As we all know, artificial intelligence changes every day, every week. And you can know the names of absolutely all the models. You can say that you've really figured the system out and still use artificial intelligence for the simplest questions and not just fall short of using it to the max, but essentially never even get close to the most important things about artificial intelligence. Or you can truly get to grips with it, so that artificial intelligence gives you that same real maximum.
And at ToTheMoon, we've been talking about artificial intelligence and its development for more than two years, and most importantly — about bringing it into people's lives. And today I want to go through the ten key rules for getting the most out of artificial intelligence in work and life this year, right now, yes, in the fourth quarter of 2026. And it's important to me that you understand the approach yourselves — how to work through even unfamiliar tasks together with artificial intelligence and see them through to a result.
How to boost your life, your effectiveness. How to boost your understanding of everything that's going on. Because the models will keep changing, and you won't have to be in some kind of stress or confusion every time, especially when you see certain news headlines. And in a few days I'll release the second part: what to learn and what not to waste time on, and which opportunities you must not miss in the age of artificial intelligence. So, let's kick off. Look, here's what actually happened. What did I do? I'll show you this whole project as we go.
I analyzed absolutely every episode we've had on the ToTheMoon channel, more than 170 of them, and picked out the trends, ideas, assumptions, thoughts that I personally voiced throughout these two years and that I saw coming, or see coming, or that I consider critically important and necessary. What does "I myself" mean? It means that it's my opinion — and my opinion, of course, is also based on my contacts with different people: businesspeople, people from Silicon Valley, tech leaders, and, of course, on studying the market as a whole, and the statements and opinions of the owners of various companies, economists, politicians, anyone at all.
So I transcribed — well, I took this whole volume, this whole mass of information — and asked the system, in this case Claude Code, and partly ChatGPT too, to help me formulate these ten most important rules for work and life. In total there are actually twenty thoughts there. And once I'd done that, I also posted it on our ToTheMoon project website, and you can browse and look at the description of each of these ideas. And we'll definitely give you the link, and you'll be able to use it yourselves and forward it to other people, especially those who want to figure out, in particular, what to do with artificial intelligence.
Although of course it's always useful to listen, to listen through the videos. But what's interesting there — a neat feature — is that you can see where, in which month, when a particular idea was mentioned. Yes, I think that's an important setup. Because what kind of world are we living in now? We live in a world where as recently as a year, a year and a half ago, everyone was saying that AI needed to be built into, for example, the browser.
And everyone rushed to build a browser, including, for example, OpenAI. Everyone was shouting about how useful it was. I, for one, said right away that there'd be no browser, and that people wouldn't work with artificial intelligence in the browser in that form. And in the end OpenAI recently wound down its browser project, and now the issue is that a huge number of people invested in that opinion, spent a huge amount of time and resources, and once again they don't quite understand what's going to happen next. So, number one.
The thought I want to give you for this year. Once again, today there'll be ten of them, and all ten are important, of course. I mean, there isn't a single thought — In no way do I want to tell you that we'll go from the least important to the most important and the most important one will be at the end of the video, but all ten carry a fundamental meaning. The first thought is that you need to get the paid version of the system and switch on the strongest mode. Let me remind you, especially those who've only just joined our channel: from the very beginning, for several years now, I've been saying that using artificial intelligence has to go hand in hand with you setting aside some money for it, right.
And today the paid version, the basic paid version, usually costs $20 — be it OpenAI ChatGPT, be it Anthropic Claude, be it Grok, be it Gemini. They cost roughly that much. There's a cheaper version, say for $10, some charge $25, $30, but roughly it costs $20. Someone might say that $20 is expensive, that it means we won't be able to have one extra coffee… or buy one. For some people, of course, it's critically important. They won't be able to buy their kids a pencil case for school.
But I want to say that, from what I've observed, most people who — for whom $20 is an absolutely affordable figure — they can afford it. I see how they don't allow themselves to. I was traveling with a friend of mine, and he's absolutely fine with going and buying a coffee for $20, but spending $20 on ChatGPT — that he can't do. For him it's some kind of incomprehensible subscription. It's not clear why you'd do it. Once again: this isn't a Telegram Plus subscription, it isn't getting a checkmark on Instagram, it isn't a Netflix subscription — it's a new approach to developing yourself.
And yesterday my sister said to me: "Listen, Sasha, if a person earns a $400 salary, he can't pay $20 for a subscription to the system." And here, of course, there's always the following point: if a person can't and never will earn more than $400, and those $20 bring immense benefit to his family's development, and he doesn't have enough money to pay the mortgage or utilities, or really to buy basic food. That's just how it is. But if a person allows for the idea that he wants to grow in terms of income — and after all, when money is invested, it's also a question of earnings, right.
Or if a person wants to improve, to some degree, the effectiveness of everything going on around him, then of course he can pay $20 out of $400. It's a question of how you think, how you relate to it, right? I mean, how do you fundamentally relate to it at all? Personally, I spend a lot of money on artificial intelligence. Yes — not 20, not 100, not 500, and not $1,000. I want to say that I, personally, don't calculate what else I could have spent this money on. I spend as much as I think is necessary in terms of current effectiveness.
If you don't switch on the paid model, and if you don't choose — this is very important, this is what I want to get across and tie together in this point — if you don't learn to choose the stronger mode in these systems, you'll be losing out all the time. These days ChatGPT tends to — and does it on purpose — drop you out of the strong mode even if you have a paid subscription, right. In ChatGPT, essentially, the paid plans don't have any hard limits. Well, for example, the top-tier subscriptions have no limits as such at all. They sort of exist, and sort of don't.
And I never, ever run out of limits in ChatGPT Pro, for example. They run out in Codex — that's a separate story — but not in the chat itself. And I think that in the $20 version, too, the regular chat doesn't run out. What can run out there is, for example, voice usage and so on. So you should always use the top model and double-check it, because ChatGPT will keep dropping you down even when you've chosen a great model, you can write the very next message, and it's already dropped you to the cheapest option, because it isn't interested in you constantly using the biggest models.
And that's a huge cheat code. Many people don't see this, because today the answer from a free or basic model, from a good paid model, and from a seriously expensive top-level model — the difference in results is already catastrophic. Some questions the system will answer more or less the same way, but overall, if you want to dive into artificial intelligence at a fundamental level, the difference in results is, well, really huge. And this is a critical rule. Right? So, again, when we talk about the $200 plan in many systems, or $250 somewhere, or there's a $100 one, then probably not everyone needs it.
But my wife and I were driving — she pays for the $20 plan — and with my sister too, literally yesterday. And I say: "Listen, don't you want to get the $100 one?" — "Nah, I definitely don't want to pay 100." Why doesn't she want to pay 100? Because here, of course, she no longer sees the effectiveness directly, in terms of those $100 paying off. Whereas I would see the $100 paying off for her, because in terms of spending, life, keeping everything around running, if she used additional systems — I mean, you should still pay the $100 if you start using the internal systems, such as OpenAI Codex or Claude Code, right — that is, when you're already starting to build some kind of automation in there.
