YT YouTube2Text

YouTube2Text → Transcripts → Большой курс по ChatGPT с нуля (Октябрь 2026)

Большой курс по ChatGPT с нуля (Октябрь 2026) — Transcript

by Dmitrii Fokeev · 15,431 words · 2,292 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:01Welcome to the largest and most
0:03up-to-date ChatGPT course as of October
0:062026. If you want to go beyond just
0:10asking ChatGPT questions and actually
0:12use it for work, business, studies, or
0:14everyday life, this course is for you.
0:18In this course, we will take a
0:20step-by-step journey from registering
0:22and paying for ChatGPT from Russia and
0:25your first prompt, all the way to
0:27building websites, automation, and
0:29no-code development. We will start with
0:31the very basics. You will learn how to
0:34properly use the standard chat, the
0:36voice assistant, and the Work and Code
0:38modes for more complex tasks. We will
0:42turn ChatGPT from an advanced Google
0:43where you just ask questions into your
0:45own personal assistant. We will
0:48configure ChatGPT's personalization so
0:51it can account for information about
0:53you, your tasks, schedules, plans,
0:55habits, and work style, taking over all
0:57your daily routines. You will learn how
1:01to connect third-party services, email,
1:03calendars, cloud storage, and other
1:05apps to ChatGPT, and work with them
1:07directly from the chat. You will be
1:10able to customize the chat's behavior
1:12for specific tasks and essentially
1:14create your own custom tools within
1:16ChatGPT. You will learn how ChatGPT can
1:19operate a web browser on its own. It
1:22can open websites, search for
1:24information, fill out forms, and
1:25execute entire sequences of actions for
1:28you. Of course, we will talk about how
1:31to connect ChatGPT to programs on your
1:33computer so it can perform real tasks
1:35within them. Almost just like a human
1:38would. We will learn to create and edit
1:41presentations, Word documents, Excel
1:43spreadsheets, PDF files, and other work
1:45materials. We will cover the generation
1:49of professional images, advertising
1:50materials, and beautiful infographics.
1:52What about homework? At the end of the
1:55course, a large practical assignment
1:57covering all the course topics awaits
1:59you. Believe me, it won't be easy. If
2:02you have long wanted to master ChatGPT
2:04and truly understand what it can do and
2:07how to use it in practice, you have
2:09come to the right place. So, let's
2:12begin. Hello, friends. In this video,
2:19we will do two things. First, we will
2:21talk a bit about the plan we will need
2:23for this course, specifically the
2:25ChatGPT subscription plan. And second,
2:29we will talk in detail about how to pay
2:31for this plan from the Russian
2:33Federation amidst sanctions using a
2:34Russian bank card, so you won't have
2:36any issues. I will show you the most
2:39cost-effective method. Let's get
2:41started. Hello, friends, and welcome to
2:43the course. I want to say a big thank
2:45you right away for choosing this course
2:47and for your trust in it. The first
2:50thing we need to do in this course, of
2:52course, is to acquire ChatGPT, the
2:54official program and the official
2:55subscription. If you already have it,
2:58you can proceed directly to the next
3:00video. But for those experiencing
3:03difficulties with subscribing, because
3:05of current sanctions, unfortunately,
3:08you cannot pay for ChatGPT directly
3:11using a card from a Russian bank. What
3:14do I recommend? You can, of course, use
3:17any online services for paying for
3:19ChatGPT. There are special
3:22intermediaries who will pay for the
3:23subscription for you. But they are
3:25usually not cheap. They, of course,
3:27charge their own commission. Therefore,
3:29I recommend that you use Otopay. What
3:33is it? It is simply a virtual,
3:35non-personalized prepaid card. You can
3:38buy it with your Russian bank card,
3:41including SBP. But this card will work
3:45worldwide. They have several variations
3:47. You can choose the simplest one.
3:50Trust me, it will be enough for you.
3:52The system here is very simple, it is
3:54prepaid, meaning you buy a one-time
3:56virtual card with a specific face value
3:58. No maintenance fees, no extra costs.
4:01If you want to top up your card from
4:03time to time, the only thing you need
4:04to pay for is a simple monthly
4:06maintenance fee. And yes, they charge a
4:08small commission per transaction. It is
4:10literally just a couple dozen rubles.
4:12Believe me, this is the most
4:14cost-effective way to pay directly for
4:16any online service or subscription on
4:18the internet. You don't need to deal
4:21with intermediaries, you don't need to
4:22wait for their response, and you manage
4:24your subscriptions yourself. The link
4:26to this service, of course, you will
4:28find in the description of this video.
4:31The second thing we need to discuss is
4:33which plan to choose so that you
4:34definitely get to the right site,
4:36because there are already various
4:38fraudulent sites that create an
4:40interface similar to ChatGPT, but in
4:42reality, it is not ChatGPT. I have
4:44prepared this Word file for you. The
4:46link to it will also be somewhere near
4:48this video. And here there is, well,
4:50first of all, a little bit about me.
4:52Secondly, the link to Otopay and the
4:55official ChatGPT website. It is
4:57precisely this link you need to follow.
4:59Here you need to log in. Either via
5:02Google, Apple, Microsoft, or by phone.
5:06But a Russian phone number will
5:07definitely not work here. The easiest
5:09way, of course, is using a Google email
5:11. And yes, if you don't have one, you
5:14need to get one. Which plan should we
5:16choose? We need to choose the Plus plan
5:19specifically so that we can try
5:21absolutely all the advanced features of
5:23ChatGPT. The free version is quite
5:26limited. And yes, unfortunately, modern
5:29tools are paid, but in reality, this is
5:31not news. It has always been this way.
5:33For example, those who create graphics,
5:35edit videos, or process photos have
5:37also been paying these subscriptions
5:39for decades. And believe me, they are
5:41more expensive and not always as useful
5:43. So pay 20 dollars to check for
5:46yourself whether it's useful in your
5:49case or not. As they say, it is better
5:52to try once than to hear about it 20
5:54times. Okay. Create an account, pay for
5:58the subscription, the Plus subscription
6:00. And we are ready to take another step
6:02into the future. I will be waiting for
6:05you in the next video. Hello, friends.
6:11In this video, we will start getting to
6:13know the three operating modes of
6:14ChatGPT: the Chat, Work, and Codex tabs
6:16. In this video, we will properly break
6:19down the Chat and what it is capable of
6:21in principle. Let's get started. Well,
6:24friends, congratulations. We have
6:26acquired ChatGPT and its Plus
6:28subscription. Where should we work? We
6:30have several options. First, we can
6:33work directly in the browser itself.
6:36This is what the ChatGPT version in
6:38your browser looks like. It is almost
6:40identical to the desktop version, that
6:42is, the computer version, which I, of
6:45course, recommend you download. You can
6:48simply type "download ChatGPT" into
6:50Google, and one of the first links will
6:51give you the opportunity to do so. The
6:55application, of course, will have
6:57slightly more features than the browser
6:59version, but nevertheless, everything
7:01you write in the browser version
7:03synchronizes with your desktop version.
7:06And there is also a separate mobile
7:08application for your phone. It also
7:11synchronizes with both your browser and
7:14your desktop application. In general,
7:18all the chats you write anywhere will
7:20appear in all three places. The only
7:23difference is that the mobile app does
7:26not have this Codex tab, so we should
7:28immediately figure out these three
7:30modes. So, at the moment, ChatGPT has
7:34three operating modes. This is just a
7:37chat where you can essentially just
7:39correspond. You ask any questions, and
7:42it answers you. Here it can Google
7:45something itself and search for some
7:48information. In general, an ordinary
7:50local chat. Correspond and chat. This
7:54chat cannot manage files or folders,
7:56create code, create presentations, or
7:59work with Excel. In general, it simply
8:01gives you the opportunity to chat with
8:03ChatGPT. You can chat with it by voice,
8:07for example, start a new chat and
8:08launch the voice assistant.
8:11>> I am here, listening, how can I help?
8:13Hi, how are you?
8:14>> Hi, everything is fine. What shall we
8:16do?
8:16>> In general, a voice assistant. You can
8:18just chat with it. An ordinary chat in
8:21the form of a voice. For example, you
8:22can practice your English conversation
8:24with it. That is exactly what it is
8:26often used for. Besides that, you can
8:29still add your own files to a regular
8:32chat, for example, Word, Excel, various
8:34documents, PDFs, and so on. You don't
8:37have to select from the list here; just
8:39drag and drop them into the chat, and
8:40the chat will see everything. Of course
8:43, you can also send screenshots, photos
8:45, and any images here and ask questions
8:47about them. For instance, if you can't
8:50get something to work in an interface,
8:53take a screenshot of that spot, paste
8:56it right here, and write your question.
8:59Yes, by the way, let me tell you how to
9:01take screenshots on Windows right away.
9:04It is most likely the Print Screen key.
9:06I can't say for sure as I'm not a
9:08Windows user, but on a Mac, you can use
9:11very convenient hotkeys: Shift, Command
9:14, and the number 4 key. It allows you
9:18to create a screenshot of a selected
9:20area. Just like this. This screenshot
9:24will appear as a small thumbnail in the
9:26bottom left and will be sent to your
9:28system screenshots folder. And this, by
9:32the way, is a great opportunity to
9:34immediately ask the chat where the
9:36system screenshot folder is on Mac and
9:38how to change it. And here is the
9:40instruction for you on how to change it
9:42right away. But I mainly use not saving
9:45the screenshot to a folder, but saving
9:47it to the clipboard. This way, I don't
9:50need to go into any folders at all to
9:53grab the screenshot and paste it into
9:55ChatGPT. To take a screenshot directly
9:59to the clipboard, you need to hold
10:00Shift, Command, and also add the
10:02Control key and the number 4 key.
10:05Everything is the same. You take a
10:07screenshot, only it isn't saved
10:08anywhere; it’s just kept in the
10:10clipboard. You press Command-V, which
10:13is paste, and it gets pasted exactly
10:15where you need it. This way you don't
10:18clutter your computer, and your work
10:20becomes a little bit easier and faster.
10:22Believe me, you will be taking a large
10:25number of screenshots when you work
10:27with ChatGPT. Therefore, I highly
10:29recommend memorizing these keyboard
10:31shortcuts. Shift-Command-4 will take a
10:34screenshot to your configured folder;
10:38Shift-Command-Control-4. Takes a
10:41screenshot to the clipboard. Okay. I
10:44think we have figured out the first
10:46mode of using ChatGPT. Just a text chat
10:50that you can chat with using your voice
10:52, send it files, folders, images, and,
10:54in general, discuss them. We will
10:57discuss the next work mode, I think, in
10:59the next lesson. Hello, friends. In
11:05this video, we will take a detailed
11:07look at the work tab, which will allow
11:09us to create files, for example,
11:12Microsoft Word, and use the built-in
11:14browser and even control it using your
11:16ChatGPT. Let's get started. The second
11:20mode of operation for ChatGPT is the "
11:22Work" tab, or the "Работа" tab if
11:25we're talking about the Russian
11:26localization. What can it do compared
11:29to the standard chat? Here, it can now
11:31create files and edit files. It can do
11:35this on your computer. You surely have
11:37a question right away: which specific
11:39files and folders can ChatGPT work with
11:40? The answer is actually quite simple:
11:43with absolutely any files and folders,
11:46and even with absolutely any programs.
