Set Up Your Machine
EditedBy the end of this lesson your machine runs the full working setup: Claude Code as your AI, Cursor as your workspace, the right plan, the settings that matter, and the one habit that means you can never get stuck. Assume nothing is installed - we start from a bare computer and do every click together.
0. What you need before anything
- A computer (Mac or Windows) with Chrome or any browser, signed in.
- About 30 minutes.
- A card for the two subscriptions this stack runs on: the Claude plan (below) and Wispr Flow (~$15/month, later in this page). This is the cost of your one employee.
1. Create your Claude account and pick the plan
Go to <https://claude.ai>, sign up, then upgrade. Plans, honestly:
- Max (about $100/month) - the course plan. This course has the AI reading eighty-page transcripts and building real systems. The $20 Pro plan technically works but you will hit its usage limits fast and feel throttled right when you are in flow.
- I run the $200/month Max plan because AI runs my entire company. You will likely end up there once client work starts paying.
2. Download Cursor
Cursor is the free editor we work in - the file tree shows the whole business, the terminal is built in, and every file previews cleanly.
Download it, install it, open it once. Skip the onboarding - nothing in it matters for us.
3. Download and extract the course folder
Download the course ZIP from your course link. Extract it and put the relay-business-brain-v1 folder somewhere findable - Documents/Relay. The folder you extract is the folder you open. Not the ZIP, not Downloads.
4. Open the folder in Cursor and look around
File > Open Folder, select relay-business-brain-v1. When Cursor asks about trust, read the path, confirm it is the course folder, accept.
Thirty seconds of orientation, because this window is where the whole course happens: the file tree on the left is the business. Click any .md file and use the Preview toggle to read it like a document. View > Terminal opens the terminal panel at the bottom - already inside your folder.
5. Install Claude Code - the way that always works
Here is the honest truth about installs: every machine is different, and a fixed list of commands works until it does not. So we install the way professionals actually work - you have an AI walk you through it.
Open <https://claude.ai> in your browser, start a normal chat (no project needed), and paste:
Help me install Claude Code on my computer, step by step. I am on
[Mac / Windows]. I will run everything in Cursor's built-in terminal.
Give me one command at a time, and I will paste back exactly what the
terminal says after each step.Follow it. Paste each command into Cursor's terminal, paste the output back into the chat - including anything red - and keep going until this works in a NEW terminal panel:
claude --versionA version number means you are done. For reference, the official one-line installers it will usually give you:
Mac, Linux, WSL:
curl -fsSL https://claude.ai/install.sh | bashWindows PowerShell:
irm https://claude.ai/install.ps1 | iexRemember the loop you just used: copy the red text, paste it to the AI, do what it says, paste back the result. That loop is how you fix everything for the rest of this course and the rest of your career. You are never stuck again.
6. Sign in and run the first safe prompt
In Cursor's terminal:
claudeThe first run opens your browser - sign in with your Claude account. Then paste:
Read START-HERE.md and CLAUDE.md and nothing else yet. Do not edit,
create, delete, or send anything. Confirm the folder you are working in,
state that every supplied company and contact is fictional practice
data, and repeat the five standing rules from CLAUDE.md in one line
each.A correct answer names the folder, says the data is fictional, lists the five rules, changes nothing.
7. How we actually run it
Close that session (type exit or press Ctrl+C) and from now on start Claude Code like this:
claude --dangerously-skip-permissionsThat flag turns off the are-you-sure prompts so the AI just works. The name is dramatic on purpose: it means YOU own the outcome. Everything in this course folder is fictional, so run it this way for the whole course, like I do. If you ever want the guardrails back for a specific session - a real client's folder before you trust your setup - press Shift+Tab inside a session to switch modes, or just start claude plain.
The four commands worth knowing, all typed inside a session:
/model- which brain you are using. Default model for everyday work, the biggest model when the judgment is hard and it matters, a fast model for simple lookups. The big model for everything just burns your plan./effort- how hard it thinks. Lower for simple tasks./usage- where your plan limits stand. Check before heavy lessons./config- settings. In the video I mirror my exact setup - copy mine.