So if you're not building automation — no need. But we'll talk more about automation later.
So, the second thought and the second extremely important point, which I'd say is absolutely fundamental for people who ask to have it explained to them: what's the difference between ChatGPT and Codex? And what's the difference between Claude Code and Claude? And why should I pick DeepSeek or Gemini? And then: "Tell me, something's glitching here for me. And I'm not getting any kick out of this. And I can see this system hasn't developed all that much. And why did they release this new version? And why should I switch between these versions?
Everything already works for me in the current one. And ChatGPT isn't going to help me any more," and so on. When it comes to growth, to understanding artificial intelligence, there's only one salvation — that you start using AI every day. And, importantly — in everything: at work and in everyday life. I mean, if you only use ChatGPT for certain tasks, for example checking particular news or writing some editorial text, or checking for mistakes before posting on Instagram, or, I don't know, analyzing some particular documents, then you won't understand at all what modern artificial intelligence really is, where it's heading, and how it's developing.
You won't. To understand it, you'll have to make a huge number of mistakes in there. You need to start seeing the mistakes that artificial intelligence makes, and you need to start working out new usage scenarios for yourself. I remember it very well — we talked about this a lot on ToTheMoon, many times over the last few years — that with the arrival of systems like ChatGPT, you start applying new scenarios in your life, right, new business processes — that is, acting in ways you've never acted before.
If you and I look at how these systems have evolved, even at the coolest companies, all these cool companies are trying to offer some improvement of current life and automation of current business processes. Although in reality they're building a system that's about something completely different. What does "improving current life" mean? Things like: this system will let you order groceries from the store faster, instead of you adding them to the cart by hand. This system will let you click a button on Booking instead of you clicking four buttons. This system will let you find out a hotel's rating not by opening Google to do it, but by just going in there and taking a look.
I mean, that scenario already existed. This is very important — that scenario existed. There are other scenarios that weren't in your system, in your head. You didn't think that way, you didn't try to use something that way. By the way, here's an interesting case. Most recently, Ilnar gave one, in terms of a different kind of use. He's in China right now, and he says he uses Google Translate to communicate with Chinese people. And... But why use translators for communicating, or special communication systems, if you can open, say, ChatGPT Voice and tell it that I'm talking with a Chinese person right now: if they speak Chinese, translate for me into English or Russian; if I speak Russian or English, translate into Chinese.
And also, if there are really complicated things, explain them to me in a bit more detail, but don't add anything of your own, anything extra. And if you don't understand something I said in Russian, or don't understand what he said in Chinese, ask him again so the translation is as good as possible. So, look, in this case the translator turns into a completely different system. It's not using a standard translator, it's not an easier way of using a standard translator — it's a completely different business process, a completely different approach.
When you tell the system, for example: "Please clarify: what do I mean?" And we all know how often translators like Google Translate or Yandex Translate, whichever, or Apple Translate — how they translate some word, and you say: "I didn't mean that word." There are these fun reels where they show some English word, they say: in English there's the word "run." And then they show which word it is, how it translates into Russian. In Russian it's: begu, begushchiy, bezhal, pobezhit, ubezhit, ubezhal, sbezhit, sbezhal, sbegu, sbegut.
And off it goes, a huge number of translations. I mean, we all know that there are words that — well, it's a joke, of course. There are words that have a huge amount of context, a huge number of different meanings, and the person might have meant something else. And how important it is to clarify precisely and find out precisely. Here's a completely different use scenario for you. How do you get to this scenario? Only if you use artificial intelligence even in the things where it will complicate your current life. Even complicate it — not improve it, but complicate it.
When you try to do it and don't block yourself, the way people say: "I already do this quickly, this system already works for me, this business process is already working for me right now, it'll cost me more. Why should I spend even more money on it?" When you start approaching and treating artificial intelligence with the view that it's normal for it to make mistakes, it's normal that you need to set aside extra time for it, it's normal that some business processes will get even harder so that in the future you can solve certain tasks in a completely different way. Which is exactly what we're doing, right?
And you can see how a huge number of tasks are being solved in a completely different way. Here's what I rolled out on the site: the list of trends and the list of thoughts I rolled out. Before, to build a project like this completely, in full, it would have taken roughly six months to a year and seven to ten people.
That's a team of developers — you know, front-end coders, back-end developers, all kinds of system administrators, marketers, editorial staff, project managers. Seven to ten people. This project, what I rolled out on the site, took 30 minutes of my time in total, in terms of the rollout, right, because I did have to prepare for this episode. I really wanted to talk about the thoughts and things I've genuinely been putting out there for a long time, like, for example: use artificial intelligence all the time, right.
I've been putting this thought out there for 41 episodes over the last 12 months. Forty-one episodes. So just imagine: in 41 episodes over 12 months, I keep talking about this. There are only a few months, for example, April 2024. There, in one of the episodes — I talked about it in zero out of three episodes. Or in July 2024 I didn't mention it. For example, starting from 2025, November is the only month I didn't mention it. And after that I repeat it every month. So I keep repeating it, and you can open it and see, basically, where those episodes are. I say it all the time. And over all the years — 2024, 2025, 2026 — I've talked about it 72 times.
Look, to do this, to even count this, to count it, in the past it would have taken — I'm telling you — a cosmic amount of time. Not to mention building this project. People now who see this — I've even been showing a project like this to some people. Many people say: "Ah, well, sure, that can be done now." So that's kind of the attitude we've developed toward AI. On the one hand, most people don't know how to use it, right? And on the other hand, people keep saying that, well, it's all already known, it's all clear how it's done, it all, you know, works great.
And at the same time there's a whole community saying that AI will kill humanity. By the way, I came across an interesting thought. Right now I'm recording this with my sister, she's here with me, and Anya came across an interesting thought: that China considers — a really big study just came out — this narrative from the Americans, that AI will kill humanity, to be pure marketing to hold back competitors. So in China this narrative doesn't exist at all. So the Chinese say: "Nothing will happen. The government and the state will rein in the development of these companies." So at some point, it'll all be held in check.
So there's no way someone will get out of control and, you know, do something. This American narrative, it leans that way, and I probably agree with that to some extent, right, because if you listen to them talk, they're telling you how AI helps you order a sushi set from Instacart properly, right? Or they argue with each other over whether their AI, I don't know, solved some math problem or not, and at the same time they're telling you that it'll destroy humanity, and that it's kind of almost already happened.
And you're like: wait, so if you're saying that, then it turns out you don't understand it yourselves, right? So that in itself is a problem. Of course, I strongly recommend that all of you use it every day, put it into practice and not be afraid of it. A lot of people are afraid of it. For example, my mom — about a year ago, I think, I set her up with a paid version. She made one request on it and didn't use that version. At some point I came over and just switched it off in the admin panel. And the funny thing is, ChatGPT had been gifting me some promo codes for OpenAI, and I had promo codes for, like, three months each or however many.
