business strategist and you are going to interview me so you can help me build things this week. I’m a beginner. I may not have a business yet, just an idea.” That is totally fine. Go ahead and tweak this according to where you’re at. If you just want it to be complete bare-bones, have it set to that, and add rules for this interview: “Ask me one question at a time. Wait for my answer before the next one. Keep it friendly and plain. If my answers are vague, ask a follow-up.”
I’ve tried to make this as quick as possible. If you wanted to really take it to the next level, take this prompt and then open Claude. I’m doing this on the browser this time just so I don’t have to jump through different windows and tabs, but this is the first page you’ll see when you open Claude on your desktop. We’re going to go ahead and paste that in here, and before all of this, I’m going to say something like: “Hey Claude, I am about to do a week of building software, solutions, automations, websites, etc. They’ve asked me to share this prompt to get a basic idea of who I am and what my goals are. Can you please go and make this interview much more detailed? Find out things that you may want to know that have not been asked. Please even set as many questions as you like now, with the goal of coming back in a week or two to do a follow-up. I want you and I to be on the same page completely. Can you please write back to me the updated prompt you would give me?”
Even if it’s messy with spelling mistakes, don’t worry about it. I’m just going to hit enter and we’ll see it do its thing. Let me turn my webcam off as well to make sure it’s not lagging. While it does that, let me jump back over into our questions. Once you’ve done that, go to the next step. The next prompt is going to be setting up your workplace. Again, this is something that people just do not do. This is getting your file system and your file management set up to at least a basic organizational setup, because it’s so important to know where your files are when you’re building different things. As a developer, even though you won’t know any code, it’s so important that you have a setup where Claude can understand where things are at. It gives you a nice clean breakdown of the different things you’ll be building right out of the gate, which is going to save you so much time later down the track. When I first started, I didn’t even think that this was something I would have to do, so I just started building things and saving things randomly as files in different places, trying to drag things around. Then I’d need to start a new project and I couldn’t find where the stuff from the previous project was, and it was just a mess.
Another thing in Claude and in all AI tools: unless you set things up like this the first time, if I started a brand new chat tomorrow, it wouldn’t remember anything from this prior chat. In fact, it wouldn’t remember anything about me. It would wipe its memory every single day until I set things up accordingly so that it saves things and conversations can be remembered so that you can actually build things going forward. I’m not going to go into depth on that; I’m just going to show you how to ensure that doesn’t happen. That’s the theme with this training: it’s about getting things moving, not going through everything, but just explaining why we do it.
Next, we’re going to load the power-ups. These are little things like setting up certain rules. For example, a rule that before you finish a workday, you want to write into a specific file an update of where all your tasks are at, so tomorrow when you wake up, the AI can remember where you finished off and knows where you can start today. That is something that you can do here. We’ve also given it the ability to create a design rules file. This is looking at how to actually make things look good and premium, ensuring it understands that we’re not just building Excel-themed files where you might get an output of a comparison table that looks like a basic spreadsheet, but instead something that looks like a beautiful Photoshop presentation. This prompt is going to help you with that.
You can continue to go through them; they’re all in here for you to follow. These are just some of the basic ones to get you up and running. As we go through the week, I encourage you to go deeper on this as well. For those of you here that have been a part of my Agent Suite—which is my exact Claude setup where there are now 187 skills, 15 agents, and all of the integrations that I’ve included—you don’t even need to worry about this. I’ve done this for you plus a lot more. If you’ve got the Agent Suite, you are in the best position possible. If you don’t have it, don’t worry. This course is not about that. You can build something out without it, and you don’t need it for this to work.
Once you’ve done the interview and the AI understands who you are and what you sound like, here is where you can set these rules to sound like you. These are copy-and-paste prompts, but the whole thing I want you to learn throughout this course is that rather than just taking what I’ve said here, give it a go yourself. Often you’ll get a better outcome than just copying and pasting text. For example, instead of saying “make this email sound like me,” I wrote: “Rewrite the text to sound like me. My voice is like an experienced friend, not a guru. Lead with the biggest benefit, use short sentences, strategic caps on key phrases, no em-dashes, no hype, and no corporate filler.” This is what I sound like. You can say: “Hey Claude, I want you to sound more like me in your emails. What do you need from me to do that? Can I give you access to my Gmail so you can see how I write?” It will guide you step-by-step or even just do it for you.
We’ve got complete copy-and-paste builds, which are essentially simple products. You can build a lead magnet landing page in your niche just by entering details about your industry, and it will build it out accordingly. I’ve got a bunch of prompts in here, including ones that put money in the bank, how to get clients, how to find your next audience, and how to price your offer correctly. I’m going to add more of those throughout the week. Then we’ve got the replays, where your replays from this series will go, and we’ve got the skills and rules.
