Hello, everyone! You're on ToTheMoon. Technological news, Silicon Valley sites and the world. And today, we will be able to gather a lot of interesting news together with Ilnar and Stani. Switzerland made a very interesting statement on how the vacancy market had changed and how seven million vacancies were now represented. A very interesting topic is how artificial intelligence influences the labour market. Anthropic has provided the Senate with data that China has created millions of bots. We'll talk about this today. OpenAI makes its chip, and it should have waited. It's shown. Such a cremated plate was presented. What's more like OpenAI or Anthropic in a certain type of task? We'll have to see this today. Hundreds of millions of users in medicine, even in the GPT chat. Which I want to say before we start now is a special record. Today, when she talked to her sister, she said a very interesting thing to me that at some point in time you realize that you were so dependent on artificial intelligence and that he was so full of it. I can't hear it, and I just want to go to bed. But while we're on ToTheMoon channel. I want to start, so...
We have a lot of news today, I think. And before we go to a very expanded subject, the OpenAI made its chip, right? I was reminded of the name Meta right away, so avocado. That's probably close. Here, and Meta, what he's done in the last six months, is gonna let out new glasses. We're not gonna discuss them here today. So, the OpenAI released this Jalapeno together with Broadcom. This is the first of its own intelligent processor OpenAI. And the company, well, says it's an accelerator, yes, an accelerator to steer large numbers of people in its infrastructure. And I don't know. Remember, it was a lot of time when we were talking about the chips, talking about NVIDIA. I think we've had a year, and in the first year of NVIDIA, it's been very frequent not only because it grew in value, but, well, somehow, it was in the market, in the usual, ordinary market for ordinary people. Yeah, I don't know. I thought so. Now, NVIDIA is doing a huge amount of everything, but it is still in a layer of infrastructure. While they're still, as they've tried, I think they're out of the infrastructure, they're in a layer of infrastructure. We see that everyone keeps trying to make their chips, how cool they are, not cool, I'm hard to assess. Obviously, when SpaceX is collating with someone, yes, in this case, they're doing their chips, right? And OpenAI, for example, with Broadcom, obviously, well, chips will come and different we'll see colonizations, but I think that's getting less important to ordinary people. Yeah, I guess everyone's gonna go down to that. Question is, will the chips they can use in some mass production for this, what is called physical artificial intelligence, huh? Here, on-- now, a roll came out a few days ago, and--I was there on the Forbes list of artificial intelligence 50. They've made a rating.
I've been paying attention to physical art intellect.
Yeah, about the chips. Of course, it's very cool that OpenAI with such a big, big player is making out his chips. We already have the colonization, I can complicate now, but I think OpenAI with Cerebras, and NVIDIA with Groq when it's with a Groq that's not XAI, which is the chips that make. That's why they're moving on that side. Google has his TPU. So we're even more specializing in the market, right? Because videotapes that, uh, are not exactly on the scene, but at least they've been doing models for years. It's a general-purpose equipment, isn't it? They're not really called video cards for nothing, they're meant for the rendering, for the pictures, not for neuronets. And now the OpenAI says that we are producing our own equipment, which is aimed at inference, that is, not at training, but at work, at using our neuron networks. They know the needs, they know where there are bottles of bottles where, with the memory, where there are other calculations. And accordingly, they create chips that are better than NVIDIA chips that will be appropriate for their purposes. That is why, of course, a step forward. And from the end-user perspective, that's how many, like a month ago, I thought, or three weeks ago, it was news that NVIDIA was producing RTX Spark. And that's just the end-user story, too, that's a very interesting story in the Intel collation, unless I'm wrong. And it's a competition for the M series at Apple. So Apple has a crazy, cool processors. M5, I think it's over. And the usual Intel, they're getting a lot of them down there. And now NVIDIA is producing RTX Spark, respectively. Well, at least while we showed him, but there's no sale, I think this year comes in the fall if I'm not mistaken. And there's a very tight competition between M processors and CUDA, a library for dealing with neural networks from NVIDIA, where virtually all are tied, it's built inside. And, accordingly, the same architecture as the M-processors, a huge amount of memory. And, accordingly, on these computers, there will be a large enough neural network on these laptops. It's not really abundant devices, but at least it's a serious laptop, you know, that a major open source and local computer launch measure can be recycled.
And it might open the door to models that would be localized on the computer to work adequately, not just to turn the information on the big servers.
Well, look, this is a very important subject, of course. I really want to support the fact that such companies are involved in even joint projects. It seemed like it wasn't in their factories, was it? Because it's new approaches, even in terms of this whole code and the building of some kind of system work, well, right in the chip, yeah, it's, uh, it's super important, yeah. This is the acceleration in the future of the vast number of this. We talked to Max Grigoriev about the identification of the faces on the camera, and he said that, so, you don't need to be there for every camera to recognize the face, transfer everything completely. A video. You're gonna have to start a face-to-face algorithm, get this mark, mask, yeah, I don't remember exactly what it is, yes, I don't know what it is. Maybe you know. Here, here, there, yes. I'm just in Russian, in Russian, how is that? I guess some kind of, uh, some kind of maquito. It's not a maquit, it's, uh, it's like a mask full, yeah, man, different locations.
Maybe so. It's usually hard to interpret. It's, in general, before the end layer of neuron network, you have a set of numbers, and this set of numbers generally quite well characterizes what it is. And it's working out...
Characterizes what you can find the same person, you mean, right?
Yeah.
In another base.
Yeah.
And they're sending them, respectively.
If there's one in the database.
Yeah, they're sending this little, small, micro-size mask or embedding, yeah, right word. And I'll be more professional here. And they'll recognize it on the server. And I think that these projects, they'll be very advanced. Well, it's a very important layer, well, just a very unrealistic layer in the offline, yeah. And what, what is in the physical world, yes, not just in cars like cars, it can still be a little expensive chips, like that. And even in terms of what we call a smart house. I mean, smart houses, simple, cheap, easy realizations that connect, are probably very easily integrated and so on, yes. Still, look, it's 2026, and this is a very relatively settled subject. I want to go to what story I want.
