YT YouTube2Text

YouTube2Text → Transcripts → Full Claude AI Tutorial (Zero to Pro Beginners Course)

Full Claude AI Tutorial (Zero to Pro Beginners Course) — Transcript

by AutoDS - Build Your Online Income · 20,816 words · 2,899 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Today, I'm going to show you how to use
0:01Claude from the ground up. We're going
0:03to start with the basics, then we're
0:05going to work away into projects, custom
0:08skills, and different tools like Claude
0:11co-work,
0:11>> [music]
0:11>> Claude code, and Claude design. We're
0:14going to cover every feature that Claude
0:15has to offer, and we're going to be
0:17breaking it down in the [music] most
0:18easiest way possible. And as we go,
0:21we're actually going to be putting all
0:23of these different skills to use by
0:25building things with each one of these
0:26different features. Look, I've been
0:28using Claude for the last couple of
0:30years, and for multiple things like
0:32content, brainstorming, and building my
0:35own businesses and my apps, and I've
0:37learned a lot about when and how to use
0:40it and to get the most optimal results.
0:42So, just open up Claude, follow along,
0:45and let's get started. Okay, so to get
0:47started, the first thing that we're
0:48going to go over is actually getting
0:49started and installing Claude. So, for
0:52this, you're going to go to
0:52claude.ai/login,
0:54and don't worry about it, all of this
0:56information is going to be in our cheat
0:58sheet where I'll be covering everything
0:59that I'm talking about in today's video
1:01just in an easy-to-reference PDF format.
1:04And if you want access to that, just go
1:05ahead and stick around just a little bit
1:06longer, I'll tell you exactly how you
1:07can get it in just a little bit. So,
1:09once we first land on the Claude
1:10website, the first thing we want to do
1:12is simply log in. So, go to continue
1:14with Google, Apple, or just enter your
1:16email.
1:20Now, after you do that, you're going to
1:21go ahead and create your account, and
1:22then you're going to choose your plan.
1:24Now, for most people, the free plan is
1:27going to be enough if you're just simply
1:28talking to it, trying to get a couple
1:29answers, and really just trying to use
1:31it as a type of, I guess, chatbot.
1:34Something where you can quickly get
1:35information. But, if you actually want
1:37to use it to start managing, let's say,
1:38your files, if you want to use it as a
1:41co-worker, something that's actually
1:43going to be efficient to your day-to-day
1:45life or for your business, then you do
1:47want to go with the pro plan. That's the
1:49one that I started with, and little by
1:50little as I started to use it more and
1:52more, I realized I'm starting to run out
1:54of my pro plan credits it's little bit
1:55quick, so I switched over to the Max.
1:58But, that's me. I use Claude for
1:59literally absolutely everything. I build
2:00a bunch of different things. But, for
2:02most people, either the free or the Pro
2:05is going to be enough. To get started, I
2:07suggest start with the free. When you do
2:08reach your limits and you realize that
2:10you're starting to reach them more
2:11often, just go with the Pro. This is
2:13definitely going to give you a lot more
2:15breathing room and it's going to give
2:16you a lot more room also to test all of
2:18the different features. Now, after you
2:19do choose your plan, you are going to
2:21have two options to use Claude. You're
2:23going to be able to use either the web
2:25version, which is this one right here,
2:27or you can do the desktop version, which
2:29is this one right here. My suggestion,
2:31use the desktop version. Yes, the web
2:34version is very convenient. It's going
2:35to work and you can even access pretty
2:37much all of the different features from
2:38your phone. But, the desktop version is
2:40going to give you a lot more
2:41functionality. Like for one, you're
2:43going to be able to give it access to
2:44different files or folders on your
2:46computer. It can also take over your
2:48browser, so it can start doing even
2:50deeper research and actually start
2:52visualizing different things like let's
2:53say on certain websites. You can tell it
2:55to go to a website and scan the entire
2:58site. On the web version, it can go to
3:00that site, but it's not going to
3:02essentially visually see the website.
3:05Instead, it's just going to see its
3:06code. Whereas, with the desktop version,
3:08it's actually going to be able to take a
3:09screenshot of the page and analyze the
3:12entire thing along with being able to
3:14also analyze the back-end code as well.
3:16Not only that, but you also get access
3:17to the basic chat and co-work. Co-work
3:21is going to be what I just mentioned
3:21earlier, being able to give it access to
3:23different projects or different folders
3:25on your computer. So, that way instead
3:27of having to upload certain files to
3:29Claude, you could just tell it, "Hey,
3:30just check out my downloads folder and
3:33look for this file." Or, go to my
3:35downloads folder and organize the entire
3:37thing. And to show you what I mean,
3:39right here, I have a bunch of different
3:41files that I have loose in my downloads
3:44folder. So, what I'm going to do is to
3:45quickly show you how this can work. I'm
3:47going to give it something simple, go to
3:48my downloads folder and organize all of
3:49my loose files into folders. Go ahead
3:51and hit enter. And in just a couple
3:53seconds we're going to have a insanely
3:55messy folder just completely organized
3:57by different folders. So, once you tell
3:59it to do that, it's going to ask you,
4:00"Can I access this folder?" All you do
4:02is just click on allow. Then it's going
4:03to ask you, "How do you want everything
4:05organized?" It's going to give you a
4:06recommendation, so you can do it by file
4:07type or by type with different
4:10subfolders. So, let's say by file type,
4:12it's going to say new folders, it's
4:13going to do images, videos, documents,
4:15installers, and all of the different
4:17types of files that you have. It's also
4:18going to ask you by type with
4:20subfolders. So, it's just going to do
4:22what it just said, except it's going to
4:23get a bit more granular. It's going to
4:24have more folders within those folders.
4:27And into existing project folders. For
4:29this, let's just do it simple, let's do
4:31by file type, and let's see if we can
4:32see it in action. So, here I have a few
4:34different videos, right? A few Oh,
4:36they're gone already. That was fast.
4:38Done. All 325 loose files are sorted
4:41into eight new folders. I didn't even
4:43know I had that many loose files.
4:45Let's go ahead and clean up by date
4:46created.
4:48And we have no more loose files.
4:50Everything is officially organized for
4:52us. We have data, audio, 3D models,
4:55archives, installers, documents, videos,
4:58images, and a bit more. So, what you
5:01just saw here, you would not be able to
5:02do using the web version. So, that's why
5:04I highly suggest you download the
5:05desktop version because it is going to
5:07help you so much more. It's going to
5:09make you so much more efficient. So, now
5:10that you know the difference between
5:11chat and co-work, let's talk about
5:13everything else that's on the same
5:14screen. So, for one, you have here, of
5:16course, your chat. This is where you're
5:17simply going to talk to Claude and
5:19you're going to give it whatever it is
5:20that you need.
5:21You also have here the option to change
5:23between basic chat. This is going to use
5:24a lot less credits. And you have co-work
5:27where you're going to be able to, again,
5:28give it access to different projects,
5:30different folders, or make different
5:33projects, which we'll talk about in just
5:34a minute. Over here you have your model
5:36selection. We'll talk about these in a
5:37little bit, but very quickly you have
5:39Fable 5.1, Opus 5, Sonnet 5, and Haiku
5:424.5. Each different model is going to
5:45give you a different level of strength
5:47or smartness. So, the higher the model,
5:51the more complex it can start to think,
5:53the more complex tasks it can do. But,
5:56that's also going to mean it's going to
5:57be using a lot more of your credits. So,
6:00you're going to reach your usage limits
6:01a lot quicker with the higher models.
6:04Now, when it comes to Fable 5.1 and
6:07Fable 5 over here, these two are not
6:10included in your pro subscription, and
6:12that's very important to know.
6:14These two you actually have to pay what
6:16they call API credits, which means
6:18you're going to have to pay for these
6:19separately. And the reason for that is
6:21because they're very powerful, and they
6:23essentially would take up a massive
6:24amount of your credits. So, that's why
6:26Anthropic did that. You can use these if
6:28you want, but always remember they are
6:30going to incur an extra cost. And
6:32honestly, for most of the things that
6:33we're already doing, whether it be
6:35creating a website or even creating a
6:36web app or even a local desktop app, for
6:39the most part with Opus 5, you're going
6:42to be able to get it done. But here,
6:43you're going to be able to see exactly
6:44what the different models are for. So,
6:465.1 is for your toughest challenges. You
6:49want to build a very complex app,
6:50something that has a lot of different
6:52levels to it, something that has, let's
6:53say, a login system, a back-end system,
6:55all of that, then you do want to
6:56implement Fable 5.1. But, my suggestion
7:00for something like that, start with
7:01Fable 5.1, and then you can lower the
7:03strength to continue on with the work so
7:05you don't use as many credits. Opus 5
7:08for complex tasks, generating websites,
7:11generating small web apps, things like
7:12that. Sonnet 5, most efficient for
7:15everyday tasks. This is something that's
7:17really quick, something that's easy to
7:18use, something that's pretty good for
7:20quick research. And Haiku 4.5, this is
7:23the fastest version, but it might not
7:25always give you the most accurate
7:26answers. Not saying that the answers are
7:28going to be wrong, but
7:30it's AI, always double-check. Then you
7:32have your effort. This is essentially
7:33going to be the depth of the responses
7:38that each one of these different models
7:39are giving you. So, if you do, let's
7:42say, Opus 5 on low, it's going to give
7:44you a pretty good response. It's going
7:46to give it to you fairly quickly, but it
7:47might not spend too much time actually
7:49thinking about a certain problem if you
7:51give it a problem. So, it might not
7:53fully solve that problem for you. Then,
7:55as you start to increase the strength,
7:57then it is going to start becoming a bit
7:59more smarter, and it's going to be able
8:00to take on these extra tasks a lot
8:02easier.
8:03We'll be diving into a few of those and
8:04showing you a few of those examples a
8:05little bit later on. Now, if we look
8:07down here, we have our project or
8:08folder. This is where we're going to be
8:10able to give it access to the different
8:11folders or to a project. You're going to
8:14see the different projects over here.
8:15These are different projects that you
8:16can create for, I mean, really just
8:19different projects. You can have, let's
8:20say, a project where you're creating a
8:23e-commerce website, and then you can
8:24have a project where you're creating a
8:26video game. I'll be demoing that for
8:28you, and I'll be showing you how to
8:29create a project and exactly how it
8:30works when we get to that section. But
8:32for now, we also have the manual and the
8:35automatic approve. What does this mean?
8:38So, as you saw earlier, Claude asked me,
8:40"Can I access the downloads folder on
8:41your computer?" and I clicked on yes.
8:43And that's because it was on manually
8:44approve. If I had it on automatically
8:47approve, it wouldn't ask me that. It
8:48would just go straight to it, and it
8:50would do what I told it to. This makes
8:52Claude a lot faster, and it makes it a
8:55lot more efficient because it doesn't
8:57have to wait for you to approve things.
8:59But, I know a lot of people wouldn't
9:01feel comfortable doing that. Me
9:02personally, I don't care. Go ahead and
9:04take over my computer, do what you got
9:06to do, just get the task done. But
9:08again, I know a lot of people aren't
9:09fully comfortable with that because it's
9:11AI. They prefer to have an oversight on
9:13it, which I totally understand. Then
9:15down here, we just have a couple of
9:16ideas. So, these are just suggestions
9:18that Claude gives you that you can do
9:20with the app itself.
9:22And over here, we have our side panel.
9:24On the side panel, we can start a brand
9:25new chat. If we click on this, it takes
9:27us directly back to this page. If we
9:29click on the projects, we can see the
9:30different types of projects that we
9:31have. This particular chat, I don't have
9:33too many projects, but if I open up
9:35another Claude account that I have, you
9:36can see here I have a lot of different
9:38projects for just a lot of either
9:40different clients or different apps or
9:43different websites that I'm working on.
9:44Then we have our artifacts. So, what
9:46exactly is an artifact? It's a weird
9:48name. I don't know why they would name
9:50it an artifact, but it's essentially
9:51just an app that you make or something
9:54that you make through Claude. So, an
9:55artifact can be something like a game
9:57like this one over here where it's this
10:00little game that I made. If you want to
10:01see this, just go ahead and make sure
10:03you subscribe because one of the
10:05different videos that we have, I do show
10:06you how you can make something like
10:07this. But, an artifact doesn't just
10:09simply mean it's a game. It can also be
10:10any type of app. So, you can make a
10:12calculator app and that's going to be an
10:14artifact. You can make a spreadsheet and
10:17that's going to be in the artifacts
10:19section. Again, deeper dive a little bit
10:21later on in the video. Then we have our
10:22scheduled tasks. Over here we're going
10:24to be able to schedule different tasks
10:25like let's say checking our email every
10:27single Monday and giving us a brief,
10:29things like that. We have Claude Design.
10:31I have an entire video on this if you
10:32want to check it out, it's going to be
10:33this one right here. But, Claude Design
10:34is essentially going to be your design
10:37team. You can tell Claude how you want
10:38certain things to look and Claude will
10:40be able to help you out with that. If
10:42you wanted to rebrand, let's say one of
10:43your websites and you want it to be on
10:46your brand style, all you do is give
10:48Claude your brand style and it'll be
10:50able to apply that to your website or
10:52whatever app that you want it to go on.
10:54We then have the customize section over
10:55here. On customize we can add skills, we
10:58can add connectors, we can add plugins,
11:00and we can also access our plugins or
11:03connectors or skills and we can look for
11:05more. And then after that we have our
11:07chat history. So, we have our pinned
11:09section over here, any chats that we
11:11want to keep pinned that we want easy
11:13access to, essentially our favorites, we
11:15can have on here. And then we have all
11:17of the different chats in our history
11:19from the one that we just did right now
11:21to the very first one that we ever had
11:23with Claude. Now, if you click down
11:25here, we're going to have all of our
11:26different settings. So, we have the
11:28settings, the language that we want
11:30Claude to be in, some help options. We
11:33can also upgrade our plan if we choose
11:34to. We can get different apps and
11:36different extensions. You have a learn
11:38more, just again more information on
11:40Claude itself, but let's click on
11:41settings. And here we have our different
11:43preferences. We can choose our different
11:45themes. We can either go with whatever
11:46our computer is set at. We can do a
11:48light one, which is way too bright for
11:50me, or we can set it default to a dark
11:52theme. I usually keep everything in dark
11:54theme, but that's up to you.