Skills and rules are things you write to teach Claude how to act differently in a way you like. For me, I like it to talk like a human, not like a robot. I like to always get three next-move solutions. If I finish having it build an image, I like when it gives me three additional things it can do from there to help me improve the image or use it for my marketing. It won’t do that unless I set a rule. Another massive one is “plan before you build.” If you don’t tell it to plan a build every single time before actually building something, it will just go straight into building it. This is the same with any AI, so make sure you say: “Hey, I want to build a project, but let’s put some initial steps in before we actually launch the build.” Does that make sense? Let me know in the chat box.
Now that we’ve got that idea of how this all works, let’s go into the tools section. Here are our key seven tools. I built these out as a concept. You probably saw the one-shot builder that I placed on the sales page, which was built with this system you are going to be learning throughout this week as well. Everything you see is what you are going to be learning, which is pretty cool. Even this backdrop here on this video was built out with AI.
This is called the One-Shot Engine. It comes as the second tool and is what gives us a prompt to build out an entire software build, automation build, app idea, Chrome extension, website, or opt-in page from scratch. Let’s do an example here today live. If you haven’t done the setup yet where the AI interviews you, wait until after to do your main build. You’ll have time to create probably three or four different things if you want to.
I wanted to jump over here first to show you the updated version of the prompt that it gave us. The prompt you gave me is for a total beginner with just an idea and no business. I’m going to use my text-to-speech tool here: “Hey there Claude, yeah, I know that this is about someone who hasn’t got a business yet or doesn’t really have an idea of what type of online business they want to create. That’s why I’ve sent you the long prompt from the members area. I’m currently on a live webinar showcasing how to use this, but I wanted to show them how they could customize this and beef it up to really go to that next level in explaining their goals, wanting you to learn as much about them as possible to get the best outcomes. If you can give me that copy-and-paste option, that’d be great. Can you do one example for someone with no idea what they want to do online, a second version for someone with an existing online business they want to grow, and a third example for someone looking at a side project they want to become their main income earner in the future? Feel free to do three versions or tie them all into one master prompt where the user can chat with you.”
Just word vomit into Claude—that’s the way to work here. It’s great because you can make mistakes, give it ideas, and not be perfect with your execution. It will understand and work accordingly. There’s no point trying to write in perfect academic English here; it’s about taking imperfect action. Let’s hit send and see what it says.
Over this week, Monday, Wednesday, and Friday is when we have these sessions, because there is a day in between that I want you guys to have the ability to work on things before we go to the next step. It’s only three core sessions, so please get the work done before the next session so you don’t have to play catch-up. That way, we can all finish together. Inside the community tab, what I’m going to be doing is getting you all to upload your finished product once you have completed your build. I’m going to be giving away cash prizes for this as well. The best incentive I’m going to set up is that by submitting your project, you will gain access to everyone else’s builds and can use their tools and creations yourself, just like they can use yours. Working together as a community is going to amplify the outcomes we get a hundredfold after this week. We have three sessions this week and two bonus sessions next week, but if you need extra sessions after that, I’m here for you.
Let me get you that advanced prompt over here. It came back with a mood-based update: “Instead of these separate prompts your audience has to choose between, I built one master prompt that asks them upfront which situation they’re in. Claude does the sorting. The real upgrade is in its instructions: it knows to dig and spot patterns in their answers.” This is perfect if you have trouble talking about yourself and figuring out what is unique about you or what value you bring to the table, especially if it’s a new area. I’m going to share this with you in the chat window now, and I’ll add it into our main file.
You would take this file, copy it, and over here in Claude, paste it directly into the chat, saying: “Please run this.” Then go. Another thing to mention since we’re getting started inside of Claude itself: there is a brand new feature in the model settings. You’ll see the option to set the effort level to max, and you can turn on thinking. However, when I asked Claude a simple question and set the thinking and effort to maximum, it took a long time to respond because it ran so many simulations trying to figure it out. It thought I was asking it something way harder than it actually was. That was a big mistake. I normally recommend turning that off; setting it to “high” is more than enough for most tasks. I’ll say: “I’m sorry, I had the effort set too high. Can you just tell me in short dot points in layman’s terms what the news articles said about Claude Mythos?”
See, now it’s way faster. It says here that Mythos is dangerous because it can find security holes in software and write code to break through them, chaining multiple exploits together on its own with very little human help. It scanned over a thousand open-source projects and found over 23,000 issues, 6,000 of them serious. This Mythos model is apparently going to be released to everyone soon, and people are saying the latest models are not too far off from what we’ll be able to do. It’s the most powerful model in the world, and it doesn’t make sense not to use it if you have the option.
