This week's top-off agent rate. And the way I was surprised by Codex, he was deeply surprised that he wrote some quick thing, she was working hard, sending me messages in Telegram, making some rounds I needed. And all of this was without rereading any code. But I still think there's a story not when I'm running some agents, but when there's some single point of entry that's all that's between itself and makes a decision, who, where and where. I'm not sure how to start.
Either it turns out and it's gonna cost a lot, or we're gonna have to wait. And really good quality things will come only when we move the next stage, linked to the videotapes.
The huge amount of sophth that the world says will disappear. And a lot of systems we know about. For example, as we know the system that there is CRM, I think that in ten years CRM will not exist at all. What are the people who are capable of being so broad, fundamentally perceived and rightly challenged? Hello, everyone!
We're on ToTheMoon. Technological news, Silicon Valley sites and the world. We're going out every Sunday. So, listen, of course, we have an agent-work thing today, and we'll tell you some interesting things that were discovered this week. And even the market for conflict has become. There was a big conflict, uh, relative, but that's a fine conflict in Huang's speech. He was at the sub-cister's, Patel's, and he had a conflict like that. This shows the conflict of such NVIDIA, the opinion of NVIDIA and the opinion of the head of Antrotic. Like supplying data to China. We'll talk about it separately. But where do you want to start? So I'm gonna say that I've told a lot of things that I've been watching, what kind of things I need in my life, so life-saving, simple to start them and use some way of a Warehouse agent. Some, some, some agents, yes, or agents in the GPT chat room. Now Microsoft has released a separate jack, and we'll talk about agents in the GPT Codex. I'm here. I told you a week ago that I finally built it. I've already made a huge number of different agents inside the usual basic tasks, but it didn't have to do with what was called programming. Here's a very important story, that for me, the key was to keep me from going into any programming concept. I mean, I don't want to open and watch the source code, I don't want to move the files, I don't want to deploy anything on the guitlabe, I don't want to do it. And the surprise to me in Codex, the surprise of being so much, the profound surprise of what he wrote some quick thing, she's working hard, sending me messages in Telegram, doing some kind of rounds I need. And all of this was without rereading any code. I mean, I didn't want to get the code, I don't want to check it out. There is now a big, big story in the world that people now hire specially additional people who should check on other people who wrote the code with AI. So people check AI code. And there's a lot of discussion that code AI is not always qualitative, that it should be written by qualitative people, and so on. This story stays, I don't clean it up, I don't clean it up about the quality of the design, but I'm always interested in just developing some very simple things. I came in, I wrote two lines, and I had an agent who walks over there, I don't know, monitors the markets, and at the right time, I said, "Alexandre, in three minutes, you have to buy stock. If you buy it, you'll have growth." Like an example. I mean, if I'm doing a stock analysis with a GPT tip, and he's doing me some kind of profitable, but I'm still hand-wrapping in, I'm asking. I'd like some system that works for itself. It's so that I, well, I know a lot-- everybody knows that I'm very much of a data analysis. These are hundreds of thousands of calls that can be put into the system, in the analysis, there's 500 thousand lines of raw material on the way staff worked, some sort of technical assignments. A lot of things. And I've got so much analyst from people in this business. I just couldn't. I asked them to unload the entire CRM system to me and to unload the entire system that contains transaction data. So I've been unloaded. And when I was downloaded, I asked the analyst, and the senior analyst, who's very serious, and I've been working with for a long time, to give me a report on this database. While he was doing a database, I was thinking, and I'm gonna ask the GPT chat, I'm gonna go over to the agent now, and I'm gonna ask the GPT chat room for a data sheet? And GPT gave me a data report, okay, in an hour. And the director, well, the scotter gave me the next day. The scoring manager gave me a report of the data written, I opened it, there were numbers. I had a lot of questions right away. And GPT gave me a report on this huge filth with a bunch of pages that were written. You want these details? Here. You want these details? Here. You want these kinds? These, these, these, these, these, these, these, these, these, these, these. What I think nobody in this company has ever analyzed or seen in their lives. Well, at least I-- this is the company that my partner runs. He never showed me that data in terms of database analysis. At this point of time, it's getting hard on what plan, I'll tell you. Well, there are people who keep scaring, running channels. This is not about me and Eldar, is it? ToTheMoon is not relevant. There are channels where you're always told you're being haunted, you need to study AI, or you'll be fired, there, and everyone's gonna get fired. And AI is coming, and robots are coming, and nothing happens. Yeah, we often show that, yeah, we have to learn AI to watch, but everyone's got jobs like everybody moves and all that stuff. That's the kind of p-p-smart we're promoting, I think, right, Ildar? Let's get this idea up here. But at this point of time, when I got this report, I had a clear history of all these people to get out because I don't need them anymore. What did I do next? The next day, I asked for some more data on my face. I asked for super access to the CRM system, you know, because I'm the owner of the company and I don't need it, I'm not sitting, I need a stop, stop. I know you know what Ildar did? I came in, Ildar, created a token to myself, just a token full. I started a week at 7:00 P.M. Codex. I went into ChatGPT five, five four pros and asked him to write a C.O. for Codex, that I should get involved in the system with that, that's what I'm gonna unload. He wrote the start, I threw this TZ in Codex and told him to get me the data. I came in this morning, he said, "All of you are ready to get any data from the CRM system." I make any requests, I'm given all the samples, any, absolutely any words, any details. I say, "Take me calls in the CSV in a week, these are the managers with this long time, with these details." He's back in 30 minutes, and I've got ninety-nine percent of unspent tokens in Codex all the time. I mean, I don't know what to be-- well, it's got to be 15 right now. I, uh, I have a Codex working on three sophes this week for days. I don't have to spend the time with the tokens, but at least my version. I've put my phase on the regime. By the way, ChatGPT added the phase to the regime. Ildar, it's not a week, and OpenAI also added a fasc to the post-Atropica regime, right?