8. The speed stack
Wispr Flow (~$15/month, Mac and Windows) - hold a hotkey, talk, it types. I do not type prompts; I talk them. Typing is the fallback, not the method. Everything else in my stack arrives in later lessons when the system that needs it arrives - do not install tools ahead of need.
Next
Lesson one: four call transcripts and a parking-lot voice memo become a Business Brain.
Full transcript
[00:00:16] So before we get building with anything, what we want to be doing is
[00:00:18] ensuring that our machine has the right working setup.
[00:00:21] What we'll be needing for, by the end of this video,
[00:00:23] is a cloud code subscription for the AI,
[00:00:27] cursor as a workspace, the right plan,
[00:00:29] and the settings that matter. I'm going to walk through each one of these
[00:00:32] one by one, including a one habit that means that we can never actually
[00:00:37] get stuck, even when the computer of yours is acting differently from mine.
[00:00:42] So what I have over here is a completely bare computer,
[00:00:45] a brand new one with nothing logged in, I'm literally on it.
[00:00:48] on a fresh account right now with nothing installed and I do every click
[00:00:51] along with you.
[00:01:01] So before we get started, what I need you guys to have is a
[00:01:05] computer, a Mac or Windows either suffices along with Chrome,
[00:01:09] Safari, or any browser where you can log in to your different tools.
[00:01:12] The setup roughly is going to take about 30 to 40 minutes
[00:01:16] top and I would like you to have your card ready as well for
[00:01:20] the subscriptions that we're going to be needing. Most importantly being Claude and I'm
[00:01:24] going to walk over which plan for the Claude subscription we want in order
[00:01:27] to run our business. So step number one is we want to sign up
[00:01:31] for Claude. So, if we come over here. Go to the pricing page over
[00:01:34] here. We have a couple of options. First of all,
[00:01:37] I want you guys to never sign up for anything in Teams or Enterprise.
[00:01:41] This is not related to us and I myself do not have a single
[00:01:44] Claude subscription related to this. Now on the individual plan,
[00:01:48] we have the free, pro and the $100 plan.
[00:01:51] So let me be straight about the plans. I 100% recommend getting the $100
[00:01:55] a month plan. This course,
[00:01:58] as we're going to see later, has the AI reading 80 pages of transcripts,
[00:02:02] thousands of documents, and building real systems.
[00:02:06] So you can sign up for the $17 a month plan,
[00:02:09] which actually is $20 if you pay for it monthly.
[00:02:13] But you're going to hit its limits really quickly when you're right in the
[00:02:16] flow. And you're going to be forced,
[00:02:18] more or less, to upgrade to the $100 a month plan.
[00:02:22] I, myself, guys, run two $200 plans because my entire
[00:02:26] business and company runs on AI. So stop thinking about the subscription.
[00:02:30] What you're paying over here is for not just one employee,
[00:02:34] but multiple employees. And each of them,
[00:02:37] if not all of them, collectively, just cost $100 a month.
[00:02:41] So click over here. On Tricloud,
[00:02:44] continue with your email address and make sure to pay for the $100 a
[00:02:47] month plan. I already, on my cloud, am logged into it,
[00:02:50] as we can see over here. And the next step after that is signing
[00:02:54] up for Cursor. So Cursor is essentially a free
[00:02:59] editor that we're going to be working in.
[00:03:01] It's, as I download it and as I start using it,
[00:03:04] you will understand it better visually. But this allows us to really run the
[00:03:08] whole business. The terminal is where we're going to be interacting with AI,
[00:03:13] cloud specifically, and the file previews.
[00:03:16] So what we want to be doing is, come over to the link here
[00:03:18] and click on Download for Mac OS.
[00:03:21] Let's give it a second. Once it's downloaded,
[00:03:24] let's give it an open. If you're on a Mac,
[00:03:27] you can drag it into your applications. I already have it installed,
[00:03:29] so I'm going to just stop it and let me give it
[00:03:38] an open. So we're first greeted by a login page.
[00:03:42] What we want to do, guys, is create an account,
[00:03:44] if you haven't already, and simply log into it.
[00:03:48] I'm going to be continuing with Google, and that's what I recommend you guys
[00:03:51] to do. And I'm going to skip this section and resume once I'm logged
[00:03:55] in. And this is really important.
[00:03:58] Now that we're greeted by a cursor over here,
[00:04:01] what we want to be doing, guys, is completely ignore this entire interface.