11:50ChatGPT in Work mode, which we are
11:52looking at now, or in the next mode,
11:54Codex, has a "computer use" mode, where
11:56the chat takes control of the mouse and
11:58interacts directly on your computer
12:00with all interfaces, programs, and
12:02anything else. We will talk about this
12:06mode separately, but you should
12:08understand right away that you can
12:10entrust any, absolutely any task on
12:12your computer to the chat. And I will
12:15say right away that it probably won't
12:16be able to do some complex and long
12:18tasks. Or rather, it will try, but it
12:20will do it with mistakes. For example,
12:22if you ask it to edit a large video,
12:25add some graphics, create animations,
12:27or create a complex layout, it
12:29naturally won't be able to do that. For
12:32now, everything here is within reason,
12:35but if you ask it to create an Excel
12:37file, set up tables in a certain way,
12:39or create a presentation file, make a
12:42certain number of slides, and add some
12:44information to those slides, it will do
12:47it all without any problems. It will
12:49rename your folders, rename your files,
12:51move files from one folder to another,
12:54sort something, delete something,
12:56fine-tune something, and analyze things
12:58. In general, this is your true
13:01assistant on your computer to whom you
13:03can actually delegate tasks, and you
13:05need to get used to using it, always
13:07remembering that you can entrust this
13:09or that task to your ChatGPT just by
13:11writing here in natural language: "Do
13:14this and that" with such and such. You
13:17can choose a folder here where this or
13:19that chat of yours will live, for
13:21example, a new project. Click on add.
13:24Choose any folder; for example, let's
13:27say I have a new folder called "test."
13:31Click on open. For you, this could be
13:34anything related to your work, no
13:36matter what you do. Let the project
13:38name also be "test." And click on the "
13:41Create" button. Of course, ChatGPT will
13:44often ask you whether to trust a
13:46particular folder. I think in one of
13:49the next lessons, we will get to know
13:51these trust settings a bit better.
13:53We'll see how they are configured so
13:55that they don't bother us too much. For
13:58now, as a rule, there is a button in
14:01the chat that immediately adjusts how
14:04often ChatGPT will bother you with its
14:07questions. Can I do this or that, can I
14:11modify a file, can I create a new file,
14:13can I open a tab, and so on. I prefer
14:16the mode where ChatGPT asks as few
14:19questions as possible. So, now we can
14:22work with various local documents. For
14:25example, create a Word file with the
14:29title Hello World. Yes, choose a
14:33slightly simpler model here right away,
14:35because on the twenty-dollar
14:37subscription, the work mode will eat up
14:39your limits very quickly. And I will
14:42say right away that you can check the
14:44remaining limit right here, and it will
14:47be shown to you here. ChatGPT has two
14:50types of limits: every 5 hours and
14:53weekly. That is, the 5-hour limit will
14:56reset every 5 hours. For example, you
14:59can work very actively for 3 hours and
15:01consume the entire 5-hour limit you
15:04were given. And after that, you won't
15:06be able to use ChatGPT. At least not on
15:08high-end models or in modes like work.
15:11As a rule, you will still be able to
15:13chat on simple models, and you will
15:15have to wait another 2 hours for this
15:17limit to reset. At the same time, all
15:20your work will also consume the weekly
15:22limit. That is, for example, by Tuesday
15:25, Wednesday, or Friday, you might use
15:27up the entire weekly limit, which
15:29updates on a specific day. And yes,
15:33with the twenty-dollar subscription,
15:35you need to actively monitor these
15:37limits and use the cheapest models
15:38possible. Use Astra very rarely. As a
15:44rule, for some medium tasks, or simple
15:46tasks like talking or creating a simple
15:49file, the simplest models are suitable.
15:53Click send. Note immediately that since
15:57we created a project, a project has
15:59appeared here in the projects tab. Yes,
16:02I have some of my other projects here,
16:04pay no attention to them, but you will
16:06have a separate folder like this.
16:09Everything you do in this chat, like
16:11creating and editing files, will be
16:14edited specifically in the folder you
16:16specified. For me, it is the desktop,
16:19the test folder. It is empty for now.
16:22Well, let's wait and see. Voila, a new
16:25file has appeared. It was created
16:27directly by ChatGPT. And inside it,
16:30there will be Hello World. Thus, your
16:33project has a specific directory where
16:36it is located. And you can always take
16:39the files that ChatGPT creates for you
16:41from there yourself, and basically
16:43understand that all of this is running
16:45locally on your computer. Once you have
16:49a whole bunch of various projects or
16:51chats, you will be able to pin a
16:53specific chat, and it will go into the
16:55pinned section. And finally. In Work
16:59mode, there is this panel that allows
17:01you to, for example, run certain files
17:03directly in the chat. It is also a
17:07separate built-in browser. You can hide
17:10or show this panel yourself right here.
17:13There. Actually, using it isn't very
17:15convenient because it constantly logs
17:17you out. And, generally speaking, it's
17:19not your browser. It's ChatGPT's
17:21browser. In order for it to visit a
17:23certain site, you, for instance, log in
17:26to it, and it receives your data so it
17:28could be used, for example, in your
17:30chat. Let's say you have some corporate
17:33service where you communicate with your
17:35colleagues. You can log in here
17:38directly inside this browser, and
17:40ChatGPT will see all the inner workings
17:43of that site and use them in its work.
17:46For example, I went to the Stepik
17:48website. This is a course marketplace
17:51where I host my courses, and I can ask
17:53the chat to do the following. Open the
17:55comment from Tatyana and reply to it.
17:58Let's select the minimal model for now
18:00and run it. It gets its own cursor, and
18:03it can click right inside its browser.
18:07It wrote that the response appeared in
18:09the discussion thread. It was done a
18:12bit incorrectly, so we are tweaking it
18:14slightly. Yes, apparently, it needed a
18:19refresh. Yes, it responded; it read the
18:23page context, understood that these
18:25were students 'homework assignments,
18:27and wrote what needs to be written when
18:29a student completes their homework. But
18:32I will warn you right away that such
18:34actions are quite expensive. When the
18:36chat directly reads the entire page,
18:39analyzes all the code on that page,
18:42writes comments, and uses such
18:44third-party services, your subscription
18:46—at least the five-hour limit—can
18:49drop to zero quite quickly. Therefore,
18:52always be quite careful with this stuff
18:54. Nevertheless, you can open any site,
18:58any service, anything at all in this
19:01panel, or ask your chat to open it in
19:03this panel, analyze them, and create
19:06documents, files, summaries, or
19:08conclusions from them, and manipulate
19:10this information in Work mode. These
19:14files will be local on your computer.
19:17But we are moving on. Hello, friends.
19:23In this video, we will talk about the
19:25Codex tab and figure out how it differs
19:27from the Work tab. And what’s more,
19:32in this video, we will create a real
19:34website—a portfolio site—literally
19:36in one prompt. And I will tell you why
19:40you should only use the Codex tab in
19:42your ChatGPT and forget about the Work
19:45tab altogether. Let's get started. And
19:48the final mode is the separate Codecs
19:51tab. If we click on it, nothing really
19:55changes. Our prompt bar will move to
19:58the bottom, but all other projects or
20:01your recent chats will remain right
20:03here. Honestly, sometimes I don't even
20:06notice which mode I'm working in
20:08because there is effectively no
20:10difference when switching. You will end
20:12up choosing your various chats anyway.
20:15And Codecs is the highest level of
20:17using ChatGPT. When you use Codecs, you
20:21are using ChatGPT to its full potential
20:23, absolutely all of its capabilities.
20:26And here is what I recommend to you. I
20:29recommend that you work directly within
20:30it from the start. I never use Work
20:34mode at all. This specific mode. I have
20:37hardly ever used it even once. I use
20:39Chat when I need to quickly message the
20:42neural network, for example, on my
20:44mobile phone. When I am at my computer,
20:47I always use Codecs. Why, you ask?
20:49Because Work is certainly a powerful
20:52tool, but it has certain limitations.
20:56For instance, it doesn't handle some
20:57tasks very well, like working with the
20:59terminal, building websites,
21:01programming, and so on and so forth.
21:04And yes, that doesn't mean you will
21:06need to open the terminal, figure it
21:08out, dig into it, or anything like that
21:11. It is still the same ChatGPT that
21:13will do everything for you. It's just
21:16that in this mode, it has absolutely no
21:19limitations. For example, in the Work
21:23tab, if you ever want to create, let's
21:25say, a website for your product, yes,
21:28it will be able to make a basic site,
21:30but when you need to expand
21:32functionality, add a user dashboard,
21:35add a database, or add something more
21:37complex, it will tell you that it
21:40cannot do it. And you will have to go
21:44into the project you were working on,
21:47take all the files from there, drag
21:49them into a new project, still switch
21:51over to the same Codecs, reconnect this
21:54new project to the folder where you
21:56moved the files, ask it to re-analyze
21:59the project, and so on, because there
22:01is no direct online synchronization
22:03between different chats in ChatGPT.
22:07Therefore, I recommend using Codecs
22:09right away. You can chat with it just
22:12like a regular chat. You can talk to it
22:15just like a regular chat. It can just
22:17as easily create all these files and
22:20folders and edit them, just like Work
22:22mode can. It also has a built-in
22:25browser on the right. It can also
22:27manipulate it in the exact same way. In
22:29general, it can do absolutely
22:31everything the same way. Believe me,
22:33you won't notice the difference, but
22:35you won't run into any restrictions.
22:37Here you select a folder in exactly the
22:39same way. Let's choose the same test
22:41folder. Let it work in this same folder
22:43. Just as before, you choose its
22:44operating mode here, how much it will
22:46ask you questions, or just work within
22:49it. For example, create a portfolio
22:51website for me. It might ask some
22:54questions; since I have some history
22:56and it knows a little about me, it
22:58immediately suggests some options. I
23:00will give it a custom prompt so it
23:02doesn't reveal my data. And there you
23:05go, in just about 5–7 minutes, we get
23:08a real website. Moreover, take a look,
23:11it even hosted it on its own domain.
23:13You can actually share it. Well, I
23:16think we'll talk about websites
23:17separately. And then, in your folder on
23:20the desktop, in this case the test
23:22folder, all the website files have
23:23appeared. And here you can continue the
23:27development. As you can see, there are
23:29no scary terminals, no code, or
23:31anything like that. Everything is just
23:33the same. You looked, liked something,
23:36didn't like something, and wrote it in
23:38plain language. And for this site, we
23:42only paid 3–4%, but that's in the
23:44minimal operating mode. And we are
23:47moving on. Hello, friends. I hope you
23:50enjoy the course and that you are
23:52getting not only pleasure from it but
23:54also useful knowledge. If so, the best
23:57way to show gratitude is to leave a
23:58like and a comment under this video.
24:01Additionally, you can visit one of the
24:04platforms where I host my courses. One
24:07of these platforms is called Stepik.
24:09Stepik is a course marketplace where I
24:12have been working for over 5 years. You
24:14can go to my profile and read the
24:16reviews that students write about me. I
24:19already have over one and a half
24:21thousand reviews, and mostly, 99.9%of
24:24them are five-star ratings. And I am
24:26proud of this result. My courses are
24:29affordable and of the highest quality.