So she was just eating up that limit, and I was just waiting. Even though I'd connected her to my own payment system. It was all completely clear. And she didn't start using it. Well, I didn't pressure her, although obviously, if she'd started using it every day in her life, it would have been incredibly useful for her. Some of you will say that it's probably, maybe, not useful for your mom. No, my mom is quite active in the social space. I mean, she's present in the world, right?
And she doesn't just, I don't know, sit shut away with her flowers or her plants, right. I mean, she actually spends a lot of time on the internet. I think her time spent on the internet, or screen time, right, is no less than my screen time. I'd guess, yeah, that hers is no less; she's present. And of course, what's more, I showed her some really cool, interesting use cases. So, the third thought when it comes to artificial intelligence — let's note it down for ourselves, right?
As I said, I'll also give you all the links in full. But it makes sense to, like, fundamentally figure out what each of these means. It'll be great if, for these thoughts, you mark where you are right now, what it is for you, right, how much you use it and how much you use, what's the norm for you and what's the standard for you. It'll be great to see that. Even those of you who, well, don't like leaving comments, please do leave them, okay? Because comments are always left by some very small, select cohort of people.
So, the third thing is: work with artificial intelligence directly, without bots, without any wrappers, any aggregators or unnecessary plugins. I want to explain what it means to work with artificial intelligence directly. It means that — look, in the world today there are basically three serious leaders. And beyond that, if you widen it a little, well, five, six, seven companies, right? Obviously, we could go off and look at a whole constellation and lineup of American companies — oops, Chinese companies. But overall there are two top leaders: Anthropic and OpenAI.
A bit simpler than them today, there's a simpler system that sits there. By the way, interestingly — I hope our editorial team will pull it up for you now. Literally yesterday I found a very interesting picture showing where certain models stand. And Elon Musk sort of endorsed this picture. And there's an interesting aspect to this picture. I was just about to tell you that after Anthropic and ChatGPT comes Google. But when they ranked the systems — systems have a certain grading: there was A, B, C, D, E, and there are S-class, SS-class and SSS-class systems.
So, the SSS-class system is Claude Opus 5.5, which just came out — I was only talking about it a few episodes ago. And it really is a very cool model — cool in terms of pricing, cool in terms of what it does, in terms of everything else. By the way, at some point I was actually convinced that Fable was better. Now, judging by this thing, what does it show? That Opus 5.5 is SSS. Maybe we'll do a separate episode on this, but as time allows, right? Because tomorrow I'll be at OpenAI DevDay, and this episode will come out — so you'll see it on Wednesday. I'm recording it today, on Monday: it's Monday, already evening, and on European time it's morning for me.
Tomorrow I'll be at OpenAI DevDay. We'll see what happens there. Maybe OpenAI will say it's SSS, right? And here in San Francisco there's going to be a closed OpenAI event. And ChatGPT-6 Astra is SS, and at the third S there are Fable and ChatGPT-6 Solo. Although I think Fable 5.1 is better than ChatGPT-6 Solo. My personal opinion. Anya, probably yours too, right? 100%. Yes, definitely. For now. Yeah. See, Anya also says it's 100% so. So, let's look at level A. There you've got Opus 5, Fable 5, and, well, ChatGPT. And then further on it gets going, the Chinese stuff gets going: Kimi K3, Moonshot.
Then the American company Grok, Qwen, GLM. Again a ChatGPT model there, Muse from Facebook, DeepSeek, HUI4, Grok. So that's why I'd tell you that there's Anthropic, OpenAI, and below them there's some movement. I'm still not going to rank Google. So Google — what's the funny thing about Google? Google doesn't land in A, or B, or C, or D, or E; it lands in its own category, called Google. But that's probably, it seems, some kind of joke, a total laugh, right? So it lands there, and then after that come the Xiaoyi models, you get the idea. Like I told you, there were Alibaba models, there are even NVIDIA models sitting at level D.
So, all the same, it's Anthropic, OpenAI, Google lower down, and there Google with Grok — that's SpaceX. Lower down I'd put — I'd still put Facebook lower, I'd put it down there. Where Grok and Facebook's Muse are, that's where the Chinese models come in. And the Chinese models, of course, have already overtaken Google in some ways. It's just that Google is a big infrastructure. We'll see whether it loses the system or not. And so my suggestion to you — or rather, not a suggestion, I believe it's impossible without this — is to work with the top models directly, with the top models you're able to connect.
What does it mean not to use bots or intermediaries? It means not working through middlemen. Even Cursor used to be just a middleman. Cursor has now become part of Grok, and so you can say: okay, that was a middleman. For example, Perplexity is a middleman, right? And every day — today, again, someone sent me an email. He sent me an email saying: "I've built a cool plugin that will calculate how you spend tokens on the first system." And I realize that for him to calculate how I spend tokens in the systems, I'd have to connect all my systems to him completely and run all my requests through that system.
How can he calculate that? Because Claude itself and Codex itself can't count it properly. For example, my Codex still says that as of today I've used 49.8 billion tokens. And when I see that number, I have a big question, because just yesterday my sister and I were looking, and she had 2.9 billion tokens, and she's a serious Codex user. Well, like, an insanely serious Codex user on the Pro version, right, she works on the Pro version all the time. So 50 billion tokens looks, like, insanely serious.
I remember once looking at someone's — well, some small bloggers — and it said something like: "I personally used up, like, 500 million tokens." Or something like that. And I'm thinking: "What's 500 million tokens? In Codex alone, it's 50 billion tokens." I mean, how were they spent, what were they spent on, with what efficiency were they spent? I don't need that middleman. I mean, either I'll write my own services myself that'll tell me how I spent them. For example, there's ElevenLabs, an audio-to-text transcription service.
So, ElevenLabs or ChatGPT Voice — well, there's ChatGPT Transcribe — sometimes they don't return the correct amount of money that was spent on a specific request. I mean, sometimes they return it, and sometimes they don't. And you can't even determine exactly how much money you spent on a request — well, on transcribing audio, for example. And that's a very easy, elementary task, because they have a specific rate that's set internally. But the system can say: "You sent an extra packet here, you used some additional data there." So for now the system isn't relevant.
But the point is: you need to work with artificial intelligence directly. Why? Because that's the only way you can get the maximum functionality. I mean, you tell yourself: "I work every day, I've even paid $20 for a certain version." And to get the most out of that $20, you need to work inside that system. You don't need to pay $20 to Perplexity on the side, or pay $20 on the side to some extra aggregator. You need to work inside the most powerful, the best, the coolest systems that exist in the world and are available in your country, right.