[00:04:06] What we're going to be doing is go into IDE. We click into this,
[00:04:10] and this is a different pop-up window. Fundamentally,
[00:04:14] click on skip. Forget about the onboarding.
[00:04:16] Do not click on any of these options. Just skip it entirely.
[00:04:19] And now, this is the exact window that I want you guys to always
[00:04:24] remember where we're going to be using our AI and our systems.
[00:04:27] Not this one over here. Anytime you see new agent,
[00:04:30] search, automations, this is the wrong window.
[00:04:34] We are not using Cloud.
[00:04:36] What we're going to be using is Cloud locally installed in our own system
[00:04:41] over here. So once again, let me just walk through it once again.
[00:04:44] We click on cursor and naturally we're always going to be greeted by this
[00:04:48] section over here. What you have to do is go into IDE.
[00:04:51] And this is the place we want to be sitting at. So let's give
[00:04:54] this a close and this is where we're sitting now.
[00:04:57] Now that we've installed cursor,
[00:04:59] what we want to be doing is downloading and extracting the course folder.
[00:05:03] So there's going to be a Google document or a zip file at the
[00:05:07] bottom of this material where you can cut
[00:05:11] that. So what you, so,
[00:05:14] cut that. So now that cursor has been installed,
[00:05:17] what we want to be doing next is downloading and extracting the course folder.
[00:05:20] In this course, you're going to be seeing a zip file available for us
[00:05:24] to download and I have already downloaded myself and that is what we're going
[00:05:27] to be opening. So as we can see, we extracted and we put this
[00:05:31] folder somewhere findable. It could be on your desktop,
[00:05:33] on your documents, on your downloads. Just make sure you remember where the folder
[00:05:37] sits and you have access to it. So let's give this a run now.
[00:05:41] So what I have it, over here is in my downloads,
[00:05:43] I have the zip file. Let me give it a click where this is
[00:05:46] the relay business brain we want.
[00:05:48] What I'm going to be doing is I'll keep it in my desktop.
[00:05:51] And now the next step is open folder.
[00:05:55] Let me just simplify this a little bit for us.
[00:05:59] Well, let's navigate to desktop, click on the relay business brain and click on
[00:06:03] enter. Boom.
[00:06:05] There we go, guys. As we can see here,
[00:06:07] we have our business brain as clear folders and structures.
[00:06:11] We can open the template. You can see the calls,
[00:06:13] clients, the core business information,
[00:06:16] the data, the outputs and the agent and Cloud.MD files.
[00:06:19] These things have already been set up and we will be going over them
[00:06:23] later on in the next modules. Right now, the purpose of this lesson is
[00:06:27] coming back to open install Cloud Code the way that always works.
[00:06:31] So next, what we want to be doing is completely remove this section over
[00:06:35] here. New agent. Cursor will always keep recommending us to use their
[00:06:40] inbuilt agents and we do not want to be using their inbuilt agents.
[00:06:43] What we want to be doing is look over here. Toggle panel.
[00:06:46] This thing. Command J if you're on a Mac. You click on this and
[00:06:49] this is our terminal. This is where we're going to be opening.
[00:06:52] We're going to be agents where you open a new terminal.
[00:06:55] You can give it a name. Let's call this my marketing agent.
[00:06:58] So this is working on some marketing work for me and this is how
[00:07:01] we're going to be navigating multi agents throughout the system.
[00:07:05] But first, what we need to be doing is we need to install Cloud
[00:07:07] and let's give it an enter and boom. This is what most of you
[00:07:11] are likely going to be facing and seeing when you type Cloud because simply
[00:07:14] said, guys, We have an Instagram. We haven't Cloud yet.
[00:07:17] We have the Cloud consumer app, but what we need is to install Cloud
[00:07:21] locally on the computer. So let's walk through this one by one now.
[00:07:25] So number one,
[00:07:27] here is the honest truth about installs guys.
[00:07:30] Every machine is different. You might have a Mac.
[00:07:33] You might have a 2025 Mac, a 2023 Mac.
[00:07:36] You might have a windows 2021 or 2026 windows and a fixed list
[00:07:40] work of commands works until it doesn't. So the way we want to be
[00:07:44] installing Cloud is actually using Cloud itself.
