Full Claude Tutorial: Beginner to Advanced in 19 Minutes

Claude has been my most used tool for the past year. Now, everyone else is getting on board. It has projects that remember everything about your work, skills that execute your repeatable processes automatically, connectors that

pull in your actual tools and data, and much more. I'll cover every feature, and

much more. I'll cover every feature, and more importantly, how to connect it all into workflows that unlock its full power. This is the pricing. They do have

power. This is the pricing. They do have one of the strongest free plans of all the LLMs. You can get a lot from it.

But, if you use AI regularly, I recommend investing in the paid plan of whichever LLM you prefer most. It makes

a big difference. For me, that's Claude, although I of course have a paid plan for all of them. For most people, this $20 per month plan will be plenty. In

general, you'll only need to move up when you start using Claude Co-worker extensively or Claude Code. And those

are separate tools beyond what I'm covering in this video. Okay, let's get into the actual product. Once you've set up an account, you'll be here. A similar

chat interface to all the AI models. We

have our prompt box in the middle, the model selector on the right, file uploads on the left. Then, the left panel opens up and has all your past conversations, plus some amazing

features which I'll cover later, like skills, projects, and artifacts. One

other thing I want to mention is under settings. There's some very common

settings. There's some very common advice people give that actually makes your responses worse. So, there are these fields where you can tell Claude about yourself, select your work function, and personal preferences Claude should consider in responses. You

tell Claude about yourself and give it a list of how you'd like it to respond.

That sounds great, which is why everyone recommends it. But, if you actually use

recommends it. But, if you actually use Claude across different topics, it's bad advice. I use Claude for a wide range of

advice. I use Claude for a wide range of things, and no custom instruction is going to apply to everything. It can

lead to some weird responses. I leave

these completely blank and set them up in projects instead. These are folders you can give their own instructions and custom knowledge bases. They're scoped

to each context, so the right instructions apply at the right time.

It's a much better system, and I'll cover all the details later. But, since

this advice is so common, I wanted to cover it right off the bat. Now, back to the prompt box. You can use this for all the things we're used to with AI by now.

Research coding problem-solving learning new things, or writing, which is what I first started using Claude for. It has the most natural-sounding

for. It has the most natural-sounding writing by far, and ability to match your style. It's been the leader there

your style. It's been the leader there for a long time, same with coding. But,

over the past few months, it's caught up and surpassed the other models in just about everything. Most people expect to

about everything. Most people expect to unlock all of that by just typing a question into this box. With a basic prompt, you'll get generic results. And

people do that, and then end up thinking that AI just gives that type of answer.

But, it's all in how you use it. So, I

have a three-part formula that changes the quality of every output. But, after

that foundation, we can get into the really powerful stuff.

The three most important parts of a prompt are what I call ICC: instructions, context, and constraints.

Instructions define the task and action you want Claude to take. Context is

setting the stage with your role, objectives, backgrounds, anything Claude should know that's relevant to the task.

Err on the side of more context. A full

context dump is fine. Then, constraints.

Specify any rules, style, tone, length, output format. And what can be

output format. And what can be especially helpful, when it applies, is including an example of the output you're looking for. Now, the order of these three isn't as important as long as you include everything. So, here's a

bad prompt. Recommend five ways to

bad prompt. Recommend five ways to implement AI in my marketing agency.

These answers aren't bad, but they're very generic. To make this better, I'll

very generic. To make this better, I'll use ICC. Same instruction, recommend

use ICC. Same instruction, recommend five ways to implement AI in my business. And for context, I give some

business. And for context, I give some more information about the business, and then also added some constraints. Make

sure they're low cost, easy to implement without technical expertise, things like that. And also format it as a numbered

that. And also format it as a numbered list with a one-sentence explanation for each.

Now, this is much more specific to the business I listed. There's actually a couple of these that I'm going to cover ways you could do right in Claude. ICC

will give dramatically better results, but even still, don't expect the output to be perfect on the first try. Iterate.