11:55We have how we want our chats to look.
11:58So, we have our different fonts. And
11:59then we have a bit more information on
12:01here, which is pretty self-explanatory.
12:02But what I do want to show you on this
12:04is going to be under accounts, under
12:06instructions for Claude. Now, this right
12:08here is going to essentially set
12:11everything for Claude, right? So, let's
12:14say we want Claude to have a particular
12:16set of instructions that it applies to
12:18every single one of our chats. We add
12:20that on here.
12:22Personally, I don't like to have
12:23anything on here, but I do know people
12:26that like to have specific instructions
12:28for Claude, so that way all of their
12:31answers are exactly the same. So, for
12:32example, in the instructions for Claude,
12:34we can simply tell it for every single
12:36time that I ask you to make a game, ask
12:40me what kind of game do you want it to
12:42be? Do you want it to be an action? Do
12:43you want it to be an adventure? Or do
12:44you want it to be a racing? And it'll
12:45give us those three exact examples that
12:47I gave it every single time. But in that
12:50case, it's going to then limit it to
12:52those three examples. It's not going to
12:54ask me if I want any other genres
12:56afterwards. It's not going to ask me,
12:57"Do you want a shooter game?" It's not
12:59going to ask me anything extra. It's
13:01going to stick to exactly what I gave
13:02it. So, I'm going to give it something
13:03pretty silly, something pretty funny,
13:04but it's going to be the perfect
13:05example. Every single time that I say,
13:07"Hey Thor, respond back to me in a
13:09manner that Thor would."
13:11So, we click on save changes. I have to
13:13remember to take that off. And I can
13:14say,
13:16"Hey, what's the weather
13:17like?"
13:18For something like what I just asked it,
13:20"Hey, what's the weather like?" Opus 5
13:21is a bit overkill. We can actually have
13:24dropped it down to Sonnet or maybe even
13:26Haiku, but that's fine. So, it's asking
13:28us what the zip code is. Let's just do
13:29it quick. 44878.
13:32As we see here, it's asking us for
13:33permission because we have it on manual.
13:36So, let's just go ahead and always
13:37allow. It says right now in Shiloh,
13:40apparently, Ohio, it's 67° F. Now, if we
13:43click on new over here, you're going to
13:44see that it takes us back to the brand
13:45new chat. Let's switch this over to
13:49automatically approve.
13:51Or even skip all approvals altogether.
13:54It's never going to pause even for
13:55unsafe actions. Let's leave it on
13:57automatically approve.
13:59And let's say,
14:00"Hey, Thor.
14:02What's the weather like today?"
14:05And let's switch it down to sonnet.
14:08Even then, I don't think we need high,
14:09but that's fine.
14:11A hearty A hearty clap of thunder echoes
14:13as I speak. Ha, a mortal asks the god of
14:16thunder about the sky's temperament.
14:18Well met, my friend Mario. So, as you
14:20see, it's talking to us in a way that
14:21Thor would. But, then it actually comes
14:23up with even Mjolnir cannot strike where
14:26it does not know to look. So, it's just
14:28asking me where are we living in. So,
14:30let's say Los Angeles. Now, it's
14:32actually going online. It's checking to
14:33see what the weather is like Los
14:35Angeles. And then it's going to reply
14:36back in Thor's voice once again. So,
14:38Mjolnir hums as I gaze westward towards
14:41the city of Los Angeles. No, sorry. The
14:43city of angels. Los Angeles this eve is
14:45calm, mortal. Clear skies, 70° F
14:48with the north wind blowing gentle at 5
14:50mph. So, that's exactly what
14:52Instructions for Cloud is going to do.
14:54It's going to go ahead and give it
14:55consistent instructions that it's always
14:57going to follow. Now, again, I don't
14:58like doing this. I prefer to have
15:00instructions rather than at a an entire
15:03chat level. I like to have it at the
15:05project level. Now, besides the account
15:07over here, we also have privacy,
15:09billing, the usage. The usage is
15:11important because it's going to tell you
15:12how much you have left in your current
15:13session. This resets every 5 hours. So,
15:17whenever you do reach your 5-hour
15:19session limit, you are going to have to
15:20wait until the end of those 5 hours.
15:22Now, this is also going to vary a lot
15:25depending on the type of plan that you
15:26have. You're going to have a lot more
15:28session limits. You're going to have
15:29higher session limits if you have the
15:31higher plans, like let's say the max.
15:33Now, besides that, what I did want to
15:34show you here was going to be the
15:35memory. So, search and reference chat
15:38and generate memory from chats. So, the
15:41search and the reference memory allows
15:42Claude to search for relevant details in
15:44past chats. Meaning, if you ask Claude
15:46something in a brand new chat that it
15:49knows absolutely nothing about, if
15:51there's any context of it in any
15:53previous chats in your history. So, if
15:55you've talked to Claude before about
15:57that particular topic or subject, then
15:59it can go back to that chat, reference
16:01that chat, and give you an answer based
16:03on that. So, with search and reference
16:05chats, it's essentially going to dig
16:06through older chats to get your
16:08information or to get the information
16:10that it needs to give you. Whereas, with
16:11generate memory from chats, for example,
16:14we can give Claude a type of
16:16instructions, right? Let's say you're
16:18creating a ad for a fan, right? You can
16:21tell it that you prefer conversational
16:22ad scripts at, let's say, 30 seconds max
16:26with no on-screen text. And then later
16:28on in the chat, you could just simply
16:29tell it to give you an ad for a product,
16:33and it'll take that information that you
16:34gave it, whether it be from one chat or
16:36another, and it's going to generate a
16:38memory from it, and it's going to apply
16:40that as essentially a type of set of
16:42instructions that it can start to
16:43reference going forward. So, that's the
16:45general overview of everything that you
16:47see when you first jump into Claude,
16:49from the chat to the side panel. Now,
16:51let's get a bit more into detail and
16:52let's talk about prompting. So, what
16:55most people get wrong about prompting,
16:57what the majority of everybody gets
16:59wrong about prompting, is that they
17:00simply just tell Claude, "Hey,
17:03do this for me." Or, "Hey, what's the
17:06weather like?" And they don't give it
17:07any context. So, you give it a basic
17:10prompt, you get a basic answer back, and
17:13then somebody looks at it and they
17:14think, "Oh, that's it? That's all that
17:15you can do? Why am I paying like 20
17:17bucks a month for this? Why does
17:18everyone say it's so good? A lot of
17:19people, if not the majority of people,
17:21are talking to Claude as if they were
17:24talking to ChatGPT when it first came
17:26out 3 or 4 years ago. They're talking to
17:28it like a basic chatbot. They're just
17:30having a basic conversation for it.
17:33They're not giving it absolutely any
17:34context. They're not giving it a role.
17:36They're not giving it any specifics.
17:38Don't worry, I'm going to show you
17:39exactly what a good prompt versus a bad
17:41prompt looks like and how to structure a
17:43prompt. But, not just that, they also
17:44expect it to be just a one-hit wonder,
17:48right? They expect to give one prompt
17:50and that's it. That's going to fix
17:52absolutely everything. It's going to get
17:53everything done. A lot of people think
17:54that if I just simply tell Claude,
17:57"Make me a Grand Theft Auto 6 game."
17:59They expect to have a fully built app
18:01out of just one simple prompt. And it
18:04doesn't work like that. You're always
18:05going to have to reiterate. And this is
18:08probably the most important thing that I
18:10can leave you with when it comes to
18:12prompting. You will always need to
18:15reiterate. It's never going to work the
18:16first time. You're always going to have
18:18to check the work that it gives you.
18:20Let's say, for example, you ask it to
18:21make you a calculator app or you ask it
18:23to make you a game or any sort of app,
18:26you're going to have to give it a
18:27detailed prompt for that app. And then
18:30once you receive your product back,
18:32you're going to have to check it and
18:33then you're going to have to keep
18:34reiterating. You're going to have to
18:35keep telling it, "Look, this doesn't
18:37work. This part right here is off. Maybe
18:41you could try it this way." And that's
18:43also something that's very important. If
18:45you can suggest how certain things
18:47should work, that's also going to give
18:49Claude a better understanding of how
18:51your app is going to work or how the
18:53functionality should be. Again, I'm just
18:55using an app as an example. This can be
18:57for really anything that you're talking
18:59to it about or any information that
19:00you're trying to get from it. Now, there
19:02is a formula for a good prompt and it's
19:05as simple as giving it context, the
19:07task, and how you want your output to be
19:10formatted. That's really all there is to
19:12it. So, as an e-commerce store owner,
19:14let me give you an example of a bad
19:16prompt and a good prompt. So, right
19:18here, I'm going to give you
19:19what I'm going to give Claude, write an
19:21ad for portable fan. That's it. This
19:24right here is what most people are
19:25actually doing, and they're expecting to
19:27have a five-star ad that's going to sell
19:31millions, and that's not how it's going
19:33to be.
19:34So, this is what it gives us, right?
19:36Stay cool anywhere. Summer doesn't wait
19:37for air conditioning, neither should
19:39you. Meet the pocket-size fan that goes
19:41wherever the heat does. It's light
19:42enough for your bag, and it gives us a
19:43bit more information, some details that
19:45it just assumes, up to 12 hours of
19:48cooling on a single USB charge, and the
19:51rest of this, right?
19:52Now, this right here is
19:54it's all right, it's not bad.
19:56But, if we give it something a bit more
19:57detailed,
19:59like let's say this, write a 15-second
20:01video ad for a portable fan. Right
20:04there, we're already giving it more
20:05instructions and more context. We want
20:07this to be 15 seconds. Over here, it
20:09assumed that it was going to be 15
20:11seconds. Actually, it didn't even give
20:13us a time limit for it. So, we're not
20:15even sure of how long this is going to
20:17be. This could easily just be 7 seconds.
20:19Then, we're giving it the audience that
20:20we want it for. So, we're giving it a
20:21bit more context. We're telling it that
20:23the audience is going to be people who
20:25walk or take public transportation to
20:27work and hate arriving sweaty. Then,
20:29we're going to give it even more
20:30context, the product facts. If it's in a
20:33work bag, has three fan speeds, and
20:35charges through USB-C. And it's telling
20:37it what not to do. This is something
20:39that is very important. Don't invent
20:42other features or performance claims.
20:44This is also something that we always
20:45want to do because, as you see over
20:47here, it invented a couple of things to
20:49just fill in the space. Then, it's going
20:51to have a creative direction. So, we're
20:53giving it the direction, we're giving it
20:56the way that we want the output to be.
20:58So, start with a frustrating moment the
21:00audience will recognize, not an
21:02introduction to the product. Where right
21:03here, it actually had an introduction to
21:06the product, which is something that
21:07nobody sticks around for when it comes
21:09to an ad, right? When it comes to an ad,
21:12you're looking for something that is
21:13going to stop the scroll. So, instead of
21:15telling somebody, "Hey, look at this
21:17fan."
21:18you're starting with something like a
21:20pain point, like somebody is saying,
21:21"Oh, I'm really hot." And that's the
21:23type of thing that you want. You want
21:24something that's actually going to work,
21:26right? So, with the creative direction,
21:28it gives us that. And then it's also
21:29giving it a bit more context. Use
21:31conversational language, not just
21:33regular basic phrases like, "Oh, beat
21:36the heat." Or, "This is a must-have."
21:38Or, "Your perfect companion." Which,
21:40funny enough, is actually right here.
21:42Order today and beat the heat before it
21:45beats you. And then we have the output.
21:47Again, we're giving it more context on
21:48how we want the end result to be. Only
21:51the spoken script, 35 to 45 words,
21:53ending with a simple call to action.
21:55Here, I'm getting very granular with it.
21:57You don't have to do this. The 35 to 45
21:59words is usually more or less what's
22:00going to fit within a 15-second window
22:02for what we're trying to do. But again,
22:04you can get as specific as you want, or
22:06you can just give it enough context and
22:08then start reiterating afterwards. Now,
22:10let's go ahead and send this off and see
22:11what the difference is going to be.
22:13All right, perfect. Now, for this
22:15example, we did put the output to be
22:17only the spoken script. So, it gave us
22:19just that. But it phrased it the way
22:21that we wanted everything to be. Now, we
22:23can go ahead and do this one more time,
22:25and we can remove the output because we
22:27already have some creative direction on
22:29there, which kind of starts to push it
22:31towards the output that we want it to
22:32be. Let's go ahead and send that. Then
22:34we're going to see the difference now.
22:35So, now gives us an even more detailed
22:36video script. So, 15-second video ad for
22:39a portable fan. From 0 to 4 seconds, you
22:41have the visual over here. A commuter
22:44awaits at a sunny bus stop in work
22:46clothes and dabs her forehead with her
22:48sleeve. The voiceover, you showered, you
22:51ironed, and you're already sweating at
22:53the bus stop.
22:54Then between 4 to 8 seconds, you have
22:56another visual. Inside a packed bus, the
22:58air isn't moving. She reaches into her
23:00work bag and pulls out the fan. The
23:02voiceover: So, keep one of these in your
23:05work bag. As you can see, this is a lot
23:07more detailed and it's a lot more
23:09essentially scroll stopping than what
23:11you see up here with the simple written
23:13ad for a portable fan. Now, one more
23:15thing that I'm going to leave you with
23:16when it comes to the prompting is going
23:17to be prompt chaining. Now, what is
23:20prompt chaining? Simply put, it's giving
23:23one massive prompt versus a bunch of
23:25little prompts. Personally, I do like to
23:27chain different prompts together. I'll
23:28give it a fairly detailed prompt with a
23:30bit of context. I'll see what the output
23:32is so we can start adding onto this. So,
23:35we already started off with our fairly
23:37detailed prompt over here and now we can
23:39add, let's make it a bit more dramatic.