[00:07:47] And I'm going to show you how exactly I would do it because this
[00:07:50] is actually a pretty technical setup by common conventions.
[00:07:54] And what we need to be doing here is really come into this,
[00:07:58] open your Cloud model over here, your local chat,
[00:08:00] nothing special. We just come, we open a brand new chat.
[00:08:03] I'll simply put the model on Sonnet because I just don't want to get
[00:08:07] it done quickly. I will turn off thinking mode as well so we can
[00:08:10] just move quicker. And what I want, what I want us to do now
[00:08:15] is back and forth. We're going to copy things one by one.
[00:08:17] So as we can see, start a normal chat and copy this over here.
[00:08:22] So I'm simply going to copy this. Help me install Cloud Code on my
[00:08:24] computer. I'm on a Mac, so I'm going to say I'm on a Mac.
[00:08:27] If you're on a Windows, please make sure it says you're on a Windows.
[00:08:31] I will run everything in Cursor's built-in terminal.
[00:08:33] Give me one command at a time, and I will paste back exactly what
[00:08:36] the terminal says. Let's go.
[00:08:39] And this is exactly how we want to be using AI,
[00:08:41] guys. For my business, while this is working,
[00:08:44] for the longest of times, when I was growing my LinkedIn account from scratch
[00:08:47] in 2024, I have never done copywriting.
[00:08:50] How did I pick it up? How did I learn it?
[00:08:53] The way I picked it up, guys, is I would always use AI to
[00:08:55] help me debug things, reverse-engineer things.
[00:08:58] So in LinkedIn's case, I was using it to reverse-engineer.
[00:09:01] Why are these people in my industry,
[00:09:04] in AI, why are they getting so much engagement on their posts?
[00:09:06] What am I doing wrong? And the same thing goes here.
[00:09:09] We're not reverse-engineering, but we're asking Claude to help us over here.
[00:09:12] And all we do is just copy things now, one by one.
[00:09:14] So let's set up our workflow.
[00:09:16] Let me minimize this,
[00:09:18] and let me open Cursor. And this is exactly how I would do it.
[00:09:22] So I come back, I copy this command,
[00:09:26] and I paste it. Now let's give it a second.
[00:09:29] As you can see,
[00:09:47] And in case, let's say guys, I cut that. And in case, let's say
[00:09:48] you're getting some friction, the AI is, it's annoying,
[00:09:51] it's confusing you. So what you can do is,
[00:09:53] you can simply open the entire terminal like this,
[00:09:58] screenshot it, and just dump it back into AI.
[00:10:01] And tell me, help me out.
[00:10:04] Just say, I'm confused.
[00:10:08] You can even misspell it like I did. And you can see,
[00:10:12] in my case, it installed it in one go.
[00:10:15] Paste this exact command. So let's paste this exact command over here.
[00:10:20] Let's paste it once again, and boom,
[00:10:22] nothing happened. and now I'm gonna give it a try once again let's say
[00:10:26] I click on Claude And there we go.
[00:10:30] This is what it's gonna look like guys. If this is the first time
[00:10:33] you're interacting with AI in a terminal, it might look a little bit intimidating,
[00:10:38] but do not worry. Trust me, over time,
[00:10:40] as you use this system for a couple of days,
[00:10:42] you'll pick it up really, really quickly. And this is exactly how we're gonna
[00:10:45] be moving with rapid pace now. I'm just,
[00:10:49] over here, do not install cursor,
[00:10:51] do not install code. Neither, just say, never show again.
[00:10:55] You can come over here, you can choose the mode you like. I personally
[00:10:57] just pick light mode. And now, this is important once again,
[00:11:02] what we wanna be doing is we wanna use the subscription.
[00:11:05] We don't wanna be choosing API usage billing.
[00:11:08] Very important guys, please make sure to select option number one,
[00:11:12] because this allows us to use the significant amount of energy.
[00:11:16] Usage credit, that already comes with the subscription.
[00:11:20] So I click on enter. It opens it on my browser.
[00:11:23] I say authorize. And there we go.
[00:11:26] If we come back over here, we're logged in. Now I simply click enter.
[00:11:30] Cloud can make mistakes, that's perfectly fine.
