Transcripts
1. Introduction to the Course: Have you ever spent hours
trying to clean up a dataset, maybe organize some information, or turn raw data into
something more useful? What if you could let AI handle most of
that work for you? Welcome to the clot Code
Fundamentals Master Class. My name is Rosa Kuchii and I'm a visual designer and creator with over five
years of experience. I'm excited to show
you how clot code can make working with files
and data a lot easier. We're first going to
get started by getting familiar with the
clot code interface. You will learn how it all works, how to write instructions, what context windows are, what permissions
you need to give, and how to make changes
to your files safely. Then we're going to jump
straight into our project. The project is taking a messy
dataset and turning it into a clean and organized datasheet
with charts or dashboard. And along the way, you're going to learn how to set up projects, which includes installing
CloudMD files, working with skills, connectors, plugins, and review the
changes that Cloud makes. And the best part,
you don't need to be a programmer
to follow along. By the end, you
would have completed your first practical
cloud code project, and you would also
know how to apply this knowledge to other
datasets and other projects. So let's go ahead
and get started.
2. Welcome & What You'll Build: I'm sure you have heard the word cloud a lot
throughout the years, and recently it's been dropping some really
powerful tools. Now, clot is another
large language model, but its capabilities
have really proved to be more advanced compared
to the other models. So not only can
you chat with it, but you can build different
websites, presentations, you can edit videos with it, you can build live applications, do version control with it, and do pretty much anything. If you're new to clot,
especially Cloud Code, which I'll tell you
what it is in a bit, this course is just for you. I will be going over
the foundations of Cloud code, what
it's good for, and then we're going
to do a project together where we take a messy dataset and we organize
it before visualizing it. So that's going to be a real application
of your knowledge.
3. Installing Claude Code: Now, there are
different ways that you can access Cloud Code, and Cloud Code is basically the place where you get
to program with Cloud. This could mean that you
could program a video, you can program a presentation, program a live application, and there's a lot of
things you can do with it. But basically, on
the desktop app, it's just a matter of switching over to
the different tab. So we're going to be on this interface for the
remaining of this course. Now if you don't have
the application, you may be wondering
how you could get it. So if you go over to
clod.com slash Download, you can easily download
the application, install it for either
Mac or Windows, and you are going
to need an account. So if you want, you can make
your account here first, then download the app or make an account through
the application. To make an account, you're going to need an email and your password,
and that's it. Now, when it comes to using it, there is a free
version for Cloud, but you are going to have to pay for a different
subscription. The thing that we're going
to do is going to be using certain features within
the more advanced plans. So if you use it for free, you are able to chat with Claude and build
different things. The only problem is
that it's limited. So at some point, you're
going to be stopped. You can't chat with
it any further, and you would have to wait for your tokens to be refilled. Have the pro version, which is popular
for individual use. Over here, you can do
pretty much anything. You can build different
artifacts without any limits. You can chat without
limits as well. And then for something
more advanced, someone who does everything
Cloud every single day, they can use the MAX plan. That's going to
be higher output, and you could do a
lot more in building. So if you want to use Clot to build five apps
at the same time, the pro plan may not
be the one for you. So if you want to follow along, you should at least
have the pro version. Now, if you're within a team, you can switch over
to the team version. So that's going to be
for an organization. That's what I'm on currently. But if you have P, you can do everything that
I'm going to show you. So purchase your plan. Then you can cancel
anytime as well. Now, when we go into Cloud, you're going to
sign in, and then this will be the
interface that you see. As I mentioned, there
is the chat feature, so this is probably something
you're familiar with. Every other language model
has a chat box like this. Here you can chat directly,
maybe ask a question. You could have it
build you a document, build PDFs and all that, but it's not really suitable
for developing, as I said. So we're just going
to switch over. Now, I mentioned
earlier that there are different ways to access clot. One of them is the application that we're seeing right now. But you can also bring
clot into your terminal. You can bring it into VSCode. You can bring it into cursor. Different IDEs where you could use clot to program
different things for you. Now, I will show you how to get those different clots basically integrated into the
programs that I mentioned. But if you're completely new, you can just download the clot application like
we did right here, and you can continue with
the rest of the course. So if you go over to
developers Cloud code, that's the forward slash
that I H. You can read the documentation on how to implement Cloud into
your different IDEs. So we have VS code here. If that's what you have and that's what you want
to use Cloud for, you can install it right here. If you have cursor, you can download
it here as well. So when you click
on this, that's going to bring you to a
whole different page. You're going to have to
provide some permissions, set up some locations
for these saved files, and then it's ready to go. The same thing can
go for terminal. If you're on Mac or Linux, you can use this
line right here. There's also Power
Shell for Windows so you can use the
native install. You can grab it with
Home Brew or Wingt. So if you have any of these platforms and you
prefer to code on them, you don't need to adapt
to the application. You can easily continue to use, VS code or whatever
that you're using, but only have clot within it. Again, I'm going to be
focusing on the application.