23:41Add cues for zooms and sound effect
23:43placeholders. So, now we're chaining
23:45onto the previous prompt. And this right
23:47here is what's very important. This is
23:49what I mentioned earlier. You need to be
23:51able to see what it gives you, assess
23:53it, and then reiterate, and then add
23:55onto it or take away from it. Whatever
23:57it is that is going to give you what you
24:00want at the end. So, now it's giving it
24:02a bit more dramatic, right? So,
24:0415-second ad again with this is going to
24:07be the dramatic cut. The visual is going
24:08to be the harsh sun over a bus stop. A
24:11commuter in crisp work clothes checks
24:13her watch. Then the camera slowly pushes
24:15in on her face. A bead of sweat rolls
24:18down her temple. And the sound effects,
24:21there's some sound effects in the
24:22background. You have some cicadas
24:25buzzing and then you also have another
24:26sound effects, heat shimmer hum,
24:29building. So, it's kind of starts to be
24:31a bit more dramatic. You showered, you
24:33ironed. Then for the next one, it has
24:35the zoom to the damp spot on the shirt
24:38collar.
24:39The sound effects is going to be the bus
24:41door. It's going to go
24:43like when it opens. You have the sound
24:45effects of the crowd murmur. You have
24:46people talking in the background. And
24:49the dramatic bass hit on the zoom. So,
24:51whenever that zoom comes in, you hear
24:52the
24:53How are my [snorts] sound effects? Do
24:54you like them? If you do, make sure you
24:56hit the like button. And then it gives
24:57us the voiceover, and then it adds that
24:59to the rest of this.
25:00Now, it's starting to take more and more
25:02shape, as you can see. And from here, we
25:04can either keep it the way it is, or we
25:06can keep on adding more prompts. We can
25:07keep on chaining more prompts to make it
25:10more detailed until we're finally happy
25:12with the end result. Now, remember
25:14everything that I'm talking about in
25:15this video is going to be in the cheat
25:16sheet for you. And if you want access to
25:17that cheat sheet, I did promise you I'm
25:19going to tell you how to get it. So, all
25:20you have to do is drop a comment with
25:22the hashtag Claude course, and let me
25:24know what your favorite part of the
25:25video is so far. I know we're only a
25:27little bit into it, but the little bit
25:29that I've given you so far is 100% going
25:31to be helpful for you to be able to come
25:33up with better prompting skills and to
25:36get better outputs from Claude itself.
25:38Besides that, make sure you hit the like
25:39button. Also, hit that subscribe button,
25:41and let's move on to our next section,
25:43the models. So, as for the different
25:45models, we already talked a little bit
25:47about this earlier. We have Fable 5.1,
25:49Opus 5, Sonnet 5, and Haiku 4.5. Now,
25:53what I told you earlier is pretty much
25:55exactly what we're going to be covering
25:57now, right? So, for Haiku 4.5, this is
26:01essentially going to be the fastest and
26:03the most lightweight model for really
26:05just everyday work. So, an example of
26:08what you can use this for would be to,
26:09let's say, summarize an article and to
26:11give you a brief description on
26:13everything that it covered. Again,
26:15simple, straight to the point, very
26:17lightweight, barely uses any credits.
26:19Then we have Sonnet 5. Here, it's
26:21labeled as most efficient for everyday
26:23tasks. This is really going to be your
26:26balanced and everyday tool that you can
26:28use for things like coding, writing, and
26:31data analysis. And of course, a lot
26:33more. An example for this could be
26:34something like, let's say, coming up
26:36from complete scratch with a video
26:39structure for a YouTube video, or with
26:42building a simple app. Then we have Opus
26:455, which is for more complex tasks. Now,
26:48this is going to be for some more harder
26:51coding tasks. Now, I'm saying coding,
26:53you don't have to worry about actually
26:55using coding in any of this. You don't
26:57do any coding yourself, absolutely ever,
26:59even with Claude Code, which we'll talk
27:01about in a little bit. Claude does all
27:03of that for you. When I say coding, I
27:05just mean straight up building apps for
27:07you. So, with Opus, you can start to
27:09build more complex apps. So, if Sonnet
27:12can build you a basic calculator, Opus 5
27:16is going to build you a scientific
27:17calculator,
27:18>> [snorts]
27:18>> or a calculator that can just take a lot
27:20more things into consideration.
27:22Something that's just a lot more
27:23complicated and more complex. And not
27:25only that, but it can also figure out
27:27what's wrong with something that isn't
27:29working. So, if you built an app, let's
27:31say with Sonnet, and something just
27:33isn't working, no matter how much you
27:34prompt the Sonnet model to fix it, you
27:37can switch over to Opus, and Opus will
27:40most likely figure it out a lot faster.
27:42And then you have Fable 5.1. And this is
27:44really going to be for the toughest
27:46challenges, for the very complex
27:48applications, for building software. You
27:51can essentially give it a small brief on
27:53a project that you want, and Fable 5.1
27:56is going to start building that project
27:58for you, and then it's going to start
28:00checking its own work. Something like
28:02that is something that Opus is not going
28:03to do. Opus will take that brief, it's
28:06going to read through it, and then it's
28:08just going to build it. It's not going
28:09to plan, and it's not going to check.
28:11And if it does, it's going to be
28:12something very quick and something a lot
28:14more streamlined, and of course, not as
28:16detailed. Now, everything that I just
28:18told you, these are not strict rules,
28:20right? You can code with Haiku 4.5. It's
28:24just not going to be as complex or as
28:26detailed as Opus or Fable. And you can
28:29also use Fable or Opus to just simply
28:32give them an article and tell them,
28:34"Hey, what does this mean? Give me a
28:36quick brief or a quick summary on what I
28:39just gave you." And that really just
28:41brings up the age-old question,
28:43"Why don't I just use the most powerful
28:46model?" And the answer to that is
28:48simple.
28:49The extra capabilities doesn't always
28:51mean it's going to be the most useful
28:53model for what you're doing or for what
28:54you're using it for.
28:56I can give Fable 5.1 an article and tell
28:59it, "Summarize this for me."
29:01But for one, it's going to take a lot
29:02longer to do it because it's going to be
29:04thinking through absolutely everything
29:06that it's reading. And two, it's going
29:08to be a lot more expensive because the
29:11higher models are going to be using more
29:12and more credits. So, it's not always
29:15the smartest idea to have it at the
29:17highest possible setting. But then on
29:19the flip side, having the lightest model
29:22isn't always going to give you the best
29:23results because it's not always going to
29:24be enough for what you want it to do, of
29:27course, depending on your project. Now,
29:29you do have different effort levels. So,
29:32this is going to
29:34increase its level of thinking by a
29:35little bit, but it's not going to be as
29:38powerful as, let's say, switching from
29:39Haiku to Sonnet or to Fable. You can put
29:42Haiku 4.5 at, let's say, max, and that's
29:46going to start rivaling Sonnet at about
29:49low. It's not going to be on the same
29:51level, but it's going to be getting
29:52close. So, the effort is going to
29:54control how thoroughly it works through
29:56your problem. So, to show you what I
29:58mean, I'm going to go ahead and take one
29:59prompt for a video idea tracker, and
30:03we're going to build it two times. We're
30:04going to build it once with Haiku, and
30:06we're going to build it once with Opus.
30:08So, you can see what the difference is.
30:09I'm not going to be using Fable because
30:11this isn't a very complex or complicated
30:13product to make. And for the most part,
30:16most users are going to be using
30:17anything between Haiku and Opus.
30:20And once you actually know what you're
30:22doing and you start getting a few
30:24different apps or a few different ideas
30:25under your belt, then you're going to
30:27naturally know when to progress to Fable
30:295.1. So, for this, let's go ahead and
30:32switch it down to Haiku 4.5. And then as
30:35far as the effort, it's going to stay
30:36over here. This one has a little bit
30:37different. This one has extended, which
30:40is essentially extended thinking. So,
30:43while Sonnet, let's say for example, is
30:45going to have low all the way to max,
30:48Haiku is going to have just extended.
30:51It's the most basic one.
30:53So, let's go ahead and end send this
30:54off.
30:55And then while we're doing that, let's
30:56also go ahead and open up a
30:59new chat.
31:01And we're going to use Opus for the same
31:04thing.
31:06For this, we are going to leave it on
31:07high as the default and we're going to
31:08send this one off. Now, the first thing
31:10that you're going to notice is going to
31:10be the speed. Haiku is going to finish a
31:12lot faster, or it should finish a lot
31:14faster. Okay, so Opus already stopped
31:17and it's asking us to publish an
31:19artifact. Let's go ahead and publish
31:20that. Actually, sorry, Haiku. So, the
31:23Haiku one's already pretty much done. It
31:25finished in just a few minutes. So, your
31:27video idea tracker is ready. It features
31:29a dark modern design with these
31:30capabilities. It has ideas, markers
31:32filmed, live counters, delete ideas, and
31:34persistent storage. Whereas Opus 5 is
31:37still thinking through it and it's
31:39working on the problem itself. So, let's
31:41see what this one comes out to. Okay, so
31:44now Opus 5 just finished up and it's
31:46asking us to publish our artifact. And
31:49this is what it gave us. So, let's do a
31:51quick compare and contrast.
31:54So, Haiku 4.5, it worked for about not
31:58even a minute. I want to say maybe 30,
31:5945 seconds. It did all of the code and
32:01it told me it's ready. And this is what
32:03it gave me, nothing.
32:05Like it thinks that it worked, right? It
32:07made the effort, it made code, it did
32:10everything it needed to, but
32:11unfortunately, it just gave me something
32:13that wasn't working. So, now we can go
32:15back to it and say it's not working.
32:17Now, if we check out what Opus made for
32:19us, it worked I want to say for about
32:21two to three minutes.
32:22And this is what it gave us. So, it
32:24actually gave us something that is real,
32:26something that we can actually use. So,
32:28what's the idea for our next video? I
32:30don't know. You let me know down in the
32:31comments below what you want to see in
32:32the next one. But, let's say chat GPT
32:35full e-commerce course, right? We can go
32:37ahead and add the idea. It actually adds
32:39it over here and it updated the one of
32:43five filmed. We can go ahead and delete
32:45this.
32:46It switches down to zero of four filmed.
32:49Let's say I unboxed my best seller or a
32:52best-selling product, something, and one
32:55of four have been filmed. Perfect. Very
32:57straightforward, very simple, and it
32:59works. Something like this is something
33:02that Claude should not be having any
33:05issues creating, but because Haiku works
33:08so quickly, it doesn't actually
33:10double-check its work. But now that I
33:12told it it's not working, it fixed it.
33:15And this is what it gave us. So, it's,
33:18as you can see, a lot simpler. It's not
33:20as complex as the other one. Not that
33:21this one is very complex to begin with,
33:24but it gave us already a few examples
33:26included. And it gave us examples for
33:29both to film and that were already
33:32filmed. Now, if we switch over here,
33:34let's see if this Let's see how this
33:35works. So, full e-commerce course. We
33:39can add this.
33:41Perfect. Okay, so it works. How to start
33:45AI drop shipping, and
33:48how to
33:50make YouTube videos.
33:52Perfect. So, it does work. As you can
33:54see, it's definitely a lot plainer.
33:56And then we can check this off. It shows
33:58that it's filmed, and we still have two
34:00left to film. Delete, and that works.
34:02Okay. So, it worked. But as you saw, it
34:05did take us an extra prompt to get it to
34:08actually work. And the reason behind
34:09that is simple. Haiku 4.5 isn't checking
34:12its work, and it's not necessarily meant
34:14for more complex tasks. Doesn't mean it
34:17can't do it, just means it's not going
34:19to be the best output. So, let's go
34:21ahead and click back to our regular chat
34:23window. Now, one last thing that I do
34:25want to leave you with is that these are
34:27the current models out right now.
34:29Anthropic is constantly coming up with
34:32brand new models. Fable 5 actually came
34:35out I want to say maybe 3 or 4 months
34:37ago and Fable 5.1 is already out. Every
34:39single time that Anthropic comes out
34:41with a new model, another rival company
34:43like let's say OpenAI comes out with a
34:45new rival model as well and then they
34:47just keep on battling each other trying
34:49to see who can win the AI race. So by
34:51the time that you watch this video, it
34:53could be at Fable 5.1 or it could be at
34:55Fable 5.2, Fable 5.5 or even Fable 6.
34:59But just remember everything that I just
35:00told you, the principles of everything
35:02that I just told you and the different
35:03types of models, that's going to be
35:05consistent no matter what different
35:07types of models Claude or Anthropic come
35:09out with. So now that we know how the
35:11different models work and more or less
35:13when we should be using them, let's talk
35:15about the next tab down on the list
35:17which is going to be the projects.
35:20So when it comes to having your
35:21different projects, one major thing that
35:24the project section excels in is being
35:26able to separate your basic chats with
35:29actual well
35:31projects. Now when we talk about
35:32projects, we're talking about specific
35:34areas of work or specific jobs rather
35:37than just a hey, what's the weather like
35:39today or let's say what products should
35:42I sell for Christmas, right? Each
35:44project is going to have its own set of
35:46rules, it's going to have its own tone
35:48of voice and it's really essentially
35:50going to have its own quote unquote
35:52personality. They can have different
35:55branding guidelines, they can have again
35:56different rules and they can also have
35:58different constraints. Now not just
36:00that, but every single project that we
36:01start, Claude can access every single
36:05chat within that project. I'm going to
36:06show you exactly what I'm talking about
36:07in just a second, but when you start a
36:09new project, you can start multiple
36:11different chats within that project and
36:13each chat can be for a specific part of
36:16that project. So let's say you're
36:17starting an online store and you start a
36:19project for an online store. In that
36:21case, you can start a chat that's going
36:23to dictate the tone of voice or the
36:26overall look and feel of the website.
36:29And then you can start a second chat
36:31that's going to help you out with
36:32finding the right products. And then you
36:34can start a third chat that's going to
36:36be to optimize it for SEO purposes and
36:39so on. Now, on my screen right now, I
36:41have all of these different projects and
36:43I'm going to show you one which I have
36:45this one over here. And this is a small
36:46project that I started and it only has
36:48currently two different chats. And the
36:50way that this one is working is here I
36:52have the AI UGC creator project. In this
36:55project, I have specific instructions
36:57which I have laid out over here which
36:59simply says use the Auto DS winning ad
37:01generator skill. We'll talk about skills
37:03in a little bit as the primary workflow
37:05for all e-commerce ad generation tasks.