Go back and forth until it's perfect.

The best results come from treating Claude like a collaborator. In this

case, I asked for just one sentence each, so I could skim through it and then ask to elaborate on the ones that sound the most promising, rather than reading through a full paragraph on every single tip. And that actually leads me into a pro tip on prompting.

The reason the list still isn't perfect is, although I added some context, it can be hard to think of all the context you need to include. You'll often end up leaving things out that would have helped. To solve that, at the end of

helped. To solve that, at the end of your prompt, ask Claude to ask you for any additional context it needs to best achieve the task. I call this a context interview. It will know what it needs

interview. It will know what it needs better than you do. Then, it will ask a series of questions to gather the rest of that information. After it has everything, it will give you an answer that's much more helpful because it's

tailored to your specific situation.

Now, these answers are more customized around client communication and reporting because I answered that that was my biggest time drain. It also asked what tools I'm already using, so I could tailor it more to my setup. Once you

combine the three-part formula with that context interview, your first output is almost always usable. If you want to go deeper into using Claude for work, I have a free resource in the description called The Complete Guide to Claude AI:

Revolutionize the Way You Work. It has a lot of what I'm covering, but in written form, and also a lot more example use cases. Plus, each section has a prompt

cases. Plus, each section has a prompt you can just copy-paste to get started right away. Like in artifacts, you can

right away. Like in artifacts, you can create a full quarterly marketing campaign analysis artifact and a marketing performance interactive dashboard. Then, there's a full section

dashboard. Then, there's a full section on setting up your AI executive assistant in Claude. That goes into way more specifics than I can cover in a video like this. So, make sure to grab that. It's a free resource down in the

that. It's a free resource down in the description.

I mentioned web search, which you can find under the plus button. It's usually

on by default, and I recommend leaving it enabled most of the time. But, it

doesn't always use it by default. It

will anytime it's obvious, but sometimes it isn't, and you'll need to ask it to search the web in your prompt. One

example is when I'm having it help critique a script. Sometimes I mention Nano Banana for generating images, and I've had it say, "I see you have a placeholder here," or even just tell me I made up the name. Since critiquing a

script isn't something it intuitively feels the need to search the web for, I have to ask it to search. But, if I just asked it, "What's the best text-to-image model right now?" In that case, it would know. But, this gives up-to-date data

know. But, this gives up-to-date data and can also help reduce hallucinations.

Something I do a lot is separate the task of searching from the task of answering a prompt. Ground first, ask second. Like, if I need help with

second. Like, if I need help with something in an A 10, I'll first run a prompt that says, "Research A 10 and make sure you're up to date on all its capabilities." Then, I can ask it to

capabilities." Then, I can ask it to help with my issue. Grounding it in that context first changes the quality of every answer that follows. Another

example of ground first, ask second is, say I was launching a product. I'd ask

it to research Alex Hormozi strategies and extract all the most important information and tactics.

Now, it has his full strategies, value equation, Grand Slam offer, pricing philosophy, market selection rule, risk reversal tactics. Now, when I ask it to

reversal tactics. Now, when I ask it to help me with my launch plan, it will use all of that knowledge to inform its answers.

Using the plus button or just dragging them in, you can upload files to work from or analyze. Drop in PDFs, images, spreadsheets, a 200-page contract, a competitor's landing page. Claude reads,

summarizes, extracts, and analyzes. And

one note, although it can understand images, it can't generate them, which is kind of crazy when every other model, like ChatGPT or Gemini, can generate amazing images, but Claude can't. It's

definitely a downside to Claude if you want to be able to generate images. But,

for uploading, as an example, I'm going to come here and take a screenshot of my channel, be the 12 most recent videos.

Drop that in, and I'll say, "Here's a screenshot of my YouTube channel.