If you are using Claude for the first time, you can go into your ChatGPT and say something like: “Hey ChatGPT, can you give me a full markdown file with every key bit of information you think would be useful to know about me, updated from my work interests, projects, etc.? Just respond with a generic response, not actual private information.” It’s going to do that, and you can take that text, paste it into Claude, and it will instantly know about you as well.
Whenever you are researching, you can ask for facts and say: “Hey Claude, when I’m doing research or asking about a specific event, if you can respond where you feel it’s fit depending on the question, please give me clean, easy-to-read dot points in layman’s terms so that it’s easily digestible. For example, if I ask for fun facts about the Tasmanian tiger, I don’t need a 10-page response. Just give me some dot points on when it was last seen, how rare it is, and maybe a funny piece of evidence of a recent sighting.” For those who don’t know, the Tasmanian tiger was a dog-like animal in Tasmania, which is an island at the bottom of Australia that broke off from the mainland about 10,000 to 12,000 years ago. Because of that, it built an untouched landscape with some of the rarest wildlife in the world. Even though the tiger is considered extinct, people still believe it exists in the wild and move there to find it.
Anyway, I’ll paste this prompt in here, and it will create a custom rule. It says: “Yes, I will match the depth of how invested you are in the conversation so you get clean dot points in plain English, scaling to how deep you actually want to go. No more 10-page essays when you just want the gist of something.” To make this preference stick, you can create a custom rule. I’ll share this copy-and-paste rule in the document for you guys. These are the kinds of rules that are included in the Agent Suite, which installs all of my pre-made rules and setups with one click to save you a massive amount of time.
One of the most important skills to add to your Claude setup is to make sure that Claude does the work for you when it can do the job itself. When it comes to the desktop app, you have Claude Chat, Claude Co-work, and Claude Code. Standard Claude Chat is a linear conversation back and forth. This is where I recommend starting every project to plan the build, get the basic overview, and work on the visual design. Claude Code is the actual coding side of things.
There has always been a disconnect between marketers and developers. If you want a two-story house built with a bathroom, a garage, and a swimming pool, you might picture the garage on the first level, the bathroom upstairs, and the pool in the backyard. But a developer might build the pool in the middle of the house, make the second story float off to the side with stairs going down to the toilet, and put a drive-up ramp to the garage on the roof. They both built the exact same list of requirements, but there is a massive disconnect in how it was executed. This is why developers hate working with other people’s code—it turns into “developer slop.” The beauty of having AI now is that it doesn’t code like this if we connect the two sides with a strong user interface and a clear plan. That’s why front-end and back-end development need to meet in the middle, and we do that by planning the user interface first.
Let’s look at an example. I asked Claude to build a mockup version of “Where’s Wally?” (or “Where’s Waldo,” as they call him in America). I wanted to create an app where I could add our top affiliates as characters in a scene, and they could find themselves in it. My concept was to gamify our JV page. The affiliate would log in, complete four or five tasks—like joining the Facebook group, adding the launch dates to their calendar, and grabbing their affiliate link—and then enter the hunt. If they found themselves, they’d get a 2% increase in their commissions. It was a fun way to get them engaging with the launch page.
Claude gave me prompts to generate a futuristic background in Google Gemini, and then I used a prompt to make a 3D cartoon character of myself using a Facebook photo. I gave that to Claude, and it segmented me from the background. Let’s see what it created: “You are on your summer in the launch. Find yourself.” You complete the tasks, enter the hunt, and there is a timer running. If you need a hint, it says: “Look for the giant person three times bigger than everyone else with no legs.” Sure enough, there I am. When you click on it, it pops up. It’s a silly piece of software, but creating something like that for a business really wows them. It shows them that if you can build something fun and custom like that, you can build anything.
Now, let’s look at editing photos. I had a photo of my partner and me in front of the Eiffel Tower, but a bus was blocking the view and there was scaffolding on the building. I can take this photo, paste it into ChatGPT, and say: “Hey, can you make this photo better by removing the bus and the people, and optimizing it for the ultimate Instagram picture?” ChatGPT’s image editing is incredibly powerful. We’ll let it run that and come back to see what it creates.
Back to Claude. In the top left of the desktop app, we have Chat, Co-work, and Code. Standard Chat is linear. Claude Code is where we do our building. You speak to the lead agent, it reads your input, makes a plan, and dispatches sub-agents to handle different files and tasks in parallel. You have the option to set the model to Haiku, Sonnet, or Opus. If you’re on a budget and want to get the most out of your token spend, Sonnet is a great place to start. It gives you about 90% of the capabilities of Opus at a much faster speed. Haiku is best for high-volume, very basic tasks, while Opus is the highest level for the best designs and cleanest code.