37:08So, it's telling it to take a specific
37:10skill and apply it to every single one
37:12of the chats in here. So, whenever I
37:13start a new chat, Claude is going to
37:15reference the instructions and then it's
37:17going to go ahead and start the task
37:18that I give it. Now, I did mention that
37:20you can have different tasks or
37:21different chats. In this case, the way
37:23that I was using it is that I have one
37:25that's specifically to make an AI UGC ad
37:28for an eye patch review or for an eye
37:30patch product. And then I have another
37:32one that's for a makeup mirror. So, I
37:35have different products as different
37:37chats. So, each one is a different chat
37:39within the same project and each one is
37:42giving me a different result, but it's
37:43using the exact same instructions to put
37:46everything together. So, essentially I'm
37:48using this project kind of like a
37:50factory where I feed in what I want it
37:52to do. The factory takes the
37:54instructions and composes or compiles
37:56everything that I need it to compile on
37:59what I'm giving it and it gives me the
38:01end result which in this case is an AI
38:03UGC ad. Now, if you do want to see
38:05exactly how I create AI UGC ads, if this
38:08is something that interests you, you can
38:09check out this video right here. In it,
38:11you see how I set up this project as
38:12well. Now, the way that you're going to
38:13start a new project is by going to your
38:15projects tab over here. You can either
38:17click the little plus button or you can
38:19simply just start a new project. You're
38:21going to have a new project button over
38:22here, and then you're going to have new
38:23project here if you don't have any
38:25created. Now, organization and shared
38:27with you, you don't have to worry about
38:29those. That's if you're on a company
38:30Claude account. So, for you, you're only
38:32going to be seeing your projects and new
38:34project. Now, when we create a new
38:35project, it's going to ask us what are
38:38we working on? This is just simply going
38:39to be the title. This is going to be for
38:41our reference. So, let's say something
38:42like an ad script generator. And then,
38:44what are you trying to achieve? Again,
38:46this is going to be just for descriptive
38:48purposes. This is not going to affect
38:50the project itself. This is not the
38:51instructions. We will be setting those
38:53after we create the project. Now, one
38:55glitch that I've found on Claude is that
38:57if you leave what are you trying to
38:59achieve empty, sometimes you're not
39:02going to be able to be able to click
39:03into the project to then use it. So, I
39:05suggest you always fill this out. It's
39:07something that I found. It's happened to
39:08me. I don't know if it's happened to
39:09anybody else. If it has happened to you,
39:11let me know down in the comments below.
39:12So, just as a quick example, we want to
39:13create viral YouTube videos using proven
39:15script guidelines. Now, down here, we
39:17can go ahead and click on use a folder,
39:19and then you can choose a folder that
39:21has any relevant files on your computer
39:23that you want this project to be able to
39:25access. Now, this doesn't have to be
39:28used. You don't always have to give it a
39:29folder. But, if there's any assets that
39:31you have on your computer that will be
39:33helpful for it, or something that you
39:34want to keep using within that project,
39:37like let's say take for example, if
39:39you're creating something like a
39:40PowerPoint or a Google Slides
39:41presentation, and you have specific
39:43images that you want added to that
39:45presentation, you can add those to a
39:47folder, give Claude access to that
39:49folder, and it'll then be able to use
39:50those assets in the presentation. But,
39:53for this, we don't need anything. And
39:55then, as for visibility, you don't have
39:57to worry about this one. This is for the
39:58most part going to be just available to
40:00you. Again, I have a company account, so
40:02this one is not going to be relevant.
40:04So, if we can click on create project,
40:06we are then going to have our project
40:08dashboard over here. Now, from here, we
40:10have our chat window. We can start our
40:12chat. We can either do a basic chat
40:14where we just to to Claude, and we don't
40:16use too many credits, or we can switch
40:18it over to Claude Work, and we can start
40:20to change the effort level and the model
40:22that we're using, and then it's just
40:24going to make it a bit more advanced.
40:25Besides that, we also have down here the
40:27ad script generator. This over here is
40:29the project that we're currently working
40:31on, which is this one over here. You're
40:32going to see that by default, and then
40:34you're going to have the different
40:35folders if you want to add any. Again,
40:37you have your manually and your
40:38automatically approve options. You have
40:40your chat and tasks history, the
40:42activity, and then you have the
40:43instructions to add that will apply to
40:45every chat that you start after you add
40:47those instructions, and you also have
40:50any context. Now, the context is
40:52essentially, in a way, the same as the
40:54instructions. It's something that Claude
40:56can reference, but instead of you simply
40:59just writing out what you want it to do,
41:01if you click on here, you'll be able to
41:03add either text context, or you can also
41:06include a file. Like, let's say if you
41:08have a PDF with written instructions,
41:10you can include that on here, as well.
41:12Now, I already have a set of
41:13instructions for Claude over here. This
41:16is an entire YouTube script guideline,
41:18but I'm not going to use this just yet.
41:19So, just to show you the difference, I'm
41:21putting "Give me a 5-minute YouTube
41:22script or structure for a Claude course
41:24video." So, we're going to compare this
41:25to one with instructions. Okay, so now
41:28it's just asking me a couple questions.
41:29What are we teaching? Let's just do
41:31Claude Basics 101. "Claude beginners to
41:33sign up, existing users to activate,
41:34general YouTube growth." So, who's
41:36watching this? Let's do Claude beginners
41:38that want to sign up, and what format do
41:40we want? Let's just do both of these.
41:42So, after we answered the few different
41:44questions that it asked us, Claude
41:45actually started working, and it's
41:47figuring out what it needs to do. It
41:49doesn't have any extra context. It
41:51doesn't have anything that's pointing it
41:53in the right direction of, let's say,
41:55what would make a good YouTube script.
41:57It's simply just looking online, trying
41:59to figure out what sources it can use,
42:02what it can use as a guide, and it's
42:04coming up with everything together. So,
42:05that's why it's taking a little bit
42:06long. So, in just a minute, we'll redo
42:08the same prompt, and we'll be giving it
42:10our project instructions and see what
42:12the difference is. Okay, so after a
42:13couple of minutes it gave us this over
42:15here, which is to publish an artifact.
42:17Now, I didn't expect it to publish an
42:18artifact. I just thought it was going to
42:20give us an entire YouTube script, but
42:22that's fine. It says here beat sheet
42:24plus the full word-for-word voice-over.
42:26Okay, so it's giving us a structure,
42:28which is the beat sheet. I've never
42:30heard of that.
42:31It gives us our cold open, the mental
42:33model,
42:34and it's giving us the actual voice-over
42:37over here. Now, this doesn't really have
42:39much structure, right? It has a promise.
42:42It has the one sentence they'll remember
42:44for a week. That's interesting. Bad
42:47prompt versus good prompt, the reveal,
42:49generic AI becomes their AI, retention
42:51beat, and one specific instruction.
42:53Okay, I mean it is pretty quick cuz it's
42:54a 5-minute video, but overall as I'm
42:57looking through it and as I'm quickly
43:00skimming through everything I'm reading,
43:01it really doesn't have much structure.
43:03Now, if we go back
43:05to the ad script generator, we click on
43:07instructions and we paste in our
43:09instructions here and save them.
43:11Now, let's take the original prompt that
43:13we gave it.
43:14Just so we can have everything exactly
43:16the same.
43:17We have our instructions ready.
43:19Let's paste that in here. Everything
43:20else is the same. We have Opus 5 high,
43:22manual, everything is good. Okay, so we
43:24go ahead and submit it. And now let's
43:25see what it gives us. Okay, so it's
43:27asking us again one more time just what
43:29the video specifics are. So, for this
43:30I'm going to do a Claude 101 intro.
43:34Total beginners full script, total
43:35beginners outline, the full script, or
43:37already using AI. So, for this I am
43:39going to go with the outline rather than
43:42the actual script. And now it's giving
43:43us everything directly in the chat. And
43:45it's giving it to us in a more
43:46structured manner, which is what the
43:48instructions told it to do. So, in the
43:50instructions it tells it to give hook
43:53and introductions. We have our hook over
43:55here.
43:56We then have a structure and substance
43:59and demonstrations.
44:01So, that's showing over here what Claude
44:04is actually good at. And it's also
44:05giving us a little timing over here.
44:07We have the interface only what matters.
44:09That right there is going to be the
44:11substance and the demonstrations. We
44:13have a credibility section, which is
44:14what the instructions are calling for,
44:16and that's going to be the demo over
44:17here along with the two mistakes. And
44:19the instructions also called for an
44:21output format. And the output format was
44:24what it asked us in the beginning. Do
44:25you want everything to simply be a
44:26voiceover? Do you want it to be a
44:28script? Or do you want it to be an
44:30outline? And that's what we cleared up
44:32at the beginning. We told it we wanted
44:33this to be an outline. So, this right
44:35here is clearly a lot more structured,
44:38and it fits more Well, it actually fits
44:40into the instructions that we gave it.
44:43So, when it comes to projects, the one
44:44thing that I really want you to take
44:45away is that each project is going to be
44:47its own entity. Each project is not
44:50going to interact with any other project
44:53or with chats outside of that project.
44:55So, now that you saw how projects work,
44:57we can move on to the next section,
44:59which is going to be what Claude
45:00actually created for us in this project.
45:02And so, we're going to move on to
45:04artifacts. Now, when it comes to
45:06artifacts, you already saw a few
45:07different examples early on in this
45:09video. But now we're going to dive a bit
45:11deeper into them. So, when it comes to
45:12artifacts, really what it is is simply
45:16an app or an asset that Claude makes for
45:20you. As you saw earlier, our YouTube
45:21script, that was an artifact. Why?
45:24Because it created an HTML document for
45:27us or it created a type of website. Even
45:29though it's not live online that people
45:31can simply go to this website, an
45:33artifact is going to be any type of app
45:35that uh Claude creates for us. And these
45:37apps can also be published and be given
45:40public access, so other people can
45:41access them as well. Now, just taking
45:43this same one that we saw earlier into
45:45consideration just as a quick example,
45:47if we click up here, we have a few
45:49different options for our artifact. We
45:51can either rename it, we can duplicate
45:52it, we can share it so we can send it to
45:54somebody else, we can pop it out from
45:56this chat. If we do that, all it really
45:58does is it just pops it out in its own
45:59little chat like this. We can move it
46:02back to the chat if we click on that
46:03little cloud icon.
46:04We can refresh it if we want to. We can
46:06pin it for quick access. We can report
46:08it if there's something wrong with it.
46:10And over here on the side, we can show
46:12any comments. These are going to be
46:13comments that either we put for Claude
46:15to have it change anything.
46:17And over here we can share it. So if we
46:19share this, we can go ahead and send
46:21this to specific people. And we can also
46:23change the privacy of it. So we can show
46:25this only for people we invite or anyone
46:27with the link. Now if we look over here
46:29on the side panel under artifacts, we
46:30can click on that. Now here we have a
46:32few different ones that I created
46:33myself. So we have something like let's
46:34say a mobile app layout. We have a
46:37design system. This design system is an
46:39HTML document really that's going to
46:41give different design guidelines or
46:43branding guidelines. We can click into
46:45it. And then it takes us to Claude
46:46design. Claude design, we're going to
46:48get to a little bit later on in this
46:49video, so make sure you stick around.
46:51But this is all of the information for
46:53our design system. It looks a little bit
46:54complicated, but don't worry about that
46:56because we actually don't have to look
46:57through any of this or read it really.
46:59But this is all just something that
47:00Claude uses. But if we go back, let's go
47:02ahead and check out another one that I
47:03have here. So if we click into this
47:05setup checklist. Now here what it's
47:06giving us is essentially a checklist in
47:08the form of again an HTML file. So here
47:11we have what I needed to do in order to
47:12be able to launch or ship this app. Now
47:14here we do have a couple of other
47:16options to create different artifacts as
47:19well. So we have docs or different
47:21documents. We have slides like I
47:22mentioned earlier like Google slides or
47:24PowerPoint presentation. And design,
47:27which is done directly by Anthropic or
47:29by Claude. But let's do something quick.
47:31Let's do something pretty easy. And
47:32let's simply tell Claude, we can keep
47:34this over in chat right now. Still keep
47:36this on Opus 5. Let's just say build me
47:38a simple calculator app. Now while
47:40that's running, let's go ahead and start
47:41a new chat. Let's go ahead and ask it
47:43for a PDF invoice generator. And let's
47:45do one more which is going to be a
47:46little bit more complicated. So create
47:49an online store breakdown artifact that
47:50will take the link of a store and give
47:52us the good and bad parts of it. Like
47:54what would need to be optimized.
47:56Stephanie misspelled optimized.
47:58And let's send this one as well. Now our
48:01calculator app is ready. This was
48:02actually really fast. It's pretty
48:04simple. And here we have the artifact.
48:06So we can go ahead and click on it and
48:08there we have our little calculator.
48:09That's all there is to it. So 25 * 2 50
48:13perfect. Everything works. 50 minus 49
48:18one perfect. Okay, so this is good. So
48:19the artifact was created. Now we can go
48:22ahead and share this. We can click on
48:24anyone who has link or who has a link.
48:26We can share publicly copy link and if
48:28we open in incognito mode
48:31it'll take us to the artifact. This is
48:32how anybody can access to the different
48:34artifacts that we create. And again, it
48:37works the exact same way. Then let's
48:38check on one of the other ones. Our
48:40store breakdown is still being created.
48:43But our invoice generator already
48:44finished. And as you can see this was
48:46created as an artifact as well. Why?
48:48Because this is a app that we asked it
48:50to create. This is something that has a
48:52function that we asked it to create and
48:54something that it needs to create
48:55externally. This isn't something that it
48:56can give us within the chat that we can
48:59use. So if we click on open here, well
49:01this is actually really good. I might
49:02end up using this for myself. So let's
49:04go ahead and
49:06John
49:07[email protected].
49:09Everything updates live on here. I like
49:11that. 58 58 North
49:1558 Street.
49:16That also updates. We can build to
49:19John Walker. Can add all of the
49:21different informations. Let's say what
49:23did we give them? We gave them a pod
49:25artifact PDF generator. And we're
49:28charging him
49:29$500 for this. If we add another line,
49:31let's see if that works too. Website
49:33design
49:34$58. Perfect. So now if we download a
49:37PDF, it should in theory download a PDF.