Analyze the title and thumbnail combinations across these videos, identify what's working visually and messaging-wise, what isn't, and any patterns you notice. Be direct and specific. I want honest feedback, not

specific. I want honest feedback, not flattery." And then, looking at that, it

flattery." And then, looking at that, it gives me some good advice. The clearest

failure, fact that I often just restate the title rather than extending it.

That's a good tip. List titles

underperform, they feel like homework. I

should use identity or emotional framing. Yeah, there are definitely some

framing. Yeah, there are definitely some good notes in here. Something else I can do is actually export my analytics as a CSV from the last 90 days. It can work with all sorts of formats, so I can just

drop in that CSV. Then, I'll say, "Here's my YouTube analytics data.

Analyze my performance trends and visualize the key metrics. What's the

data telling me about what's working?"

And it gave me an answer on the left, and went above and beyond, and decided to create a full analytics dashboard on the right. Claude just comes up with the

the right. Claude just comes up with the best way it thinks it can portray information. Sometimes, that's a

information. Sometimes, that's a dashboard, sometimes that's just a little visualization or a chart. And

coming back to this Alex Hormozi prompt, when it broke down the whole strategy playbook, it has all this formatting here. These are called inline

here. These are called inline visualizations, and they are actually a newer feature. I can actually click

newer feature. I can actually click between each of these tabs. This had the whole Grand Slam offer, there's the lead gen sales content scaling and mindset. This is a super helpful way to

mindset. This is a super helpful way to get this information across. Otherwise,

it would have been just this massive wall of text. And this comes in all sorts of different formats. You can also have Claude create actual downloadable files. Now, here's a chat where I had a

files. Now, here's a chat where I had a folder on my phone where I was just saving random cool spots I might want to go in Utah. I just took screenshots of those, and then asked it to organize everything. It put together this awesome

everything. It put together this awesome document, grouped by region and difficulty, and I can download that right to my computer. Or in a previous video, I had it create a full acquisition deck as a PowerPoint presentation. All sorts of formats that

presentation. All sorts of formats that you can view directly inside Claude or download to your computer. And with

these, we've moved on from just chatting, now we're creating. But,

there's an even more powerful form of creation available called artifacts.

That's what this analytics dashboard was.

Artifacts are standalone interactive outputs that Claude creates in a dedicated window alongside your conversation. Anytime you ask for it, or

conversation. Anytime you ask for it, or it just thinks producing something substantial will be beneficial, it generates it in a dedicated side panel.

And you can iterate on it without losing it in the scroll.

I made a few more examples really quick, just to show the range. Here's a

flowchart of a customer onboarding process. I've got an interactive expense

process. I've got an interactive expense tracking dashboard, a landing page for an energy drink with our logo if we ever want to create future fuel. Another

really cool one is it can use p5.js for all sorts of interactive graphics and animations. I literally just one shot at

animations. I literally just one shot at each of these for this demo. It's

incredibly easy to do. Just prompt

whatever you want into existence and it can do way more than what I just showed.

Or another common one I use is writing.

You can just ask it to open up in artifacts and you'll have the document on the right with the chat on the left so you can ask for changes while viewing the whole document. I'll show how I actually use that in a minute it incorporates skills which I haven't

covered yet but are amazing. And if

Claude doesn't automatically create an artifact when you expect one, you can explicitly ask create this as an artifact.

One more I haven't covered yet in this realm is research. This completely

changes how Claude finds and analyzes information. I'll say research how to

information. I'll say research how to use Claude co-work for business and surface what most people are missing.

Something I could just send in a chat but I'll click this plus button and come down to research. And now watch what happens. Instead of a single search,

happens. Instead of a single search, Claude operates agentically. I can open this up in the side panel to see exactly what's happening. It comes up with a

what's happening. It comes up with a plan first then conducts multiple searches that build on each other just adapting its path and exploring different angles of your question all on its own. So Claude spent 5 minutes, went

its own. So Claude spent 5 minutes, went to 169 sources, synthesized all that information down into this document that's fully cited. It can take longer for more complex investigations but it

does work that would typically require hours of manual research in minutes and provides all the citations to make verification easy. Now here's where

verification easy. Now here's where things get different. Everything up to this point has been about what Claude can do in a single conversation, chatting and creating. This next part is

about building a system which needs to remember and connect things.