There was news that, uh, xAI, I mean, I remember that this is SpaceX because in the Russian-speaking space, not all people can know. And the xAI has a Grok, it's the artificial intelligence of Grok. And the Grok is developing very much now, and in particular we know that they have an integrator-- a deal with Kursor, yes. I don't know if she's gone through or not. I just-- I got the market all over these. And it is clear that the HAI will compete in strong competition, including wanting to go there, and he won't, he'll be with Antropic and with OpenAI, and with Google, in particular. And they're gonna let them out now...
And they're now, so they're letting some integration out this week. One of the integrations was in Word. I didn't want to tell you anything about it, and I didn't want to raise this subject because I think we should be getting Grok in Word? And then I decided to raise the subject here today. What kind of thing? The thing is, all these systems were int-integrating, right? Cloud was passing, ChatGPT was passing, Gemini was passing. And yet Word and Excel are the most mass soph-- well, the softs that exist, right? And I, for example, never use word in Word or Excel. I can't say I have a lot of my life in Word and Excel, it just dropped sharply, it's virtually gone, but I don't open these systems. You're opening these systems because these integrations, they're in the end in the corporate business. Because I think that ordinary people in the real world are about to be here for years, but I don't know, even this year, everything goes to that if you're a user of artificial intelligence, There, the users of any model, you need this integration as long as you need it. Here. Or is there integration? Tanya, you've got a lot of stuff in this part.
Well, I don't have a fan of Word or Excel in this part, and I'm trying, you know, to avoid any way. But when someone sends me, there, suppliers, vendores, it's always Excel. And I'm like, "Why, okay, okay, let's get your Excel." But naturally, I never use them and try to do something more adapted to the design of the program. I don't know, I think it's something that's already outdated. Well, that's for me. Well, here, you know, plus that I'm a Mac user, I used to use the numbers, all that stuff to make tables.
I have a special program.
Why?
Do you have any of these scenarios?
No, I don't have these scenarios. Maybe I just didn't, like, try, but why again? I don't see the point that he's automatically editing the text. Well, maybe it's one step less than writing this text in ChatGPT, and then sending it in the form of a document. But I don't know, you can do anything at ChatGPT or whatever, in Claude, in Gemini. Why go to a separate program? I don't really understand that.
Well, I'm, like, connected to different things, because I'm in many companies going in there or as a owner or boards of directors differently. I'm looking at a lot of reports. All accounts, for example, I'm looking there, Google Dri-- in, Google Sheets, Google Excel, parole, yes, Google Excel and Google, Google Docs. And, well, Google Sheets. And huge tables with very big data. But if I want to make any conclusions, I do or do a scrub of this table and redirect it to ChatGPT, or I do, download the file with all the sheets, there's XLS, XLS-form, and I'm putting it on the next floor. analysis. I was just thinking, okay, if I was a financialist who works in these companies, in the team, and there, and the finances, they're still making these documents and making these reports, for example, often. And they need analysts, yes, some, and they need to use-- work with baseline data. They may, in theory, write a different formula for this. But here's a very interesting thing, and if it's convenient to write formulas through such integration, or I think these integrations are always a little bit short, they're not really good at the quality of the model, because they're in the middle of something. These integrations usually have limitations on the choice of model quality. I mean, I don't know if I can pick five-five Pro, like, right? And how long will I wait for a task like that? And accordingly, I think that even those people I know who create such documents, they still make them directly in, there, in Claude or in ChatGPT, or in Gemini et cetera.
But in Gemini, there's a lot less people who make. And even, by the way, Sasha Masharov was always very much for Gemini. Well, I support him in terms of Google growth and Google stock purchases and so on. And that's a good position. But from Gemini's point of view, he was literally dating him a few days ago, and he said, "Who does Gemini even use?" I said, "I knew you were saying."
Sasha is always categorical.
I don't know. Well, that's, that's really a tone of the market. There was a time, and Ilnar, I remember, came when he talked about Gemini. I even bought some versions, I don't think I remember the biggest or not. Well, I understand there's a Pro version of Gemini automatically, there, with Google different tariffs, but I just don't use it. I don't use this infrastructure, though I wish that, because Google is in my life, there is Google Card, there is Google Post Office, there are Google Documents and everything. But I don't understand why. I mean, what is the essence of this related infrastructure if it works very much by the end. But it's interesting for our audience to hear it, too. In any case, Grok is gradually making all these integrations. I think these are some of the pyard releases. It's not like Elon Musk talking about them, it's more like the Groks on the market, and I think it's something I need to see.
I'm not gonna... I can't, I don't know how you are. You have something better out there, which is exactly worse?
Sas, this question is about once a week in our chat rooms with friends. Like, tell me, what do you use? And what's the best thing for Claude, and for what's the best, there's Codex, and there's Gemini or something. And I never have the only right answer that I'm like this. Everybody's using, everybody's trying to compare. And how you juggle the limits because some limite ends, you go to another one.
I wonder how you're talking, I guess. I have a basic, you know what the subject is? What I think, first of all, systems change all the time. Well, I noticed that ChatGPT five points was much more cool than it was. Well, I don't know, I, I, I don't even know how he's doing it. I mean, I still don't understand how he could. He's been working on a six hundred fifty thousand or eight hundred thousand lines of a huge text, you know, like a huge piece of text, right? And there's a million and a half, there, records. I mean, I don't understand what these context windows are like, and how it connects. I mean, I think the version is like there's no idea of a context window, right? And if it is, I mean, it's, I see. But the system did it quite accurately. Like if I said, like, the code, there's a record, a story. I mean, you know, there was a situation before, you're gonna load the text there for 5,000 lines, and he's looking for fragments there, right? Or looking for some parts. Then it came when there were 5,000 or 5,000 pages, not a fold, but 5,000 pages, or 3,000 pages, it could be handled, you know, like a pack, like, clearly. But I'm seeing the result now, and I'm falling short of the current results. And, accordingly, I'm saying about the code and about Claude that I feel like nobody else knows, but how each system works, and what other structures it has and what it's ready to do, and what it's ready to do, I'm sorry I didn't. I mean, sometimes that kind of system is stupid, just really stupid and doesn't do some basic simple things, and sometimes in one session, it just sort of makes some space stuff. You want me to raise the subject quickly? Now the subject is very much debated on the grass, isn't it? Well, you're reading it there, too. What's so funny now is saying you shouldn't miss writing and running a session, but you're supposed to do a series of development agents that will make you something, right? And if you write different assignments there, through ChatGPT about these butts, you can see how many of these systems resist. They say, "No, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, Too much, so big session." Or, on the other hand, they say, "Why not, let's rot, so that's the way it is." I mean, she's just as ready to do it, and she's not ready to do it. I am not yet fully aware of the system ' s willingness to do so, but how much the system is willing to do. I have some sessions that I've just been using for weeks, just for weeks, and they're all fucking working. And then you start a session, and she just dies a little bit soon and doesn't want to do anything. Well, or make some restrictions. Here. So, if I'm an adapter of Claude, if I'm an adapter of Claude, I can't tell whose adept I've become because I'm periodically happy and coded, and Claude, and I'm sick of Claude and Claude and the code. But at the same time, I went on, Ildar, but for my normal life, I used ChatGPT. So I didn't migrate on-- you migrated, did you? Na--- not migrating. I mean, there's a lot of people who say that and so on.