49:42And it did. Perfect. Now again, we can
49:44go ahead and take this.
49:46We can copy this link.
49:49And switch it to anybody who has a link
49:50so share publicly so we don't get that
49:52last error that we got for the sign-in
49:54message.
49:56And anybody can use this PDF generator.
49:58Perfect. So, now that you saw how easily
50:00you can create different types of
50:01artifacts and share them with anybody,
50:03and even publish them online, let's move
50:05on to our next section, which is going
50:07to help us create some of these
50:08different artifacts, and really just
50:10optimize any niche work that we're doing
50:12on Claude using skills. Now, what
50:15exactly are skills? The easiest way to
50:18put it is simply a playbook that you
50:21write once, and it just keeps working
50:23over and over. So, it's like a set of
50:25instructions that you can call upon
50:28whenever you need those instructions, or
50:30when you need your project or your
50:32prompts to go a certain way. Now, when
50:34it comes to your skills, the way that
50:35you're going to access them is through
50:36the customize section over here. So,
50:39you're going to click on the sidebar
50:40under customize, you're going to have
50:41skills, connectors, and plugins. So,
50:45what's the difference? Well, under
50:46skills, you're going to have the set of
50:48instructions that Claude is going to
50:49reference. A connector is going to be a
50:51way that you can connect Claude to any
50:53other tool. So, in this case, as you can
50:55see, I have it connected to AutoDS, have
50:57it connected to Higgs Field, Shopify,
50:59and a few other apps. And a plugin is
51:01simply like a type of app or a package
51:04of different types of capabilities that
51:06you can install directly onto Claude.
51:08So, for this, we're going to stick to
51:09skills, and under skills, you can create
51:12your own skill, or you can simply search
51:14the different skills that Anthropic has
51:16available to you.
51:17Now, for this, all you have to do is
51:20under skills, just go ahead and click on
51:21add, and you can either upload a skill,
51:23you can create a skill directly through
51:25Claude, or you can create a skill
51:27directly on here, or again, you can
51:29create one with Claude. So, you can
51:31create a skill by text, you can paste in
51:33all of your instructions, or under add,
51:35you can also create with Claude, which
51:37is going to bring up the chat where it's
51:39going to start with, let's create a
51:41skill together using the skills skill.
51:44So, Claude actually has a skill to
51:46create skills. Now, under skills, if we
51:48click on yours, you're going to see all
51:49of the different skills that we've
51:50created or any that we've uploaded. And
51:53then if you click on discover, you can
51:54see any other skills that are already on
51:57the Claude or the Anthropic website or
51:59on their database that you can simply
52:01add to your Claude chats. Now, from
52:03here, if you find any that you like, so
52:05let's see if we could find any Let's do
52:07a files and documents. Let's see what
52:09they have here. They have a PDF viewer,
52:11they have a design. This is directly by
52:12Anthropic, all of these are or the
52:14majority of them are. Some you have by
52:15different partners like let's say Adobe.
52:17And you also have a quick description as
52:18to what each one of them are. So, let's
52:20say Adobe for creativity. It brings
52:22together Adobe Creative Cloud tools for
52:24images, vectors, designs, and video. So,
52:26if you click on this, you can edit
52:28multiple assets at once, adapt for
52:30different platforms, and complete
52:32multi-step creative workflows for fast
52:33results. So, this actually is pretty
52:35cool because this seems like what it can
52:37do is that you can add this skill and
52:39it'll be able to use your Adobe tools
52:41like let's say Photoshop to optimize or
52:43edit your different images. So, this is
52:45pretty interesting. This is one that I
52:46didn't know about. Now, if you do want
52:48to add this to your skills list, all you
52:49have to do is click on add
52:50and that's it. You can either turn it
52:52off like this or simply turn it back on.
52:55Now, let's run back to our basic skills
52:57page over here. Let's go ahead and click
53:00on add and let's create a skill. Now,
53:03this is going to be a messy notes to
53:07checklist skill. The description, create
53:10a checklist from messy notes. Something
53:13simple and straightforward. Now, on
53:15here, this is where we're going to add
53:16our instructions for this skill. Now, to
53:18do this, I suggest you make one that's
53:20very detailed and it's going to be kind
53:22of hard if you write everything out
53:23yourself. So, the easiest way to do this
53:25is going to be to actually create a
53:27skill using Claude. So, that way it gets
53:29every single point taken into
53:30consideration and everything that you
53:32want it to do. Your prompting versus
53:34AI's prompting or Claude's prompting,
53:37there's going to be a huge difference.
53:38Your prompting can be very good, but
53:40when it comes to Claude, it actually has
53:42a skill to create these skills. So, it's
53:44going to know exactly how to structure
53:45the wording, how to structure the skill,
53:47and how to structure everything so that
53:48way you get the end result that you
53:50want. So, because of that, let's
53:51actually make this a lot easier. Let's
53:52go ahead and take this. Let's run this
53:54out.
53:55And let's discard this. And then let's
53:58do add create with Claude. And let's
54:00create a skill together using the skill
54:01creator skill. First, ask me what the
54:03skill should do.
54:04Let's go ahead and do that. So, it gave
54:06me this. What should the skill do? A few
54:08things that would help me build it. What
54:10task it handles. So, what's it going to
54:11do? In this case, I'm telling it, I want
54:13the skill to take my messy notes and
54:15make them into a checklist. Then, when
54:17it should trigger, what you'd say that
54:20should invoke it. So, we actually have
54:21to tell Claude use the skill. And in
54:23order to do that, we give it essentially
54:25just a little keyword or a trigger word.
54:27And in that case, it's going to be messy
54:28notes to trigger it. And what it
54:30produces. A document, a plan, a set of
54:32edits, or a checklist. Well, we already
54:33have here that we want a checklist.
54:35So, we're pretty much good to go from
54:37here. Let's go ahead and send it off.
54:39Now, where would the messy notes usually
54:40come from? They can either be pasted
54:42into the chat, a file that I point to,
54:44or both. I would say both, actually.
54:47That makes a lot of sense because
54:48sometimes you could just be writing
54:49something out, and then you could feed
54:50that. Now, how should the checklist come
54:53back as? Directly in the chat, that's
54:55going to be the fastest. No file is
54:56created. A task artifact, or a markdown
54:59file. A markdown file is going to be
55:01simply just un- a file that you can
55:02download. So, for this, I think the
55:05right in the chat is going to be the
55:06easiest one and the fastest one. Now,
55:09how should it treat notes that aren't
55:11actions? Separate a notes section, drop
55:14them, or turn everything into items. So,
55:17I think we can do turn everything into
55:19items, and then anything the checklist
55:21should always capture per item. Just the
55:23action, the action owner, due date when
55:26mentioned, action and priority grouping.
55:28Let's do a due date when mentioned. Now,
55:30Claude is going to go ahead and invoke
55:32its skill skill, and it's going to
55:33create this skill. Okay, so after After
55:3515 minutes, it created the entire skill
55:38and over on the right-hand side, you see
55:40the skill.md.
55:41Now, MD is simply the file type that
55:44skills are created in, so just so you
55:46know. And then over here under the
55:47description, you have all of the
55:49different instructions that Claude
55:51created to make this skill work. So, as
55:54you can see, it's pretty detailed. It's
55:55very organized. It looks like something
55:57that I probably would have taken an
55:58entire month to write myself. And then
56:00over here on the left-hand side on our
56:02chat, we have the instructions on how to
56:04use it. So, pasted client call notes
56:06with a name and a deadline buried in
56:08them. The three test prompt Okay, so it
56:10ran a few test prompts and this is what
56:12it what it did. So, all we do is simply
56:13click on save skill if we wanted it to
56:16just go live.
56:17And if we click on manage down here
56:20under skills or if we go to customize,
56:23click on skills,
56:25we can look for our messy notes to
56:28checklist skill, this one right here.
56:30Now, if you look over here under the
56:31description, it says messy notes. That's
56:33the phrase that's going to trigger it.
56:35So, let's go to new and let's do
56:37messy notes.
56:39And then let's just type in a bunch of
56:41different notes, right? So, here I have
56:43ton of notes, no structure or anything.
56:46Let's go ahead and send it off. Over
56:47here, it should say that it's going to
56:49invoke the messy notes skill. There you
56:51go. So, it's reading the skill and
56:54there you go. So, look at all of these
56:57crazy notes over here,
56:58completely unstructured, just typed in,
57:01and it gives us this over here. So, we
57:04can totals. This is how much we sold,
57:06and then it gives us a to-do list right
57:08at the end where we can go ahead and
57:09check these off. Now, we can't check
57:11them off directly in here, but we can
57:12copy all of this and put it into like a
57:15Word document or something like that
57:16where we can go ahead and start checking
57:18these off. Now, if we didn't have that
57:20skill, let's see what it would look
57:22like. So, if we just type this in like
57:24that, I'm actually let's turn off the
57:26skill first.
57:32Turn that off.
57:33And then we just tell it to organize
57:35these notes.
57:37Let's see what it gives us. So, it kind
57:38of does give us more or less what we had
57:40asked it before, but it doesn't give us
57:41the actual checklist, which is what we
57:43wanted originally. So, the whole point
57:45of the skill previously was to make a
57:47checklist, and that's what it gave us.
57:49In this case, it didn't do that. Now,
57:51there are lots of different skills that
57:52you can create. One of the different
57:53skills that I created, that's actually
57:55one of the most useful ones, was a skill
57:56to create AI UGC videos. And again, you
57:58can check out the video I referenced
57:59earlier, but that set of instructions
58:01will take our raw prompt, and in that
58:03skill, it tells it to connect to
58:05multiple different services, like let's
58:06say Higgs Field, to generate the assets,
58:09and then put everything together to come
58:10up with an actual AI UGC video. Now
58:13again, if you want to check that one
58:14out, just check out this video right
58:15here. But then that brings me to our
58:17next section, which is going to be our
58:19connectors. Moving on to connectors. So,
58:21you already saw how we can access the
58:22connectors. All we have to do is go
58:24ahead and click on customize, and then
58:26click on the connectors tab over here.
58:28Now, this is a list of all of the
58:29different connectors that I have
58:31currently available to me that I've
58:32used, but if we click on discover, we
58:35can find lots more connectors that we
58:37can add to our Claude arsenal. Now, what
58:41exactly is a connector? So, a connector
58:43is a way to be able to connect your
58:46Claude account to any other app that
58:48supports it. So, for example, I can
58:50connect Claude to Shopify, and I can
58:52tell Claude, "Go to my store, and give
58:55me the last 10 sales that I've had. Or
58:58how much money have I made in the last
58:59month? Or what are my top-performing
59:02products?" I'll show you exactly how
59:03this works in just a second, but that's
59:05just a general overview of how a
59:07connector works. It essentially gives
59:09Claude access to different tools and
59:12different services that it can't offer
59:14on its own. One perfect example is image
59:17generation. Claude cannot generate
59:20images on its own. It can generate
59:21certain shapes, or it can generate
59:23things like SVGs. Now, really quick, for
59:25those of you that don't know, an SVG is
59:26kind of like an image type, except it's
59:28a different file format and it doesn't
59:30really work or behave the same way that
59:31images do. So, Claude doesn't have a
59:33native ability to generate images or to
59:35generate videos. So, what we can do in
59:37this case is we can connect it to a
59:39separate app like Higgsfield. And to do
59:42this, it's very simple. Most websites or
59:45most tools that have or offer an MCP
59:48support or a connector support are going
59:50to have this over here, which is called
59:52MCP. Now, an MCP all it really is a web
59:56address that you can give your Claude
59:58account or you can give your Claude
1:00:00assistant to connect to the service. Of
1:00:02course, you are going to have to have an
1:00:03active subscription for something like
1:00:05Higgsfield, but this is how you're going
1:00:06to be able to connect it. So, for one,
1:00:08you go to the website and you look for
1:00:10anything that has to do with MCP or
1:00:12Claude support. And then in this case,
1:00:15for Higgsfield, we have MCP, which is
1:00:18this right here. This is a web address.
1:00:20All you have to do is go ahead and click
1:00:21on this, copy the MCP URL, then go back
1:00:24to your Claude account. We can click on
1:00:26add, add custom connector. We'll go
1:00:28ahead and name it. So, Higgsfield, and
1:00:30then we type in the web address here. Go
1:00:32ahead and click on continue. It's
1:00:34already existing on my end, so that's
1:00:35why I can't re-add it. But, that's
1:00:37pretty much all you have to do. After
1:00:38that, it opens up a browser window where
1:00:40you simply log in and you're good to go.
1:00:43Now, you can actually call this
1:00:44connector and you can use it. But,
1:00:45before we do that, let's go ahead and
1:00:47just look at the ones that are currently
1:00:48on Claude already that you can simply
1:00:50just install without having to do any
1:00:52extra steps like connecting the MCP. So,
1:00:54under discover, we have things like
1:00:56Google Drive, we have Gmail, we have
1:00:58Google Calendar, we have Canva. And if
1:01:01you keep scrolling, you have a lot of
1:01:03other different types of plugins. Now,
1:01:05this one right here, Sonos, that's
1:01:06pretty interesting. I can control my
1:01:08entire home theater system through
1:01:10Claude. That's pretty crazy. And I never
1:01:11even thought about that. But, if we want
1:01:13to add, let's say, our Google Calendar,
1:01:15we can give Claude access to our Google
1:01:17Calendar by simply clicking on this
1:01:18little plus sign over here, and then
1:01:20it's going to take us to a separate tab.
1:01:22We can go ahead and continue connecting.
1:01:24We choose the account that we want to
1:01:25connect to it.
1:01:26Continue one more time.
1:01:29Here you do want to select all so it can
1:01:30see everything on your Google Calendar
1:01:32or whatever app you're connecting it to.
1:01:34And just like that we connected two
1:01:35different apps through two different
1:01:37connectors. So if we click on new, we
1:01:39can simply just tell Claude generate an
1:01:40image for me of a flower and a mall. I
1:01:43don't even have to tell it to use Higgs
1:01:44Field. It's already going to know to use
1:01:45that because that's the only way that it
1:01:47can create it. So let's go ahead and
1:01:49send that off.