24:32For instance, I have courses, of course
24:34, on development using neural networks
24:36and, in general, on how neural networks
24:39work. Check out my profile; perhaps you
24:42will like something. If you want lower
24:45prices, there is my separate personal
24:47website, my platform. Prices there are
24:51about 30%lower. And additionally, you
24:55can use the promo code "Sale" to get an
24:58extra 20%discount. All links will be in
25:01the description. And we are continuing.
25:07Hello, friends. In this video, we will
25:10talk about two operating modes: Code
25:12mode or Work mode. That is working on
25:14your computer, working remotely, and
25:17working in the cloud. So, we will
25:19discuss these three options together.
25:20We will figure out why you need to keep
25:24your computer turned on whenever you
25:26want to work on it remotely. Let's get
25:30started. It doesn't matter which mode
25:33we are working in, the work mode in
25:35this tab here or the codex mode, we
25:37have several operating modes. Operating
25:41mode on a computer, operating mode in
25:44the cloud. We have been working on the
25:47computer until now, meaning we took a
25:49directory and created files and folders
25:51inside it. It doesn't matter what it is
25:55: a Word document, presentations, Excel
25:57spreadsheets, our websites, code,
25:59anything at all. It is stored directly
26:02on our computer locally. The same goes
26:06for codex. It is located right here.
26:09And we have several options here.
26:11Computer, cloud, and remote. We will
26:14talk about this button in just a minute
26:16. So, we have been working locally.
26:18What kind of problem does this cause?
26:21If, for example, we work on a desktop
26:24PC at home, or we just have a laptop
26:27but don't take it with us on the road,
26:29and we need access to those files and
26:32folders we created while on the go,
26:35then if remote control is not set up
26:37for this computer, we won't be able to
26:40access those files and folders. How do
26:44we set up remote access to our computer
26:47, to our ChatGPT? We need to go
26:50directly to the mobile app, find, for
26:52example, codex, and click on the "
26:54Connect Computer" button. Click right
26:57here. And we have a short instruction
26:59here. On the computer itself, we need
27:03to go to settings, go to connections,
27:05toggle the switch here; a QR code will
27:07appear, and we scan that QR code with
27:10the phone. This way, all our projects
27:14will appear right here, and we will be
27:17able to remotely manipulate all the
27:20files and folders located on our
27:22computer. And it's super convenient
27:26because you can now literally develop
27:28websites from your phone, view them
27:30there, refine them there, and do all of
27:32this on the road. You're sitting on a
27:35bus, just staring at your phone, but in
27:37reality, you are developing a website.
27:40Of course, not all of us are
27:41programmers and not all of us create
27:43websites. Perhaps someone needs to
27:45process some information, for example,
27:47prepare a presentation for a speech. It
27:50is all the same. Your phone connects
27:53remotely to your computer and
27:54manipulates your files while you are on
27:56the road. But there is one big problem
28:00here. The problem is that your computer
28:03must always be turned on, and it
28:05shouldn't go into sleep mode. As soon
28:09as your actual computer at home turns
28:12off, remote control ends, and you lose
28:14access to your files and folders. That,
28:18by the way, is exactly what this remote
28:21access is. You can connect one computer
28:24to another, just like we just connected
28:27the phone to this computer. But take
28:31note, we now have a Cloud tab as well.
28:35It’s available both here and in work
28:36mode. But since we agreed that we are
28:39using Code interpreter, we won't talk
28:41about work mode. So, when we work in
28:44the cloud, all those files, folders,
28:47and documents we created locally are
28:50now saved directly on an OpenAI
28:52computer, somewhere far away on a
28:55server, perhaps in a city like
28:57Cupertino. And by doing this, we remove
29:01the limitation of having to constantly
29:03maintain a remote connection to our own
29:05computer. And the need for our computer
29:08to always be on and not go to sleep. We
29:11are simply working, you could say, on
29:13someone else's virtual computer that
29:15never sleeps or turns off. And that is
29:17quite convenient. I should mention
29:19right away that this feature appeared
29:21just 4 days ago. Before that, you could
29:24only connect to a computer remotely
29:25from a phone. Actually, there’s no
29:28problem with that. You leave your
29:30computer at home, set it in the
29:31operating system not to sleep, and
29:33continue working remotely. It is
29:35convenient. But now you know that you
29:37can choose between different options. I
29:40prefer to keep everything directly on
29:41the computer so I can see all my files
29:43and folders exactly and manipulate them
29:45myself. But when I need a remote
29:49connection, I simply connect remotely
29:51to my computer through my phone. But
29:53perhaps that’s just a matter of habit
29:55. Most likely, the cloud is simply more
29:57convenient. You don't need to worry
29:59about whether your computer is on or
30:01off. And we are moving on. Hello,
30:06friends. In this video, we will get to
30:09know the basic settings of ChatGPT,
30:11tell it about ourselves—that is,
30:13personalize it a bit—adjust the
30:15appearance, and, in general, talk about
30:17the most important things here. Let's
30:21begin. Well, since we've started
30:23talking about settings, let's continue.
30:26Let’s finish configuring our ChatGPT.
30:28Let’s go to settings here. And the
30:30first tab is General. Of course, the
30:32first thing we need to change is the
30:34language. Well, surely you’ve done
30:36that without me and figured out that
30:38the setting is here. Second. One of the
30:41most important things on this tab is
30:42permissions. First of all, I definitely
30:45recommend setting the default
30:46permission here. This way, your chat
30:48will be able to read and modify files
30:51in its working directory without extra
30:53prompts. For example, if you provide a
30:56directory, it will be able to work with
30:58those files without bothering you with
31:00extra questions. There is also a "Full
31:02Access" toggle. If you enable this
31:06setting, a "Full Access" button will
31:08appear when selecting the operating
31:10mode. Unrestricted access to the
31:13internet and files on your computer.
31:15Whether to use it or not is up to you.
31:18Honestly, I really don't like all these
31:20notifications because ChatGPT asks
31:22quite a lot, and you usually don't even
31:24read them and just click "OK" anyway,
31:26and you only face the consequences
31:28after it has already done something.
31:31What are the risks here? I have only
31:33ever had one risk: it might
31:35misunderstand what you are asking for.
31:38It could accidentally delete some
31:40important files or folders from your
31:42computer. And as a rule, it doesn't
31:44move them to the recycle bin for later
31:46recovery, it deletes them permanently.
31:48But even so, even with full access,
31:50OpenAI itself restricts ChatGPT. In
31:54fact, they have been experimenting a
31:56lot with these features lately.
31:58Deciding exactly what can be allowed
32:00and what cannot. Especially since there
32:03are major debates in America regarding
32:05privacy and similar issues that we
32:07aren't particularly aware of. Therefore
32:11, if you ask your ChatGPT to make a
32:13payment on a site, it will likely still
32:15ask you 10 times, and sometimes it will
32:18just refuse anyway. It will say that
32:21only a human can perform such actions.
32:24That is, at the system level, there are
32:26some "red lines" that ChatGPT will
32:28simply not cross. They are configured
32:31directly by OpenAI itself. The second
32:33thing you need to set up here is "
32:35Notify only when Codec is inactive."
32:38Why is this necessary? For example,
32:40when you have given it a task, such as
32:42creating a website, which took us 7
32:44minutes last time. You naturally won't
32:47just sit and watch it type out the text
32:49. You will switch to another program,
32:52for example, a browser, and go about
32:54your business there. So, when it
32:56finishes, a notification will sound. It
32:58is very useful. You can configure
33:00notifications more precisely right here
33:02in this tab. The next thing you can
33:04configure is your profile. I don't know
33:07if ChatGPT intends to become a social
33:09platform, but you can change your
33:10avatar, change your name, and generally
33:12, your activity is shown here. On the
33:1520-dollar plan, configuring this is
33:18pretty much useless. But there is the
33:20100-dollar-plus plan, which has a
33:23separate tab for team collaboration.
33:25That’s where it probably makes sense
33:28to set things up, because other people
33:30on that team tab will see your name and
33:33avatar. You will have shared files
33:35there that you can edit together, for
33:38example. All of this is actually still
33:40in beta, so don't be upset if you don't
33:42have team features yet. In fact,
33:44there’s not really anything useful
33:45there right now. It’s just setting
33:47the stage for the future. Of course,
33:49you can customize the look of your
33:51ChatGPT. Light or dark mode, and adjust
33:53the background or fonts as you like.
33:56Parental controls, I think, are
33:57self-explanatory. If you trust ChatGPT
33:59and your child, you can set some
34:01restrictions. Trusted contact. If you
34:04ever start discussing suicide with
34:06ChatGPT, it will notify your trusted
34:09contact. Voice. These are the voice
34:11settings. For instance, when you talk
34:14to your voice assistant, you can choose
34:16which voice it uses to speak with you.
34:21>> Nothing unusual, just standard
34:22customization. Configuration. There’s
34:25nothing really worth touching here, so
34:27we’ll move on to personalization.
34:29There are some important things here,
34:31important toggles. Enable memory and
34:33allow it to save memory from chats and
34:36your actions. Why is this needed? For
34:39example, from time to time you will
34:40talk to it about various topics. Let's
34:43say you're learning English and you
34:45tell it your current level and what
34:47level you want to reach. When it hears
34:51that kind of text, it saves it in its
34:53system folder in a text document and
34:56uses it later when needed. In other
34:59words, it saves important information
35:02about you. If you want it to save
35:04important information about you, then
35:06turn these toggles on. This way, it
35:08will understand you better. For example
35:11, when we were creating a site, I just
35:13wrote: "Create a portfolio site." It
35:16immediately knew that I teach, that
35:18I’m a designer, what things I’m
35:21into, and so on. So, without asking
35:23extra questions, it immediately created
35:26sections that were tailored exactly to
35:28me. If you want to expand its memory,
35:31instead of waiting for it to pick up
35:33information on its own, you can simply
35:35start a new chat and tell it a little
35:37about yourself: "My name is such and
35:39such, I am this old, I have this
35:41education and experience, I am
35:42interested in this and that, and so on.
35:44" By doing this, it will save these
35:47settings and get to know you better. It
35:49is genuinely useful. There is also a "
35:51consider writing style" feature here.
35:53Actually, I have experimented with this
35:55feature, and it doesn't really change
35:56much. The problem is that LLM systems,
35:59like ChatGPT, are quite precise, but
36:01not very good at being creative, so
36:04this feature won't change much. Minis
36:06and pets. I haven't really used this
36:09thing, it is purely cosmetic. You can
36:11choose an avatar that will appear
36:13somewhere in the bottom right corner of
36:15your screen. Here you can click to hide
36:18or show it. I think we will talk about
36:20some keyboard shortcuts separately. And
36:22, of course, usage and limits. Here you
36:25can simply familiarize yourself with
36:26them. And there are certain credits
36:29here. That means you can pay in dollars
36:32once your limits run out, top up your
36:35balance, and use credits directly, so
36:37to speak. I never recommend you do this
36:41. It is always more cost-effective to
36:43just buy a higher-tier subscription.