They added the fasc to the regime.
It's a trail of the trail.
Yeah. What's this regime? It's a fun regime, by the way, and I think he'll be here soon and in ChatGPT as a regular human being. It's a regime that allows you to get answers faster. So we'll get him soon, right? I mean, I have answers now about where the system is three hours old and where I can say, "A, waste a piece of my resource. I need that answer, like, in three minutes, right? And he's going to be. And somewhere I say, "I can wait there, night." And I'm starting the night. But I do, because I have algorithms working around five, seven sessions with a lot of data, all the time. They give me, I write next immediately, I write the next prompt, and he leaves. I mean, it's an endless process. Yeah? Well, we're going to this regime now. And when I got it, Ildar, this access is easy. And by the way, Ildar had a subject like that, I don't know. Write it, the audience, if you had that question. I didn't want the system working on my computer. I wanted her to work online. But when I came to Codex online, there was a problem. I was asked to connect GitHub. I was told that GitHub was not very good at being a listener of some kind of external systems. Well, he's listening to messages from Telegram or listening to messages from some system. So he's good at time and time and time. Well, shorter, he started writing some complicated words that I should set up some kind of server clouds and everything. I can do that. I'm technically well-trained, but I realized I didn't want to. I mean, I want to have a system I come in, write, and she's gonna start me up and do something.
And now Ildar and I are going to tell you. He's out, he lets Claude Codex Desktop and he shows up.
Yeah, yeah.
Yes, Claude Desktop. And it could be compared to the Codex OpenAI. But it's different things. Yeah? Still, Claude Desktop is a code work. And that's when you can analyze the code, and he specializes in it, at least for now. Maybe it's changing. What I saw.
No, Sas, they've done a lot of work for other users, too. When the files are there, the files are working, they just expanded it a little from the software.
A little more of it was from the programmers. But still, by yelling that anthropic products are purely for the
programmers and are primarily made for them.
Yeah. And you and I can see that ChatGPT has released two apgreids now. He's the one who let corporations out. We'll tell you right now, yes, you'll tell me. And the other one, he released for-- further updated the code. And now, by the way, there are five four codes that are the strongest. Oh, she's not a code, but it's just five-four. There's a model of five two, five two codes. As always, you can't figure out what it is, what it's about. But five four is the newest. There's a model inside, it can still include this high level of liening, it can include fast-mode. I'm saying I don't even have a flower. But the point is, I went back online to the offline. They're even pro-- now, by the way, like ChatGPT planning on doing it online. They'll do it, I'm sure. I'm sure they'll do it when they can run it all on their servers and work it out, listen to how-- how does a Russian stander say right in the programming? The software product that listens to another environment is constantly studying, yes, and is built upon. Suddenly, he sees, for example, that he's always there, I don't know, checking stock every second, like, watching, changing, or seeing some kind of trigger. There's a message that you received from the WhatsApp or Telegram or some system you had a call. He's responding very quickly and easily at this point of time, yes, here's the system. I think they'll turn it on. And for me personally, of course, that's what it has, has a plus. I'm leaving the computer completely on the computer now, and we're thinking back to Carpathova, yes, which, like this, has betrayed this case with the Mac mini purchase, although Mac mini, obviously, has bought a lot of people when He turned it on that computer then, really, Open Clo.