4. How Claude Code Thinks: Tokens & Context Windows: Now, let's talk about
some things that are very important when it comes
to working with Claude. The first one is the
prompt that we mentioned, and we're just going to look at a very simple example
so you can get a better idea of what
a generic prompt looks like and how you could
make it more specialized. To say, I want to build a
website for my small business, a generic prompt is
going to look like this. Build me a website
for my business. This is a task for sure, but we're not really saying
what my business is. What's the name of my
business? What is the product? Who is the audience, and
all the other details that someone needs to know in order to
build your website. If we put Cloud
aside for a second, and let's say we
hired a developer, they need to know
everything about your envisioned website in
order to build it for you. Try to think of it
that way as well. We want to provide Cloud with as much as context as possible. Context could be, what
you say in your prompt. It could be documents
that you attach. It could be a link, an image, a video, or maybe
another website as well. So if we were to provide more
context for this prompt, it would look
something like this. So build me a website. You can say an
informational website. Already telling Claude
what the purpose of the website is,
and what it's about, it's about my candle business, let's say, called, let's
say, Starun or something. And I could provide
even more context here. So you can say, use the attached documents to provide information
about my business, products, and
contact information. And say I have some documents that I made for my business, I can easily attach them
here as a reference, and that way, Claude can have something to
get started with. So going forward,
we are going to be working on more
structured prompts. Now, depending on what you want, the prompt system is going
to be a little different. So with websites, you want to mention the
different sections, you may want to do a wireframe. You may want to include
different APIs. If it's a presentation, you will probably
need some images, some sort of brand
identity document. And as you can imagine, things are going to be slightly different for every purpose. Let's talk about
what context is. Now, context is the
knowledge limit that Cloud can have for
every single session. You can think of context as the memory limit for every session that
you have in Cloud. So if you have a session where you're building 20 websites, at some point,
you're going to run out of that context space. Meaning that once
you hit that limit, Cloud may not be able
to remember things about the information you're providing. It may hallucinate. It may not do the things
that you're asking for. Every session has
a context limit, and depending on which plan you're on and what sort
of task you're doing, that limit may be different. Now, tokens are
basically the unit that Cloud is going to use
to perform these tasks. So think of it as AI cost. So if you're doing
a very heavy task, such as building an entire ecomers application
that is going to take a lot of tokens compared to building simple presentation
with two slides. So tokens are dependent on
the amount of work you want from Cloud depending again on
what sort of plan you have, your token system
may be different. So you can check that when
you're buying your plan. But as we move forward
with the lessons, you're going to get a more
visual representation of what the context window and what the token spending is
going to look like. But clot does have ways that you could clear up your
context window. So basically
clearing your cache, and it doesn't always have to be deleting everything
and starting all over. There are some commands
within clot embedded. Of course, you can bring
in your own commands. But if you simply
do a forward slash, then do compact right here. You can basically summarize
everything in your session. Say I'm in a really
heavy session that I've been using for two
weeks now, it's full. There's a lot of information. I could do this command to basically squish all
that information into a summary and then clear up some of my
cache, basically. So there are ways
that you could keep that session and basically allow clause to be
fully functional. And again, we're
going to look into that as we move
forward in the course. Now you just need to
know what tokens are, like the concept of it,
what the context window is, and what prompts are. Now, let's take a
deeper look into our interface and see exactly what it is that
we're dealing with here.
5. Interface Tour & Safety Basics: So when you first
enter the application, you are probably brought
to the chat section. So this is again, where
you have a regular chat. You have cow, you have different
projects and artifacts. But for more complex things, we tend to move towards
the code tab instead. So I'm not going to
get into all of these. We do have a separate course for that if you wanted to know. Over here, the interface
looks a lot more simpler, and that's because a lot of the work gets done on
your own computer. And so we basically have an interface that's
similar to a compiler one, meaning that we have root files, we have directories, we have folders on the side,
right over here, and all the projects that
we're going to do is basically going to build
files in these folders. So it's not going to be all visible to you when
you first look at it. Now to start a new session,
we have the new button. We have the artifacts section, which allows you to do specific
things with clot code, for example, building documents, building slides, designing, this is a whole different
interface as well. And then we have two
things that are in Beta. If you are on the P plan, you're not going
to see these two. But because I'm on the
Teams plan, I can see them. They are currently in Beta, meaning that they're
being tested right now, but the final polished product is probably going to be
available very soon. And once it's available, you can use it on other plans. In your artifacts,
you can have your own and you can have the
ones shared with you. So you simply switch
over to different tabs, and you're able to
see everything. Now, the on call tool is available for teams
and enterprise. So again, you may not see this. But basically,
it's when you have clot B integrated into
your everyday tools, such as Slack as you're seeing, Gmail, Google Meet and
other platforms like that. Customized tab is very
important because this is where all of your
Skills and connectors live. Now, Skills are basically
specific instructions that are compacted into a file and clot can easily grab them
to do a certain thing. So it saves the time for you to explain everything
in one giant prompt. You can easily write
that skill once and then call that skill in
your one sentence prompt. Going to get into this
later on as well. But, for example, if we want to maybe let's go down to