1:01:50You're going to see here it's going to
1:01:51start to look for how it's going to do
1:01:53it and then it's going to use Higgs
1:01:54Field. Right there. Higgs Field generate
1:01:56image. So it's going to start to
1:01:57generate the image for us. It's going to
1:01:58ask us for some permissions. Let's go
1:02:00ahead and allow. And now while it's
1:02:02doing that, let's go back to new and
1:02:04let's see what meetings I have in the
1:02:05next week. So
1:02:07what meetings do I have this coming
1:02:09week?
1:02:10And it's going to check our calendar.
1:02:12It's going to connect to the Google
1:02:14Calendar through the MCP connection or
1:02:16through the connector that we set up
1:02:17earlier. Perfect. So it's giving us our
1:02:20week. On Monday we have a daily meeting.
1:02:22We have the same one for Tuesday. And we
1:02:24have a few other meetings throughout the
1:02:26week. That's all we needed to do. Now if
1:02:28we go back to the other one, you're
1:02:29going to see here that it's currently
1:02:30generating and processing our image. So
1:02:32it called Higgs Field and within Higgs
1:02:35Field it's using the GPT image 2.5 model
1:02:38to do this. There's lots of different
1:02:39models. Claude is going to choose the
1:02:41best one, the one that it thinks is the
1:02:42best or the one that's most relevant. If
1:02:44you want to change it, if you want to
1:02:45give it specific instructions to use,
1:02:47let's say Nano Banano, you can do that
1:02:49as well. And the same thing goes for
1:02:50your weekly meetings. I can simply just
1:02:52tell it reply yes to all.
1:02:54And it'll reply yes to all of my
1:02:56meetings to let that everybody know that
1:02:58I'm going to be attending all of them.
1:03:00Here it's going to ask us again for some
1:03:01permissions. This is going to be to
1:03:03allow it to actually respond to those
1:03:05calendar invites. Now it keeps asking me
1:03:07to all of these to give it access
1:03:10because I have it set to manually
1:03:11approve. If I do automatically approve,
1:03:13it would have just done it all for me.
1:03:14But perfect. Accepted all five
1:03:16outstanding invites and the whole week
1:03:18is now confirmed. Good to go. Give me an
1:03:20update. And while it's giving us an
1:03:22update, let's check out the flower
1:03:23image. There we go. So, here we have our
1:03:25flower growing in the mall in between
1:03:28Sephora and Zara. That's pretty cool. I
1:03:29didn't think it would give us those two
1:03:30actual real stores. Now, from here, if
1:03:32we want, we can go ahead and download
1:03:33it, recreate it, or whatever else we
1:03:35want. But then under weekly meetings, we
1:03:37have all five RSVPs went through.
1:03:38Everyone on your calendar this week now
1:03:40shows you as accepted. Perfect. We can
1:03:42go ahead and verify that. And here we
1:03:43have accepted accepted. These two are
1:03:46accepted and so is this one. This one
1:03:48does not show as accepted. Actually, it
1:03:49does show that I did accept. So, now
1:03:51that we have a good understanding on how
1:03:53skills and connectors work, we can move
1:03:54on to our next section, which is going
1:03:56to be web search. Now, let's talk about
1:03:58web search because this is probably one
1:04:00of the most important things that Claude
1:04:02actually needs to have turned on for it
1:04:04to be efficient. If you don't have it
1:04:06turned on,
1:04:07then you're not going to get the most
1:04:08up-to-date, if any up-to-date, pieces of
1:04:10information. So, let's start a brand new
1:04:12chat. And to start, let's go ahead and
1:04:14have web search turned off. I'm going to
1:04:16show you one of the most basic examples
1:04:19that we can use for this. Actually, you
1:04:20know what? Let's turn it on. So, let's
1:04:22just ask it, "What's the weather like
1:04:23today?" It's going to think. It's going
1:04:25to search the web. And it's going to
1:04:27take my location and it's going to say,
1:04:29"In Miami, today it's hot and humid
1:04:30around 83 to 87° F." Perfect.
1:04:34Now, if we start a new chat and we turn
1:04:37off that web search and we ask it the
1:04:39same exact question, "What's the weather
1:04:40like today?" it's not going to be able
1:04:42to answer. Why? Because it can't search
1:04:44the web to find the information that it
1:04:47needs. It's being restricted to its own
1:04:49environment, to its own database, to its
1:04:51own knowledge base that it has within
1:04:53our computer. So, that's why it says, "I
1:04:55can't check current conditions from
1:04:57here. I don't have like data and I don't
1:04:59know where you are. If you tell me the
1:05:01city, I can at least give you a sense of
1:05:03what's typical for late September." So,
1:05:05even if we tell it where we are,
1:05:07if we tell it, let's say
1:05:09we're in Miami, it's not going to give
1:05:10me accurate information. It's going to
1:05:12tell us more or less what it's like
1:05:14around this time of the year, which is
1:05:16pretty close to what it actually is, 88
1:05:17to 89° Fahrenheit, but it's not the
1:05:20actual current weather. Now, this isn't
1:05:22going to restrict it completely from
1:05:24making logical answers or logical
1:05:26decisions. We can still ask it, let's
1:05:28say, basic math questions, right? So, so
1:05:30what's the square root of 100? It's
1:05:32going to think about it, it's going to
1:05:33do the equation, and it's going to give
1:05:34us the answer, 10. But this is something
1:05:36that doesn't require web search. This is
1:05:37something that I can simply do based on
1:05:39its own intelligence. Now, another
1:05:41example, let's say, is something that's
1:05:43related to the news or current events.
1:05:45So,
1:05:46what happened
1:05:48yesterday in
1:05:51Washington? Right now, web search is
1:05:53turned on, searching the web, and now
1:05:55it's assuming that I'm talking about
1:05:57Washington D.C. That's fine. The main
1:05:59story out of the capital on Sunday was
1:06:00President Trump's announcement about his
1:06:02proposed triumphal arch. Okay, perfect.
1:06:05Now, if you go back to new, paste in
1:06:07that same exact question, and turn off
1:06:09web search, it has no way to know, and
1:06:11it's telling us right there, "I don't
1:06:13have any way to know. My knowledge runs
1:06:14to the end of May 2026, and I don't have
1:06:18web search turned on in this
1:06:19conversation, so I can't see anything
1:06:21from yesterday." So, it does have a
1:06:24certain amount of history that it can
1:06:26reference up until, in this case, May
1:06:282026. Now, my suggestion to you, always
1:06:31keep web search turned on. Don't turn it
1:06:33off. There really is no reason to have
1:06:34it turned off, and if you want the most
1:06:36accurate information, the most
1:06:37up-to-date information, just always
1:06:40leave this by default turned on. So, now
1:06:42that we know how web search works, what
1:06:44it does, and how you can use it, and how
1:06:46Claude uses it, let's move on to our
1:06:48next section, which is going to be
1:06:49Claude co-work. This is one of my
1:06:51favorite ones. Now, let's go into Claude
1:06:53co-work. And Claude co-work is pretty
1:06:55straightforward. The way that you use it
1:06:57is simply by giving access to your
1:06:59computer or to certain files to Claude,
1:07:01and letting it really just do its thing.
1:07:03So, for this, what we're going to be
1:07:04doing is we're going to be taking two
1:07:06different examples. The first one is
1:07:08going to be by giving it access to our
1:07:09downloads folder. It's going to ask us
1:07:11if Claude can change files, go ahead and
1:07:13click on allow or always allow. We're
1:07:15going to change that down here after we
1:07:17click on co-work. We're going to switch
1:07:18it in the little tab down here and we're
1:07:20going to choose the downloads folder.
1:07:22Now, if that's not there for you, you
1:07:23can give it access to any folder that
1:07:25you want to just by simply clicking on
1:07:27add folder
1:07:28and giving it access through here. So,
1:07:30right now if we look at my downloads
1:07:32folder over here, I have it full of just
1:07:35a bunch of random videos, images, files.
1:07:38So, what I want to do is just clean this
1:07:40all up. Now, I do have a lot of
1:07:42different folders on here as well. So,
1:07:43I'm going to go ahead and tell Claude to
1:07:45organize this for me. So, go to my
1:07:46downloads folder and clean up all of the
1:07:48files, please. There are a lot of
1:07:50folders on there, too. If the loose
1:07:52files are relevant, move them into those
1:07:54folders. So, if there's an images
1:07:56folder, take any of the images and move
1:07:58them into there or the loose images and
1:08:00move them into the folder. Now, for this
1:08:02I'm going to switch this over to
1:08:02automatically approve so that way it
1:08:04could just get it done and send it. Now,
1:08:07what I'm going to do is I'm going to
1:08:07open up my downloads folder so you can
1:08:09see
1:08:10exactly what's going to happen and how
1:08:12fast it happens. You can see back here
1:08:14in the background it's starting up and
1:08:15it's working on it. Just give it a
1:08:16couple seconds. If we move this over,
1:08:18we're going to see that it's looking
1:08:19into the existing folders to understand
1:08:21the organization scheme, if there really
1:08:23is any organization in here. And it has
1:08:25so many loose files, mostly videos,
1:08:27images, and 3D files. Oh, it's asking us
1:08:29for a question. And 20 plus existing
1:08:32folders. So, how should it sort the
1:08:34loose files into the folders? By file
1:08:36type, by project where obvious, or by
1:08:38type but keep recent files out. So,
1:08:40either by directly file type, so just by
1:08:42images, videos, documents, and 3D
1:08:45models, or by projects where it's
1:08:46obvious. So, if I have a project for,
1:08:49let's say a specific app that I'm
1:08:51working on, it'll move those files into
1:08:53that app because it'll know that those
1:08:55files belong to that app. Now, this is
1:08:57pretty good because it is by where
1:08:59obvious. So, So, to do that, but I'm
1:09:00going to tell it, let's do by project or
1:09:02obvious, but anything else by file type.
1:09:04Now, it found several duplicates. Ooh,
1:09:06okay, what do we want to do with those?
1:09:08Move them to a to delete folder. Let's
1:09:10go ahead and do that so I can review it
1:09:12myself. And should I only tidy the messy
1:09:14folders themselves or only the loose
1:09:16files? Let's just do loose files only.
1:09:18And now we're going to see that
1:09:19everything starts to just disappear. And
1:09:22there we go. It just organized
1:09:23everything for us. Okay, so if we look
1:09:25through everything, we have
1:09:27the data over here. This is a new folder
1:09:31with just a bunch of JSON files. We have
1:09:34audio files.
1:09:36We have different installers and a bunch
1:09:39of different images in our images
1:09:40folder. Perfect. So, it did what we
1:09:42needed to do. Now, let's take it a step
1:09:43further and let's tell it to So, in my
1:09:46downloads folder, check a folder called
1:09:47receipts and make all of those receipts
1:09:49into a spreadsheet on Google Sheets. So,
1:09:52now I have a bunch of different receipts
1:09:53in this little folder that I'm giving it
1:09:54access to and I want to see exactly what
1:09:57my breakdowns are, what I've been
1:09:58spending it on, and what I can file from
1:10:00those, let's say, on my taxes, right?
1:10:02So, what Claude's going to do is it's
1:10:03going to take those receipts, extract
1:10:04all the data, and it's going to make for
1:10:06us a workable and usable spreadsheet.
1:10:08Okay, so it doesn't have a connector
1:10:10just yet. It does need to connect to
1:10:12Google Drive, so let's go ahead and
1:10:13connect it to Google Drive. So, all we
1:10:14do is simply click on connect. It's
1:10:16going to bring up this page over here to
1:10:18finish connecting connector. Go ahead
1:10:20and continue. Choose the account.
1:10:25Click on select all and continue, and it
1:10:27is connected. Now, it's asking us for a
1:10:29quick question. I can't write to Google
1:10:31Sheets directly. How do you want the
1:10:32spreadsheet delivered? Oh, that's fine.
1:10:33So, we can do connect to Google Drive.
1:10:35That's fine. Then, which columns do you
1:10:36want? Let's do standard expenses. We can
1:10:39do add line items as well. And this one
1:10:41right here, business or personal flag.
1:10:43Perfect. Okay, so we see over here, it's
1:10:45picking up all of the images that I have
1:10:46in that folder and it's analyzing all of
1:10:48them. Now, what we can do, if you want
1:10:49to see exactly what it's doing, you you
1:10:50could click on the receipt and it shows
1:10:52you exactly what it's looking at. All
1:10:54right, so it just finished and And gives
1:10:55us a quick little preview over here. I
1:10:57can see what's going on, but let's go
1:10:58ahead and click on this option here for
1:11:00Google Drive. Now, this is what it gave
1:11:01us. So, here we have the merchant, so
1:11:03where I spent the money at, the
1:11:04location, then this case most of these
1:11:06were in Orlando, how much I spent, any
1:11:08taxes, the payment method, the category,
1:11:11and this is what I needed right here,
1:11:12the business or personal. So, these
1:11:14right here are showing as business. And
1:11:16this was at the Orange County Convention
1:11:17Center, yep. And my taxes, that's
1:11:20correct.
1:11:21Now, these are personal because these
1:11:24were at supermarkets, understood. USPS,
1:11:26that was definitely business. Tap 42
1:11:29Kitchen & Bar, it's set to review. I
1:11:31remember this, this was a business lunch
1:11:32that I had. So, I am going to go ahead
1:11:33and switch this over to business, and
1:11:35everything else here also looks like it
1:11:38is pretty accurate, perfect. Now, that's
1:11:40pretty much it for that, but then it did
1:11:42also give us uh two other tabs. It also
1:11:45gives us the line items down here, where
1:11:47it gives us a breakdown of what was
1:11:48actually in those receipts. So, prepare
1:11:50electronic filing fee, soda, what we had
1:11:53at the restaurant, and everything else
1:11:56that it could find from those receipts.
1:11:58And then it gave us non-expenses, which
1:11:59were not receipts, these were just
1:12:01things that were on there, like let's
1:12:02say different invoices, or a few images
1:12:05that I guess would have been duplicated.