[00:11:32] And enter. Use recommended settings?
[00:11:35] Yes, use recommended settings. And you would also wanna say,
[00:11:38] I trust this folder on here, over here. So now Cloud can freely navigate
[00:11:42] this entire folder and do what is needed. I trust it,
[00:11:45] and there we go guys. We have Cloud installed over here.
[00:11:49] Now coming back to the document, what we wanna do is,
[00:11:52] you can even test it out if you want. You can see what version
[00:11:54] of Cloud do we have. So, what I would do is,
[00:11:58] press enter to continue. And we can say for a second,
[00:12:00] I'll just show you hello who is this?
[00:12:04] You're basically chatting to Cloud over here now.
[00:12:07] You can see, this is the same chat,
[00:12:10] simply in a terminal that you would be having over on Cloud over here.
[00:12:14] It's the exact same thing. Like, we can test it out.
[00:12:17] Some really cool things. You can simply screenshot something,
[00:12:20] and you can just dump it into this and say,
[00:12:23] what, what is this?
[00:12:26] And let's give it a second.
[00:12:29] It will read the screenshot, and it will tell you. So,
[00:12:32] over here, it's going to say, do you want to proceed? Yes, allow reading
[00:12:35] during this session. Let's give it a second,
[00:12:38] and you can see, that's a screenshot of Cloud, installing Cloud.
[00:12:40] Exact same model, and the exact same system we have been using on the
[00:12:44] consumer model on the app on Cloud.
[00:12:47] The only difference here is that it's using through the terminal.
[00:12:49] So, a couple of tricks, guys, I want to show is,
[00:12:52] you can click on slash model.
[00:12:54] If you click on slash model, now you can choose which model you want
[00:12:57] to be using. By default, it is Opus 5.0.
[00:12:59] You can switch to Fable, you can use Sonnet. Let's say we want to
[00:13:02] be using Fable, switch to Fable.
[00:13:05] Then, you can click on slash effort.
[00:13:08] Effort allows you to choose the thinking level of the AI.
[00:13:11] You can go to UltraCode, which is maximized.
[00:13:13] You can go to max, extra high, and whatnot.
[00:13:17] Then, now let's come back to the document. As we can see,
[00:13:20] we have skipped this process, or we have really completed this process,
[00:13:24] if I were to rather say. But,
[00:13:26] what you really want to be doing, guys, if you fall,
[00:13:28] or if the terminal is giving you some complexity,
[00:13:33] or if it's asking you questions that you don't really understand the answer to,
[00:13:36] do not fall. Do not sit there and try to figure it out yourself.
[00:13:38] Simply, copy, literally,
[00:13:41] let's say this is the section in the terminal when I was trying to
[00:13:44] install Claude, and it was giving me some errors. Just copy it,
[00:13:47] come back into Claude, and paste it. It will tell us exactly what to
[00:13:51] do, and all you have to do is just copy things one by one,
[00:13:53] left, right, back and forth, until you finally have Claude installed the way I
[00:13:57] did. Now let's go into the next step.
[00:14:00] So, So we have signed in, we have used Cursor's Terminal,
[00:14:03] and we have typed in Claude, and we can see.
[00:14:06] Let's paste this test command over here. Copy.
[00:14:10] We can come back over here, and just paste it.
[00:14:14] Just to test it out and see if it can truly read our documents
[00:14:18] over here, which is the StartHere.md file, and the Claude.md file,
[00:14:21] which we can see. The .md file,
[00:14:24] once again, this is a markdown file,
[00:14:26] which is simply a, the formatting document that allows the AI
[00:14:31] to pass through information in a relatively seamless manner.
[00:14:34] We can see both files read, nothing edited, created,
[00:14:36] or deleted. Fictional data, five standing rules from Claude.
[00:14:40] Never send a message, email, or text, and never contact anyone.
[00:14:43] Now we will go into the exact structure of these files in later lessons.
[00:14:46] Right now, the purpose of this lesson is simply to install Claude,
[00:14:50] and install now the next couple of tricks that will allow us to move
[00:14:53] really quickly, guys. So let's move this over here.