one of the categories. We could go into
files and documents. Say I want Claude to
help me with contracts. Instead of writing
every single time, you're going to
compact my contracts. You're going to point out
to any sort of anomalies, any sort of problems
within the contract, I could have this forwardslash
legal Skill B there. Basically, it's
going to put Claude into the role that
it needs to be. Have something for
FIGMa in design. We have research. We have different skills for job searching, research online, doing certain types
of documents, building datasets, and really
anything that you think of. So this is where all your
skills are going to be. Now, what you see here is
what's default on Cloud, but you can also make
your own, as I said, by clicking on the ad we're going to look
at this later on, but you have the
option to either use what's available or create
something yourself. Connectors are basically
external things that you get to
connect to Cloud. For example, I could
connect Google Drive, have Cloud research on my topic, make a document, and then export that document to
my Google Drive. I'm basically providing
permission to Cloud to do a certain task on my
account in different ways. Again, we have files
and documents, websites, designs, and
the list just goes. You're not seeing a
connector that you want, again, you could
add it right here. But usually, you
should be able to find the biggest ones and
the most popular tools. We also have plugins. So this is like a
package version for a certain type of task. It's basically a
more advanced skill, if you think about it. And again, we have
categories down here. So a plugin can do everything from start
to finish for you, whereas a skill can do one
thing most of the time. So let's say we go to Design and Media and we go into design. You can see that within
the design plug in, we have seven Skills
and nine connectors. So again, this is
a whole package, and it differs from the two. So that's our plug ins. If you have anything yourself, it's going to show up here, and if you want to
find something, you can go to Discover. When you click on Me, we have
something called routine. So these are going
to be scheduled items that happen
at a certain time. So for example, I get a lot of emails and I want them
to be summarized for me every Monday at
9:00 A.M. That's going to be a routine that
Claude can do on my behalf. You can build the routine here or you could use one
of the templates. There are some pretty
good templates here, so you could check these out and then implement them
for your day to day. Again, we could add
our own routine, just like we can add our
own Skills and connectors, but it's a good feature to have. Now, anything that you do is
going to show up down here. These are your most
recent projects or most recent sessions, and you can easily
grab one, toggle them. You could add another
chat within that session. You could group them, sort them, and the list just goes. Whatever conversation you had, whatever project that you built, they will show up right here. We also have a search bar here if you wanted to look
for a certain thing, you can hide the sidebar. Now when you make a new session, there are a couple
of things over here. The first thing is the bar
where you talk to Claude. You can ask for certain things, ask questions, clarifications. Any sort of conversation you have will be right in this box. If you want clot to look at something or to add
constraints to Cloud, you can use the
Plus button right so it could reference
files or photos. I could reference
an entire folder. Now, the permission bar
right here is basically how you tie Claude down from doing certain
things on your behalf. So we can have Claude ask us permission for
every single change. And when I say change, it could be something
like reading your emails. Do you want Claude to
access your emails? You can give the
permission here. If Claude wants to open up a
document on your computer, do you want it to ask
permission for that as well? So depending on how much control you want to have
over these changes, you can change the
permission setting. I would always put it on something where you're
aware of what Claude is doing because you never want
to let Claude do whatever it wants without you giving permission or without you
knowing what it's doing. Now, manual is too intense. Sometimes it may even ask you permission to access
a public website. You don't need to
know about that. So I usually go for auto
or I could make a plan and tell Claude the conditions as to when it should
ask me for permission. I could only limit it to
whenever it wants to access my files and not ask me permission when it's
accessing URLs. That's an example. We have our models over here. These are all the
available models as well as their efforts. So how much reasoning
do you want? How much time do you want the models to spend
on your task? If they spend more time, you're going to get
a better result. But sometimes you may want something that's
quick and easy, so you may want to
lower that effort. Now we have voice mode as well if you don't want to
continuously type. And then on the
right side is where basically all the files that Cloud makes
is going to live. Everything is going to
be on your computer unless you attach a
cloud environment. But the directory and the folder structure is
going to be in files. So all my works are
going to show up here. A sort of output that Cloud has for that session,
they will end up here. You can go onto
these, download them, share them, do whatever
that's needed. You can search for
your files over here. You can sort them, and then, you know, organize the
files as they come. Now, over here, we have
a built in browser, which is cool because when you're dealing with
URLs all the time, let's say you're building
a web app or a website, you don't have to always go open up Chrome, test that URL. You can easily do
it within clot. So this is a fully functioning
URL type Apple here. You can see it brings me to
Google, and I get Apple. So you can use this to
test different things. If you want to do a quick
search or something, that's what you get to do here. We also have a
terminal within Cloud, which is also very
cool Cloud Code. So if you want to access
something from your computer, you don't have to exit Clot Code and open up
your terminal to do that. You can access Webroo here. You can get different files, build environments yourself dot ENV files all through the
Cloud Code interface. These are going to
be really helpful if you're doing a
full on project. Where there are layers
to building it. Alright, and that
is our interface. Now, if this seems a
little intimidating, don't worry because
in the next lesson, we're going to get
into practice. We're going to apply the
concepts that we learned about and turn a messy dataset
into an organized one. We're going to build a system. We're going to review changes, and see clod code
work in action.