When Claude Code goes to do a job, it can launch a head marketing agent to scrape 100 competitor pages, spawning 100 individual sub-agents to do the scraping, feed that data back to an analytics agent, and pass it to an ad agent to create 50 different Facebook ad variations to split test. Once it’s all done, the information gets fed back to the head agent, and you get the completed results. You can have it schedule content, give you morning briefs, or monitor the web to inform you the second an event happens, like a new model dropping.
Let’s set up a morning routine. I’ll ask Claude: “Can you write me a quick prompt that will schedule a morning plan of attack? It should go through my Gmail, rank important emails at the top, draft responses in my voice, flag urgent messages, and provide a daily summary. It should look through my personal profile file in Claude to understand who I am.” This turns Claude Co-work into your personal chief of staff. All you need to do is copy this prompt, paste it into Co-work, and connect your Gmail.
I also want to add a custom rule: “Whenever you suggest something to me, or if I ask a question and you respond with steps, please always follow up with: ‘Would you like me to do this for you?’ if you have the ability to do that job. Often you’ll tell me how to do something manually that you could easily automate or do yourself. If you ever find that I’m struggling with a task, see if there’s an automated way to handle it in Claude Code or Co-work.” This is a massive value-add prompt that will save you weeks of manual work.
Now let’s get into our actual build. We’re going to use the Idea Engine. What’s my name? My name is Luke, but let’s do an example for my father-in-law, Chris. He runs a commercial fit-out business that builds, rebuilds, and retrofits Domino’s, Nando’s, Australia Post, and Cheesecake Shop stores across Australia and New Zealand. I want to help him save time and systemize his business. Currently, he is the only one in the company who knows how to quote jobs, manage builds, and price profit margins. One in five jobs blows out because there are too many moving parts managed by phone calls and trust—for example, if the flooring team isn’t finished on Tuesday, the cold room can’t be installed, putting the entire timeline back. I want to build a system that allows him to scale, free up his time, and turn his knowledge into an asset so the business can run without him. He has tried migrating invoicing tools before, but found it too hard and is resistant to change. He has a computer, a phone, and a lifetime of knowledge in his head.
I’m going to run this in Claude. It came back with three brilliant ideas:
- Blowout-Proof Quoting Tool: A web-based quoting calculator where a team member enters the project scope, and it spits out a fully costed quote with a built-in risk buffer and profit margin based on Chris’s historical knowledge.
- Job Coordination Board: A visual sequencing tool where Chris inputs a store type and gets a preloaded sequence of all trades and deliveries in the correct order, with dependency locks so no one can schedule the cold room until the flooring is signed off.
- Fitout OS: An interactive digital playbook where Chris can record quick voice notes or videos from a job site, and the system transcribes and structures them into step-by-step checklists for new site managers and subcontractors.
Let’s do another example for my sister, Georgia. She owns a boutique wedding dress shop and wants to transition to an online-only model. She specializes in custom, unique patterns like black wedding dresses, which she gets manufactured overseas to keep costs down. In the next 90 days, she wants to generate buzz and drive online sales via social media.
Claude suggested:
- Dress Personality Quiz: A short interactive quiz that matches a bride’s vibe and venue to one of Georgia’s custom styles.
- Black Dress Finder: A dedicated mini-web tool where brides input their body type and wedding theme to receive a curated mood board and instant price estimate.
- Virtual Fitting Room: An AI-assisted experience where brides upload a photo of themselves and get a generated visualization of what a custom dress would look like on them.
Let’s go with the Dress Personality Quiz. I’ll use the One-Shot Engine to generate the prompt. I’ll select the features we want, choose an elegant black-and-white theme, and hit “Generate my prompt.” Now, we’re going to take this prompt over to Claude Code, create a new project folder called “wedding-quiz-example,” paste it in, and let it run. It’s going to build out the entire functional prototype for us.
For more complex builds, we are going to use Linear to manage the tasks. If you download the Linear app, you can create a team for your project. Back in Claude Code, you can connect your workspace by typing: “Connect to my Linear team and break down this build.” Claude will automatically generate all the parent and child tasks—breaking them down into database structure, front-end design, and back-end logic—and organize them into clear implementation waves on your Linear board. As Claude builds, it will check off the tasks one by one. It is a seamless way to build software cleanly and efficiently.
I have also created a brand-new community for us on Go High Level. I’ll put the invite link in the chat and the Google document. Come join, introduce yourself, and share your build progress, screenshots, and links in the dedicated channel. I’ll be jumping in there to do live Q&As, share bonus content, and help you with your projects.
Your homework for tonight is to get your desktop environment set up, run through your business brain interview with Claude, connect your third-party tools like Gmail and Google Drive in the connectors settings, and start building your first simple tool. I can’t wait to see what you guys create. Get cracking on your builds, and I will see you all for our next core session on Wednesday!