36:45Credits are used up four times faster
36:48than the limits in ChatGPT itself. That
36:51means that by sending one message on a
36:53Plus subscription, the same message
36:55will cost four times more in credits. I
36:58do not recommend it. Also, if you are
37:00not actively using your ChatGPT,
37:02meaning your limits are preserved, you
37:04will have the opportunity to reset your
37:06weekly limit a few times so you don't
37:08have to wait for the current limit to
37:10refresh. This is also a quite useful
37:13feature. Therefore, if you have hit
37:15your weekly limit to the max, come here
37:17. You might have a reset button
37:19available. The remaining things are
37:21quite practical, and we will get to
37:23know them separately, so let’s move
37:25on. Hello, friends. In this video, we
37:30will start getting acquainted with
37:32plugins, analyze what they are, talk
37:34about the pros and cons, and, of course
37:36, connect your email so that ChatGPT
37:38has access to it. Let’s get started.
37:41Okay, it's time for us to get
37:43acquainted with three things. These are
37:46plugins, skills, and so-called MCP
37:48servers. We will talk about these
37:51separately. Let’s start with plugins.
37:54To find the plugins, you need to click
37:56the button here and go to plugins. This
37:58panel will open up for you. Here you
38:01will have access to a huge number of
38:03different services. It all looks very
38:06cool, actually. As if ChatGPT could
38:10connect to, say, your Photoshop and
38:12work there directly for you. But in
38:16reality, that's not quite the case. The
38:19first thing you need to know about
38:21plugins is that they are just a way for
38:23ChatGPT to retrieve some data from a
38:25particular service. And that is
38:28basically all plugins do—they allow
38:30you to retrieve data. Sometimes a
38:33plugin can manipulate a service, also
38:37in a limited way. The simplest example
38:41is, of course, Gmail. You can add this
38:44plugin, select some permissions, what
38:46else besides email will be accessible,
38:47for example, your calendar, Google
38:49Drive, and so on. Install your
38:52selection, go to your browser, log in
38:55to your email, and grant all the
38:57necessary permissions specifically for
38:59that service. Most likely, they will
39:02pop up, and then you can return. Now we
39:05have the ability to select this
39:07specific plugin directly. How do we
39:09select it from the list? Usually, you
39:12need to press the "@" symbol and choose
39:14from the plugins presented here. But
39:17actually, if you already have this
39:20plugin connected, which is a connector
39:22to a particular service, you can just
39:25write requests in natural language.
39:27Check my email for today. It will find
39:30which email clients are connected and
39:33analyze your mail by itself. And yes,
39:35it gave us the results. This way, you
39:38can manipulate some data. You can ask
39:41it to draft an email or reply to a
39:44specific message. In short, your real
39:47assistant. You can ask it to make an
39:50entry in your Google Calendar, remind
39:52you of something, and so on and so
39:54forth. There is a small problem here:
39:58all these plugins shown here work only
40:01with foreign services. That is, you
40:04won't find a mail.ru plugin here, for
40:06example, if you use mail.ru. But how
40:09can we solve this problem, for example,
40:11with mail.ru? We can go to our email,
40:15here it is, mail.ru. Find the settings
40:18at the bottom, go to all settings, go
40:22to filters, and click on add forwarding
40:25. Enter your Gmail address, and all
40:29emails that come to your mail.ru will
40:31be automatically sent to the Gmail you
40:34specify here. Thereby, your ChatGPT,
40:37for example, will be able to see all
40:40those emails. What other useful plugins
40:42are there? Of course, the GitHub plugin
40:46is useful for checking what is
40:47happening with your repository, sending
40:50your files to your repository, having a
40:52set of commands, and so on and so forth
40:54. This is, of course, if you are a
40:56developer. I also like Supabase for
40:59developers to manage databases. You can
41:02go to the popular section and see what
41:04else is here. I will say right away
41:06that there probably aren't any super
41:08cool universal collections. Firstly,
41:10the plugins here change often, and
41:12their functionality changes too. It’s
41:14a useful thing. It really allows you to
41:17connect to certain services you use,
41:19take data from there, and manipulate it
41:21. But it is important to understand
41:24that this is not a full-fledged program
41:26built directly into your ChatGPT. When
41:29you connect Gmail, you don't connect
41:32Gmail inside ChatGPT; you are simply
41:34reading all the data that Gmail allows
41:37you to read based on its privacy
41:39settings and the plugin's settings.
41:42That's all. For example, if you work
41:45with Figma and connect a Figma plugin,
41:47you won't be able to create beautiful
41:49designs there via ChatGPT, even though
41:50that's what it says here. As a rule, at
41:54the moment, this plugin is used to read
41:57layout data: width, height, what
41:59elements are used, and the like. In
42:02order to, for example, then take this
42:05precise data from a connected layout in
42:08code and recreate an exact copy, but
42:10using code. How applicable this is
42:12needs to be seen in reality. But we are
42:15moving on, friends. Hello. In this
42:21video, we will talk a little about
42:23butterflies, or more precisely, we will
42:25study how skills work. This tab right
42:29here. Additionally, we will talk a bit
42:31about plugins and understand the
42:33differences between skills and plugins.
42:36And at the end of the video, I have
42:38prepared a large collection of the best
42:40plugins and skills for you that you can
42:43use for any occasion in your life.
42:46Let’s figure out this complex topic.
42:49Well, friends, we have basically gotten
42:51to know plugins. Now we will get to
42:53know the second tab, which is called "
42:55Skills". By default, ChatGPT only knows
42:59what its creators, that is OpenAI,
43:01taught it up to a certain date, for
43:03example, a month or two ago when the
43:06last ChatGPT update was released; the
43:08information given to it by the
43:10developers initially is what it knows.
43:15For example, if you ask it to write a
43:17post for your Telegram channel or an
43:19article for your website, for example,
43:22like this. Write an article about
43:25butterflies, and it will do it in this
43:28specific form, because the OpenAI
43:30developers set this structure. First,
43:34create a headline, then some
43:35introduction, then in the middle, make
43:37the text in a certain way, draw some
43:39conclusions at the end, and so on. If
43:43you ask it to write an article on
43:45certain topics day after day, it will
43:48always follow the exact pattern
43:50initially set by the OpenAI developers.
43:53So, Skills allows you to set up a
43:56certain personality for your ChatGPT.
44:01For example, you can create an
44:03article-writing skill, and it will
44:08simply create a text file for you with
44:11a description of how articles should be
44:13written. But for some reason, it
44:15decided not to even ask us what exactly
44:17we want to change in this skill. So, we
44:20ask it to do so ourselves. Change the
44:23skill. Make it so that every article
44:25starts with an interesting fact and
44:26ends with the signature Dmitry Fakeev.
44:29Let's choose something, maybe this
44:32model, so it’s a bit smarter and
44:34balanced. And yes, besides the main
44:38text and instructions on how to write
44:40the article, it wrote that, for example
44:43, at the end, I need to add Dmitry
44:45Fakeev. You can edit the rest of the
44:48instructions completely, meaning you
44:50can rework this file entirely, defining
44:52every nuance of how your ChatGPT should
44:54create articles and where to look. You
44:57can specify links here, or where not to
45:00look. For example, if you don't want to
45:02use articles from Wikipedia, you can
45:04explicitly forbid it from doing so
45:05right here. What style to write in,
45:07what persona, how long the text should
45:09be, and so on. Write out everything
45:11down to the smallest detail. And now,
45:14when we do the same thing—new chat,
45:16write an article about butterflies—it
45:19should, in theory, pick up this skill
45:22automatically. And yes, it writes: "
45:25Applying article writing skill." And
45:28yes, it immediately did something
45:30interesting at the beginning and wrote
45:32Dmitry Fakeev at the end. And that is
45:35what Skills are. It's just a way to
45:38customize your ChatGPT exactly the way
45:41you need it to be. Besides
45:43customization, you can provide any
45:46instructions here. And now your skill
45:48is located right here, in this folder.
45:52I don't remember what it was called. I
45:54think it was called this. You can
45:57always open it, re-read it—yes,
46:00that's it—delete it if necessary, or
46:02edit it. How do you edit it? The same
46:06way. Just go to the chat and write: "
46:09Edit this skill." By the way, to
46:12specifically tell the chat to use a
46:14particular skill, you need to use a
46:16slash. Remember, we used the at-sign to
46:20find a specific plugin, and we use the
46:22slash for specific skills. Here they
46:25are, the main ones are listed here. You
46:27can start typing the name of a skill.
46:31There it is, for example, "article
46:33writing." And after that, make the
46:34request: "Write me, for example, an
46:36article about butterflies." Then it
46:38will definitely use this skill. But as
46:41a rule, ChatGPT will always check with
46:44Skills, this list of all skills, for
46:47every request. And if it sees a
46:50relevant skill for the request, it will
46:52automatically apply it itself. That’s
46:55why I usually never use that slash to
46:58select any of the skills. And yes,
47:02essentially, Skills are just a text
47:04document with instructions on how to
47:06perform a certain task. That’s all
47:09there is to it. It’s not some kind of
47:10magic, it’s not a program under the
47:13hood or anything like that. It is just
47:15a text document with various
47:17instructions. And it can contain
47:20anything: links, it can even have some
47:22embedded code or a piece of code. There
47:25can be various instructions here.
47:27Basically, anything that can be
47:29described in text to be used as an
47:31initial instruction for a request. And
47:33you can create such slashes yourself
47:35for any situation in life. Now, an
47:38addition about plugins. What are
47:40plugins, anyway? A plugin is two things
47:43. It’s access to a particular program
47:46or service, for example, Google Mail,
47:48and Skills on how to work with it. By
47:51default, your ChatGPT only knows some
47:53basic information about a particular
47:55program. Yes, of course, ChatGPT in its
47:57knowledge base knows what Gmail is, it
48:00knows what Notion is, and so on. But a
48:03plugin allows, besides getting data
48:05from that service, the plugin also
48:08contains instructions—the very same
48:10skills—on exactly how to retrieve
48:13them and what to do when a person asks
48:16to write an email, change a letter, and
48:19so on and so forth. Therefore, a plugin
48:22is essentially just a connector to a
48:24specific application and the same text
48:26file of skills with instructions on
48:28what ChatGPT should do with this
48:30connector. That’s all. No magic at
48:32all. And as I already said, every
48:34plugin works in its own way. It’s not
48:36something universal. One will just
48:39retrieve data, another will allow you
48:41to manipulate that data in a service,
48:43another will let you edit something
48:45somewhere in that service, and so on.
48:48But let me say right away, if one of
48:50your plugins isn't working the way you
48:52imagine it should, that’s normal.
48:54Look, even if we take a simple plugin,
48:57let's say one with chess, I was
48:58experimenting here. Let’s try to
49:01launch it. This plugin allows us to
49:04play chess right in the interface. Here
49:06is the chess board that appears for us.
49:09We make a move, and look at how long
49:11ChatGPT will think about its move. Yes,
49:16as you can see, it's easier to open a
49:18standard online chess site and play
49:20there. I don't know how long you have
49:22to wait for it to understand that it
49:23needs to move a pawn. And finally, I've
49:26created a Word document for you, which
49:28will be attached to the course
49:29materials, either in the video
49:31description or attached to the video
49:33itself. In short, it will definitely be
49:35somewhere nearby for you. For now, I've
49:38simply compiled a list of the most
49:40relevant and frequently used plugins
49:42and skills that you can use right away.
49:45Plugins are a fairly simple story.