But you're still turning some kind of agent on it, and this agent environment is starting up, depending on the level of quality you want an agent to build or what you want to do, what kind of analysis you're looking at. You want to do it. It's starting a little extra work there, and you can write a program there. That's what I like right now, that there's a program that I don't like. You still have to agree on some technical stuff. And you have two regimes or a full access or a negotiation. And what you're agreeing, you don't even know you're gonna agree on. And that's what I think, even a bunch of programmers don't understand sometimes that they're in there. Oh, well, that's how life works. But I like the system now. I asked you to re-examine the fact that there's no leaks to my computer that he wrote this code and that code I've been checking it out a lot. I even loaded it separately and rechecked it separately in AI. It's clear, for example, in the GPT chat room either separately or in the clade. It's loaded. And you say, "Try me this code for extra leaks that might be." He's in my code, like on the computer, he'll check and do it to you. What's the problem? What he's writing is technically, when he's making this application, software is creating you to, like, get you on the system, or you're gonna get in the house to the system. video cameras, or you're connected to some, I don't know, system, some sensors, right? So he's still telling you what he's doing technically. Although, I think, here, Illnar, I want to start a discussion on this subject that this week's highest rate on agents. Microsoft made a controller for the management of AI agents.
It's a very clear title. Google, he said we're turning a separate environment for the management of agents. ChatGPT has already turned into a code before, has issued a code, says it's gonna be like some kind of system for managing agents. Although it's not a system for managing agents, it's some sort of project files and chat chat. I mean, 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, no But I still think there's a story not when I run some agents, I go in, I'm getting a separate name, and when there's some single point in the entrance that's all between itself. It distributes and makes decisions where and how to start. And he knows where the timers, where the lens, where the functions are started at times, are not at the time where the power is needed, where the extra server needs to be removed or where it doesn't have to be, where it's more or less advantageous, that pay where to wait or day. I mean, what we're going to do is, like, a whole different story. Actually, that's not what we see right now. I mean, what we see now is more like the increase in the shares of all these Microsoft, Google, Apple, including who said, made a new statement.
XAI, by the way, is planning to buy a code. Look. Well, I guarantee.
Yeah, but for them, 60 billion is not a question, because they're planning to estimate more than a trillion. And they seem to me to be able to pay the shares of sixty billion, you know? They can take the money of a sixty billion people, give them the shares. Look, they never do. Look, the most interesting thing is that SpaceX is not gonna kick anyone out of its sixty billions.
Sas, there's a funny guy on this subject, you know, I got a friend here, so, let's just say, I know someone. He's doing the construction of housing units. And there they built a gym in their housing complex, some tennis ships, something like that. And you go in there, and there's almost no one out there, you know, because there's still a lot of people in there, not everyone's here. And in general, there's a gym in the middle of town. And he was asked, "Look, you spent so much money in the void, built a gym, it's definitely not paying off, it's definitely working for you at a minus." He says, "Well, you know, we just built this gym in our housing complex, with the cakes, increased the price of square meters and two billion above it just came to its existence. I'm told he's got money or not, but he's gonna be there sometime. It doesn't matter. The same thing, SpaceX can do now, just adding to himself inside, like you say, they're on the IPO now and so on. By adding a cadet to themselves, they'll just raise the price of their shares so much that out of the sixty billions, they'll even be unmarked for their spending, even if they get them out of their pocket and give them away. Here. And about the whole story you told me, I haven't seen you so excited in a while. Second, I remembered Carpathians right away. Remember, we talked about three or four months ago, probably just when one of the next updates of ChatGPT happened, he was just telling me that the code was not coded in the near future, the programmes would not be available in the near future. I'll buy, and they'll be generated, they'll be generated, and they'll be created right now. You've got something to do, you're making it, you're generating it. Then we have audiences in the commentaries a little sceptical about it, but in general, that's what you're saying right now, that's what you're talking about. You should have done some connection with the CRM system, collect data from there, analyse them, and so on. I'm sure you could be there, I don't know, buying some product from, uh, CRM system publisher or paying them money to finish something or something.
I mean, you could have bought a software outside, but now, with these signatures and all that, you can already generode yourself, there, in the evening, so that it works and does everything.
Ildar, look, Ildar, it's very important for all people here to tell, all people to get a little bit of a generosity in the evening. Again. First, that I was out of years, I didn't generate anything in the evening, I wrote a request.
I wrote a prompt, yes.