I mean, I'm not going on these topics, I'm not.
Anthropic. I didn't try it, but they say there's a hole in the anthropic, and you can start a fairy like that, you start a computer session, then you put it in remote control on your phone, and there's a phone in it. The possibility of choosing a fairy, and you will have a computation and a work-by-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-
Fable calculations and work are carried out. It's probably gonna close this thing soon, but that's because she just got upset. Wix writes and there is a general public saying that. Have you tried?
There's a story on this, and if you ask the session, and this is Fable, he'll tell you it's not Fable, right?
Yeah.
Or sometimes, like, Fable says. I was just throwing it, remember, just last week, just like Fable earned on the phone.
Yeah, yeah, yeah.
And I have a private, well, in the States, that I see that sometimes Fable is included and that the session goes through and sometimes does not pass. I mean, it's kind of weird...
They must have been a stew-- they've been patching up web-coding, claude just made, and they forgot to fix IOS too.
Look, I don't know. I think they're not supposed to have her on the server. And I'm not sure it's Fable working.
Yeah, just Wix wrote that the limits were very good to fly, as if it wasn't Opus four eight, it was Fable. So I'm guessing that it's probably him. Well, it could be.
If one of our viewers has tried, write your comments, it'll be interesting.
Cyber, something's security.
Cyber, yes. Yeah, they started chatting, ChatGPT five Cyber. And I don't know how much this model is really as cool as the models, well, that's just how it's a blocked Anthropic family. We'll see. Let's see what's going to happen, huh? They said it was-- oh, it's expansion, too. They have a program that's called Daybreak, which is on software protection. And this complete version of Cyber, there, like, a cool, specialized, less complex, tending to refuse complex cybersecurity requests. Well, let's see, let's see what happens next. But we see that ChatGPT is silent in its development. By the way, the fun subject was that we told you many times that Chinese models are trying to distil American models.
And Illar even made the terms online. I have always argued that distillation will not help in any way create the same system of cool and such infrastructure, such models, etc. And then, that's the interesting news that, uh, officially, from Anthropic. I mean, I don't think Anthropic's cheating, do I? Because it's a big trial in the world. Anthropic said that, well, that's so interesting, that's what it means. The point of what? Anthropic sent a letter to American senators, in which it accused the operators associated with Aliba and Qwen Laboratory. So there was an outrage attack... to gain Claude's opportunities. And there was, uh, a familiarity with this letter. We understand that this letter is not seen with you. But the point is, according to Anthropic, this company was on the data extraction from April 20 to June 5. And now, attention, attention! It has created more than twenty-eight million, twenty-eight points eight, if more accurately, millions of exchanges with the model through almost 25,000 fraudulent academic and accounting records. Do you know the level at all? So Anthropic has serious enough, I'm sure there are systems to prevent these illegal scouts, illegal workers, all right? And of course, it's a shocking amount because 25,000 scouts and 28,000 records, well, it's a big volume, it's a very serious volume. And the proceedings will be going. I don't know what American senators are going to be getting at. I'll remind you that the opinion in the U.S. is shared. I mean, there's a Jensen Huang line that says that China doesn't have to close completely and, on the contrary, we need to supply American equipment models.
There's a line this, with Anthropic, that says Chinese people need to be very rigid, so they're gonna be isolated, yeah, and go. You're losing it on their lines sometimes. There is, we know, one senator in the American government, and he's issued a proposal for a new law. Have you read the proposal for a new law? Trump just gets nervous, so... They're in a coalition. So the new law has been saying that all major American companies in the artificial intelligence zone or in the entire world are working in artificial intelligence, and that there is no such big profit. They had a size-based approach, and they should issue new shares to transfer the proportion of the companies to the public. How much is that? Fifty percent of the share. Well, I understand that in America, it won't happen. In America, this is not a good thing to do, but I'm a little confused about history. I am a little embarrassed because if parallel parties are increasingly issuing such laws, there may be something in theory where the State wants to own a part of artificial intelligence. And the problem is, they just own artificial intelligence, they show it that it would both make it possible to give US citizens a thousand dollars a year or a time. I think there's a fundamental story, well, that they're putting their people on board, right? And that they are on board boards of directors, that they are starting to run these companies. Well, I don't-- I don't believe in running companies through government structures. Again, I don't think that's gonna happen. It's just a story that such a discussion is. And it is clear that this is the subject that China is dealing with and, well, they have passed to the Senate and transferred what China is, therefore, distilled the model.
What's going to happen to that? I don't know.
Partly because of these distillation stories, there are rumors that Anthropic is just planning to introduce user passport verification. So, this is the KYC, you know your customer, when, accordingly, you need to get a little faith first, and only after that, you'll get access. Well, not all models, but some good models are likely to be accessed only for the faithful users. Not the best normal story.
That sounds normal, isn't it, Elnar? You think you're not?
On the one hand, yes. On the other hand, well, I don't want to be a passport all over the place. As if I could not light it, I wouldn't be lighting it. That's if, of course, yes.