[00:14:57] Now that we have this, what we want to be doing is,
[00:15:00] let's come back over here, and, it's a really simple command,
[00:15:03] but on, on the Mac, if you press,
[00:15:06] let's say we are already, as you can see here, nothing has been installed
[00:15:09] here, correct? We have an open Claude. But if we open Claude,
[00:15:12] boom, there we go. Now we have Claude installed.
[00:15:14] But let's say we want to quit Claude, or we want to close it.
[00:15:17] On the Mac, what you can do is Ctrl-C,
[00:15:20] and press Ctrl-C again, and you can see,
[00:15:23] we have exited back to the base terminal.
[00:15:26] So this is the terminal, and whenever you type Claude,
[00:15:28] that is when Claude opens once again. You can simply delete it.
[00:15:33] If you click on click or the delete icon. The trash can,
[00:15:35] you close it. And let's say we've come back to Scratch.
[00:15:39] Once again, we open this, the terminal,
[00:15:42] and we type Claude. We open more agents.
[00:15:46] You can call this, if you're having different workflows over time.
[00:15:49] Let's say you're working on client MedSpa.
[00:15:53] This is how you would format it. So you have clear visuals into all
[00:15:56] of the agents that at work. Working concurrently and doing work for you.
[00:15:59] You can also right click. You can change the color.
[00:16:02] This is just something I do at times visually because at times I have
[00:16:05] many different running systems or terminals. And you can simply change the logo.
[00:16:09] Cut the entire section of when I was talking about the colors and the
[00:16:12] logos. Or right clicking because that did not work.
[00:16:15] Please cut that. Now
[00:16:24] the next thing we want to be doing is we want to install something
[00:16:28] known as Claude Dangerously Permissions.
[00:16:31] And the reason being guys is if you noticed and paid attention during the
[00:16:35] previous section when I was pasting this prompt into Claude,
[00:16:38] it asked me for permission for something. It asked me for permission.
[00:16:41] Can I see or can I specifically get access to this file?
[00:16:45] Confirm it with me.
[00:16:58] First I'm gonna go over, the different settings in Claude.
[00:17:03] I'm about to start. So
[00:17:10] here are a couple of tricks besides model and besides effort that I
[00:17:15] use myself every time when I'm using Claude. So,
[00:17:17] one of them is, specifically,
[00:17:20] What the f***? Okay, cut that out as well.
[00:17:23] What the f***? Besides
[00:17:33] model
[00:17:39] and effort, here are a couple of other slash commands that I typically do
[00:17:43] use with Claude. One of the important ones being Usage.
[00:17:46] So guys, as we, start using Claude quite a bit,
[00:17:48] and as you start using it to go through massive files and information,
[00:17:53] you want to be keeping track of your usage over here.
[00:17:56] And this is a simple one that, lets us know that within our $100,
[00:17:59] $20, or $200 plan,
[00:18:02] how much have we consumed. That is a simple one.
[00:18:05] You can always press escape to cancel it.
[00:18:07] Then what we have is Configurations. Configurations are simply all of the settings.
[00:18:12] I'm going to walk over the ones that I believe are pretty important.
[00:18:15] By default, the best ones are already going to be installed.
[00:18:19] But, this, all of this looks fine to me.
[00:18:21] We want to keep ThinkingMode turned on. Ultracode keyword trigger,
[00:18:25] keep it on. Dynamic workflow size, that's all good.
[00:18:28] Terminal. Default permission.
[00:18:31] Accept edits. How are we going to put it at?
[00:18:34] Don't ask. But we're going to make a better version of this.
[00:18:36] Shortly after this. And then what we have are.
[00:18:41] Okay, honestly, just, Cut the section about configurations as well.
[00:18:44] That was useless.
[00:18:54] So now, if you were paying attention to the previous prompt that
[00:18:59] we had entered into Claude, you would have noticed that Claude asked me for
[00:19:02] permission. It asked me that,
[00:19:05] for this session, am I allowed to get access to this file,
[00:19:08] or am I allowed to do this. And guys,
[00:19:11] that is great at first, however,
[00:19:14] over time, very quickly, you're gonna see that as you start using it.