There are, after all, systems that may not be available to you. Well, so it connects to the systems, the ones that are available. And the world today is split into certain parts, right. Some people have access to Chinese models, some don't; some have access to American models, some don't. You get to choose what you — what to use inside. There are, of course, a huge number of nuances. For example, there are corporate rules where you can't use certain systems at work, but at home you'd like to use your own systems.
That's already a matter of, well, your own settings and choices. And so sometimes there can be several systems. I, for example, am a user — a professional, constant user of two systems. I mean, I use Claude, I use OpenAI, right. Anthropic and OpenAI, essentially. But I still do all my chats in ChatGPT. Whereas development, automating systems, I do in Codex and in Claude Code, right. To some people this terminology seems confusing, unfamiliar. It's completely unclear what it is in terms of code and everything else.
Be sure to stay with us on the channel, stick around, don't forget to be here, and be sure not to forget to leave comments under this video and support our channel, right, leave likes. It really helps us grow. So, the next thing is not to be afraid to give data to artificial intelligence.
And a lot of people often kind of hold back, saying: "We won't use AI because it'll get too much data." And there's this funny thing. Just yesterday I was also out walking with my sister — well, I mention her a lot simply because we've been spending a lot of time together these past few days. She flew in from Europe to see me, and I told her that all people — you start telling them some idea, and right from the start they begin saying: "In our case, for example, I can't do that within our business processes, say, or within our company, or within my life, and so on.
I mean, I can't hand over some document." And I keep saying: the whole world violates intellectual property every day, violates confidentiality, and generally doesn't understand what's going on. Today there's no difference whether you saved your test results in Gmail — they came to you by email — or uploaded them into artificial intelligence. I mean, logically, you have to understand very clearly that if your medical provider sent your test results to your email, that email — often you don't even know where that email's server is located, and, well, that data is stored in the cloud.
Or you uploaded them to Google Drive, or to Yandex Disk — it doesn't matter where that data is, who even has access to it. Because you'd also need to understand the legislation of the country where that service is located. How it treats its own citizens or foreign citizens, residents or foreign residents, what rules it has for handing over data, details and everything else, what applies there to medical, personal, work data. Look, there are a huge number of nuances and questions, and people don't fully know this and don't fully grasp at all what's actually going on.
They just don't fully understand it; they're kind of not ready to come to terms with it, or something. You know, it's an interesting story: I remember when they started blocking — not blocking, but in Belarus and in Russia they started checking phones at the border, and it caused a huge stir, like, why are they checking our phones at the border? And when I tell everyone that in America this rule has existed — that at the border they have the right to search absolutely anyone's phone, no matter whether you're a citizen of the country or not a citizen, you've arrived at the border.
Why can they search a phone? People just don't understand this rule. Because a phone is something that's physically present at the border. You brought it with you, just like oranges or tangerines, or meat, or gold, or money, or a suitcase, or clothes. I mean, it's physically there with you. It's logical that they can search a phone. And by the way, I didn't fully understand this topic either, until a customs officer told me: "But we don't look at the phone in cloud mode. I mean, we don't connect to the internet — not because some law says so, but because everything that's in your cloud, you didn't bring with you to the border, so we don't have access to it.
But if you did bring it to the border, and it's on your flash storage, then we have access to it, we have the right to look at it." Which, on the whole, is logical from an organizational point of view. But it turns out that we don't fully understand the laws, we don't fully know the rules, we don't study them properly. And that's why we often think that uploading information to ChatGPT means someone will get access to that information — to your, I don't know, medical data or your banking data, or your tax data, or your photos with your husband or your wife, or, say, some birth certificate — that someone will make use of that data.
But at the same time you keep that same data in cloud notes, or keep it in your email, or keep it on Google Drive, casually sending it to each other… keep it in WhatsApp, it doesn't matter, cloud WhatsApp or Telegram, which syncs across all your computers. And in general that's the cloud, it's stored in the cloud. I mean, everyone saves all of this, everyone has this data sitting there, and people don't worry about it. I really went through this myself. I started running the company Megaplan in 2010.
It was a cloud operator in the Russian-speaking space, the largest in terms of task management systems. And I remember that resistance, when people would say: "We don't want to move corporate data to the cloud." I'd say: "But your server is much easier to hack than our cloud. Because on our cloud, on protecting our cloud, we spend millions of dollars, while for protecting your server — you've just got, so to speak, a key to the door." And this is, you know, an inability, I guess, to make sense of it, right — it's somehow impossible to get your head around how it works and how to go on functioning with it.
So, at the same time, on this point I want to say that you should always understand that this concerns not only artificial intelligence, it concerns email too, right? That there are boundaries. You need to check the settings in these systems. And one of my favorite settings, which exists in both ChatGPT and Anthropic: by default, the system is set to train on your data. I mean, you have to — like, yesterday I turned on the new Siri that Apple released, and the first thing they said to me was: "Allow us to train on your data."
And I don't mind it training on some of the data I have on my phone, so that — I have phones, computers, that run developer versions. I get Apple's operating system versions early, right? But I don't want my voice messages — I don't want the system training on them, right? So I click "I don't want to."
So, the defaults in ChatGPT, the defaults in Anthropic — you can switch all of that off. I recommend you figure out how to switch it off, and at least understand in your own head how it works. That if you haven't deleted a chat, it's saved in the cloud. If you have deleted it, then ChatGPT, for example, keeps it for a month. Google, I think, keeps it for up to three months. No, I can't tell you for sure. I mean, you have to read the rules in each system, and they may have changed, too. And they also depend on the plan: free, paid, corporate.
So there are a huge number of nuances there, right? But the fact remains that you can forbid training on your data. There are models where you can't forbid training. For example — this is my opinion — that's why I don't use Chinese models. I'm sure that with Chinese models, even if I tick the "don't train on this" box, it doesn't mean they won't train on it. In America, after all, the law is such that if a company has manipulated something somewhere, it will later face insane lawsuits from a huge number of people.
And so this kind of thing can at least come out at some point, or you'll find out that they trained on your data. In China there basically, well, just won't be cases like that. You just won't find out about it, and you won't understand at all what's going on — who's training on your data, who isn't training on your data, whether they had the right to do it or not. And in general you basically need to understand this, you see? People don't dig into it and don't understand what's going on. So, the next thought — the fifth, number five.
We've gone through four, four. Every time I sit here talking, I realize how incredibly significant each point is. I've repeated them, and I'll keep repeating them. But still, I really like that we — well, our editorial team, our team at ToTheMoon — got to the point where these episodes are coming out for you, right? And today, these ten thoughts that will let you — not even thoughts, but key rules for work and for life — will let you get this effect in the fourth quarter. Well, they're very important, right? So, the fifth rule is: tell AI about yourself and give it your data.