6. Your Class Project: Clean & Visualize a Datasheet: The class project is going to be centered around a
very simple exercise. It may look simple, but this is something you can apply
to anything else. So we're basically
going to take this very dataset and clean it up. We're going to fix
all the errors, we're going to reformat it. And then at the end,
we're going to visualize the corrected information into
an interactive dashboard. So this is basically a
fake student survey where 25 students put in the information about
their weekly study hours, the amount of times
they studied for the exam on the exam day, attended lectures and
different variables that we have recorded here. Now, this is the original file, and right off the bat, you can see that there's
a lot of errors. And we're basically going
to fed this into Cloud, write our own MD file
or markdown file, and have Cloud
basically switch it out and give us the
corrected version. Then you could use that same
tool for other datasets. Let's say it's about
a different topic. As long as you set
up cloud right, you can use it for
the same process. So let's take a look over here. First of all, we are going
to download this file. It's available for you
guys in the resource pack, and I want you guys to just take a look and look at the errors. So what needs to be changed and what the final output
needs to have. I'm just going to write
it here, but again, you can do it on
a piece of paper, you can keep it in your mind
or whatever you prefer. First of all, I could see that the rows here are not
ordered correctly. It starts from 14, and then one is all
the way down here. So this needs to be changed. So we're going to do
like a to do list here. So the first one is going
to be order of rows. Second thing, we are
going to take a look at the range for
each of the columns. So weekly study hours, it can't really be
negative, right? It's either zero or
a larger number. 95 seems like a lot, too. So we're going to have
to define a range for the weekly study hours so that if anything
is out of that range, it gets corrected into
either null or zero. So the weekly study hours
average is about 15 hours. So I can see that
this is crazy amount, and this one is a little
bit over the top, but it's not impossible. So maybe this student
really put in the time. So for the second column, so this is the first one. Let me just write it
here. Student ID. Second thing is weekly hours. We're going to let me
just remove the numbers. Alright, so weekly hours
is going to be 0-25. I'm just going to put that in. If you are analyzing data, you would know the ranges. So for this exercise, we're just going
to max it at 25. So for the day of the exam, I'm going to cap
it at ten because obviously there's going
to be less time to study, and most of the time is going
to be spent on the exam, transportation,
or anything else. Attended lectures. Over here, I already set
it to true or false. So basically, clot needs to reject anything that's
not true or false. Now, it can accept null fields. Basically, there were no inputs, but it can't accept yes because
that's not true or false. And if something is true instead of we can have Cloud basically
merge them together. So these are some rules that we're going to set up together. So attendance, it's
going to be T or F. Another thing is
that empty doesn't necessarily mean they didn't
attend or they did attend. It just means that
there were no inputs. That's another important thing. Sleeping hours, there's
24 hours in a day, so obviously it's
going to be 0-24. And we can accept the
seven H that could easily be turned
into a 714 is fine. For whatever reason this
person slept a lot. Meals per day, I have
the range at 123, but we can do zero to three because maybe the
student didn't eat. And again, we always have
that no input option as well. Exam score is out of 100. I can't be negative. So
score, I'll do it like that. It's 0-100 and
satisfaction. I'll do SATs. This is going to be 0-5. This is basically things that
I need to tell Claude to bear in mind when it's actually
analyzing the datasets. If I don't give
Claude this context, it may do something else. It may assume that empty
field is a no or yes, and me as the organizer
of this whole process, I need to make sure that everything is actually
going to make sense. So these are fields and the ranges that Claude
needs to bear in mind, and then we're going to
think about the output. So once I send this into
Claude, it's a messy file. What do I want to
be brought out? The output, it needs
to be this file, and also I'm going
to need a copy. By default, Claude is going
to rewrite that document. But we want a copy. Let me just make
this a little nicer, output this format and then
it's going to be a copy. Third thing, it
could be ordered. So only numbers
and true or false. So I don't want to see,
like, written numbers here. I don't want to see
like H over there. I don't want to see like a yes. That's basically what I mean. Number five, it's
going to be the fact that empty values are
going to be set to null. You can set it to X or no
value, whichever you prefer. So empty equals null. That's going to be my rule. And we can maybe have Clad
do a average at the end. So we can say average
of each column. And then we can say like an additional thing that I want average for every student. Another thing, I want
to make sure that the names all start
with a capital S, and there's no space
between that ID. So it should be SX. I'll put X for the number. So these are going
to be the output that I'm going to
expect from Claude. We also listed the ranges over here as well as what the
output needs to have. I'm going to add this as well, no negative numbers
because that's not going to make sense for the
layout that we have here. If one of the fields do
allow some negative numbers, that could be an exception. Once you have your file and
you've written down the following either on paper
or in the same document, we can now move on to actually set up our project inside clot.