[00:19:18] It becomes fundamentally very unproductive
[00:19:23] and unusable when you have multiple agents who you want to be running autonomously,
[00:19:28] completing big workflow tasks, pulling all of the call transcripts from your
[00:19:33] Fireflies, ingesting that information with your CRM,
[00:19:37] then doing some data segregation,
[00:19:39] some data sanitization, creating reports out of that,
[00:19:42] all in one workflow. It is simply impossible to get something like that done
[00:19:47] if the AI is going to keep asking us for permission every 5 seconds.
[00:19:50] Because what happens is, traditionally,
[00:19:53] you open Cloud, you give it instructions,
[00:19:55] it says, okay, wonderful, I'll get working,
[00:19:57] and then you open another agent, let's say, and now you're working on another
[00:20:00] part of the business. Maybe it's a bit of outreach.
[00:20:02] Maybe you're working on some content creation.
[00:20:06] But what happens is, you are already mentally expecting that
[00:20:10] the agent over here, let's call it client,
[00:20:14] let's call it client scoping,
[00:20:16] you're already expecting the agent to be scoping the entire,
[00:20:20] system for a client along with the pricing and create a proposal at the
[00:20:23] end of it. And you expect it to probably be,
[00:20:26] probably be like a 30 minute long session after that.
[00:20:28] When you check back into it, and now here you are creating,
[00:20:32] let's say, doing some outreach, okay? This is traditionally how I use AI,
[00:20:35] and this is what my workflow for my businesses looks like on a day-by-day
[00:20:39] basis. And now, let's say you've done your outreach,
[00:20:41] you've been doing it for 30-40 minutes, and you're like, okay,
[00:20:44] cool, I got done with this task, outreach is going out,
[00:20:46] now let me come back and see if my proposal is ready.
[00:20:49] And you come back over here, and literally on the first step of even,
[00:20:52] like, ingesting the calls from Fireflies, the AI completely stumbled and asked you,
[00:20:57] oh, can I get permission for this? Guys,
[00:21:00] that is the most frustrating thing that can ever happen.
[00:21:03] Because you were expecting the AI to do 30 minutes of valuable,
[00:21:06] deep work. And when you came back, you realized that it hasn't even started
[00:21:10] the process. It's not even started getting, using the API to get the call
[00:21:14] conversations. So that is why, what we want to be doing,
[00:21:17] is we want to be turning on a feature in Claude that is known
[00:21:21] as Dangerously Skip Permissions. Now,
[00:21:24] it sounds scary, but the simple meaning is,
[00:21:27] it simply stops asking you, are you sure?
[00:21:29] And just works. And you own the outcome.
[00:21:33] So everything in this folder that we created over here is fictional.
[00:21:37] So run like this.
[00:22:02] And I'm going to show you how to install it.
[00:22:04] So let's just copy this over here. And inside of Claude, here, we can
[00:22:04] just tell it, Thank you for watching. Paste it and say, Can you install
[00:22:09] this and make it the default setting
[00:22:14] when I launch Claude?
[00:22:17] As you can see guys, I made a bunch of typos, but it's perfectly
[00:22:20] fine because the AI will understand exactly what I'm trying to communicate regardless.
[00:22:24] Take a look. I gave Claude the instructions to install
[00:22:28] dangerously, skip permissions, and it came over here telling us that I'm blocked.
[00:22:33] The session is running in a mode where every tool I call is being
[00:22:36] denied. Okay, cool. But either way,
[00:22:38] we're going to tell it to just get it done. And it tells us
[00:22:41] over here, go ahead. Let's simply just copy this and see what it comes
[00:22:46] out with. And
[00:22:51] it seems, for some reason being, Claude is still struggling to do it,
[00:22:55] and it gives us a permission over here. Let's give it a read.
[00:22:59] The fastest fix is to run the command yourself by typing it in this
[00:23:02] session. Cool. So what it's telling us over here is open a terminal,
[00:23:07] and copy this instruction. So I'll just come back over here,
[00:23:11] and I'll copy the entire thing. Command C,
[00:23:16] and paste it. Where,
[00:23:19] it seems to have installed, and, and once I've entered it,
[00:23:23] it seems to have installed the exact instruction.
[00:23:25] And to test it out guys, I'll just come into this.
[00:23:27] Type Claude once again, and we can see,
[00:23:30] warning Claude code running in bypass permissions.
[00:23:33] Yes, I accept it, and there we go.