The more it knows you, the more precise the answer. I remember that time, the last 15 years or so of the internet's development, and the huge discussions about how Facebook trains on our data, how Google trains on our search trail, how Yandex learns how we click on ads. And huge arguments over whether it's OK to give them this data or not. And the European Union is probably the quintessence of that kind of hell when it comes to personal data laws. And then along came ChatGPT and Anthropic. And they showed that, as it turns out, a person will share with you not just which little button they clicked somewhere, or what age they chose on their Facebook, or what messages people wrote to each other — but it will know data right down to how you peed and pooped, because you sent those test results somewhere or generated them from Apple Health, right?
From how you walk every day, where you walk, why you walk, what you do, who looked at whom and how, what you're interested in, what color dog you like, what kind of cow you like, what clothes you like and where. It'll know which device you picked where, why you picked it, why this person picked that device. Your earnings, your tax returns. It'll know a hell of a lot of data about you, all of it. And in any case, you won't be able to shrink that knowledge. I mean, you sort of — the resistance to handing over this knowledge — it's already a thousand times beyond what Facebook and Yandex knew.
I have a question for all the people who talked about how advertising remembers things, and resisted it, and because of that held back a great deal of development in companies. I know a huge number of companies that banned installing certain systems, afraid that someone would take something, and so on. And now, today, the whole world is taking photos and screenshots of reports and screens on their private phones, feeding them into ChatGPT, running recognition everywhere, passing them on, and so on.
And if before, Google sort of didn't analyze that email when people personally sent each other messages, or WhatsApp didn't analyze them, now in WhatsApp there is News, which will analyze every single file. Now Apple has launched the new Siri system, which is going to start analyzing your files. Or you have to exclude everything, shut everything off. And then someone comes and says: "Sasha, can you explain something to me? I've got all the services turned off, everything's off. But you and I were just talking about the puma that came to your place, came onto your property, and Instagram served me a reel with a puma.
Can you explain to me how that happened, if everything on my phone is blocked, nobody's listening to anybody, there's nothing? And besides, we were in the forest with no internet, and we didn't even have a phone with us. And yet it still found out about that puma." And you sit there and sit there, and you genuinely don't understand what's going on. And at the same time you realize that even right now I'm in a room, and I've got one phone here, a second phone, my sister's phone, I've got a Mac Studio computer here, a MacBook lying on the desk.
That's serious hardware sitting there. And there's an Alexa too. Serious hardware inside what's just one room. But I've also got a smart device connected to the internet, there's an air conditioner connected to the internet. There might be something else in here, in the room. I can't even tell what anymore. Some light bulbs, maybe, smart ones, or some smart plug. Ah! I've got — here, in these spaces at least, there's an ironing cabinet, which is also connected to the internet.
These days, what isn't connected to the internet? You no longer even understand what's listening to you, or where you are. Maybe on the ironing cabinet you just didn't set that option — not to listen to you. And there's a microphone standing there in the room, listening, training on the data. It says: "I'm only listening to this so I can figure out how best to freshen up your jacket." Right? And you don't know what to do. So what's my take? That if you live in a sane world, if you personally are in a sane system, then nobody needs your health data. And if somebody does want to get it, believe me, they'll get it without you.
Your tax data — they know it anyway. You upload it into the tax systems anyway. By the way, I want to say that OpenAI and Anthropic are much better protected than any tax service in the world, I think. I don't think there's a single government system today, except maybe some military institutions, that's better protected than Anthropic and OpenAI. Although I think even the military is less protected than Anthropic and OpenAI, because the data that's in OpenAI and Anthropic is humanity's data.
Can you imagine what that is? There's a project that's being discussed a lot right now — that Anthropic is scanning all the books in the world. And now there's a request to buy up 100,000 Japanese books. And they say they cut off — so, they take a book, cut off its — I forget what it's called.
The spine.
They cut off the spine here, and these pages get scanned automatically, all the literature and books in full, and then, well, they go into the 'shirver', right, and get destroyed. And there are whole debates about how much time people spent on this. And all you're doing is training a system, right? I mean, these systems — the volume of information inside them is colossal, and it has to be protected. And it's clear that these systems obviously pose a big threat to any country today. Access to these systems, in general.
Imagine connecting to, I don't know, Brazil — to the people, so to speak, who use ChatGPT. Well, that's very serious, right? There was that case when OpenAI's American office handed over data to, I think, a Brazilian agency — I think Brazilian, it was in South America — or an Argentine agency, because some person was preparing to kill other people. They passed it to the police, the police went and dealt with it. I mean, that's data of a very serious nature. And so your data about what you think of your wife, or what you think of your husband, or what you think about cheating, or what you think about raising kids and all that — well, nobody needs it.
At the same time, of course, you have to understand that there are some things that, in your country, in your location, it may be in your interest not to share. And you need to learn the laws. I mean, there are laws, in the US for example, that are objectively normal. You'll write in ChatGPT: how do I properly optimize this tax system. And there are countries where you'll write the query: how do I properly optimize the tax system. The very presence of the word "optimize" says that you want to, I don't know, evade something, rather than properly sort out your taxes, right?
And that leads to certain consequences. So, of course, you have to understand that. By the way, I made an episode on this topic, and in that episode I talked about how the system reacts very strangely to certain countries and certain laws. Even if the system is American or Chinese, they're very strange. I mean, American systems, for example, restricted data use in certain countries — in China, or in Vietnam, or in Russia. Even though people actually just wanted to understand something. The system seemed to be helping the state, yet it has no agreements with the state to do that.
Well, I mean, it was a somewhat strange story. So, it would be great if you actually wrote in the comments what you're not ready to tell AI. It would be really interesting to do a quick roll call like that. So go in now, hit pause, write it, then keep watching, so you don't forget to do it — what you're not ready to tell artificial intelligence, specifically in your location. Someone will think right now: hold on, if I write what I'm not ready to tell, then YouTube and the systems and the comments will keep tabs on me, my government will see it, and I'm toast.
I think it's also really important to explain to people what training on data means. Because the phrase "training on data" itself can be interpreted in extremely different ways and scenarios. You could even make a separate episode about it. What does training on data actually mean? What specific information does the model take, how exactly does it use that information, what data is kept, is that data anonymized or not anonymized? Maybe there's value, after all, in strengthening artificial intelligence, in its more competent and higher-quality… well, how to put it, its higher-quality and more competent development, when artificial intelligence has the chance to train on the data of competent, smart people, professional people.
So here's the thing: what you're saying now, Anya, is exactly the point that artificial intelligence would then develop more properly.
How can artificial intelligence develop if it only knows and analyzes information about how to peel tomatoes for soup? But if it really sees the whole range of questions that concern a person, certain solutions that work out, or certain solutions that don't work out, then on that basis it can grow stronger and really go on to help a person, humanity, anyone, any individual in general. So I think that, first of all, we don't fully understand what "the model trains on our data" means, what specific information it takes and what connections it picks up.
After all, training on our data may have nothing at all to do with my identity or me as a person, with the fact that Anya Volchek asked this question or uploaded this information. I mean, we have, we're dealing with really fine lines here. What does personal data mean? What does training on our data mean? What's the difference between using personal data and training on the information in chats? So, basically, you need to understand what's behind all of this. There's a lot of fear, and it's based on ignorance, as always. Right now there's actually a paradox: people create content with the help of models, and then they themselves forbid the models from training on that content.