7. Setting Up Your Project: Alright, so this is where we left off. Everything looks good. I know what I want
from clot and I want to get a certain
type of output. I don't want another
random dataset. So the first thing
you have to do is build a empty folder
on your computer. You can name it
whenever you want. And I'm going to
drop the original, the file that you
guys download from the resource pack
into this folder. So there's my original. I could even name it OG, so I don't mix it up later. And now we can go ahead
and open up clot. And now we're going to start
a new session in code, and we're going to choose
that very folder down here. Trust the workspace, and now it should be this computer,
your folder name, and then this one permissions, you can go ahead and
switch it to bypass permissions so it can do anything it wants
without asking you. I'm going to set this
from Auto to accept edits so that Claude is not asking me about every
single small change. Because we're doing
something very simple with a fake datasheet,
I'm able to do that. But if this is for your job, I do recommend having
it on manual or auto so that clot doesn't
mess up your dataset, and then you have to
work with it later on. In terms of our model, I'll keep it to Opus 5.5 with a medium effort because this
is not that complicated. So now we have the
project setup, but clot doesn't have
any instructions. So what is it supposed to do exactly with this
inputted dataset? So in the folder here, you can see that when
I click on this, we're able to see
exactly the things that we have in our
current folder. So we can basically kind of come over here
and see exactly what's being made
and what's being altered in the folder
bar on the side. Now, over here, we're going to first give some context to clot. So what is this datasheet about? What is the goal, and what are the things
that I want done? So we wrote that in the
Excel sheet right here. We're just going to translate
it for clot right now. So I'm just going to
move it to the side like this so that I
have my reference. Alright, so we're going to
provide some context first. You're going to act as
my data analyst and reformat slash correct
the input files. So every input needs to
be basically a dot XLSX. So that's the first check
that Cloud needs to do. So I'm just going to
do a section here. Before reading any file, make sure of the following. If you're familiar with
programming fundamentals, this is basically
an if then loop. So if you document is
populated, do this. If not, exit the program. So we're implementing
that same idea into here, but it's only in terms of words. So before reading any file, make sure of the following. Number one, it needs
to be in this format. If it's not, we're
going to say or maybe if not ask for another
file, something like that. Number two, make sure
the file is populated, meaning that there is something in the file, and
it's in English. Now, that's what I want Claude to do before
starting anything. But now once I have that set, I could tell it what
to do with this file. Say it's in perfect shape and it listens to these
two conditions. And then we could say
you are going to do the following with the file
and just do another list. Change the student
ID rope to fit this format S plus ID number. Number two, we're going to make sure that's in the
correct order. Reder the rows to
ascending order, so it goes one,
two, three, four, five, six, and so on forth. Number three, it's going to mark inputs that are
false as invalid. That's going to be my rule. You can also have it
change that cell to red. Delete that input
whatever you prefer. So mark inputs that
are false as invalid. Refer to the column
name for range. Then we can say
envelope for range. So later, I'm going to give it the range one more time just so that I'm sure that Claude fully understands
what I'm going for. Number four, it needs to remove any negative numbers and set
to invalid again. All right. So we did ordered, and
then we have four. I think I'll do four later,
five t should be null. Number six is going to be an output detail, so
I'm going to skip it. The S student number, I've already mentioned it,
and eight, no negatives. And then we can say refer
to output rules below. So now I'm going to first describe the range and
then the output rules. Range for each column. You want, you can break
these down into segment with equal sign or dashes,
whichever you prefer. So maybe around here, I could do Shift Enter and
then a bunch of these so that clot knows
that there's like a segment that it's going over. Alright, Range for each column. We have student ID. So the first one student ID, there's no range for that, so student ID should fit into S, put it in quotations plus number without any
spaces between. And a being capital. So that's for student ID,
very straightforward. Weekly hours should be 0-25. Number three, exam hours. I'll put these in quotations. Let me see. I'll try to
name it exactly like this. So I just wrote down
the column names, and now I can just
put in the ranges. So zero to 25 study for
the exam d is zero to ten. Attended all lectures is T or F, and then I'll do a comma, convert any lower case, T F to upper case, and any true or false value into T F. Sleep is
going to be 0-24 hours. Meals are going to
be zero to three, and then exam score is going
to be zero out of 100, with the satisfaction being 0-5. So that's the range. Now we're going to do another section. This is going to be
for the output rules. So basically, what do I want coming out
from the other end. So first, it's going to be
make a copy of the document. Do not alter the original. Let's say I want the
original for another work. Number two, I want to make
sure that it outputs a dot. Say it again, just to be sure. Three, make sure
rows are ordered. Average for every student, I'm going to include it here. So basically what
Code is going to do. So I'll say process here. So number seven. So number seven includes
an additional row, averaging the variables based on the student input or just averaging the column
values. And that is it. So I have told Claude exactly
the role it needs to do, the process it needs to have some checklist items
before it starts, and then the output rules in addition to the range because we're dealing with numbers here. So I could go ahead
and bring this back. And now, again, with the
accept Edits permission and the Opus 5.5 medium, we're going to send this in. And let me just attach the file before I do that.