He wrote a request. And in the morning, I'm sitting in the morning, and they say, "System is ready." I say, "Get me into the CSV file with all the information on my phone in the last week." He was doing it while I had a meeting before, and I just had, I'm sorry, I was late for you. He's got it all down, he says, "It's all set." I'm off, I'm getting a CSV ready. But I-- I'm not in CSV, I didn't need CSV, I really need a friend-- other things. I mean, I have some requests I want to analyze. I mean, for example, "Can you get me to pick up the staff who called for less than two hours this week?" Right into the system, ask her not to unload the files, ask her nothing. And since it's AI with thinking, well, I mean, thinking about it, it's not a program written that writes a program, right? She'll decide where she's going to go to get her up. So this is the ChatGPT 5.4. It's not 4O. Look, he's offering me, like, a call to assess, a voice-and-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- Call, call, check, gerization, and share them, for example, in the field. But it's 4O. You're like, what four-Os are you? I want to have a logic. I'd like to go to Pro's version, too, wouldn't I? And I may be going to Pro later on, but at least 5.4 high-risoning versions, she'll be in the infrastructure myself. And it turns out I didn't connect here with some abstract , you know, I don't know if you're using or not, viewers, write ours at all, you're using. I hooked myself up in ChatGPT, there, Adobe flame, there, I don't know, Zilow flame, there, Booking plagin, I don't use them at all. So, here we are, Ilnar, and that you and I were discussing two years ago that these flames are... why are we looking for a flame? You don't need him.
And CustomGPT, and
that's all that's not necessary, yeah.
What's it called? So he made a separate environment for GPT tea, for corporate clients. Teams can create a general code, a code of committed agents who work in the clouds using various tools. And they're just saying that they used to be GPTs, and now they're agents. I mean, finally. It's weird, yeah, they're GPTs... Remember we always said that you didn't develop what you did? You could have developed those. And they're saying that in the future, there's a way to transphorus-- transform GPTs into agents, right? What's the problem with transforming them now? I can now take all the GPTs codes, copy this code and completely clean it up. And that code, well, translate it into an agent. It's a basic thing to do. But the point is, you'll have the Agenta deposit in your corporate environment. I hope she'll be in the private environment and she'll be rolled out, and she'll be rolled out on the interpranise tariffs differently, plus food tariffs, well, eduction and so on. And there's gonna be agents, impound, in this deposit, running, like, online, the creation of such agents. They are positioned that these agents can be connected to the application, to the various applications, there, working with the browsers, there, with details, with everything. See how it works, because I'm OpenAI, I don't trust their products, they're chaotic. It's very much testing. I hope they don't throw the code on this, do they? Because they've made a kind of deposit of the Agenta and, in fact, this Workspace Agents, it feels like he's not a code, but he's a code, it's a code based on the code.
He needs him first, as you say, for the Enterprise. It's just that you have, on, on, I don't know, a hundred-staffed agent manager who distributes the power that controls their ability to work, tests, changes, changes, Something like that. And you're telling the staff member that you use it, that's what you use for. Don't write it yourself. We have ready, tested. And you're the manager, so inside the company, you can access all these agents, what power you give them, how many tokens who eat and so on. If there's too much to cut, you can either cut it or something. I mean, you have complete control over it. In the private environment, as if I didn't seem to show up, I guess. First, I think it's done for the Enterprise, but as much as ChatGPT has ways to unconvinced. Let's see what they're getting to.
Yeah, but it's gonna be fun.
You want more? Yeah.
I'm just gonna add that the cash would be exactly what I'm interested if I had it inside right now. But again, I don't believe in the linearity of agents. Well, I mean, I don't believe in... agents, in the files of the lying, yeah, or that read some microchid. First, they must interact, they must interact, agents must be agents, agents, they must be between different systems, and they must be in there. I mean, it's gonna migrate sometime. And I'm a little bit of a corporate approach, like, very hip-hop.
I mean, corporations are like not a good product wanting to do, but win this information agenda so that, as I see in OpenAI this morning, the girl says, "You know, I don't know, I'm not the first person to know. Maybe that's the problem with my family, with my kids, I had a pretty picture of myself, and now I can take a picture. And ChatGPT is a new image... " ChatGPT has released an update on the picture. I don't know if he needs time now. A ChatGPT will give me this new image. And it shows that she didn't take a picture in the area across the building, and she shows that they're on the resort. I think it's like a mom of kids who wants to show that she didn't know how to take a picture and decided to change the location. I don't know that. You know, one Huawei case has released a thing in its phone that automatically, during the photo of the selphia, tells you the right thing and tells you different positions as to how to be with this object or in this. I'm here to make you look better. It looks fun, it's a fun idea. But another case she's telling you is that it's ancient world. Well, Illnar, it's an ancient world, right? How many systems can change--
It's just a confirmation that all AI companies, they're good at Hype. That's how it is. First, you remember when the first Images came out, everyone was making pictures of the Ghibli studio. It's in Japanese mults, the animation of the studio is quite well known. Then Sora began, respectively. They're up this side of the hip. Everyone was making, like, a little video at least some time, where, I don't know, Sam Althman or anyone else. Now they've released the Images two zeros and they've also gone on a Haip train, a hip track. It's just that everyone started to smash the Internet with a neuroslop, which is good enough, but it's just, you know, it was with the Ghibli studio. Now even in the demonstration, this is the story of the pictures show, they're showing off that we can draw comics now. Accordingly, the comic people start to generate the same kind of thing as it was once with the Ghibli studio. And accordingly, I don't know, I may be in this environment. I've got a friend, and on the Internet, everything I see is all Sidney Sweeney's. She's been playing the game over there and Hermione Granger. So it just came up with the possibility of generating a rather good content already inside the GPT. And, accordingly, it's just in the coming days, weeks, probably, it's gonna be a mess. Then everyone gets tired. It makes a special sense of this update for a simple user, I don't think so.