Using projects correctly is one of the most impactful changes you can make to be more effective with Claude. The

projects are self-contained workspaces with their own memory, chat histories, knowledge bases, and custom instructions. Dedicated environments for

instructions. Dedicated environments for specific workstreams. Within each project you can input different custom instructions that only apply to that project. It's far more effective than

project. It's far more effective than using them at the account level like I mentioned earlier. I use projects for

mentioned earlier. I use projects for everything. If you're new to them, just

everything. If you're new to them, just start with two or three to learn how they work. The best candidates for a

they work. The best candidates for a project are areas where you have reference materials you'll use repeatedly, consistent requirements for how Claude should respond, or just to stay organized and find previous chats

easier. Click new project, give it a

easier. Click new project, give it a name, then add a short description if you want. This doesn't affect how it

you want. This doesn't affect how it functions. Once it's created, come over

functions. Once it's created, come over to instructions and that's where you can add all those custom instructions. Now

good project instructions typically include context about what you're working on, process instructions, tone and style preferences, any other specific requirements. You also have

specific requirements. You also have project files where you can upload any sort of document you'd like Claude to have persistent knowledge of in this project. Reference documents, you know,

project. Reference documents, you know, brand guidelines, style guides, templates, examples of work you want Claude to emulate like newsletters and blog posts. So upload your brand docs,

blog posts. So upload your brand docs, tone guide, or SOPs, write a custom instruction set, and every conversation inside that project inherits that context automatically. Now you don't

context automatically. Now you don't have to re-explain things across chats, rewrite context, or upload the same files repeatedly. It's all there every

files repeatedly. It's all there every time. That's a huge improvement in the

time. That's a huge improvement in the experience. But here's the limitation of

experience. But here's the limitation of projects alone. They remember things but

projects alone. They remember things but they don't know how to do things. That's

what skills are for.

Skills are preset workflows you build once and Claude pulls them in automatically whenever they're relevant.

They're little expertise packages. They

teach Claude how to complete specific tasks in a repeatable way. And there are two categories of skills. The Anthropic

skills are created and maintained by Anthropic. They are available to all

Anthropic. They are available to all paid users and Claude invokes them automatically when relevant. Custom

skills are ones you or your organization create for specialized workflows and domain specific tasks. For example, you might create a skill that applies your company's brand guidelines to presentations or maybe it structures

meeting notes in a specific format or executes your organization's data analysis workflows. It's really easy to

analysis workflows. It's really easy to create these. Just tell Claude what you

create these. Just tell Claude what you want to create and answer Claude's questions. It will interview you about

questions. It will interview you about your workflow. Also upload a reference

your workflow. Also upload a reference materials if you have them or an example of the output you're looking for. All

these things will help Claude understand exactly what you're looking for. It will

analyze everything, come up with the whole workflow, then there will just be a button to copy it to your skills. You

can also download the dot skill file to your computer if you want. You can see all your skills under customize. Then

from that point forward, Claude will automatically invoke it whenever you work on relevant tasks. You don't have to trigger it manually. Another way I'll create these skills is if you've gone back and forth with Claude iterating to

get to a final output and that's something you think you may need to do again in the future, then I'll ask it to turn that into a skill. Claude will

synthesize it all into a more refined workflow. I'll show you a few skills in

workflow. I'll show you a few skills in action. I have four different skills I

action. I have four different skills I use regularly for YouTube. One is my script critique skill. The way I write scripts is I write out the full thing myself, I'll refine it a bit and get to

a good spot. I do this in my scripting project that has custom instructions.