I mean, they create content, I don't know, strategies, solutions, how to get treatment, which marketing strategy to choose, how to solve one problem or another. They get all of this with the help of a model and then forbid the model from using that content. So they pick the best result with the model's help, the one they currently think is best, but forbid the model from training further on that best result. But that's a paradox, isn't it?
Look, this thing is 100% paradoxical. In particular, it's very specific to basic human perception. Right now the models, look, basically all artificial intelligence systems are set up to move toward superintelligence. I mean an intelligence that's capable not just of doing some single task or carrying out some agentic work, right, or putting together some little report, but an intelligence that will be able to run a country or businesses, or that, essentially, could be a ministry, could be some kind of organization.
So, a fairly fully functioning construct. And when we say that it's an intelligence that will be able to replace a person somewhere, in certain actions, and people talk about this a lot, that in itself is actually paradoxical. First of all, you won't be able to replace a person, because a person has physical flesh, and here there won't be any physical flesh, there's no connection. On the other hand, current models are already smarter than any person in terms of knowledge, the number of questions, the number of answers, speed of work, they don't sleep and so on, right?
You can deploy computing power, they can deploy simply insane amounts of computing power. And yet the models seem limited when it comes to really functioning with a person. I'm always, you know me, I'm always saying that a model isn't capable of handling and managing, or helping to manage, all the projects I have. I mean, as of today a model can't actually analyze my life and make sense of it. So overall, I'm interested in giving it even more data, in sharing even more different issues with it.
And there's this cohort of people who say: "Nobody's going to share anything." So nobody's going to find anything out. Like the example with Amazon I often used to give, back before AI. Why doesn't Amazon find out right from the start how many kids I have, why doesn't it properly ask me how old I am, properly ask me about my particular shopping habits, collect ongoing information from me, so that it's able to actually sell me more, but on the whole help me, instead of every now and then trying to push something on me with some generic algorithm?
Because, essentially, the system didn't know me and still doesn't. And then OpenAI, Anthropic came along, and as of today, OpenAI in particular knows more than any other system. And I've only been using this system for, like, three years, right? I mean, I've been using ChatGPT for three years, and ChatGPT knows more about me than any system I've used for decades. Those systems died off somewhere, damn it, with that data, because they never asked me these questions, and they don't know anything about me, they don't ask.
I mean, I'd like to tell this system things. I'd like the system to really help me with my health. How can it help me with my health if I restrict access to my personal data? How can it make any recommendations if I've uploaded my test results to it but haven't uploaded the relationships in my family? I mean, we know that health is also affected by sleep, health is affected by how I communicate with my kids, health is affected by how I communicate with my wife, health is affected by my psychological state at work, health is affected by what else I do during the day.
So it's not just metrics like how much I walk, the way the whole world has now put on wristbands and measures 'licherv', right, or measures the number of steps. So there's a huge number of things that affect health. And that's why I'm interested in sharing with the model, right, and not restricting it. You're interested too, right?
Can you imagine what a breakthrough there could be in absolutely every area of human life? Take medicine alone. If medical data were combined, what a force that could be for a breakthrough in tackling various human diseases, or new ways of treatment, or creating new medicines and so on. But at the same time, it's a very tough, very sensitive area, which is restricted to the maximum by personal data laws. I'm not calling on everyone here to share and say that Alexander Volchek specifically has such-and-such a condition, and to put that label on him.
Well, the point, really, is for people to understand what's going on at all, and to figure out how it works. You explained that really well, yes. Let's move on to the next one, to the next rule, a critically important rule. And we can see how much… each rule here could be broken down, made into whole episodes, right. And that's why I've split these 20 rules into at least two episodes. And really, you could talk about it in detail and go off into whole discussions of it. And you should look at what you actually think about this.
They know anyway.
Or cameras, or, you know, cameras at the airport watching something somewhere. It's just...
Uber knows where you usually go. Like you say, any…
But I don't see this as… look, look, I'd really love for ChatGPT to load my trip history from Google Maps. I just don't even understand how to integrate it, how to do it right now. I'm not against it, because if it knows where I go, well, Google knows where I go anyway, look, because I'm logged into my car, I'm logged into my phones, it knows where I am anyway. I buy tickets anyway. Cameras are working everywhere, these sensors are everywhere. At least this system will be able to, I'll be able to use it going forward.
Meaning I can say to it: "Pick me some optimal routes. Or maybe I made the wrong moves somewhere. Listen, can you find me, there was some interesting place I visited, drove through, and there was a cool view there." So it'll be much easier for me to search for it and find it in the system. But people, Anya, are of course afraid of this data. Recently, here in California, they took down cameras. As you know, very few cameras are installed. Here on our street, on Page Mi Mont Street, they took down the cameras, Anya, because the city was using these cameras secretly, only for, from the provider.
Anya, do you know why they took the cameras down? Because the city wasn't sure that this provider wouldn't share the data if some feds, federal agencies, suddenly came and requested that data. So in California the city is against the government having access to this data, like, if they suddenly have the resources for it. And you kind of sit there and think, I mean, on the one hand, at the same time, there are clear patterns showing that these cameras helped fight crime.
Also, I think there's this point: a person, after all, restricts the use of the kind of data they have some notion and understanding of, and that they can somehow define. And they can create certain scenarios, rules, laws that restrict the use of data that's clear to them. But this, let's suppose, is only a small part of everything that circulates and lives in the world at large. And we don't fully understand how the interaction happens between different systems, between different companies, between different laws, between different countries, between, in general, all sorts of different components of our lives.
And that's why we restrict, say, medical data, as an example, purely because it's an understandable structure. There's a person's visit to a medical center, there's some health record that contains personal information about their own health. And it's very clear how to put a veto on that information, on training, use, publication and so on. But if we go into some other zone, like that wonderful example with Instagram, which bothers me too. Whatever I discuss, a minute later I see an ad, or some blogger with a similar scenario, which the feed throws my way.
And that's with absolutely everything turned off, everything that can be turned off on the device. So then my question is: where exactly is the protection of my personal data, or the protection of whatever privacy I have? And based on that, I conclude that we only restrict things that are very clear to us, where we understand how they work.
And here's an interesting aspect. After all, from a person's face they'll be able to, and already can, diagnose a huge number of their problems, or their state of stress, or the details of everything else, even in certain places where we don't realize that it's, say, allowed, right, to recognize a person, scan them and so on. So what do you think about this? A very big topic, a very interesting topic for discussion, by the way. And the sixth rule goes as follows.