1:12:07So, as you saw, Cloud Co-worker
1:12:08essentially gives Cloud access to any
1:12:11one of your different folders, or any of
1:12:13your different files, or any anything
1:12:14that you want to give it to on your
1:12:15computer, and let's it do whatever you
1:12:18need to do with it. Another example is,
1:12:20let's say you run an e-commerce store.
1:12:22It can take the images from a folder
1:12:23that you have on your desktop, and it
1:12:25can upload it to that e-commerce store
1:12:27through something like a Shopify
1:12:28connector, which I showed you earlier,
1:12:30and it can update all of the product
1:12:32photos for you. So, there's a lot,
1:12:34there's an infinite number of use cases
1:12:36that you can use Cloud Co-worker for.
1:12:38This is just barely even scraping the
1:12:40surface. So, I actually want to know if
1:12:42you've used Cloud Co-worker in the past,
1:12:43and if you have, what have you used it
1:12:46for? Let me know down in the comments
1:12:47below. I want to see what the different
1:12:49use cases are, and what different people
1:12:50are using it for. So, with that, let's
1:12:52move on to the next big function of
1:12:55Claude, and that is one of the most
1:12:56popular ones, Claude code. Now, for
1:12:59Claude code work, the way that this
1:13:00works is a little bit different.
1:13:01Actually, it's pretty different than the
1:13:02way that your regular Claude works. For
1:13:05this, you are going to have to install
1:13:06it through the terminal. And don't
1:13:08worry, that sounds complicated, but it's
1:13:09actually super easy. Now, I am using a
1:13:12Mac, so it's going to look a little bit
1:13:13different if you're using Windows, but
1:13:15the process is pretty much the exact
1:13:16same thing.
1:13:17So, on a Mac, you're going to click on
1:13:19the launchpad, and you're going to look
1:13:21for your terminal,
1:13:22which is this one right here.
1:13:24Now, if you're using a Windows computer,
1:13:26you're going to look for your command
1:13:28prompt. And the way that you're going to
1:13:29look for it is the exact same way. Just
1:13:30go ahead and click on your start menu
1:13:31and type in cmd, and that's going to
1:13:33bring up your command prompt. Or you can
1:13:35just type in command prompt and then
1:13:36open it up this way. Now, you're going
1:13:38to go to this website over here, the is
1:13:40going to be in the cheat sheet for you.
1:13:41Remember, if you do want access to the
1:13:42cheat sheet, just drop a comment down
1:13:43below with the hashtag Claude course,
1:13:45and let me know what
1:13:49you're taking away from this video, and
1:13:50if you're actually finding any of this
1:13:51useful, and you're going to be applying
1:13:53it to your workflow. Also, you have to
1:13:55make sure to subscribe, like this video,
1:13:56and if you see a hype button down there,
1:13:58make sure to hype this video up as well.
1:13:59So, you're going to go to this window,
1:14:00and then you're going to look for what
1:14:01you're using. So, in this case, I'm
1:14:03using Mac, which is this one over here.
1:14:05So, I'm going to get this little command
1:14:06line. If you're using Windows, whether
1:14:09you're using the PowerShell or you're
1:14:10using the cmd, which is the command
1:14:12prompt, this is the one that I suggest
1:14:13you use, it's a little bit easier to
1:14:14use, you're going to copy the relevant
1:14:16prompt. So, for this, again, I'm using
1:14:18Mac OS. Let's go ahead and click on
1:14:19copy. I'm going to open up my terminal
1:14:21over here, and I'm just going to simply
1:14:23paste it in.
1:14:24Press enter, and then you're going to
1:14:25see that it's starting to set up Claude
1:14:26code. It's installing Claude. It
1:14:29installed it successfully, and that's
1:14:31it. Installation complete. Good to go
1:14:33from here. Next, it tells us right here
1:14:34what to do. Run Claude {dash} {dash}
1:14:36help to get started. So, we can just do
1:14:38that same thing. And do Claude
1:14:41{dash} {dash} help,
1:14:43and then it gives us all of this extra
1:14:44information over here. It tells us how
1:14:46to use it. It gives us some different
1:14:47little arguments, some options, and it
1:14:49gives us the prompts. But, let's make
1:14:51this a little bit easier to do. And now
1:14:53that we're already installed, we can go
1:14:54ahead and take that off. Let's open up a
1:14:56command prompt one more time, or the
1:14:57terminal, and simply type in Claude.
1:15:01Now, it's going to give us a quick
1:15:02security warning. It just says that
1:15:03Claude Code is going to be able to read,
1:15:04edit, and execute different files. It's
1:15:06pretty much kind of the same thing that
1:15:07you're using Claude with, but it's like
1:15:09a built-in type of code work. So, from
1:15:11here, you can just click on yes.
1:15:13Then, it's going to give us a little
1:15:14security prompt. Go ahead and allow.
1:15:17And this right here is Claude Code. So,
1:15:19as you see, it's not a huge difference
1:15:21between what we're using in our chat
1:15:23window. One thing I will say though is
1:15:24because I was already logged in, it
1:15:26automatically logged me in. It will
1:15:27prompt you to log in before it takes you
1:15:29to this window. So, after you type
1:15:30Claude, it's going to ask you for your
1:15:32credentials, and then after you log in
1:15:33from there, it's going to take you to
1:15:34this window over here. Now, from here,
1:15:36there's a few things we can do, right?
1:15:37We can go ahead and give it access to a
1:15:38folder. To do that, all you do is open
1:15:40up your folders. So, for this one I'm
1:15:42going to start giving it access to one
1:15:44of my current files where I'm working on
1:15:45an actual app. And over here, we have
1:15:48the path. I'm going to go ahead and take
1:15:49that. I'm going to go back to Claude
1:15:51Code.
1:15:52And I'm going to type in the file path
1:15:55down here. I'm just going to paste it in
1:15:57and hit enter. Now, for you, it might
1:15:59ask you Actually, it will ask you to log
1:16:01in if you haven't set this up before.
1:16:03Logging is super simple. It just asks
1:16:05you to log in, you click on yes, and
1:16:07it'll open up the Claude website where
1:16:09you log in through there. And that's it.
1:16:10Now, from here, you can simply just talk
1:16:12to it like you would your regular
1:16:13Claude. So, you can go ahead and ask it,
1:16:16"What can you do?" And it's going to
1:16:18answer you back the same way that
1:16:20basically Claude would. So, I'm Claude
1:16:22Code, an agent that works in your
1:16:24terminal with real tools. Here's what's
1:16:26available in this setup. So, it can code
1:16:28and create files, same thing with docs
1:16:30and data. It can connect different
1:16:32services. It can do e-commerce
1:16:34workflows. This is awesome for that. So,
1:16:36it does things like niche research,
1:16:38product selection, and stuff like that.
1:16:40But, that's because I have an Auto DS
1:16:41connector already installed through over
1:16:43here, through the actual Claude setup.
1:16:45And it can also make different
1:16:46automation tasks. So it can schedule
1:16:48recurring tasks and give us things like
1:16:49a morning brief and different things
1:16:51like that. Now, it's asking me what are
1:16:53we working on? So I'm telling it to look
1:16:55at my downloads folder and there's a
1:16:56folder within the chat GPT folder inside
1:16:59the documents folder called Gang this.
1:17:01And let me know what that app does. It's
1:17:03an app that I created for a
1:17:05print-on-demand business that I have. So
1:17:07let's see if Claude can dissect the code
1:17:09and let me know exactly what the app
1:17:11does. Perfect. So over here it says,
1:17:14"Gang this this is desktop gang sheet
1:17:16builder for DTF transfer printing, a
1:17:18commercial product you sell under my
1:17:21name brand." So what the app does, you
1:17:23point it at a folder of artwork,
1:17:25different types of artwork folders or
1:17:27files on are listed on here. It imports
1:17:29each design at its true physical size,
1:17:31gives us a bit of information on that,
1:17:34and some key pieces. So two different
1:17:36packing models, live preview, print
1:17:38readiness checks, and big sheets. And
1:17:42there's also an AI connector, so I
1:17:43allowed it to connect to Claude code. So
1:17:46this is pretty crazy. Claude code just
1:17:48looked at the folder and it figured out
1:17:51everything of what the app does, exactly
1:17:54how it does it. Now I can actually ask
1:17:56it from here to start making changes to
1:17:58the app. I'm not going to do that right
1:17:59now because I actually have it set the
1:18:00way that I want it to be, but that is
1:18:03one of the main functions of Claude
1:18:04code. It can take existing code, it can
1:18:07read it, it can modify it, it can
1:18:08improve it. And that's the best part, it
1:18:10can improve the code to make your apps
1:18:12work faster and smoother and just
1:18:14overall be more efficient. Now that's
1:18:16not the only thing that Claude code can
1:18:17do. Claude code can I do so much more
1:18:19literally from scratch. So let's start
1:18:22this up one more time. I like to start
1:18:25in new windows. So I'm asking it to
1:18:27create for me a quick reactive and
1:18:29cinematic landing page for a watch
1:18:30brand. Make up the information and use
1:18:33Higgs field to generate some of the
1:18:34assets. I want an exploding watch to
1:18:37break apart and come together as I
1:18:38scroll. So, it's going to go ahead and
1:18:40create for us a website. So, right here
1:18:42you can see it's nucleating. It's funny
1:18:44how it changes a lot of the different
1:18:46terms. It tries to be a bit more
1:18:47personal, I guess. All right, then it
1:18:48says it's going to build it in two
1:18:49passes. It's going to generate the watch
1:18:51assets first using Higgs Field. Then
1:18:53down here you can see everything that
1:18:54it's doing in a bit more detail than you
1:18:56would with Claude, the regular chat. So,
1:18:58Higgs Field, it's going to use this
1:19:00action recommend about the Higgs Field
1:19:02models. Do you want to proceed? Yes, and
1:19:04don't ask again for Higgs Field. You're
1:19:06perfectly fine to use it. Now, the
1:19:07balance Okay, it's going to ask me if it
1:19:09can check the balance and continue using
1:19:11it. Go ahead and yes. My credits
1:19:13balance. All right, so you can see here
1:19:15it actually started creating the code
1:19:16for the request to be able to take
1:19:18everything from Higgs Field. We can see
1:19:20what it's using. It's using Nano Banana
1:19:21Pro for the image generator. It's also
1:19:23using another one here, Nano Banana,
1:19:25Nano Banana Pro. Correct. It's asking us
1:19:27do we want to proceed? Yes, and don't
1:19:28ask again for Higgs Field generation
1:19:30image batch. So, don't ask again for
1:19:33permission to create batch images. Just
1:19:36go ahead and do it. Asking us again. Go
1:19:38ahead and click on yes. All right, so as
1:19:39you see it's doing everything in stages,
1:19:40but then it's also not waiting for one
1:19:43thing to finish. So, it says the assets
1:19:45are rendering. Higgs Field's doing its
1:19:46thing, and while that is working, it's
1:19:48going to start to build the page. So, it
1:19:50really wastes no time. All right, so it
1:19:52worked for about 15 or 16 minutes and
1:19:54made a few calls to Higgs Field, and it
1:19:55built a page all together. And what it
1:19:58came up with was actually really, really
1:20:00good. So, check this out. This is
1:20:01Eclipse caliber N01. I didn't tell it to
1:20:04do any of this. I didn't give it any
1:20:06direction. I just told it I wanted the
1:20:07exploding watch, and this is what it
1:20:09came up with. So, up here we have some
1:20:11information on the website. Scroll
1:20:13through. We have the watch itself. The
1:20:15watch itself actually looks really good.
1:20:16It was rendered through Higgs Field, and
1:20:18so are the other assets that you're
1:20:19about to see. So, as we scroll, you see
1:20:21that it actually explodes into multiple
1:20:24pieces. And this I absolutely love. Now,
1:20:27we can make this a lot more detailed. We
1:20:29just have to tell Claude, let's go ahead
1:20:30and separate also the hands, the gears
1:20:33inside, everything as an explode. But
1:20:36that's going to take a little bit
1:20:36longer, so we're not going to do that
1:20:38just yet. Now we keep going.
1:20:41Gets put back together, and then we have
1:20:43information on the watch itself and our
1:20:45brand. Now this is definitely something
1:20:48that you can create directly on Claude
1:20:50itself, directly through the chat over
1:20:52here. So I'm pretty sure right now
1:20:53you're probably wondering, why not just
1:20:55use Claude instead of Claude Code? Well,
1:20:57there's actually a very distinct reason
1:20:59why. So with the regular Claude chat,
1:21:01you can talk to it, you can brainstorm
1:21:03ideas, you can give it code, it can code
1:21:06for you, it can get all of this stuff
1:21:07done. But with Claude Code, it's like
1:21:10having an actual agent, an agent that's
1:21:13dedicated for your project files, that
1:21:15can easily build, fix, and maintain your
1:21:18different files. So if there's anything
1:21:20wrong with the app that I told you
1:21:22earlier, my GangVes app, I can tell
1:21:23Claude Code, it'll go into the actual
1:21:25files, it'll look at the source code,
1:21:27it'll figure out where it's breaking,
1:21:29why it's breaking, and it'll fix it. Not
1:21:32just that, but then it'll go back and
1:21:34it'll test its work, it'll make sure
1:21:36that it actually works, it fixed the bug
1:21:39or whatever it was that it found, and
1:21:41before it gives it back to me. So a lot
1:21:43of the times,
1:21:44the regular Claude chat, it will
1:21:46update your app, but it's not going to
1:21:48update the app directly within uh the
1:21:50source code. It'll create a new version
1:21:52of it and then it'll essentially
1:21:53override it. Whereas Claude Code isn't
1:21:55going to do that, it'll go in there and
1:21:56it'll make the changes directly in the
1:21:58app. Now for the most part, this is
1:22:00going to be something that you use when
1:22:01you're creating mostly software, apps,
1:22:05mobile apps, or uh web apps, or anything
1:22:07like that. If you're using it for
1:22:09something like I just showed you to
1:22:11create a landing page and a few
1:22:12different things like that, you can and
1:22:13you'll be pretty well off using the
1:22:15regular Claude chat. But for more
1:22:17advanced coding options, for again,
1:22:19actually creating apps and software, the
1:22:22best option for you is going to be to
1:22:23use Claude Code. Now Now that, let's go
1:22:25ahead and move on to the next section,
1:22:27which is going to be, let's say, in this
1:22:30case going from Claude Code, making an
1:22:32app to branding that app, or creating an
1:22:35entire design layout or design style for
1:22:37that app. And for that, we're going to
1:22:38use Claude Design. Now, for Claude
1:22:41Design, we're running back to our
1:22:43regular Claude chat. And what we're
1:22:44going to do is over here on the
1:22:45left-hand side, we're going to click on
1:22:47Design. Now, this has changed just a
1:22:49little bit from, I want to say about a
1:22:51month ago, you no longer chat directly
1:22:53for the Claude Design directly in here.