Yeah, there's a logic here. But let's just do it, honestly. So we're still getting licenses to a car, right, we're telling who we are. And there's a subject, I think you even come to a library with a lot of people, you're giving me an ID to register. But at least, you'll get a name, at least you'll have a name checking. Right? In the library.
That's right. Sas, the Internet was just a zone like that, and you and I kind of caught up when he was really free. And now it's getting smaller and less in it. It's just pathetic that this time passes. Well, personally.
Well, the question is, I guess, the quality and limitations of the models. I mean, I kind of feel comfortable about it, but at the same time, I kind of want to be kept quiet. You think about all the other people, you know, it's about what it's about to limit the world and share society very much, and some people will get something much bigger. Yeah, that might not be a good story, I guess.
In this context, it is not very, very adequate. But here comes the subject of the British law that minors cannot use social media. And then he was just acting like Durov, and he said, "What do you think they're doing for that? They do it for other purposes." Yeah? So the main target is to keep an eye on citizens. That the restriction would not lead to anything. Well, there's some resonance in that, yes, that if you register with the social network and have to report your passports, then it's obviously smelling. Let's say it's children. And in fact, we're not doing this for the kids, but for the purpose of getting the identification of every account in a social network with a particular person. But there's a moment here that, in fact, if people are honest, their account is so easy to identify right now, right? Well, I think the accountants are very easily identified. And even imagine if Elnar doesn't show up, there's a picture in Instagram, for example, but he doesn't show up. Well, let's imagine if he is.
Yes.
I don't know if I'm eating or not. And he doesn't put his pictures in, can he be identified? Like, looking at some page at my place or Tanya at the same time. That's the connection.
That's what we found.
Yeah, how's the connection going? And then there's a connection to Taz ⁇. All right, here we are. He's Elnar Shafigullin. There are no other. Right? That's it. And even if our viewers say that he's not Ilnar, but Elnar, then-- and I think Tatarstan has the whole time, so the letters are confuse, and then--
And it's getting a new name.
Not Tatarstan, they're confused. Those who don't live in Tatarstan are confused.
No, Tatarstan is just changing the letters and getting a new name.
Yeah, a new name is coming. Why is that? Well, uh...
It's working. Why is that? Well, because. And...
Different names in different situations.
Still the system now, for now, she's--
Very comfortable.
Yeah, she's capable of identifying it. But I wonder what you think our viewers think about the identification, too. I think the I.D. was supposed to be long before. It's strange that she only showed up now, in particular, models. Obviously, models are not mail, obviously, they're not CD files, CDs, etc., right? I mean, obviously, models are a little different. The only thing I personally hate is that there are no people, there are no countries that want to deal with the laws real and real, so deep, detailed, all, uh, creature, and walk into the leg with time, together. with these companies that all make it. Yeah, because even that story still hasn't been solved, which, for example, the data you left in the model is not secret, like a doctor's secret. Well, that's weird, right? But the idea is, if it's health data, they're automatically supposed to be like this. And the model today is capable of some algorithm to determine what they're dealing with, well, that it's a doctor's secret. So, what's the secret of the doctor? She's not that it's your medical records, but that you told someone, right? This man can't tell you that, can he? Or the hospital can't share. Well, let's see what happens to this part. I'll continue to use my model in person. Maybe someone, of course, I have some friends who are so fanatic, they might not be used. We'll see. There's an interesting subject that is about, uh, Euros, I think two, two incredible, two great, you know, research, big and large news.
I always like a lot of numbers. The first is that, uh, data are provided that in Europe, it is considered that 70 per cent of companies have tried to use artificial intelligence in their work. And I think that this number is more inadequate. On the other hand, it's kind of decent, very big. And there is a certain reservation, though, that only seven percent of us are seriously using. So, 70% of the people tried to do something, 7 percent of them were serious. It looks like real data, because I would have accepted the number, there, 7 percent of those. And the question is, there was still a survey of a few companies. Did they really interview all the companies? And the next moment is left. We were in the twenty-sixth year, we were talking a year ago, I think a lot of companies and people have already tried artificial intelligence, and they're old-fashioned. I mean, people treat him like an initial chat. We even have people on YouTube Channel who write, "Oh, that's all the fairy, there, that's all the bench.
He can't do anything there. It's just chatting and still a simple model." There are still others! And it turns out that a lot of people are left with this old memory. And these companies that have tried, they may not be returning to real use artificial intelligence, try to use it every day and find new opportunities. Now, yesterday, my partner said an interesting thing, "If you don't use models like Codex or Claude Code every day," I'll give attention to it, not ChatGPT and Claude, he says you're already after. You're not gonna figure out how they work. Well, what different tools are there, where it's easy or hard to do. Says it's gonna be a lot harder to figure out. Too much of it comes out. Now, these-- a lot of this dynamic agent concept, these cycles, the buttons, and they're not, well, until they're human.
I agree.
So that it's primitive.
Absolutely.
Not that they'll be made primitive. So they can be made primitive when AGI showed up, yes, or ASI came in, yeah, I guess. But now, I think it's a very heavy-- it's really hard to do it primitively, Tanya. Well, if you have a-- a system that'll launch parallel agents.
It's not even primitive, it's primitive, it's really simple, you know, at least some more detailed instructions to move. Because again, returning to my experiments with Codex, if I had this, when I started my business, there, seven years ago, it would have saved me a couple of years. But now that what he gives me is what I needed seven years ago, and how to move on, I don't know.
Well, not to that, not to that extent, that's understandable. But still you're like you're hiding in a certain wall. You're gonna need some specific knowledge you don't have, and you don't know where to get it. Well, Tube Moon, maybe help.
It's a fact. I've been watching, watching, watching interviews, talking about, over there, with one of the Codex officers. And the interviews gave different examples of discussion. These are the examples that you've been saying, well, there's some senior person in the Codex, I don't know if he's the boss or not the boss.
Boris Black, yes, I suppose.