That was the fifth rule: tell AI about yourself and give it your data. The more it knows, the more precise the answer. The sixth rule is: double-check AI answers and ask the model to double-check itself. Look, this is a critically important area in general. Like I said, the 10 rules are super important, and this is a very important rule. You have to understand, recently, a few days ago, I shot a video, last week I think, about an experiment they ran, repeated more than 10,000 times, there were 100+ questions, financial questions to do with taxes, salaries, your financial well-being and so on, mortgages, loans.
And they saw that the system gets it wrong in more than half the cases. And one friend of mine even wrote to me and asked: "Sasha, does this mean I can keep making analytical reports for the company? Or does the system really make that many mistakes?" I want to tell you that, as of today, these are systems that can tell you things that aren't true. And that's normal. And in general, for some reason we live in a world where we often assume that if some person said something, they told the truth.
Then we check up on them five years later and see that they lied, lied completely, they deceived us. And we're fine with that. We're fine forgetting about it, we don't look. There are people who are telling the truth now. We say they're lying, and it turns out they told the real truth, and they didn't lie. And when it comes to artificial intelligence, artificial, of course, the current models, the way they look, the way they are, can give you incorrect information. They can give you information correctly, the way they understood it, but you understood it differently, and so the answer can differ. And this system can also manipulate.
I mean, we should all calmly understand that today's artificial intelligence can put across whatever point of view it sees fit. And for a huge number of questions in our lives, there are lots of different variations. And you've often noticed, right, when ChatGPT answers something, and you tell it: "Actually, I also have a son who's five." "Ah! Well, since you have a son, then this, this and this." You say: "And by the way, I also have a bush in the garden." "Ah, since this bush is red, well then, since the bush is red, that changes the whole situation altogether. Nothing else matters.
The red bush rule applies now, right?" And there are lots of stories like that, lots of them, lots of circumstances like that. A person's life is complicated, it's big. That's why it's important for you to double-check ChatGPT. It's important for you to understand that even on the simplest question it can make a mistake. And that's normal. What's more, at that exact moment they might be updating the system, testing it, it might have a glitch, and at that exact moment it can make a mistake.
Just like a person might be operating on you, and whoever is doing the operation could have a stroke at that moment. I mean, can a surgeon have a stroke in the middle of an operation? It can happen. Can he stick the knife into the sheath in the middle of an operation? Probably, he can. I mean, absolutely anything can happen, right. And…
By the way, this double-checking also helps the model, among other things, to learn and get stronger. Because the next time, the model gets scenarios, certain dependencies, that it perhaps didn't have before.
Yes, and what's more, the model can at the very least learn from you, it can understand you. You often try to hide from it or show a different opinion. It'll think it's a different person. Remember, I had an example on ToTheMoon, where I told how my mom asked me to find out some details about the garden, how to cure a particular plant. And I told ChatGPT: "This is my mom's garden, not mine." Why? Because I didn't want ChatGPT to say a month later: "Listen, you had this plant with this disease."
And she didn't have it, right, in the garden. Otherwise it would draw false conclusions, because she and I live in different regions, in different zones altogether. And we have different plants growing, and different treatments for them. That's why double-checking AI is critically important. And it's important to ask the model to double-check itself. Many people use the approach of double-checking AI with another model. That can be a plus, and it can be a minus. I want to tell you right away that there's such a thing as excess information.
You can double-check your answers endlessly, you can keep improving something endlessly. And since you have a free improver, a sort-of-free improver, it will endlessly suggest and generate ideas for you. At some point this trend of suggesting ideas appeared in the world. And how cool it is when people are creative — they come in, they're thinking people, and they suggest lots of ideas. They simply — the way a normal, healthy person says, "OK, I'll do what's been suggested." People started saying, "Come on, suggest some ideas, bring some."
Whoever started suggesting ideas — that became cool. And now everyone has got these endless, well, endless ideas. Now anyone in the world can generate an endless number of ideas. You shouldn't play this game either; you need to be very critical and careful with it. It's a kind of addendum. I might even make this a separate thought and a separate point. Maybe in the future we'll have it separately. So, the seventh rule.
The seventh rule: you shouldn't trust ratings, reviews and various experts. It doesn't matter who the expert is. Whether it's Alexander Volchek, some expert, or Sam Altman, some expert, or Elon Musk, some expert, right? Or it's a person you don't like, so you don't consider them an expert at all. You have to test AI yourself, on your own tasks, and know how to use it, and be able to draw your own sound conclusions. And if you look at our episodes, we often ask people: some really liked Astra GPT-6, some really didn't like it.
Some think GPT-6 Astra is a super-breakthrough, some think it's very bad. What's important here is to be calm about it, because in this case everyone has the right to an opinion. I mean, they did it for some reason, they reacted that way for some reason. And on top of that, a huge number of ratings and experts may talk about something sometimes out of ignorance, sometimes out of hype, sometimes out of marketing, and sometimes they may simply be wrong, as in my example at the very beginning of the video.
OpenAI was saying that browsers would become incredibly successful, that it's a super-trend, a super-cool thing, and then they were wrong, and it turned out it doesn't work. And there's a lot of that. Look, OpenAI has a lot of this, it's just endless. Tomorrow I'll be at OpenAI DevDay, and it's not that I'm interested in hearing about some new, super-endless features. I'm interested in looking at different opinions and then drawing my own conclusions, because they may come with an agenda. Maybe they'll have an IPO in two months.
They want to announce it tomorrow. Maybe they've found out that Anthropic has developed some super-cool model, and they need to adjust to it. We don't understand what game is being played where, and someone wrote in the comments: maybe OpenAI and Anthropic have been colluding from the very start, and it's no accident they split up. Recently I shot a reel on this topic and wrote: what if Elon Musk and Sam Altman agreed to create this conflict? Because when Elon Musk and the current US president — at first he was pals with him, and then he came out and said the other one was out of his mind. And that one said he should be kicked out of the country altogether.
And then, some kind of silence. They support each other. Well, you realize opinions can change, and maybe they're colluding, everyone's colluding, really, and so they can broadcast, well, whatever they want to broadcast. Or maybe there's someone higher up who arranged this collusion, and these people themselves don't realize they're taking part in it, right? And that's why you need to have your own healthy opinion about your current situation, about your family, about your life, about your work, about your company, about your profession, about the country you're in, about the place where you live, about your circumstances.
But of course, to hold to this rule on ratings, reviews or experts, you need to take an interest in all of this and be engaged in it, to learn and develop, and not just believe the hype.
So, the eighth rule for today — we've still got two left, and the tenth rule. I can just see it right in front of my eyes. It's the only thing I see, these ten rules. Me, well, and my sister. I think the tenth rule is actually very much on trend today, as an addition to this, although for many people it may be very difficult. The eighth rule is to upload a certain range of data into AI as a second opinion. Look, I've separated this out — not just uploading data, maximum information. For example, there's the area of medical documents, or your test results, or there's the area where you went to, I don't know, a psychotherapist.