We're going to send this in. Now, while this is happening, we're going to look at how
we can write our MD file, which is basically
this instruction that we gave into
Cloud directly, we wrote it in the chat, but say you want to use the same explanation on
a different computer, or you want to share
it with a friend. So that's where you
can get a MD file, a markdown file that you can easily upload
to other places, bring it into Cloud in
a different session. And basically, it makes the whole process a
lot easier because you don't have to explain to Cloud every single time what you need. From this type of data sheets. So we're going to just
do one real quick. I'll show you the order, but if you ended up
doing what we did in this lesson where we
set it directly into clot, you can have Clot basically package everything
into AMD file. But if you want to write one yourself before you
even go into clot, to save yourself some
tokens and usage limits, it will look
something like this. So basically, I've broken it down into different sections. The first thing you need to
do is give this file a name. So I'm going to say
data sheets cleanup, and that's going to
be the name that I will refer to it
within the chat. For the role, we
already wrote it, so I'm just going to scroll
up and copy exactly what we did into our text edit. I'm on Mac, so I'm
using text edit. You guys can use
a different tool. Process is going to be
this middle section. You can just pick out the
format that we put there. Range is going to
be the next thing, and then we have the output. Which is just three things. That's an example of an ED file. Now, when you go
on the Internet, there's so many
different variations. Some of them are like, you
know, paragraphs long. There are some that are
just a bunch of sentences. And when you first start,
you may be confused as to how much you need to
write for files like this. The thing is that you need to be as straightforward as possible, but do not include fillers. Because if you
write way too much, you're going to just fill
up your bar over here. And when this fills up,
Claude can hallucinate. I can give you wrong outputs, and it's going to
just not be helpful. So this could be reduced into
a couple sentences, too. But as you can see, I didn't include that much information. It was just very
straightforward. I want this, and I just
formatted it into a nicer way. So when you're writing
your MD files, don't try to over
explain anything. You don't need to
put it in examples. You don't need to overextend
every portion to, like, three paragraphs because it's only going to hurt you. For example, your MD file
was not as descriptive, you can always come in the
chat and work on it more. So say you forgot to add something in your
original MD file, you can have Cloud
write it again or maybe updated based on the discoveries you make within the session. So now to save this, again, I'm using text edit. So if you're doing that,
you can follow along. We're going to head
up here to format, and we're going to
switch it to rich text. So just do that and
then you can easily go to the name here and
call it whatever you want. So I'm going to
just copy the name. Put it here dot MD. Now, if you get an extension
like this like dot RTF, this is a setting that
we can easily fix. So go to Text Edit setting, and we're basically
going to prevent text edit from putting
that ending there. So just switch over
to open and save, and we're going to uncheck this. And if for some reason
it's still persisted, you can easily change
this inside finder, so it's not the
end of the world. Just going to go over here, right click Rename
and remove the RTF, use MD, and there we go. So this is a markdown file. That's another way that you
can write instructions. Both are going to
work just fine. It's going to close this now. And in the folder,
we can see that we have the original
and the new one. So let's first take a
look at what Claude did. So it changed a few things. Student ID, the orders are
sorted, invalid items, true or false, gray fill, and we have the column
with the average. So you can see everything
is nicely set, and from the looks of it, they are within the
range that I defined. And there were some duplicates. These were things that I
didn't tell Claude to do, but it caught it on its own. Remove some duplicates, two
rows, share the same thing, not whole numbers, meals, so we can decide what we want
to do with the duplicate. I'm going to say one invalid because there can't be two
inputs from one student. Alright, so this is going
to now edit the file, and then I'll just show you what that new
file looks like. Alright, so this
is the new file. It's currently being edited. But I could see that
everything is nicely ordered. There's the duplicate that Glaude was mentioning goes 1-25. We have an average bar, the invalid places, the null, and, you know, everything
else is the same. Now if you are not okay with the invalid things
being in the datasheet, you can go back and say change all the invalids to
just an empty space, an empty cell or whatever other metrics you
want to implement. But I'm just going
to allow this. And you can see that right now, we're able to convert our original dataset to
something that's corrected. And that is just
one application. So it made the change. Let's open the file again. So when I click on it, I can see that there is only one as 09, and the other one
is set to invalid. So that is one way to
tackle a problem like this. In the next lesson,
we're going to introduce some skills,
because, you know, see, we don't have the time to
do all of these setup and we want to use something
from the enthropic library, or we want to write a
skill that we can call across other sessions without having to upload that MD file. So we're going to take a look
at that in the next lesson.