Nah, there was another one, there was Toby or something, I don't remember. But the point is, what is it? What-- those examples that this man from the Codex gave us were with this example, like, let's use Codex texting a sophth that-- or launch a thing that's gonna order us a Instagram--- from Instacart, right? Well, I mean, the regular mail, we'll put some kind of a line in there, and in my calendar, I'll be meeting you. I think life, life in people, which in their calendar would pick up a place, that's what they got-- that's how they're organized? That's what I'm thinking about, about my four kids, about twenty different business projects, about my wife, about other countries, about a lot of flights. I think how they organize these calendars in their own hands that they can trust, tell the system: to put me in there, so, inside, find me a place and offer me some event inside. So they offer some weird boxes, very simple, yes. And that's what you're talking about now, Tola, you say, yes, the system often suggests you're going to solve some very light, light topics, very little things. But I'm telling you, that, well, because of my immersion of a two-month Claude Code and Codex, as a person who's not motivated to create, I'm not a programmer, there, I'm not an analyst, and complexities are some kind of thing. I can tell you, things, this is a phenomenal opportunity for me. Well, that's just phenomenal. I get more and more every day for myself some incredible discovery about what I can do. And by the way, I've been working more and more directly to create something in Claude Code and Codex, not through ChatGPT. Although, of course, most of me work through ChatGPT five-five Pro. I wonder, by the way, again, as our viewers have, this is all changing. We'll ask about it all the time. I think this is a very important subject, right, Elnar? We-- you raised her, I think, a couple weeks ago. She's very important. Liu--- the job came from-- the way you work in Codex all the time.
Yeah, from the chat, yeah.
I'd like to support you here, Sas, and that's what you're saying, that you're not really doing your own hand programming and you're not an analyst who keeps thinking, but whatever it is. You're good at it. But like Codex and, there, Claude CoWork, they're hand-clipping you, you're starting to do a lot more. Little bike. One friend of mine works in the company, too in the aisle, and there was a situation, an accident, you know, an accident, in the sense of the system, but a huge number of people, it was between holidays, huge. The number of people was on vacation. And it just happened that not all the staff were needed to resolve this accident. But because the company is well integrated with the use of both Claude and Codex, the remaining staff have been able to capture the roles in which they are not active. Well, the programmers can, as a date-sientist, make some sort of a challenge, can go to the base, get some data from the base. Well, there's a programmer who's working with SQL, okay, but it's supposed to be that, like analysts, they can sort out the structures and so on, and vice versa. The date-sientists can do some work there to write a code at an acceptable level and to fill the hole before the main staff members are, respectively, removed.
That's-- that's even a T-shaped tech. It's when you get to some level of competence, you know, roughly, there, on a date-sayens, on a programming or something. And then there's a senorary-level parole, you can do the rest of the job, too. So, like analyst to do something, and as a programmer, write the code, and as a tester, cover all the tests, and so on. And the proliferation of these systems is just the same, and there's Claude and Gemini, and--
code and Gemini, and there, ChatGPT from OpenAI. It's, uh-a, most of the specialists do T-shaped by specialists, yes, when you're not only in your field of expertise, but with some experience and some knowledge, much more fields are beginning to close and close. I'm so much faster. In fact, this is the same thing with you when you tell me that your business experience can also, and accordingly, close all the other tasks, because of the availability of these tools. And that, in my view, is the main, perhaps, of the last six months, of the year, thanks to the development of these systems.
Well, there's a story about the first thing, uh, if people are really gonna be down this stage of the current Clod code and code or at some point of time, some migration will happen. What people, uh... like this branch would pass. Well, most people, I mean, and for most people, they'll just give you some kind of little interface, and a lot of people will miss this area, you know, the real creation of something or fundamental. There's no change for yourself, and it's not gonna be understood that this artificial intelligence is a little deeper. Well, in particular, there are a lot of people who have learned how to print and print blind. Or even if they print, not ten fingers, but two or four, they print very quickly and have the ease, so much ease. And there are people who can't do it. Well, for example, my wife, she can't use desktop at all. So she can use the usual phone. She doesn't know what desktop is. So it's past her life, and she, if she has to do something, uh, whatever's not done on her phone, it's a big problem for her. And she finds a way to do it on the phone, even when I don't know. So she'll find it sooner, like sending a phone off. There was a funnel on Apple Fax, I think, sending some sort of accounting document here. She'll find a way to settle any question from her phone and fix any document. I mean, I'm not gonna open a doc file and fix it on the phone. I mean, I just can't do it. Yeah, why would I? I'll open the computer, I'll fix it quickly, and I'll open the same file right away from the phone. So she doesn't have that skills, and maybe this is a skills, and it won't be acquired by a lot of people at all. So, and that, by the way, raises the question, uh, youth economy, uh, Anthropic, OpenAI and, uh, uh, is a big question.
So they might have a problem. Because when we say there's a billion users, is there gonna be a billion users for this, I don't know, 100 million or 200 million to treat artificial intelligence as electricity, right? When they really spend a few hundred dollars a month in peace and treat it naturally, not as it is now, and when people say I'm not gonna buy 100 bucks there or I'm not gonna buy it. I'll buy you $20,000 and so on. I mean, and, uh, they keep doing that. The topic of the European Union, which is interesting, is more about Switzerland. The Swiss portal Jobsin.ch, he analyzed, which is where it is important to have seven million vacancies. The number is so decent. And concluded that the proportion of entry points, uh, that, well, the opening positions of the 25th year were thirty-two per cent lower than the average of years, well, years earlier, yes, and They used it in their own view, uh, well, until the power of genetic artificial intelligence spreads. And it seems like, uh, thirty-two percent, and that thirty-two percent, maybe there's a change of attitude. Uh, really, really, really, really, really interesting how the market is going to evolve, in particular in the intellectual domain. There, or at least where there are computers. Uh, is this the number of market entries now, or is there some sort of a time-lapse, yes, in many companies? I mean, I still think that artificial intelligence is helping to develop, and there are people like it should grow up with it. Unless you've changed the whole business model. Uh, but here's the question, what is now, in a large number of companies, that they are more optimizing than growing. And as part of this optimization, they reduce and do not hire people they did not hire before. Well, I don't know, junior analysts, there, junior marquetologists, junior designers, and so on. Or vice versa, maybe they hire junior designers, but they don't hire a junior designer, but they don't, they don't hire a lot of other people. I think that's a problem with that, too, yes, programming is very big. Who should I hire? I mean, people are, like, hiring any man. He says even that he doesn't know it, and he says I understand programming. We're not talking about hiring companies that are cool and good at programming. I'm talking about companies that don't know how to program. Well, Tanya, for example, will decide to hire a programmer. She's got a little company, like, and she needed a programmer.