Or there's the area where you've made a choice at school, for example, listened to a teacher about which club your, say, child should go to. Today you've got AI as a kind of second-opinion system. You know, it's like nowadays you're driving in a car or walking, talking, someone says something, and you say you're sure that's how it is. And the other person says, "No, it isn't." Now there's ChatGPT, you can settle any dispute, right? Open ChatGPT that very second and double-check.
And the one who's used to being wrong says, "Oh, cool, I was wrong." The one who isn't used to it keeps getting hit over the head, right? So now there's a system to quickly double-check everything. By the way, I keep seeing Anya in meetings with people. You fact-check them, and they look at you as if you'd lied anyway. You can tell them, "Well, it says here that two plus two is four." And he sort of acts as if two plus two is still three. No matter what! He just pretends. He comes to the next meeting and says, "Two plus two is three." So — but in any case, ChatGPT gives you this very strong help, and it gives you the help of a kind of second opinion.
Just remember this. It will let you develop new use cases, new scenarios within your day. It's a separate rule of sorts, a direction to move in for your life together, right, in terms of whether you dress right, whether you communicate with someone the right way. Choosing a route or choosing a restaurant to go to, or which trip to take, or buying something in some store. It's a great additional second opinion. The important thing here is that someone may say, "Alexander, then we're moving away from a system where we trust someone; we'll have to double-check everything."
Keep this rule for yourself for a year of your life, right? That's not so much. You lived for decades when this tool wasn't available to you, and there were a huge number of things you couldn't check. For example, you and I didn't know that the continent of Africa, it turns out, can fit several Russian Federations inside it, right? We just didn't know that. I mean, we thought the Russian Federation was very big, and Africa was sort of smaller, right? Or Greenland — it's very big when you look at the map, and Africa is sort of smaller.
It turned out Greenland is, what was it, 18 or 14 times smaller than it's actually drawn on this standard, commonly accepted map. And that's why we now have the opportunity to actually learn to think critically. The ninth rule — once again, be sure to stay, the tenth rule is a super-important topic. By the way, it would be cool if, for the eighth rule, about double-checking with a second opinion, you shared some cases; they'll be useful to people. You see, people see these cases, they read them. It's very useful. Far more people read the comments than write them, right? Well, for other people to read these things.
Write in the comments about the cases where you double-checked with a second opinion and they were important. So, the ninth rule is: give AI the whole task and demand a finished result, not hints.
Here's one of the things that happens in life: say, I don't know, there's some ordinary day, and a person tells themselves, "I lay out various tasks, and AI does some micro-things," right? Or, for example, a person says, "I have an automated system, and a human will handle some small process." Today we're heading toward a point where AGI will appear, ASI will appear, right? And there will be — we'll arrive at the technological singularity. There's a certain sequence of steps drawn up — it's imaginary, but it has been drawn up. You can watch about this — I had an episode on AGI, ASI and the technological singularity.
And these systems have to work on a task holistically, and you'll have to decide. I'll talk a lot about this in the next episode, when I'll be talking about the ten — the ten main things you mustn't miss when it comes to the future. Namely, where — what to learn and what not to waste time on. There, in the age of artificial intelligence, one of them is about automation. But the point is that the task facing you and me — is not to try to automate some micro-processes, but to rethink your everyday routine and your life altogether, to approach it a bit differently.
And for that, of course, you sometimes even need to set aside a bit more of your time in ChatGPT and tell ChatGPT the whole picture of everything that's going on. That is, if you're handling, I don't know, some legal — or you're dealing with some legal issue, don't try to discuss just one paragraph with ChatGPT or look up just one law. From the very start, tell it the whole story and then keep discussing as you go, attach those documents. This will let you come back to it after some time has passed.
It will let you, by asking the system again, learn more broadly about some aspects where you even thought, for example, there were no mistakes. This is a very important rule.
And the tenth rule, which stands at the dawn of, probably, 2022… it's a very powerful, modern thing. It began in the spring of 2026, and it's fundamental, and I believe it's fundamental not for programmers — not only for programmers and not only for people who do automation. It's fundamental for all ordinary people. And probably that's the only way to truly get a feel for it and understand: what is OpenAI today? What is Anthropic today? To get a feel for what Grok is today. To get a feel for it. What is Gemini today, right? That means starting to use systems which, at OpenAI, aren't called ChatGPT, but a system called OpenAI Codex.
And the time will come when all of this will be called something — ChatGPT or OpenAI. The time will come when it's called, well, not Claude Code, but everything will be called, I don't know, just Claude or Anthropic. All these systems — not Claude Design, right, but everything will be called by some single name. These systems, Codex or Claude Code: starting to automate certain processes in your life when you've never done that before. By the way, for my wife, in fact, who basically runs the home and looks after the kids — she doesn't work for money in the usual social sense, right?
And she has a huge amount of other work. To train herself in how to create such business processes at all. I mean, she has a very clear business process in her life. She knows she has to get up at six in the morning to take the kids to school, leave at 7:30, drop off some of them, then come back, take, well, another daughter at — we have four kids — I don't know, at 8:40, and so on. So there's a certain well-oiled business process, but of course she doesn't use such terms. And in order to — she has certain terms when it comes to payments, there are terms with banks, with legal systems, with medical systems, when it comes to social communication, interaction and so on.
But for that you need to start using Codex and Claude Code. And for business, look, these are things where you'll start doing many processes yourselves and migrate the very notion of development. Today a huge number of companies are looking for people who understand AI and can program something for you; they're looking for people who know how to implement various AI projects. I want to tell you that this is a losing proposition. The only possible win, the most effective one, is if all employees start using artificial intelligence.
It's like — imagine you have a small ERP system or CRM system, and 20 people in the company use it. And when all 20 people start using Claude Code, or all 20 people start using Codex, they'll start making use of it. And the ERP and the CRM system, and Google Drive, and Docs, and Google Team, and Zoom, and Slack — all of it — CRM and everything — all of it will be a part that gets automated by Claude Code or Codex. So Codex and Claude Code will become the central system that communicates with everything, and all the employees will work with it. But of course, people need to get there.
And I think that today OpenAI and Anthropic, and many companies, are afraid of this and don't see this market, in particular that sales managers will work — this is my well-known case, right — that there won't be any CRM systems, and sales managers will work out of ChatGPT or out of Codex, or out of Claude Code; that marketers will work out of Codex or Claude Code, that architects will work out of Codex or Claude Code, that editors will work from there. Why should an executive sit in Excel or Power BI? Why be in there? Why should finance people be in financial and accounting systems?
We talk about this a lot on our channel. Be sure not to forget to forward this episode to the people who need to watch it. Get ready, stay tuned. On Friday a new episode will come out — well, or the next one, if you're watching a bit later. The next episode is about ten advantages ahead, further into the future, so that we end up with this set of twenty. And on Friday we'll also give you the information and the link, and show all these 20 listed things in terms of how we did it in our project on our website.
See you in future episodes. Bye, everyone.