49:48Basically, there are several services
49:50from Google here to create calendars,
49:52track emails, and so on. Notion, for
49:55those who use it, Figma for designers,
49:57of course, and GitHub for developers. A
50:00standard set. You can install them, and
50:04if necessary, your ChatGPT will use
50:06data or skills from them that are also
50:09installed by default. And also the top
50:1310 skills plus another twenty. I'll say
50:16right away that some skills may already
50:19be installed for you. For example,
50:21Skill Creator is likely installed by
50:23default in every new ChatGPT. But
50:26nevertheless, there are links here for
50:28everything. You can follow these links.
50:31It might be a repository on GitHub, for
50:33instance. It will look something like
50:35this. You can open GitHub, you don't
50:39actually need to download anything,
50:42just take the link, go to your ChatGPT,
50:46start a new chat, and tell it: "Install
50:49this skill." And, in principle, you
50:55don't even have to download them
50:57individually like that; you can just
50:59take this whole file I've attached,
51:00upload it here, and write. Analyze the
51:04skills and plugins that are installed
51:06and add the ones that are missing from
51:08the list. It will read the Word
51:11document itself, follow all the links,
51:13compare what is already installed with
51:15what isn't, install everything needed,
51:17and tell you if anything didn't work
51:20out. And yes, it's better to do this
51:22right here in the code interpreter so
51:24that it does it via the terminal. What
51:26do we have here in terms of skills?
51:28Here we have, well, first of all,
51:30working with all Microsoft programs:
51:32Excel, Word, PDF, presentations, and so
51:35on. There is a skill for frontend
51:39design here, which guides ChatGPT on
51:41exactly how to format frontend design.
51:45Designs for Canva, for presentations,
51:47file and folder organization, and
51:49getting accurate YouTube transcripts.
51:52Among the useful ones, I use Humanizer.
51:54It removes robotic phrasing and signs
51:56of machine writing. If you often create
51:59posts or emails via ChatGPT, you will
52:01see how obvious those errors are.
52:04ChatGPT often tries to mimic a human,
52:06and it doesn't quite succeed.
52:09Brainstorming skills for when you need
52:11to flesh out an idea, for social media
52:14copy, SEO, and so on and so forth. In
52:17general, for all occasions in life. A
52:19fairly useful little set. You can
52:21install them as a package or just pick
52:23what you like best. Well, let's move on
52:26. Hello, friends. In this video, we
52:31will break down one small nuance. The
52:34difference between global skills we've
52:37set for our ChatGPT and specific
52:40instructions we can give to a
52:42particular project when we create it,
52:45for example, in Cursor. Let's talk
52:48about specific instructions for a
52:50particular chat. This is a very useful
52:52thing. Let's get started. Well, we've
52:55figured out the skills. It’s just a
52:58text file with instructions on how
52:59ChatGPT should perform certain actions.
53:02But what do I often use instead of
53:05these skills? As I mentioned, I usually
53:08work in Cursor. In Cursor, I have
53:10various projects. For example, I have a
53:13Telegram assistant. It helps me
53:15sometimes write articles for my
53:18Telegram. And here, I don't use a
53:20specific skill for it; I didn't create
53:22one. Instead, I just understand that I
53:27have this assistant in Cursor, it's a
53:29project, and it has its own local
53:32folder on my computer. That's where it
53:36stores all the files, folders, code,
53:38programs needed for it, and so on.
53:40Visually, this space looks like this.
53:45On my desktop, I have "ChatGPT Projects
53:48," and here is this Telegram project.
53:52And each of these projects usually has
53:54an "agents.md" file like this. It's
53:59also a text file with instructions for
54:01ChatGPT, and it creates it itself. Also
54:04, there is even a "README" file here.
54:07It was probably created after the
54:09transfer. This "agents.md" file is used
54:12in the project you created in Cursor to
54:15provide initial instructions to ChatGPT
54:18in the same way. And, for example, when
54:23I ask it to draft a post, it will
54:25always, for instance, add an "@" symbol
54:27and my channel name at the end. It will
54:31also, for example, bold the headline
54:33and write the article just the way I
54:35need it. It will definitely include a
54:37link and so on. But it did all this
54:39without a formal skill instruction. How
54:41do you create such an "agents.md" in
54:43your project? It's very simple. You
54:46just talk to your chat and ask it to
54:48remember certain things using natural
54:51language. "Remember that you need to
54:55use this and that." And it just goes
54:57ahead and fills out this agents.md file
55:00in the project folder and adds the
55:02instructions you write here. And this
55:06logic is more obvious, simpler, and
55:08clearer, because you usually just have
55:10chats or projects, it doesn't matter.
55:13In essence, it’s practically the same
55:14thing. And each chat simply contains a
55:17specific task. One chat is a Telegram
55:21assistant, another chat is a web
55:22developer, a third chat works with
55:24Excel tables, and so on. As a rule, you
55:27will have exactly this kind of
55:29architecture. Each chat carries out its
55:31own task. That is why it is much easier
55:34to create an instruction for a specific
55:36chat on how to behave and what to do,
55:38instead of creating a global skill that
55:41works for absolutely all chats. In my
55:44opinion, this is more manageable, but
55:47you can decide for yourself. Skills are
55:49a useful thing. They will, in any case,
55:51suggest certain skills for certain
55:53requests, even in your projects. That
55:56is, this agents.md does not replace
55:59skills. It's just instructions for a
56:01specific chat. To create one, you just
56:04need to ask in a particular chat where
56:07you have a project to remember certain
56:09things so that the chat can use them in
56:11the future. Few people know about this
56:13feature. And, in general, few people
56:15understand how skills work or how a
56:17given project works. Well, now you know
56:19. And now, let's move on. Hello,
56:25friends. In this video, we will look at
56:28browser technology, which will allow
56:30your chat to directly control your
56:32Google Chrome browser. It will have its
56:36own mouse and will perform any actions
56:38you ask it to do. For example, our chat
56:43will create a brand new course, fill in
56:45the description and all the fields
56:47itself, and start creating content on a
56:50site for which the chat has no
56:51connectors, no skills, or anything
56:54similar. It will figure it all out on
56:57its own and give us the result. Let's
57:00get started. The next topic is "browser
57:03," using your browser. Not the browser
57:07built into ChatGPT. Of course, it has
57:10direct access to it, and theoretically,
57:12you can work here, but trust me, it's
57:14not very convenient. You will still end
57:17up using the browser you use by default
57:20. For me, that's Google Chrome. And
57:22I'll say right away that this thing
57:23definitely works in Google Chrome. I
57:26haven't tested other browsers, so if
57:28you are an Opera fan and it doesn't
57:30work for you, I advise you to switch to
57:32Google Chrome after all. So, here is
57:35what you need to do. You need to click
57:38on this extension management button,
57:42find the ChatGPT plugin here, install
57:45it, and log in. This is a window that
57:49is now built directly into your browser
57:52. This ChatGPT sees everything
57:55happening on your page and can now
57:57directly manipulate various pages.
58:00Moreover, you never even need to open
58:03this plugin here in your browser at all
58:05. You can now give requests and
58:08instructions directly in your ChatGPT
58:11to use your browser. How do you do that
58:14? Just write in plain text: "Use my
58:16browser" for such and such tasks. To
58:19make sure everything works properly,
58:22click on settings here, go to browser
58:24—no, not browser, but computer use—
58:26and check that the Google Chrome
58:28extension is selected, enabled, and
58:30running. Oh, and by the way, if you
58:33want your Mac to always be used by your
58:35ChatGPT, even when it is in sleep mode,
58:37you can toggle the switch here. So, now
58:41your ChatGPT can manipulate everything
58:43that is on the internet. And this is a
58:46really useful tool, especially for some
58:48complex tasks. For example, you need to
58:51file a tax return. And, for instance,
58:54you need to file a tax return for
58:56another party. All you need to do in
58:59your browser is go and log in to the
59:02site where you submit your tax return,
59:04and give the task to your ChatGPT. Just
59:08say: "Fill out my tax return for me
59:10right here." It will ask for all the
59:13necessary data itself, perhaps your
59:15passport, some numbers, and so on and
59:17so forth. And believe me, it will do it
59:20well. It works great with all sorts of
59:22things like that where you need to fill
59:24out various forms. And let's experiment
59:27. On the left, I have my chat; on the
59:28right, I have the Stepik site where I
59:30host my courses. I logged in here,
59:33clicked the "New Course" button, and
59:35I'll ask it to create a draft for a new
59:36course. And look, a cursor has appeared
59:42, and it is clicking and filling in
59:44various fields by itself. Anyway, I'll
59:48stop it so it doesn't create a whole
59:50course for me right now. But look, if
59:53we go to the description, it filled in
59:56most of the fields: title, short
59:58description, what you will learn, about
1:00:01the course, text, what you get, and so
1:00:04on. And it even started creating the
1:00:08content. The only thing is, of course,
1:00:10I should have written more detailed
1:00:12instructions at the beginning so that
1:00:13it would at least ask me something,
1:00:15like what the course is about, what
1:00:16units, what lessons, if there are any
1:00:18materials, and so on and so forth. Well
1:00:21, that's just an issue with the prompt.
1:00:23I just want to show you that in this
1:00:25way you can perform any work in your
1:00:27browser. And with some minimal models,
1:00:30it has become really quite inexpensive.
1:00:33Yes, I spent about 15%of a five-hour
1:00:36subscription, which costs only 20
1:00:40dollars. But, believe me, how much it
1:00:42sometimes makes such tasks easier. For
1:00:45example, I was developing this very iOS
1:00:48application. And look, everything in it
1:00:52was written directly by ChatGPT itself.
1:00:55All the text, all the images, it
1:00:58cropped the screenshots from this app
1:01:00itself, all the declarations that had
1:01:02to be filled out to publish this app,
1:01:05because the App Store is checked by
1:01:07real people. All the forms for
1:01:10connecting payments and a whole
1:01:12mountain of things that had to be done.
1:01:15And I didn't do this myself at all. I
1:01:17didn't even read the site, didn't read
1:01:19the forms, didn't read what was written
1:01:20there. I just confirmed some things and
1:01:23checked a little to see if it did it
1:01:25correctly or not. And yes, it can take
1:01:27some time, but you don't necessarily
1:01:29have to sit at the computer. Don't
1:01:31forget that you have control over your
1:01:34projects. You can give a task and check
1:01:37its progress on your phone while
1:01:39walking outside. And additionally, with
1:01:42this function, you don't need any
1:01:44special plugins, connectors, and so on.
1:01:47As a rule, you can access any service
1:01:49through a browser. Whether it's a
1:01:52Russian service or one from another
1:01:53country, an email client, anything at
1:01:55all, it can read everything. And this
1:01:57is real magic. Don't forget to use it.
1:02:00And see you in the next video, friends.
1:02:05Hello. In this video, we will talk
1:02:07about what the computer-use technology
1:02:10is. It allows us to manipulate our
1:02:12operating system, in principle, as if
1:02:15another user were sitting there instead
1:02:17of us. It has a keyboard, a mouse, and
1:02:20it can press any keys in any unfamiliar
1:02:23applications, figure them out,
1:02:26manipulate them, and achieve a result
1:02:28that we specify for it. And we will
1:02:31experiment. We will give our ChatGPT a
1:02:35task to use Premiere Pro, which is an
1:02:37editing program, to cut a lesson,
1:02:39remove pauses, repetitions, and bad
1:02:41takes, and we will look at the result.