I don't know, I'm thinking about it. Real.
So how do you hire him? And you hire a programmer. He says, "I know." He also says it's cheap, like, yeah. You're like, "Oh, cool! So he knows and costs cheap." And he's hired-- what's he doing up there, how does he do, how much he
handles?
And the most important thing is, how do I assess it?
Or expensive people. I see it now.
What do you mean?
I'm seeing a lot of directors right now. Yeah, it's hard to assess. I'm seeing a lot of more cool people right now. When people are very serious, they say we're dealing with it, we're gonna get the teams right there, and they're gonna do a lot of things. But then you're looking at the facts, see that man himself is not using artificial intelligence that much. I mean, he's not himself, he's not-- he's an adept of it. So he doesn't use the code, he doesn't use Claude. He tells people to use it or use some different systems. You sit there, you see, you think, "Hey, and he's... he's, like, using ChatGPT to tell someone what system to use?" That's obviously getting, Here's a big question.
Yeah, I mean, I'm getting a little more and more surrounded, you know, a good tone rule when you go to a manager or something, especially, first of all, with some idea. I don't know, five or ten minutes with, uh, chat-bots. If they didn't stumble you out there, you're wrong, that idea won't work or something, you're just saying that. And if you didn't, you know, like it was a disrespect, something. You're wasting the time of other people, though you could spend, there, three or five minutes and, there, and you know why this idea isn't gonna work or something. I think this is a good tone rule, and it's a business ethics. If there's any idea, go ahead and eat it, smash it, hmm, chat-bot. If it passes, then you can keep it.
Look, it's not a big deal if there wasn't a chat-bot, is it? Go first, her shuttle, just somewhere with some stuff.
It's just that it's cheap now. You should have, there, thought, looking, spending, there, day, something else, and then you're gonna send two or three requests and, yes, no, no.
It's a question, how much model would really be positive and correct, and how she understands how you're going to develop. Given that we live in the world, in very different ways, someone in Deda Moroza believes, someone doesn't believe, does it? Here. And, uh...
There's a feeling that any idea can be destroyed, well, if it's desired, and it could work.
Yeah.
Yeah.
Or develop.
And who's gonna take this-- take responsibility?
I'm doing it, Sasha.
imagine.
Or a critic.
Or a little bit of a scream, yeah. You know the approach.
Actually, it's already present because it used to be criticized by neighbors, friends, two-way aunts. A-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a- ChatGPT still doesn't know how to measure size, nothing. And I sometimes throw it away in advance to be ready... hear... know what they're gonna tell me. Because now, especially in BL, everyone checks through ChatGPT. And I'm telling them in advance that you might have thoughts like that later, but that's not true. Because this, this, this. Yeah. This is a new reality.
Well, that's a problem. Yeah, that's a problem, of course, right? What you said when you were on the one hand-- llinar said there was a good tone when you could re-check yourself, not send, deal with and then come to someone before you speak. And there's another story when you, uh, you're gonna use ChatGPT now. And I see many meetings when people jump and say, "We can, we did the research, and we looked there." And you realize they're, uh, how cooler than they are? The fact that there's a man sitting there, he's got a simple role, and he has this supercomputer, and he's supposed to find someone in this supercomputer.
How can anyone find any answer and recommend business do anything? And when these ideas come, well, I mean, the leaders are very hard on it. Well, it's really hard right now. So you have to figure out more, you've actually got the idea of the man. Even if it's not a good idea, you know he tried, like, to do something or something, or to discuss it, or you got a man. The idea is good, but it's made through the model. Or this model that was originally built into this model. I see it all the time, everywhere, just in crazy volumes. And very often, people do it like kich is so kind. I mean, not to make a company really good, you know, or you got a good score, but for yourself to move out, to be at a meeting somewhere, to say, "I'm gonna say, I imagine what's happening in a corporate world, in a big corporate world." It's a thin man, of course. It's just that in a big corporate world, I'm a chat-- a part of the projects that touches on such a big, you know, relatively large, anyway, where thousands or tens of thousands of staff, or hundreds of thousands of staff are there. I'm not sure it's hundreds of millions of clients, but still.
And you know, of course, that this is a serious, serious problem. I've got an interesting analysis, by the way. OpenAI reported that every week more than-- interesting, by the way, number. I'll remember that they're about to have a billion weekly users. They reported that they had over two hundred and thirty million people on health and well-being every week. Two hundred and thirty million. I'm here to show, in particular, that this is not an issue that is relevant, we all understand that the subject of health is relevant. It's just that a decent number of people, yes, two hundred and thirty million, even from this billion. I mean, people, two hundred and thirty million people are asking about health. Poor doctors. And I got my wife and my daughter, and I got a baby girl to show her. I think I was in some sort of graduation saying that I was still taking pictures of a coin a year ago. And then she says, "Tell ChatGPT again, what's with the mole." I found the same old chat a year ago, took a picture with the AirPods headphones. The system says, "Come on, show her to the doctor." And when they were at the doctor's, my daughter takes and says, "And my dad at ChatGPT asked about the mole, so we came to you." That's it. Who's 11 years old as a daughter. My wife is exactly like, "Aaah! You can't do that!" Yeah, like a blacklist. You can't tell doctors that. And look what happened, by the way. By the way, I think it's a really bad, bad thing. I mean, it actually works, I-- I was in surgery right now, and I was very serious at ChatGPT doing a research on everything. And it helped me to get a doctor and get involved, and understand the hospital, understand the details, and even the doctor, don't touch him after surgery or before surgery. So you have the opportunity to take any picture, anything to ask and calm down, like, or, on the contrary, not to calm down and write an unnecessary letter. When you're told, "This is not a safe place, write to a doctor as you need." Or study the hospital rules in advance, which you don't understand. You get some descriptions, you don't know them, you're helping. I mean, I think ChatGPT is making incredible aid to everyone, so health facilities, doctors, etc. Help. I was gonna tell the insurance companies. Then I realized he didn't do any help. No. So, yes, he's insurance, the other way around, he can help you. But the point is what? That's a complex question. And it turns out you can't tell a doctor that I've done a research. I've got this because I'm supposed to say, "What are you doing with ChatGPT?" Although it's okay. Why? Because in the world, it's considered that all people are stupid and doctors, and doctors are just the same story, very relative. It's a story you have, like, a diploma or you've been certified, and you're smart. Although I know a lot of cases when doctors are wrong, and we all know these cases when doctors in the virus call antibiotics, like, there, without seeing blood or making a mistake, and So on. And, respectively, but the trend is that these people are smart, these stupid. And now everyone's told that ChatGPT is also supposedly stupid. And I think it's a very big problem in society. A growing problem, because at some point in time people should blow up. Because if I were to see a doctor now and I would have a doctor openly lie, I'd say, "Wait a second. I just did a research." He said, "You can't do this research." I said, "No, I'll sue you then.