8. Using a Skill from the Anthropic Library: We're now going to
integrate something that you've probably heard about
and those are Skills. So Skills are shorter
instructions that you can easily call within
the application itself. So you don't need to spend so much time writing
the instruction or giving context or writing a
separate file to then import. Everything can be done
from inside the program by simply accessing the
Skills from the library. Now, there are different
ways to get a skill. One of them is to
download it from the library here and
customize Skills, but you can also
download some from external websites or write
some Skills yourself. We're first going to focus on the many skills that
we have right here. Again, customize
Skills discover. Since we are working with data, I'm going to scroll down to the categories and then
find data. There we go. Now, these are basically plug ins that have multiple
skills within them. So I could easily look into
one that match what I need. And inside that, there's
going to be multiple skills. And then I can see exactly
what each skill is good for. For example, we have
traffic change diagnosis, so it sees why website
traffic changed, and there's also some
more descriptions that will tell you right there exactly how it's going to operate and how it's going
to be useful for you. So that's one example, and then we can go around
choosing our skill. Now, I'm going to go with this
plug in just called data. When you go into the plugin, there are ten skills, and these sets of skills are
more suitable for what we're doing because it's a very
basic set of instructions. For example, analyzing
data context extractor, data visualization. We can do explore data
and even some analysis. So these are not specific for one type of data or one
sort of application, but it's just data in general. So I could just take this,
bring in the instruction, and then tailor it to let's go ahead and choose
the one that we want to use. Our goal was to analyze our student inputs and then visualize all
of that information. We also need some conclusions about what those
numbers can tell us. So based on that, those goals, I can choose whichever
works best for me. The first one I'm going to use is for getting
those conclusions, so I will go with this one. So that's the first one.
When I go inside of it, it just confirms what I
thought this is going to. All right. Now, to
apply the skill, depending on whether
you got it from inside the application
or outside, we are going to head back to
our session and essentially just click on the Plus plug
in and get our data plug in. Within here are all the skills
that we just looked at. This is just another
way to access them. So I'm going to start with Analyze, actually. So it just does
forward slash Analyze. And now I need to
ask a question. So let's say, since we're
working with students, what do the numbers suggest in terms of what helps
students with? So what do the numbers
suggest in terms of what helps students with high
satisfaction scores? So I'm trying to see if there is correlation here between, you know, the amount
of meals they have, the amount of sleep they have, and I'm going to have
our data analyst basically tell us what
the numbers are showing. Over here, it's only asking
me to access the file. I'm going to allow that.
We have the folder, as you recall, and then it's going to report
the findings. So that's our first skill. You can see it's
data, Colin, Analyze. So the data is the plugin. The analyze is the actual skill. Alright, so we
have two findings. First one is that exam score is the strongest link to
satisfaction. So that's one thing. Second finding is that several habits separate
high and low satisfaction. And these are the factors. There's also a third
one here, and that is. Then the natural next step
is to build a dashboard. Now, I'm just going to
accept what Claude gave me, but if you want to
make some changes or ask additional questions, you can just ask
them right here. But I'm just going to
do build it so we can get that visualization that
we've been looking for. And this on its own
saves me of hours of having to look at the numbers myself
and draw conclusions. I could have Claude do that
for me, and then as a human, I could either confirm or deny
it and work based off of. So that's a helpful way of using the Skills to basically speed
up some of the processes. So that is how we can access
Skills from the library. It was very straightforward. You can find the category. You can go into plugins, call the skill with
a forward slash and go about working
with your dataset. The next lesson,
we're going to focus more on the
visualization aspect. So far, I have started that command because it
was the next step here. But we're going to go
into more details in the next lesson and see
how we can make edits. We can add additional
components, maybe remove some stuff, and see how we could get a good interactive
tool for our numbers.
9. Build It: Fix, Organize & Visualize: Mm hmm Here is our
final dashboard. We have all the components
that we have asked for, as well as the final cleared student dataset at the bottom. We can see that there is
also some caution here. So some sales were
invalid or null. So that's something that Claude
took into consideration. We have satisfied versus dissatisfied students and
what the factors were. We saw how each factor
tracks satisfaction, so we have a little
comparison here. And then we also have
two different charts. So that's something that
Claude provided us with. We go to sources, we
can see that Claude is listing exactly what it fit
into this visualization. And again, you can
always make some edits. Now, that is the base
work for the dashboard, but now I'm going to add
the interactive element. So I'm going to add
another section where I could see the what
ifs come into play. So we're going to
add a new section where I could interact with a slider to see the factors affecting the student's
satisfaction score. So this could help if I'm
presenting this dataset and I want to show what factors we
should focus on as faculty. And this slider will
allow me to do, very small adjustments, and all the information needs to change with my
mouse movement. So that's the interactive part. And then the dashboard
is going to be updated. If you want, you
can also mention to Claude to change the fonts, the color, the shapes, or the type of data visualization.