1:02:45We'll talk about how much it costs and
1:02:47whether it's usable in principle at the
1:02:49moment. Let's get started. The next
1:02:51thing we'll talk about is computer-use
1:02:54technology. In the last video, we
1:02:56worked with the browser. The browser
1:02:59allowed ChatGPT to directly see the
1:03:02entire page using a plugin, use a
1:03:04cursor somewhere, that is, click on
1:03:06certain fields and, under the hood,
1:03:09using code and this extension, fill out
1:03:11or somehow edit various blocks on
1:03:13various websites. But, in principle,
1:03:18for the last six months, "computer use"
1:03:20technology has been developing, which
1:03:22allows you to use absolutely any
1:03:23programs. Not just your Google Chrome
1:03:27browser, but, for example, the Opera
1:03:29browser or any program that no one
1:03:31knows about at all, with no plugins,
1:03:33connectors, or anything else. This
1:03:36technology allows your ChatGPT to
1:03:38create a virtual mouse and control your
1:03:41operating system and any of its
1:03:44programs. How does it do this? Upon
1:03:46your request, it opens a particular
1:03:49program. For example, let's say it's
1:03:51Premiere Pro. Premiere Pro is a program
1:03:54for professional video editing. It
1:03:57takes screenshots of different screens,
1:03:59understands that this is Premiere Pro,
1:04:01sees the coordinates of certain buttons
1:04:03, clicks on them with its virtual mouse
1:04:05, and, essentially, executes your
1:04:06instructions. Once again, without any
1:04:10connectors, as we had in the browser,
1:04:12it may know nothing at all about this
1:04:14program, yet still work in it like a
1:04:16regular person, using screenshots and
1:04:18clicks. And, for example, I created a
1:04:22separate project in the folder, just a
1:04:24separate folder, named it "editing,"
1:04:26and I will ask it to edit a lesson for
1:04:28me that I filmed previously. Create a
1:04:32project from this lesson. Which one is
1:04:35this? I open my file system, find the
1:04:38lesson that I want to give it. Here it
1:04:41is. I right-click here to copy the path
1:04:44to this folder. I paste it here. And
1:04:47now it knows which lesson I am talking
1:04:49about. From this system folder in
1:04:50Premiere Pro. I already opened it.
1:04:53Opened Premiere Pro. After that, remove
1:04:56all the failed takes and pauses, use
1:04:58computer use. I will explain a little
1:05:01later why I wrote that specific line
1:05:03using computer use, we select the model
1:05:05again. You pick something smaller, but
1:05:08I will say right away, the worse the
1:05:09model is here in this specific mode, in
1:05:11the computer use mode, the worse the
1:05:13result will be. And the second thing
1:05:16I'll say is that the result will most
1:05:18likely still not be very good,
1:05:19especially for any creative moments.
1:05:22For now, ChatGPT is quite bad. at
1:05:25working with editing, with other
1:05:27programs unfamiliar to it, without any
1:05:29internal extensions like in Premiere
1:05:31Pro. And the third, most important
1:05:35thing you need to hear: this is the
1:05:36most expensive thing that exists in
1:05:38ChatGPT so far—the computer use
1:05:40technology itself. because in the
1:05:42process, it will take a lot of
1:05:44different screenshots, try to figure
1:05:46out what kind of program it is, and
1:05:48constantly analyze the results of its
1:05:50actions: whether it clicked correctly
1:05:52or not. often this process will glitch.
1:05:55For example, it will constantly miss
1:05:56this or that button. It will take quite
1:05:58a lot of time, but nevertheless, the
1:06:00technology is still in its infancy,
1:06:02though I think that in the future, this
1:06:03is exactly the technology we will use
1:06:05to control computers. Well, let's see
1:06:07how it handles this task. Let's note
1:06:10right away that I had 78%of the
1:06:12five-hour limit and 91%of the weekly
1:06:14limit. I, most likely, will not carry
1:06:17this experiment to the end, otherwise
1:06:19we will burn through all our limits. It
1:06:22does this itself: creates a new project
1:06:25, navigates to different folders, and
1:06:27creates those folders. As a rule, in
1:06:29Premiere Pro, this is exactly what it
1:06:31has a problem with. For some reason,
1:06:33its system stubbornly fails to see this
1:06:35context menu. Well, friends, with your
1:06:40permission, I won't wait until the end
1:06:42because it eats up quite a lot of
1:06:44limits, and I might not have enough
1:06:45left for future lessons. But
1:06:47nonetheless, you should understand what
1:06:49kind of result such work can lead to.
1:06:52Yes, it created the files for me. Yes,
1:06:54it edited my video, removed unnecessary
1:06:57pauses, and removed unnecessary
1:06:59repetitions. How did it do it? Simply
1:07:02with screenshots and by using a text
1:07:04transcript. It literally took and
1:07:07looked at where the text was repeated.
1:07:09It highlighted these moments on the
1:07:11timeline by time and deleted them. But
1:07:13I will say right away that its work is
1:07:15not perfect. It deletes some moments
1:07:17correctly, others incorrectly, and
1:07:19sometimes it messes up the cutting. For
1:07:22example, it might do something like
1:07:24this. We will talk about this. Let's
1:07:26start with, it cut this part of the
1:07:30video exactly one second too early here
1:07:33. Start talking about this. And you
1:07:36will still likely have to take it and
1:07:38fix the whole thing yourself so that it
1:07:39doesn't jump around like that. But
1:07:42nevertheless, it can still perform some
1:07:44basic functions. for example, create
1:07:47the file itself, prepare the editing
1:07:49file, put everything on the timeline,
1:07:51prepare subtitles, do something with
1:07:53them—something super-duper simple.
1:07:56But don't forget that computer-use
1:07:57technology is quite expensive; it will
1:07:59quickly eat up your limits. And don't
1:08:02forget that this is one of the top and
1:08:04most advanced developments, and I think
1:08:06it will improve every day. And with its
1:08:09help, you can use absolutely any
1:08:11program. Just don't expect miracles
1:08:13from it. it will do the basics, but
1:08:16anything real, especially professional
1:08:17work, certainly not. And we are moving
1:08:21on. Hello, friends. In this video, we
1:08:26will quickly talk about what an MCP
1:08:28server is. This is likely the final
1:08:30video regarding the integration of
1:08:32ChatGPT with any environment. An MCP
1:08:35server is a special program that allows
1:08:38your LLM system, for example, ChatGPT,
1:08:40to use specific functions of a specific
1:08:43program instead of Computer Use
1:08:45technology. An MCP server is much
1:08:48cheaper and sometimes more accurate
1:08:50than Computer Use. And the most
1:08:52important word here is sometimes. We
1:08:53will talk in this video specifically
1:08:55about how to find them for a particular
1:08:57program and how to use them. Okay. The
1:08:59last thing we need to figure out is the
1:09:02so-called MCP servers. If we go to this
1:09:05tab again and click here, at least in
1:09:07this version, on this little lock icon,
1:09:10we will see these divisions: plugins,
1:09:13applications, MCP, and skills. We have
1:09:16already figured out what plugins are.
1:09:19One service or another creates skills
1:09:21for its service and provides access to
1:09:23that service. It decides for itself
1:09:25what data to provide and how to
1:09:27manipulate it. Skills are the
1:09:28capabilities that a particular service
1:09:31has provided to us. They are also
1:09:33stored here, or they are your personal
1:09:36skills—simply a file with
1:09:37instructions on how the chat should
1:09:39behave when performing a given task.
1:09:42Everything is very simple. There is
1:09:43also a new "applications" section. For
1:09:46now, these are system items that are
1:09:48typically added here by ChatGPT itself
1:09:51or created via plugins. Essentially,
1:09:53this is just a foundation for the
1:09:55future. Just a few days ago, there was
1:09:58a presentation that, for example, Adobe
1:10:00programs are already being integrated
1:10:02into ChatGPT, and more than likely, new
1:10:05applications will appear that you can
1:10:07use directly inside ChatGPT so that the
1:10:09actual interface is displayed rather
1:10:11than just having an integration. But as
1:10:15this is all under development, you
1:10:17won't be able to find any new
1:10:18applications; it is all just shown to
1:10:20you as a list. But there is also an
1:10:22interesting tab here called MCP.
1:10:25Perhaps some of you have heard of it.
1:10:27What is MCP? MCP is a technology that
1:10:30appeared just a year ago. It allows an
1:10:34LLM, an artificial intelligence, to
1:10:36connect to a specific program, for
1:10:39example, Premiere Pro or Figma or some
1:10:41other real program. But it does this
1:10:45without using "computer use," where
1:10:47your ChatGPT simply takes screenshots
1:10:49of your program and tries to figure out
1:10:51what the program is, what buttons it
1:10:53has, and spends a whole lot of
1:10:55resources on it. MCP is a special layer
1:10:59, a special program between your
1:11:01ChatGPT and the program you are using,
1:11:04and it already contains all the buttons
1:11:06that ChatGPT can use, all the
1:11:08instructions that can be used when
1:11:10working with that particular program,
1:11:12and so on and so forth. So, it's a set
1:11:16of control buttons for some program for
1:11:19your ChatGPT. And such MCP servers can
1:11:22be created for almost any program. I
1:11:25can't tell you the exact limitations,
1:11:27but I will tell you how to check them.
1:11:29Let's take Premiere Pro, for example.
1:11:31We can create a middleware program, an
1:11:33MCP Server, that will immediately
1:11:36transmit information that there is a "
1:11:39New Project" button, an "Open Project"
1:11:41button, and an option to create
1:11:44subtitles. The subtitle creation button
1:11:46is located in such and such a place.
1:11:47There are certain fields to fill in,
1:11:49and all this will be transmitted as
1:11:51code, not through screenshots or the
1:11:52neural network's own guesswork. Sounds
1:11:54pretty good, right? And yes, it really
1:11:57is. How can you create or find an
1:11:59already existing MCP server? Because
1:12:01for popular programs, for example,
1:12:04After Effects or Premiere Pro, other
1:12:06developers or just people have, of
1:12:08course, already created their own MCP
1:12:10servers, and you don't need to create
1:12:12them yourself. To check, all you need
1:12:15to do is go back to Cursor, click on a
1:12:17new chat, and ask a question. Yes, you
1:12:20don't have to select a folder, it's not
1:12:22necessary. Search the internet for an
1:12:25MCP server for Premiere Pro. And there
1:12:29you go, it gives you the results. It
1:12:31found this specific MCP server,
1:12:33explains what it can do, what it can't,
1:12:35and so on. It provides installation
1:12:38instructions. It says, "Open the
1:12:40terminal and install it." You can
1:12:42actually just tell it, "Install it
1:12:44yourself." And it will do it. I won't
1:12:47install it, because I already have this
1:12:48MCP server installed. You can also
1:12:51refine this MCP server just as easily
1:12:53using plain language. Right here, you
1:12:56can ask ChatGPT to test it. As a rule,
1:12:59an MCP server won't work perfectly, and
1:13:02you can, so to speak, finish
1:13:03programming it. Of course, your ChatGPT
1:13:06will do the actual programming, and you
1:13:07will just be asking it to do so. Well,
1:13:10I will say right away that MCP sounds
1:13:12like, of course, a great option for
1:13:15doing higher quality work in complex
1:13:17programs like Premiere Pro. In the last
1:13:20video, we discussed that when using
1:13:22Computer Use, ChatGPT still makes
1:13:24mistakes. So, an MCP server won't
1:13:26provide a perfect result either,
1:13:28because ChatGPT still needs to evaluate
1:13:31complex work. It needs to somehow
1:13:33understand if it made the right cuts,
1:13:35and it will still take screenshots or
1:13:36draw similar conclusions. It will still
1:13:39spend tokens on analysis and so on. But
1:13:42what is the important nuance? An MCP
1:13:44server, unlike Computer Use, is
1:13:47significantly cheaper. It might even be
1:13:49ten times cheaper than the same thing
1:13:51with computer use. But the quality, as
1:13:53a rule, does not improve significantly.