With regard to these data and this study, about your incompetence." Well, I guess it's the idea. What do you think about this?
I think it's a story that looks like, on the one hand, an expert with a-- a smart enough model-- becomes a T-shaped specialist, yeah, when you're not just in your field, but in your field. You can appreciate the fact that you have a model that speaks different ways and works with it. And there's another story when the so-called neuroslops start to generate ideas. That's how you say business starts making more money, and when you don't even know how it works. I got something, I got it, and so on. And this doctor-related story, maybe they just see this neurosop more often than proper use. But I'm fully supportive of the proper use. I think it's a huge help for people. It's access to a good expert tool that can help you anytime, yeah, there. The picture, according to the description, suggests that there's something else to be found. But as if it were totally, obviously, it doesn't replace a doctor. But when you're used properly, he's only helping you with it. Well, the doctors are probably the same.
I think it's a huge plus, yeah. And it's not just about medicine, it's about, I think, the whole life. It is very important for people to be quite adequate here because it's different, a man asked ChatGPT or a man has studied certain scientific work. I had a doctor situation about a month ago, by the way. I came and offered to do something that another doctor told me to do. And I was told by a Russian doctor. I came to my doctor in America, and she says, "No, no." I say, "Trust me, please." I didn't see ChatGPT at the time. She checked her system, she's running fast. They have something, I don't know, some sort of vector search or something. She started and found such a addiction. She says, "Look, I really found something that's addictive. I'll do you a research." And then when some conclusions came, I already sent ChatGPT. ChatGPT confirmed that American doctor rather than what I was told to do by a Russian-speaking doctor. Although the Russian-speaking doctor said, "Tu-Tu-t-tou-tou, I disagree with that. And then with the results, and he kept saying something. And here, the question is, the healthy soundness of healthy debates. That says, Elnar, more about people being open to listening to other opinions, right? And be calm, calm to hear other opinions at all. Because there are so many possibilities, so many different topics. And now the system allows, I have, for example, an operation that she was, I have, you know, a doctor who offered me this, you know, a menu of all possible types of surgery, and said, "I'm doing this in person." Well, I've been offered me a menu. And then ChatGPT explained to me the details of this menu that this doctor had not explained to me personally. Well, or maybe because I wasn't super-speaking, maybe because ChatGPT might go deep, and you can ask even the stupidest questions you're like, well, you're not gonna touch a doctor anymore, You're not gonna get him any more bullshit, like this. I think it's a very cool auxiliary story. And here's the question of how fast medical systems are going to be differently embedding these, you know, those things that are called assistants, right? Or they'll be so cool with their own systems to recommend that. Well, in the idea, it's only possible if there are joint projects, that's how OpenAI has a joint project on health, very big, large. And there's this health bench, right? This is a set of medical dialogues and criteria on which doctors assess responses from models different and so on, yes.
When different red flags are compared. I think that's very interesting. So very-- that allows very much such systems to develop and move quickly.
I'd be another little question. Clearly, we're pretty big on video, just crazy interesting. I'm periodically comparing Claude and Codex, which is more about what we were talking about at the beginning on the same tasks. And I noticed the very interesting behavior of Claude. You must have been facing him, too. Remember when you said models would be much better when they started asking questions? So Claude learned to ask questions. I needed one table that was so slab, not very well made, and there were just some slang marks left. And I had to describe her. I threw it in Claude and Codex. Codex, without asking, thought it was a mark, and it wasn't right everywhere. And Claude, on the contrary, started asking. Said he was just about to start collecting a description. I need to make sure that these are...
I need to make sure that these marks mean. And he offered some options, which could be chosen or written. And now I'm dealing with it periodically, and I think that anthropics in this direction have made a very good move. Did you get into this and see if our viewers were in the process? I'm just saying, I really liked this story.