[00:23:36] The indicator to tell that it is turned on is this,
[00:23:39] over here, bypass permissions. If you click shift plus tab,
[00:23:42] so I'm clicking shift and tab, we can see that it rotates now.
[00:23:45] Auto mode on, manual mode on,
[00:23:48] accept edits on, plan mode, and bypass permissions.
[00:23:51] I personally always keep bypass permissions on,
[00:23:55] however, if this is something that you want to be moving with a bit
[00:23:57] of caution with, I'd recommend you to definitely start off with manual mode,
[00:24:02] accept edits on, and auto mode.
[00:24:04] Try them out, and you probably are,
[00:24:07] like me, going to come back to bypass permissions on.
[00:24:10] Cause, in order to truly have a native agentic running business,
[00:24:15] you need the AI to be able to make decisions for itself,
[00:24:18] and move ahead with tasks. And that is exactly what we're going to be
[00:24:21] teaching over the future modules, on how you want to be ensuring,
[00:24:25] and creating the rules, guardrails,
[00:24:27] SOPs, for the agents to reliably use something,
[00:24:32] such as bypass permissions on, and take actions,
[00:24:34] and get work done within the company. So
[00:24:44] now that we have bypass permissions turned on,
[00:24:47] we went over slash model, this is the model that we're going to be
[00:24:50] using, it is, totally up to your descri- it is totally
[00:24:55] up to your- which can rotate between Fable,
[00:24:58] Opus, and Sourcetree. the effort is how hard the AI model thinks,
[00:25:03] the usage is where our plan limit stands.
[00:25:06] I would definitely recommend you guys to check and do a quick slash usage
[00:25:10] before heavy lessons, or before heavy tasks and projects that you're expecting the AI
[00:25:15] to run through. And finally, we have configuration.
[00:25:18] Now,
[00:25:20] The final tool I want to go over is something known as WhisperFlow.
[00:25:25] WhisperFlow, on a high level, is a AI dictation tool that allows us to
[00:25:30] go really, really quick. I often joke around this with my friends and I
[00:25:34] tell them that we do not type anymore.
[00:25:37] You know, in 2026, we do not type,
[00:25:39] all we do is talk. So I,
[00:25:42] myself, I never type prompts, I never copy things.
[00:25:44] I never type personally. What I do is I use this tool known as
[00:25:48] WhisperFlow. It costs $15 a month and the way it really works is you
[00:25:52] go into your settings and you put something known as a hotkey.
[00:25:56] In my case, it is FN and when I click on FN,
[00:25:59] take a look at this. So now the AI is listening to me.
[00:26:02] So what I do is I'm simply going over information,
[00:26:05] parsing through it and once I'm done, I tap F in the again and
[00:26:09] boom, you can see the entire piece of information is boom,
[00:26:13] like instantly dictated. I'm probably one of the longest running customers,
[00:26:17] 3 million words dictated and typing is definitely the fallback,
[00:26:22] but this is not the method I would recommend,
[00:26:25] especially as you get deeper into this course and improve your AI proficiency,
[00:26:29] working with deeper and bigger systems that require more context,
[00:26:32] cause what a tool such as this allows us to do on a fundamental
[00:26:35] level, like why am I here dictating 3 million words over the past year
[00:26:39] and a half? It is simply because AI is all about the context you
[00:26:44] give it and as you work and see the lessons later on,
[00:26:48] the more context you give the AI, the more you can explain to it
[00:26:51] how you think about the things, why you asking it to do things a
[00:26:54] said specific way, the better it is going to respond and give us outputs.
[00:27:07] All in all, I highly recommend you to download WhisperFlow and begin
[00:27:11] using it and this lesson on a high level shows us how to set
[00:27:15] up our Mac, or sorry, the machine that allows us to run AI natively
[00:27:19] internally. Once again,
[00:27:22] if there are any issues that you face when you're installing Cloud locally,
[00:27:26] simply screenshot that, copy whatever you have in the terminal and paste it back
[00:27:30] into Cloud and copy what Cloud gives you back into the terminal.
[00:27:34] Within 20 minutes, you will have your entire clod running internally as well and
[00:27:38] that's it. This allows us to create the fundamentals and the starting level to
[00:27:43] now begin creating the business brain and working through the next modules and making
[00:27:47] different AI systems.