1:13:56There will still be flaws in complex
1:13:57programs that you will have to fix. And
1:14:00one last thing. If your chat hasn't
1:14:03found a specific MCP server for a
1:14:05specific program, you can also just ask
1:14:08it directly in the code to create that
1:14:10MCP server. And it will tell you how
1:14:14useful this MCP server will be, which
1:14:16functions in a given program it can
1:14:18actually use and access via code, and
1:14:21which it cannot. There is no magic here
1:14:23either, but it's worth a try. And we
1:14:26move on. Hello, friends. In this video,
1:14:33we will take a very detailed look at
1:14:35the image tab and learn how to edit our
1:14:38images, for example, taking a photo
1:14:40like this and turning it into a studio
1:14:42one like this. Also, for example, we
1:14:45will learn how to create covers like
1:14:47these and, in principle, any
1:14:48infographic. We will also talk more
1:14:51about how to edit images. For instance,
1:14:54creating something like this from this,
1:14:56changing specific elements. And at the
1:15:00end, we will talk about how to get rid
1:15:02of, or at least try our best to avoid,
1:15:04the redrawing artifacts that inevitably
1:15:07appear. Like these spots and all sorts
1:15:10of scary lines. Let's get started. Well
1:15:13, of course, we need to talk separately
1:15:15about image generation. Currently, GPT
1:15:17is the best image generator on the
1:15:20market, the highest quality and most
1:15:23accurate. People use ChatGPT to make
1:15:26YouTube thumbnails. They make banners,
1:15:28they make menus. I have noticed that
1:15:31many restaurants have recently simply
1:15:33generated various dishes for their
1:15:35menus. And they are even creating
1:15:36entire menus in this way. And we, of
1:15:38course, need to briefly talk about how
1:15:40to achieve maximum results. So, you
1:15:43have this special image tab here. First
1:15:47, you can simply choose some examples
1:15:49here, play around with them, and
1:15:51randomly type in what you want to
1:15:52generate. There are templates. For
1:15:55example, you can create a business
1:15:58portrait of yourself, take one of your
1:16:00photos, for example, let it be this one
1:16:02, and attach it here. It asks
1:16:05clarifying questions. Let's summarize.
1:16:09Realistic photoshoot, soft studio
1:16:12lighting. And yes, that is the result
1:16:16we got. Pretty good. Of course, besides
1:16:20the templates that ChatGPT itself
1:16:22offers, you need to have a library of
1:16:24examples to get a specific result. For
1:16:27example, you are preparing a YouTube
1:16:29thumbnail, and you need examples. You
1:16:32need to show something to ChatGPT,
1:16:33because it is precisely when you show a
1:16:35specific example that ChatGPT
1:16:37understands exactly what it needs to do
1:16:39. Using only words, you will get a
1:16:41rather poor result. And, for example, I
1:16:44want to get a thumbnail in a style like
1:16:47this one. I go to the Pinterest website
1:16:50. It’s a board where you can find a
1:16:52bunch of visual examples like this. I
1:16:55enter "YouTube Thumbnail" here and
1:16:59choose what I like. You can simply take
1:17:02a screenshot. As a reminder, on a Mac,
1:17:05it's Shift + Command + Control + 4.
1:17:08Let's take a screenshot. Paste it here.
1:17:12Insert our photo again and ask it to
1:17:15make a YouTube thumbnail. Make a
1:17:20YouTube thumbnail for a free course on
1:17:22ChatGPT. First photo. Here is an
1:17:25example: instead of the old logo, use
1:17:27the ChatGPT logo and style the person
1:17:29for YouTube. Lighting, shadows, color
1:17:32correction, and glow. Let's give it a
1:17:34bit more time to think. And yes, it’s
1:17:37not bad at all. Actually, there isn't
1:17:41much for me to correct here, but notice
1:17:43that you can click the edit button, and
1:17:45ChatGPT opens a separate menu where you
1:17:48can, for example, remove the background
1:17:50, erase something, or highlight
1:17:52specific elements for it to remove. I
1:17:55won't do that for now. You can also use
1:17:58the marking tool; for example, you can
1:18:01select an element like this and write
1:18:03what you want to adjust within it. I
1:18:06won't be doing that for now either. And
1:18:08, of course, you can change the size,
1:18:10selecting whatever you need for a
1:18:11specific social network. It’s super
1:18:13convenient. But what problem might
1:18:15arise here after re-generation? Let’s
1:18:18experiment. For example, I give it some
1:18:21adjustments. Move the ChatGPT cube a
1:18:23bit toward the center of the hand. It
1:18:26doesn’t matter where I do this, in
1:18:28this chat or in a separate editing
1:18:30window. Okay, we got the result. It
1:18:34moved something slightly. Next. Add a
1:18:36headline or a note saying "October 2026
1:18:40". Let’s make it small. Yes, it added
1:18:43the caption. Okay. Let's do one more.
1:18:46Add a caption saying "5 hours and 15+
1:18:50lessons". Perfect. But look at the
1:18:53problem that emerges. If we compare it
1:18:56now to the very first image, this one.
1:18:58Look at how high-quality it is without
1:19:01any artifacts. And now, if we take the
1:19:04very last image, look at how strange it
1:19:07has started to look. Some spots have
1:19:10appeared on it, especially on complex
1:19:13objects. Everything seems to have
1:19:15become low-quality, blurry, and stained
1:19:18. Why does this happen? The fact is
1:19:21that the LLM doesn't just redraw a
1:19:23single element and insert it precisely.
1:19:27Every time you edit an image this way,
1:19:29ChatGPT redraws it entirely from
1:19:31scratch; that is, it takes the previous
1:19:34version, looks at it, and tries to
1:19:36recreate an exact copy. And since this
1:19:40is an LM system, and an LM is not a
1:19:42calculator, it makes tiny, pixel-level
1:19:44errors from time to time. And every
1:19:48time, with each redraw, little
1:19:51artifacts appear that blur the final
1:19:54image like this and create these spots.
1:19:58AI spots. Look at how visible they are
1:20:00if we zoom in. Some lines just appear
1:20:03like this, and so on. And this is a
1:20:05problem that many creators don't combat
1:20:07at all. What do you need to do to solve
1:20:10it? Once you have gathered the final
1:20:12result, you need to ask ChatGPT to redo
1:20:14it from scratch. Okay. And upload the
1:20:19original image here once again. "Okay,
1:20:22redo it, but use the source files,
1:20:25don't redraw the last result, so we can
1:20:28get rid of the redraw artifacts." And
1:20:33look, it might not be perfect, but it's
1:20:35a little bit better. The serious
1:20:39artifacts we had over here are no
1:20:41longer noticeable. But the only thing
1:20:44is, make sure you double-check with it
1:20:46so it understands it needs to redraw
1:20:48from scratch. Take this result as the
1:20:51composition goal, but draw from the
1:20:53sources instead of redrawing this
1:20:55specific version. Because each time,
1:20:59these artifacts will stack on top of
1:21:01each other with every redraw. I hope
1:21:03this was useful. And now, we are moving
1:21:06on. Hello, friends. In this video, we
1:21:11will finish up two tabs. The "Scheduled
1:21:14" tab, which is quite simple and
1:21:16intuitive, and the "Library" tab. Let's
1:21:19see how these two tabs can be useful
1:21:21for us. Let's get started. Of course,
1:21:23we need to spend a couple of minutes on
1:21:25the "Scheduled" tab. In fact, it's
1:21:27quite simple, and I think you could
1:21:30have figured it out without me. What
1:21:32can it do here? You can schedule
1:21:35recurring actions—that is, set an
1:21:37instruction and choose when to execute
1:21:40it. A trivial example. Show the Bitcoin
1:21:44price daily. Let's say, I don't know,
1:21:47at 11:00 AM, create. And yes, every
1:21:49morning, every day, it will send me the
1:21:52Bitcoin price. Google it and send it.
1:21:55What else can it do? You can ask it, I
1:21:58don't know, at 10:00 AM to analyze your
1:22:00mail and provide a summary. What do you
1:22:03need to do? For those who actually work
1:22:05in large companies, this is a very
1:22:06relevant feature. Simply put, you can
1:22:09set up a kind of assistant here that
1:22:12will provide you with specific
1:22:13information at a specific time. It's
1:22:16all simple and trivial. Since we
1:22:19touched on a new tab, let's finish up
1:22:21this "Library" tab, which also appeared
1:22:23just a few days ago. What is it for?
1:22:28Now, whenever you chat and use any
1:22:30files. whether you create them, edit
1:22:34them, or upload them here, it doesn't
1:22:36matter if they are Microsoft Office
1:22:37documents, some other types of files,
1:22:39or images. They all end up here in the
1:22:42library from all your chats. Here they
1:22:45are, all the files we’ve essentially
1:22:47been working with in this course. You
1:22:49can download them, rename them, discuss
1:22:52them separately, create folders, sort
1:22:54them, and so on. But what is this
1:22:56library for? Now, to access a specific
1:23:00file, you don't need to hunt through
1:23:02dozens or even hundreds of your chats
1:23:04to find it. For example, a week ago I
1:23:07created this cover right here. And I
1:23:10just don't remember which chat I did it
1:23:12in. Before, I would have really needed
1:23:15to find that chat. Now, I don't have to
1:23:17do that. I just need to write: "Send me
1:23:21the last YouTube cover we generated."
1:23:27And yes, it sent me directly to the
1:23:29chat where I made it. You might have
1:23:31noticed that it didn't search in the
1:23:33library, but actually searched through
1:23:34the chats. Yes, it has that
1:23:36functionality too, but in theory, it
1:23:38should have just looked in the library.
1:23:41Since this feature is new, apparently
1:23:43OpenAI hasn't fully polished it yet.
1:23:46But just the day before yesterday at
1:23:48the presentation, I heard from the
1:23:50creators that every chat, when
1:23:51searching for files, should indeed look
1:23:53in the library first. I think they will
1:23:56finish tweaking this in the next few
1:23:58weeks. And yes, you don't actually need
1:24:00to search for anything in a new chat;
1:24:02you can just come here and see what
1:24:04you've been working on. All your files
1:24:06and documents are now here. Well, let's
1:24:09move on. Hello, friends. In this video,
1:24:15we will look at how the $ 100
1:24:16subscription differs from the $ 20 one.
1:24:20We will talk about the AI assistant,
1:24:22which is only available on this
1:24:23subscription, and about collaboration
1:24:25and teamwork. Let's get started,
1:24:27friends. This video is for the wealthy.
1:24:30I'll say right away that I don't
1:24:32consider myself one, but nevertheless.