Look, the whole thing about questions, I think Claude, Cot asks ten times more questions than the Codex, and he's clearing them out. And I think that-- and he also loves to ask them in the interface, yes. When you're asking in the interface, I don't know why he's sometimes asking in the interface, sometimes not in the interface. There's a lot of technical questions to ask. Super, I say, "Don't ask me technical questions, ask me in a normal human language." Actually, I think it's one of the biggest problems I'm using now in GPT chat rooms in common, Codex, and in Claude, it's endless for me to say, "human language, human language," Business, business, business, business, logic, human language, no technical stuff. So explain everything in a simple language, explain all the terms. It's just impossible, right? Don't use it, so that's the cuts, don't use those things. It's a good word at first. He offered me that in the staff assessment yesterday, and in the lid evaluation, he proposed that this notion was 20-5% per cent, that's where it is. I'm sitting there, I think what a normal person is, even if I'm even more or less dealing with it, some kind of... like ordinary people, like this manager or a macro, how do they know what that means? So there's no people sitting there that will figure out, so that's the term you're talking about. And I think asking questions is very important. Illnar, I remember you know me, I think that the model and the artificial intelligence that man's gonna ask questions will win. And if the systems weren't starting sessions today, you wouldn't start immediately on complicated things forever answering what they're smart, how they all understood how they've been all over the time, even when you've done them. You're asking me to re-check everything. And if they said, "Look, we want to ask you questions, we'll ask you questions carefully, patches or different interfaces, and so on. Quick, they asked, asked different questions in simple mode, they gathered these questions, even when Proversia clarified 10, 15 questions. Remember when Deep Research was working, he asked you a question. Just made a great interface. You're getting any details on GPT-5, 5 Pro, then you're 15 minutes or ten minutes waiting. And he thought he saw a quick question, and he asked you in the course of his session, you answered it very quickly, and you could write those clarifications quickly. By the way, the system of clarifications was so normal and did not. Remember how many systems they tried to do, the cool refining systems? Claude Codex, by the way, I'd rather like the system when you can get in the session, much better, yeah. The chat, I think it's very curved. He feels like he's interrupting everything, and it's still doing it in parallel, and so on. Here's the story that the subject is still incredibly important, asking when the system is going to be asking questions, asking questions, asking questions, asking questions, asking questions, asking questions, asking questions. And that's a good story about what we're talking about now, especially in American health, because they use all these systems very much automated. When you come to a doctor and get into a certain area, he always asks you, there, forty-eight questions, and so on, the algorithm. I mean, all the questions are computerized. It is now developing very much in the world, in different countries, but it is not everywhere. And I really like it when I hear things coming up. So you're being asked a lot of questions ahead of you: what if? What if? And if that's the situation, if it's the details and so on. And the ability to ask more questions, to know what a person is going on, not to make any abstract conclusions. Well, for example, I'm always advised by all systems not to use alcohol the next day. But put it in your card that I never use alcohol. I had a very interesting situation. I was, uh, I was dealing with the VT, and I had a lot of doctors asking, what kind of food you got there, and so on. I always said, I'm very on the nutrients, and I don't use alcohol, no substance, nothing at all. And they asked me repeatedly. Then, you know, I was given a test. I'm back, and my sister says, "Sash, listen, you have one of the tests in there, whether you're drinking alcohol and how you've been using it for the last few years." Well, there was some serious test. I think he should be cheap, by the way, yeah. And they still decided to re-examine me, so they re-examine, cheat or not. And they probably did the right thing because I'm sure a lot of people will be cheating. But the idea is, if you're over-checking, you have this test, you have to stand in the system, and you have to keep not using it. Well, or not. When you write me these proposals, like, most importantly, don't drink alcohol the next day, or we recommend it don't drink. It looks a little weird, yeah. I mean, you know it's not about you that it's about such a crowd of people. It says we-we-- the day before the operation don't do this, that's it, that's it. And I'd like to see you sleep with the dog at night. Here. You know, you're not allowed to, I don't know, touch your baby, but with a dog like, well, we're recommending you not to sleep, yeah. And you're sitting like that, you think you're going on. Why don't you-- question me about 100, two, three hundred, there, four hundred questions in my life with you and not do what we did once, Ilnar, remember, with the students, yeah, when you were, You're asking students, and, yes, and you're gathering information about him, and you know, this man, like, in their family, and we're talking a lot about artificial intelligence here. It would be great to know, and what number of people have their own computer at home, yeah. And when people don't have their own computer, and there's a computer alone on their family, that's another situation. Or there's no desscratch. We're talking about Codex and Claude, and there's a big difference between a lot of people thinking they're starting a phone, but it's gonna be a little bit of a job on the phone. I can, of course, by the way, be me. I'm in the Cloud Code myself, by the way, not the code has done a lot of phone calls, but it's still a big question. These are, like, surveys, such systems, they're very interesting. Maybe someday we'll do some research on this channel. It's just, uh, usually a little bit of an audience, yeah, that's, like, YouTube. But I think that there would be a lot more people involved in the health system. I always said it was super cool. I wonder how the audience will say, yes, in terms of what you've raised in the Cloud Code, you've raised the technical subject, yes, it's more, it's also extensive, very interesting. Who's facing? Because I don't like the code in this respect. I don't like the code in this respect. But let's see what's going on.
You've been throwing your own limit in the code for once or not?
I saw that there was a pattern. I'm happy to see her, but I still have the limits until I get to the edge, so I'm gonna dig these dumps. I have two or three of them now, I don't think I remember how much.
I'm the one, I, uh, dropped one time, I had two. I dropped it once. There's something else that could be appointed for gifts if people were registered. But it's some kind of nonsense. And I do have periodic access to the Cloud Code and the Code even at maximum tariffs. But it's not easy. I mean, when the system is so very, you know, it is.
We've got to try, yeah.
In the pros, yes, the maximum versions should be tried. And I feel like they're still fighting for competition, and they're expanding, they've expanded, and, uh, they're probably expanded. It's like there was a moment when, you know, there was a number of 18 or when June or whatever, they were supposed to be VOPs. I'm running out of it like it was. And there was a moment when Cloud Code was sort of over, and now I had a feeling that Cloud Code was ending much later, like. But I think it still depends on what you're doing, what kind of data you're doing. By the way, this is my job to take my chat off, she, she started making Fable, I'll remember. And four days there or three days, it was Fable. She's still in this semi-semi-semi-semi-semi-semi-semi-semi-semi-semi-semi-semi-semi-seasy stage. And what the system did, I can't say I liked it. One more time, I saw, of course, this is the complexity of analysing a large analysis of the dispersed data. What are people getting at? That people, without knowing that the system can't understand that, that's all, that's all, that's all. They say at some point, "Oh, some kind of nonsense, oh, some kindergarten, oh, something's not working." Because of course the system isn't so easy to do such a cool, big, big research. For real, I mean to create an external system. I mean, if there were four and a half hygigabaytes to download at ChatGPT five, five Pro, maybe she would. By the way, I'm not surprised if I'm four now, five gigabytes are split into ten archives, a ChatGPT five, five Pro, and she can analyze it herself. I'm gonna try to do this, because I recently dropped four hygigabates on one ERP system, and she's completely analyzed. I wonder if she can analyze it here. Well, ToTheMoon Channel. Send our channel to your friends. Sign up for those who haven't signed up yet. You'll leave a comment. I'll see you on the canal in the coming days. We always have different episodes. I recommend.
We always have different episodes. So do, including the topics. Uh, well, uh, uh, I'll see you in the subcaste in exactly a week.