But I didn't do that. So if you wanted to just
add some information here. So the change has been made. Once we make this full screen, we can see that there
is a triad section. What satisfaction score
do these habits suggest? And I could just drag the
slider and, you know, imagine a different scenario, and that is going to change
the predicted satisfaction. So this is a what if situation. And then I could switch around
different factors here, like the exam days
study, meals, sleep. So if a student has
4 hours of sleep, didn't attend the
lectures, didn't eat, didn't study at all, they're going to have a
predicted satisfaction score of 1.7, and that makes sense. The other hand, if
someone's doing great, they will have a
predicted score of 4.3. So you can see how the original
data is being fed into this prediction
because it's not a perfect five or a perfect zero. And then you can just go
back to the survey averages. And now we have this
dashboard that we could share around and study
our students further.
10. Recap: So this was a very simple
yet good example of seeing how clot code could be integrated into
your daily work. We fed it a very
messy datasheet, and we asked it to clean it up, reorder it, reformat it, give us some conclusions, and then make an
interactive dashboard. You could apply this process to any other type of dataset. It could also be like a
document if you want. It could be videos,
it could be images. And as long as you're giving
clot the right instructions, it's able to read that type of file and then draw the
conclusions that you need. Here we had Cloud basically read our information and
draw conclusions. But there's also times
where you can have Cloud completely
generate things for you. So it would be just one
input, multiple generations, and that's going to
be something that we will try in the next course. We're going to be building
a static ad generator, where you will describe
your business once, the type of ads that
would work in your niche, and then Cloud is able to
generate multiple ads for you that you can easily post around and
advertise your business. I hope you guys were
able to follow along and end up with a
dashboard similar to this. The more you work with it, the better you're
going to get at it, and you'll naturally understand the applications that you
can have with this tool. And try to add even some
additional sections inside this dashboard. For example, a slideshow.
You could do that. You could do like a
live feed of a channel, you know, whatever
that you want to put. Anything that you
want to see basically at the end, you can
include it here. So try this out with your own projects and see
what you can make.
11. Class Project Clean and Visualize a Dataset: Now it's time to put
clod code to work. For your class project, you will be turning
a messy dataset into something that's
organized and clean, and you could visualize it via a dashboard or just
by simple tables. We will provide you with
that messy dataset, but you're free to use
your own if you prefer. First, create a project folder
and open it in clod code. Then it's time to write
the instructions. You need to describe
the context of that dataset and also what the cools are
for your project. Then create a simple clot at MD file to keep your
instructions organized. Next, you're going
to take a look at the enthropic library and install the skills that
are relevant to you. Depending on your goals, these skills may be different, but we went over some of
them that were important, so you could use those or
something additional as well. Then you're going to ask Claude
to examine that dataset. Wanted to find the errors, fix those errors, and report to you about every
single change. The goal here is to take
these messy numbers and put them into a
nicely useful format. And remember to review every single change before
you give permission to them. You don't want clot
code to do a bunch of random things to your dataset because it needs to
match your goal. AI can make mistakes,
as you know, so it's important for
you to be a part of this process and overview
everything that is happening. Once you have turned
that messy dataset into something more organized
and nicely formatted, you can now transform that
into a chart or a dashboard. Doing this so that the
dashboard or the chart can communicate something
important about those numbers. So the final submissions for your project needs
to have two things. The first one is the clean
and organized datasheet, and the second one is
your representation. Again, this could be
charts or a dashboard. When you're finished,
you can upload your files into the
class project gallery. You can include the files or screenshots,
whichever you prefer, alongside an explanation
of your process, what your goals
were and where you struggled or where you
found it very interesting. Don't worry about having to create something
very complicated. The goal here is to
see how clot code can be implemented
into a real workflow. So good luck with the project, and I look forward to
seeing what you create.
12. Congratulations! What’s Next?: M Mm hm. Congratulations. You have
reached the end of the course and have completed your first practical
project in clod code. We started by
installing clod code, learning how we can
attach our folders. We learned about some of the
fundamentals like skills, context windows,
instructions, prompts. We also learned
about ClodD files, which is going to
be something you will implement in any project. Then we put those skills to practice by building
our first project. But what is more
important is that you now know how to approach a
real problem with Cloud. And instead of accepting everything that Cloud
was trying to do, you learned how to review the changes and provide
adequate permissions. From here, you can
try this very process with different documents
and different datasets. The more you use
these workflows, the more helpful they're going
to be in your day to day. And if you're ready
to go further, our next Cloud course builds on these foundations and goes
into more advanced things. If you have any questions, feel free to ask in
the discussions. And if you found
the class helpful, feel free to drop us a review. It really helps us out.
Thanks for watching, and I'll see you guys
in our next course.
Skillademia Academy, Creative Skills for the Future