1:24:33Not everyone can afford a $ 100 +
1:24:36subscription. Every month, when they
1:24:39send me the bill, I chew my hands off
1:24:41up to the elbows. Then they grow back
1:24:43again. So, I switched to my main
1:24:46account where I have the $ 100 +
1:24:48subscription. And I just want to show
1:24:52you how it differs. You can draw your
1:24:54own conclusions on whether you need to
1:24:56use it or not. For now, it differs in
1:24:59two things. First, I have this button
1:25:02here with this cute icon and the name
1:25:05Gleb. This is a so-called AI assistant
1:25:09that is always online. It was also
1:25:12introduced just 4 days ago. I have a
1:25:15separate, full-length video about it on
1:25:17my channel. What can it do, and what is
1:25:20its intended function? First, it always
1:25:23runs in the background and tracks your
1:25:26current tasks. For instance, you
1:25:28connect it to your email, and it reads
1:25:31all your mail. If it sees things that
1:25:34need your attention, like a meeting
1:25:37tomorrow at 4:00 PM. Or if someone
1:25:40wrote to you about it, or you have
1:25:42something scheduled in your calendar,
1:25:44and so on, it will tell you about it
1:25:46from time to time as the situation
1:25:48requires. You can write to it directly,
1:25:50just like a real assistant, to set
1:25:53tasks or reminders. Remind me of this
1:25:56at such-and-such a time, and so on. You
1:25:59can even call it like this. In reality,
1:26:02it’s just an emulation of a call. In
1:26:06fact, it just switches to a standard
1:26:09voice mode. I’m currently testing
1:26:15this, trying to shove absolutely
1:26:17everything into it: my thoughts, plans
1:26:20for future videos, my expenses,
1:26:23tracking emails here, and so on and so
1:26:25forth. I’ll say right away that
1:26:28it’s currently super-duper useless.
1:26:31It doesn’t really change much; it’s
1:26:33just a layer on top of regular ChatGPT.
1:26:36Overall, you can just as easily create
1:26:38a new project in Code/Custom GPTs and
1:26:40write out all these instructions
1:26:42yourself so it monitors your mail and
1:26:44sends you information. You can send
1:26:47that information there yourself, and so
1:26:49on and so forth. We’ve already
1:26:51discussed Skills, Agents, MD, and
1:26:54various settings. You can now fine-tune
1:26:56absolutely anything you want. Yes,
1:26:59unfortunately, the only downside is
1:27:01that you won’t be able to change the
1:27:03name or have a nice avatar like this
1:27:05one. And the second thing is this tab,
1:27:08the "Spaces" tab. In one of the first
1:27:11lessons of this course, I explained
1:27:13that teamwork in ChatGPT isn't
1:27:15currently implemented in the best way.
1:27:18And second, it’s only available on
1:27:20higher-tier subscriptions. For now,
1:27:22that means a subscription starting at $
1:27:24100. What can you do here? You can
1:27:27create a shared space where you can
1:27:29create things like presentations—
1:27:30which are coming soon, by the way—
1:27:32spreadsheets, websites, images, and
1:27:34organize everything into folders. And
1:27:37another person, also a wealthy person
1:27:40with a $ 100 + subscription, will be
1:27:42able to collaborate on all the
1:27:44documents here if you grant them access
1:27:47. Basically, that’s all there is to
1:27:50it. The last useful thing here is Pages
1:27:53. I am sure that someday, and I think
1:27:57quite soon, they will add Pages to the
1:27:59twenty-dollar subscription as well. But
1:28:01for now, it looks like this. You can
1:28:04create a separate space for a page. And
1:28:07here you can create, for example,
1:28:10headers, bulleted text, some images,
1:28:12and so on. And do it all in a specific
1:28:15view. You can even create nested pages
1:28:18here. For example, create a table,
1:28:21create some text, create a nested page.
1:28:25In general, those who love Notion, I
1:28:26think, will appreciate this
1:28:28functionality. This is, roughly
1:28:29speaking, something built directly into
1:28:31ChatGPT. And here you can also write
1:28:34prompts, and it will somehow edit this
1:28:36page. But once again, there is nothing
1:28:39supernaturally useful here. You can
1:28:42quite easily use a separate Notion and
1:28:44just attach a connector to it, here it
1:28:47is, and use it separately in your
1:28:49program. But if you like having
1:28:51everything inside ChatGPT, then it
1:28:53might be worth considering. In the end,
1:28:55with the hundred-dollar subscription,
1:28:57you currently have a rather useless AI
1:28:59assistant and a workspace tab where you
1:29:02can work with other wealthy users. And
1:29:05additionally, in this tab, you can
1:29:06create these kinds of beautiful-looking
1:29:08pages. And that is all the differences
1:29:10for now. Well, of course, besides the
1:29:12increased limit. 100 dollars just gives
1:29:14you five subscriptions of 20 dollars
1:29:16each. That's it. And we move on. Hello,
1:29:22friends. This is one of the last videos
1:29:25of this mini-course, and we will talk
1:29:27about a large practical assignment to
1:29:29practice everything we have learned so
1:29:31far. Let's get started. Well, friends,
1:29:34it is time to put all our theoretical
1:29:37knowledge into practice. And therefore,
1:29:40in the same file that was attached
1:29:42either in the comments to this video or
1:29:44at the beginning of the course, where I
1:29:45left skills and plugins for you, there
1:29:47are practical assignments for the
1:29:49ChatGPT course at the bottom. In
1:29:52reality, everything is quite simple.
1:29:55First, you need to set up your ChatGPT
1:29:57and tell it about yourself, as we did
1:29:59in one of the first lessons: who you
1:30:01are, what you do, what your hobbies are
1:30:03, and so on and so forth. Even send it
1:30:07your photo. Let it know what you look
1:30:09like. Second, you need to work in Work
1:30:12mode or Code Interpreter mode, create a
1:30:15project, and ask it to create a
1:30:17presentation, for example, in
1:30:19PowerPoint. In this presentation, you
1:30:23need to put together a few slides about
1:30:25what exactly you learned in this course
1:30:27, what you can now do, and how you can
1:30:30use it. I should start by saying that
1:30:33you need to create the presentation
1:30:35specifically through ChatGPT, meaning
1:30:37you should ask it directly: "Create a
1:30:39presentation in this style, with this
1:30:41title, and this information." I have
1:30:43studied this and that, and now I
1:30:45understand this and that. Create a
1:30:47beautiful presentation from this. Feel
1:30:50for yourself how you can now manipulate
1:30:52other programs with the help of ChatGPT
1:30:53. Third. Create and edit your photo, as
1:30:58we did in one of the previous lessons.
1:31:00Choose any style you like. You can go
1:31:02to Pinterest; you don't necessarily
1:31:04have to make a YouTube thumbnail. Just
1:31:06look at how people edit their photos.
1:31:09Give an example to ChatGPT. Give your
1:31:13photo to ChatGPT, and it will produce
1:31:15the final result for you. By the way,
1:31:18you can also ask it to include this
1:31:20photo directly in your presentation.
1:31:22Fourth. Create a personal website. Ask
1:31:25ChatGPT to create a website about you.
1:31:29It now knows information about you, it
1:31:31has your image, it has your
1:31:32presentation, and so on. I remind you
1:31:36that you can create a site right here
1:31:39in the websites tab or just ask it to
1:31:42do so in any chat. Also, for
1:31:45inspiration to create some cool visuals
1:31:48, you can go to Pinterest, type in "
1:31:50websites," and find an example of a
1:31:53site you like, in a style you like.
1:31:56Take a screenshot, send it to ChatGPT,
1:31:58and it will use that style as a
1:32:00reference. The only thing is, don't
1:32:04forget that while ChatGPT can create
1:32:06all the visuals, some text, vector
1:32:08lines, and certain icons, it cannot
1:32:10create images itself. Meaning it won't
1:32:14be able to create a site like this
1:32:16without such images. Therefore, if you
1:32:18need images on your site, ask it to
1:32:21generate those images or send your own
1:32:24photos. It will use them directly on
1:32:27the site. And separately on this site,
1:32:30create a dedicated section with a brief
1:32:33overview of your presentation and
1:32:35learning experience, and add a
1:32:37clickable link so that anyone
1:32:39interested can download the
1:32:41presentation from your site. And fifth,
1:32:44share your results. All of this is, of
1:32:46course, optional. In fact, the entire
1:32:48practice is optional. No one is forcing
1:32:50you. Nevertheless, if you'd like, you
1:32:52can share a link in the comments under
1:32:54this video. I think it would be really
1:32:57great for me and the other students to
1:32:59see your results. This is our big
1:33:02practical assignment covering all the
1:33:05topics we have studied in ChatGPT. See
1:33:08you in the next video. Well, friends,
1:33:15we have reached the end of this course.
1:33:18And if you have completed it entirely,
1:33:20you have done a great job. Learning and
1:33:23applying the knowledge you've gained is
1:33:25hard work, and not everyone can carry
1:33:27it out effectively to the end. And if
1:33:30you are one of those people, I am proud
1:33:32of you. Of course, I want to say a
1:33:34special thank you for choosing my
1:33:36course and for your trust in it. I hope
1:33:38it has been useful to you. I really
1:33:41tried to gather here everything
1:33:43important you need to know about
1:33:45ChatGPT today and to show not just
1:33:47individual features, but how all of
1:33:49this can be used in real life. I hope
1:33:52that after this course, ChatGPT is no
1:33:55longer just a chat where you can ask
1:33:57questions, but a full-fledged tool that
1:34:00can help you with work, information,
1:34:03documents, your browser, apps, images,
1:34:06code, and a huge number of everyday
1:34:09tasks. But now, the most important
1:34:11thing is practice. Do not try to
1:34:12memorize absolutely everything from the
1:34:14course. Choose a few tools from this
1:34:17course that suit you best and start
1:34:20using them today. And be sure to
1:34:22complete the final practical assignment
1:34:25. Practice is exactly what best helps
1:34:27to reinforce everything we have covered
1:34:29together. If this course turned out to
1:34:31be useful for you, please support it
1:34:33with a like. For you, it is literally
1:34:36one click, but for me, it helps a lot
1:34:38to promote such free materials and to
1:34:40understand what to do next. Also, be
1:34:44sure to leave a comment, tell me which
1:34:46ChatGPT feature surprised you the most,
1:34:48what you have already started using,
1:34:50and what topics you would like to see
1:34:52in future videos. I read the comments,
1:34:55and I often choose topics for new
1:34:57content based on them. Subscribe to my
1:35:01channel so you don't miss the next
1:35:03videos about artificial intelligence,
1:35:05ChatGPT, and the new tools that appear
1:35:08every day. I will leave all the links
1:35:11to my social media in the comments
1:35:12under this video. There, I also publish
1:35:15useful tools, instructions, and
1:35:17materials that don't always make it to
1:35:19YouTube, of course. Well, with that,
1:35:22our big course on ChatGPT comes to an
1:35:24end. Thank you for watching it to the
1:35:27end, and thank you for your time. I
1:35:29hope the course was truly useful and
1:35:32will help you use artificial
1:35:33intelligence not just out of curiosity,
1:35:36but for real-world tasks. This has been
1:35:38Dmitry Fakeev. See you in the next
1:35:40videos. Bye. M.

About this transcript

This page contains the full transcript of Большой курс по ChatGPT с нуля (Октябрь 2026) by Dmitrii Fokeev, generated from the public captions YouTube serves with the video. The transcript has 15,431 words across 2,292 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 translate the transcript into 100+ languages.

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, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.