1:22:55Now, you actually choose one of these
1:22:56different options. Or you can click over
1:22:58here where it says view the standalone
1:23:00home page. So, let's go ahead and click
1:23:02on that. It's going to bring us to this
1:23:04new page over here, where we're going to
1:23:05be able to start our designing. Now,
1:23:07what exactly is Claude Design? So,
1:23:10Claude Design is essentially a tool
1:23:11within Claude that helps you turn any
1:23:14ideas that you have into something
1:23:16that's actually visual. So, you really
1:23:18don't need to know how to design
1:23:19anymore, or you don't need to know how
1:23:20to match different colors for things to
1:23:23actually look good. In the past, that
1:23:25was probably one of my worst traits,
1:23:27being able to match different colors.
1:23:29And Claude Design gets rid of all of
1:23:30that, because it knows what actually
1:23:32works together, and it keeps those brand
1:23:35assets, as we're calling it. So, it'll
1:23:37keep your brand visuals, your brand
1:23:38colors, and it keeps everything
1:23:40consistent throughout all of your
1:23:41branding, whether that be on your
1:23:43products, on your labels, on your
1:23:45website, or whatever it is that your
1:23:47brand consists of. So, to start, you can
1:23:49start with the chat over here, and you
1:23:52can ask it to improve the colors of a
1:23:54certain website. Or you can ask it to
1:23:56create something like a branded social
1:23:57media post for you. Now, for this, you
1:23:59are going to have a few different
1:24:00options in terms of choosing your design
1:24:02system, which is this is the most
1:24:03important part. The design system is
1:24:05going to be what incorporates all of
1:24:08your brand colors, or your design
1:24:09colors. And if we click over here,
1:24:11you're going to have already a few
1:24:13pre-made ones. So, you have industry,
1:24:15broadsheet, nocturne, organic,
1:24:17classical, and modernist. But then you
1:24:19also have the custom ones over here. And
1:24:21these are the ones that you're going to
1:24:23be using the most if you have your own
1:24:25brand colors or if you have your own
1:24:27style. And in order to add these, all
1:24:29you do is you click on this little plus
1:24:30sign over here, and you can either
1:24:32reference another project. So, it's
1:24:34going to take a project that you already
1:24:35have existing, and it's going to take
1:24:37those brand assets. You can attach a
1:24:39folder or a file that's going to have
1:24:41those different brand assets for you.
1:24:42For this, you can have, let's say, a PDF
1:24:44that has all of your branding
1:24:45guidelines, or you can give it your logo
1:24:47and a few screenshots of your website
1:24:49where it can then take those assets,
1:24:51analyze all of them, and create a design
1:24:53system for you. Now, if we look over
1:24:55here, we have a few different templates.
1:24:57So, this is going to be for the
1:24:58different types of products [snorts] or
1:25:01the different types of assets that you
1:25:02want to generate. So, these two are two
1:25:04that I created myself for an Instagram
1:25:06carousel. If you want to learn how to do
1:25:07that, if you want to learn how to create
1:25:09an entire design system from scratch to
1:25:11be able to create branded social media
1:25:12posts, then you can check out this video
1:25:13right here. It's also going to be in the
1:25:15cheat sheet for you. But, just as a
1:25:16quick walk-through, what we're going to
1:25:17do is we're going to go ahead and choose
1:25:19down here one of the already existing
1:25:21color palettes. Let's do the broachie. I
1:25:22like the broachie. And it's going to
1:25:24take this design system. It's going to
1:25:25take the branding colors and everything
1:25:26that's in there, and it's going to apply
1:25:28it to whatever we want to apply it to.
1:25:30So, we're going to give it something
1:25:31easy, something that's actually
1:25:32referencing one of the last projects
1:25:34that I made for the pet health carousel
1:25:36in the video that I just mentioned. And
1:25:37we're going to tell it, "Create a
1:25:38presentation for why it's important to
1:25:40cut a pet's nails every few weeks and
1:25:43some tips on how to do it." You can
1:25:45choose your model over here. We can
1:25:46change it if we want to. I'm going to
1:25:47keep it at Opus 5.5, which by the way,
1:25:50earlier when I was showing you the
1:25:51different models, they had up to Opus 5.
1:25:53The next day, we have Opus 5.5. So, as
1:25:56you can see, models come out every few
1:25:58weeks. So, let's go ahead and create.
1:25:59And now, this is the page that it's
1:26:00going to take us to. This is going to be
1:26:02the design page or the design layout
1:26:04where we're going to be able to change
1:26:06the different assets as it's making
1:26:08them. So, first it's going to ask us a
1:26:09few questions. So, who's the audience?
1:26:11You know what? Let's do new pet owners.
1:26:13That's a pretty good one.
1:26:14Which pets? Oh, this is actually a
1:26:16pretty good one. Let's just stick to
1:26:17dogs. The length, let's do a short six
1:26:20slides, and the tone is going to be
1:26:22let's do playful. So, what do we want to
1:26:24include? Health risks of long nails, how
1:26:27often to trim, the tools you need, and
1:26:30we did say six slides, so let's see what
1:26:31else. A step-by-step how-to, finding the
1:26:34quick, what to do if you cut too short,
1:26:36calming anxious pets, and when to see a
1:26:38groomer or vet. Let's do what to do if
1:26:41you cut too short, and when to see a
1:26:43groomer or vet. Photos, we can use
1:26:45placeholders, that's perfectly fine. And
1:26:47what is a brand name or the logo to
1:26:49feature? I don't have a brand name for
1:26:51this, so let's do playful paws. And for
1:26:55speaker notes, we can go ahead and type
1:26:57this as no, cuz we're not going to be
1:26:59presenting it, which is something that
1:27:00we're going to show. All right, let's go
1:27:01ahead and send the answer. Over here, we
1:27:04have a new chat on the side, which is
1:27:05similar to the one that we have in our
1:27:07regular side chat, but this is specific
1:27:10to design now. Okay, so it says that it
1:27:12built the playful paws deck, and then it
1:27:13gave me that really strange coloring
1:27:16that you saw. But then, it actually
1:27:18started to check its own work to make
1:27:19sure that it actually looks good, and it
1:27:21fixed it. So, let's continue letting it
1:27:22do what it's doing. It already came up
1:27:24with the different slides. It gave us
1:27:26seven different slides. Actually,
1:27:27technically six, because one is the
1:27:29intro, that's fine. And it looks like
1:27:30it's done. So, on slide two, the bullet
1:27:32list ran past the bottom of the slide
1:27:34and ran into the playful paws footer. I
1:27:35shortened all four bullets so each fits
1:27:38on one line. Okay, that's fine. All
1:27:40right, so this is where we're supposed
1:27:40to have a close-up of the dogs. Okay,
1:27:43that's fine. Here we have a few strange
1:27:46little I don't like how that looks. So,
1:27:48what we can do is we can actually edit
1:27:50these by clicking over here where it
1:27:51says edit, and then we can click on the
1:27:53asset that we want, and then we can
1:27:54change, let's say, the different colors.
1:27:56So, we can change this to all black. As
1:27:58you can see, there's actually a few
1:27:59different ones, so let me delete two of
1:28:01those, and let's just keep it like that.
1:28:02That makes a bit more sense. Now, we can
1:28:04actually change it to black. Okay, so
1:28:05that looks good. Now, through everything
1:28:07else, we have a lot of different options
1:28:09to be able to change what we're looking
1:28:11at, and we can be very specific with all
1:28:13of this. Let's go ahead and click on
1:28:14save. We can see everything that it came
1:28:16up with, so it has a trimming kit, the
1:28:18step-by-step, oops, when to get help,
1:28:21and happy trimming. Perfect. So, this
1:28:22looks pretty good. Now, what we can add
1:28:24on top of this. So, we're telling it to
1:28:25use Hex Field to generate images for
1:28:26each slide and make some natural tone
1:28:28brand colors to add a little bit of pop
1:28:31of color here and there. Right now, it's
1:28:32just too plain for me.
1:28:34Okay.
1:28:35All right, so after a couple of minutes,
1:28:36it generated all of the different images
1:28:38that we wanted to use by using Hex
1:28:40Field, and this is what it came up with,
1:28:42and it looks so much better, and it
1:28:43looks like it would belong in an actual
1:28:46brand now, or as part of a company's
1:28:48presentation. So, over here, we have the
1:28:50short nails happy paws intro. If we
1:28:53hover over this, it does clear up the
1:28:54image. We can ask Claude to get rid of
1:28:57that. I'm not a huge fan of that
1:28:58transitioning between that RGB, or yeah,
1:29:01it looks like it's RGB, red, green,
1:29:02blue, and the regular image. So, I would
1:29:05ask Claude to get rid of that, but let's
1:29:06just see everything that it came up
1:29:08with. So, it has this over here. It has
1:29:10a nice image of the dog, the little
1:29:11golden. Long nails, big problems. The
1:29:14nails could be a little bit longer. We
1:29:15could ask Claude to generate a little
1:29:17bit longer nails for this, but it does
1:29:19work. Gives us some info on why it's a
1:29:22problem. They throw off how the dog
1:29:23walks. They can curl into the paw pad.
1:29:26They snag, split, and bleed, and the
1:29:28quick grows with them. I'm not entirely
1:29:30sure what the quick is, but I would
1:29:32assume it's part of the paw. We have
1:29:33here the next image for playful paws.
1:29:37Three to four weeks between trims. Bit
1:29:39more information. We have the trimming
1:29:41kit. So, we have the clippers or
1:29:42grinder, the styptic powder, and treats.
1:29:46Lots of them. We have a step-by-step on
1:29:48how to do it. We have the little beagle
1:29:50over here being a good boy. And oops,
1:29:52when to get help, when to go see the
1:29:54vet, and just an outro, happy trimming.
1:29:56So, overall, everything looks really
1:29:57good. I do like the color scheme that it
1:29:59chose. It went with the nice little
1:30:01green and this beige over here. So, it
1:30:04has those nice little pallet colors.
1:30:06Now, from here, we can so Claude to save
1:30:08this design style and when it saves the
1:30:11design style and when it saves the
1:30:13design style we'll be able to choose it
1:30:14from the intro so it keeps this same
1:30:16consistent branding, the same colors and
1:30:18the same style along with the same fonts
1:30:20and everything. Now if there is anything
1:30:22else that we want to change we can click
1:30:24into any one of these so let's say this
1:30:26one over here and we can always click on
1:30:28edit and we can edit anything that we
1:30:30see on the page itself simply by
1:30:32clicking on it and making our changes on
1:30:34the side panel over here. There is a
1:30:35simple way to do it. There's also a pro
1:30:37way to do it. There's the code option
1:30:40which I don't touch. If you don't know
1:30:42anything about code don't touch it
1:30:43either and you have some tweaks where
1:30:45you can just simply just describe a
1:30:46tweak and it'll make that little change.
1:30:48Now we can go ahead and click on save.
1:30:51It saves the slides if we made any
1:30:52changes and we also have the option to
1:30:54share this and then publish it as an
1:30:55artifact like you saw earlier we already
1:30:57went over artifacts where it'll publish
1:30:59it as its own piece.
1:31:01You can also export it to PowerPoint or
1:31:03Google Slides which is the connected or
1:31:05create a PDF from it. So as you saw
1:31:07through Claude one thing kind of leads
1:31:10to the other where one function or
1:31:12feature reaches its capabilities Claude
1:31:14has another one that's ready to take on
1:31:16the workload and continue what you're
1:31:17working on. So today you learned when to
1:31:20essentially switch from the free to the
1:31:22paid version, when to switch over from
1:31:24let's say chats to project, when to use
1:31:26Claude co-work, when to implement Claude
1:31:28design and you've learned how to
1:31:30optimize your use for Claude so that way
1:31:32you don't simply just burn through
1:31:33credits and most importantly the
1:31:35functionality and the answers really
1:31:37that you are looking for not simply just
1:31:40asking Claude what's the weather like
1:31:42today and getting basic answers. You've
1:31:44learned how you can actually use it, how
1:31:46you can implement it and how to optimize
1:31:48it. So with that I'm going to leave you
1:31:49with three different tasks today. The
1:31:52first one is going to be to sign up for
1:31:54Claude. The second one is going to be to
1:31:56download the Claude desktop app and the
1:31:59third one is going to be to simply
1:32:01create a project. Create a project that
1:32:04is going to help you create a website,
1:32:06something simple. And within that
1:32:08project, add a connector. Once you can
1:32:10do that, once you can create a project,
1:32:12have some sort of functionality behind
1:32:14it, something that it actually does, and
1:32:16have something connected to it, an
1:32:17external service, then you pretty much
1:32:19have the basics of Cloud down. And at
1:32:20that point, you can really just keep on
1:32:22going, keep using it, and you're going
1:32:24to keep learning as you go, and you're
1:32:25going to find your way of using Cloud,
1:32:28and how you can optimize it for your
1:32:30workload or for your use. Now, remember,
1:32:32if you do want the cheat sheet with
1:32:33everything that I covered in today's
1:32:34video, all you have to do is drop a
1:32:36comment with the hashtag Cloud Course,
1:32:38and let me know what you thought of this
1:32:39video, what your takeaway is, and how
1:32:41you're going to be using Cloud for your
1:32:43day-to-day use or for your workload. And
1:32:45with that, my name is Mario with Auto
1:32:46DS, and I will catch you all next time.

About this transcript

This page contains the full transcript of Full Claude AI Tutorial (Zero to Pro Beginners Course) by AutoDS - Build Your Online Income, generated from the public captions YouTube serves with the video. The transcript has 20,816 words across 2,899 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.