I'll paste in my full script, then I'll just say critique this script and that will invoke the skill. Now my

instructions specifically ask it not to rewrite anything. And the output format

rewrite anything. And the output format is as an artifact. It starts with a paragraph with a high-level critique, then it has the full script intact but after each section it lists issues for me to review. I have a big list of

things to look for and different techniques to use like identify areas where things are unclear that could benefit from a better definition or analogy, areas where viewers may be left with unanswered questions, spots that

are just too boring, or areas where I use hedging language too much and some other things. After that, I can go

other things. After that, I can go section by section refining it in the artifact. I don't really have Claude

artifact. I don't really have Claude rewrite things. It just gives me

rewrite things. It just gives me suggestions on how it could be improved.

The only section I sometimes end up with a big rewrite on is the intro. I have a separate skill for that. So it uses my existing intro plus the whole script as input. Then it will craft three

input. Then it will craft three different options for intros based on a big document of guidelines and examples.

Then I can take inspiration from those.

I also have a title skill that will write titles based on the script and also allows me to just input a title idea and it will explode that out into new ideas and variations on the original. I also have a skill for

original. I also have a skill for writing the summary that I put in YouTube descriptions. That's the one

YouTube descriptions. That's the one part I just copy and paste as is.

The four YouTube skills I just described save me hours every single week. And

every workflow you have regularly is a candidate for the same thing. They are

the most underutilized feature in Claude and they work really well when combined with projects like this. Projects are

persistent, reusable knowledge. Skills

are persistent, reusable processes.

Connectors allow Claude to access different apps to read information and perform actions on your behalf.

Depending on the connector and permissions you grant, Claude can search your files, retrieve documents, analyze data, create new content, update records, and execute tasks across your

connected applications all from within your conversation. Google Drive, Google

your conversation. Google Drive, Google Calendars, Slack, and Asana are some common ones but there's way more in here. So here's one I use. I have

here. So here's one I use. I have

Granola connected. After any meeting, like after a YouTube strategy call, I can ask Claude to pull the notes in and begin expanding on those ideas. I also

use Gamma to create better presentations. There's tons and tons of

presentations. There's tons and tons of others. Whatever tools you use, you can

others. Whatever tools you use, you can connect them and bring that information into your workflows in Claude. Then

Claude becomes the command center for your entire workflow.

Now there are different models you can use. These change all the time so I

use. These change all the time so I saved them for the end. Opus is the most powerful model for complex tasks. That

does take longer and uses more tokens which are basically Claude's currency for how much processing a conversation uses. And you get a limited amount

uses. And you get a limited amount depending on which plan you're on.

There's a 5-hour rolling window on tokens and also a weekly cap. Sonnet is

much faster and is still very powerful especially with extended thinking turned on. That lets Claude work through

on. That lets Claude work through complex problems more carefully before responding. Haiku I never use. It's very

responding. Haiku I never use. It's very

fast but not as good. The Sonnet with extended thinking is my default then whenever I'm building anything I switch to Opus. There is so much in here beyond

to Opus. There is so much in here beyond just getting better at prompting. Go

from chatting to creating to connecting it all together to make it more powerful. If you do one thing after this

powerful. If you do one thing after this video, set up a project. Then build one skill. That's the fastest path to a

skill. That's the fastest path to a system that actually works for you. I

also want to mention there is a desktop app for Claude as well. When you

download that, it has three tabs. One

for chat which is the same interface and all the features we just covered. Once

you've mastered that, there's another tab for co-work which is an agentic AI assistant. It operates directly on your

assistant. It operates directly on your computer to handle multi-step tasks.

Things like managing files, organizing documents, and summarizing data. It

works as a co-worker rather than a chatbot. It autonomously executes tasks

chatbot. It autonomously executes tasks to deliver finished results. It is

amazing. I have an entire video on that.

I'll put it right here. And there's also Claude Code which is an agentic coding tool that handles entire development workflows through natural language. It's

much easier to use than people expect and requires no coding knowledge or technical ability. I have a full guide

technical ability. I have a full guide on Claude Code coming soon so make sure you're subscribed for that.