Play funny, but it's like it's not that much.
Yeah.
Here. And the P.E.'s generation was about to finish the thought. Yeah, you started saying that, like, everybody said that you should watch what the AI is, and you can't trust him, he can make mistakes like that. It's just important to understand the value of the error. And you're building a MVP or you're making a sales product where a lot of users, and the price of mistakes is you've got a server lying there--
And you're taking money for it. And you're taking money for it.
Yeah, yeah, yeah, yeah, yeah. So how much I'm not getting out of that level, huh? I mean, there, I don't know, some very important systems with hundreds of millions of users, he can't be slacked into the sale, and mistakes can be. And when you're just making MVP for yourself, there's no problem. You're doing it peacefully. If you know how much the price of the mistakes isn't so high. I mean, if there's any error in the code that makes you unannounced, there I don't know that the stock is crawling up or something, you're not gonna get really upset. Also before, I guess I just didn't see that, did I? There, or do some other notice or something. But as many as 100 million users will not suffer. Here. And in that sense, you just have to split it. When we talk about MVP, you can do it. Many of the start-up creators are now, there are one, two, three people just with agents, and they do all this with the HI chat-bots. They get to MVP, raise money, and only after that, they start pulling a team.
So, this is a perfectly normal development.
What do you think about agents? That's what I said, that there's still this linearity of agents, or that's even what Microsoft does there's a system for managing agents. When I read about it, I think that for-- that's Microsoft's style, creating something, you know, it's the way they're approaching it, it's a kind of hardness. So now we're gonna create, and we've got system administrators who run laptops or computers, and now we're gonna set up a system for managing agents. Still, where Anthropic is moving and where OpenAI, XAI is moving, they are still moving into completely different spaces. Like, you know, how I was-- it was information that 40 per cent of the music that is being plunged into the Internet today is music that is built with artificial intelligence. Yeah? That's a big point. But I remember Ilona Mask, who, a year ago in May, my buddy said that we knew-- we believe that in XAI, the whole content would be generated by AI. All of it. Not AI people are sidelines, but some AI, probably, well, in this case, AI XAI. I mean, XAI generates itself. Why would XAI have someone who's doing something? XAI generates the whole content itself, right? I mean, that's what makes him a man. And that's a completely different paradigm. So it's not a problem with who's getting a content now. It's a question of where you're going. I mean, these companies, they're moving, I hope they're still moving in to create what's called AGI, you know, conditionally, something, some kind of myth of another, exactly when we're... We understand that we're going into some environment where it's not what it is, it's some kind of environment when you have ten D environments at the same time, isn't it? And it's one chat, and it's a chatty. He's the one who's trying to figure out what to do. You don't do that because it's a crazy thing to do. When I walk in, I'm in now, I'm in, I'm-- I'm opening a code folder, and there's a bunch of files, a bunch of written code, descriptions, readme there, just a nutcase, just a nutcase like that. He's the only one who tells me when I'm in the code of ki-ki-kid with some thicken or something, he says, "So, this co-cos-tock is compromised. You're gonna have to put it down here, this file and create a new one. Because if you put me in the chat...
That's funny, by the way, Ilnar.
Yeah, yeah, yeah, yeah.
He reads his own chat, scopro-
Your own chat with compromising token. And I like it.
I am. Yes, it is.
Yeah. But I'm not, like, that-- yeah, that's not the idea, because they don't have to learn or watch my data, they're supposed to be removed there, but at the same time they are, right? Because he still got somewhere. And it's cool that he reminds me so much. But, obviously, a future... What do you think of this layer?
I want a little entrance because you said that music was generated with AI. On the one hand, it's bad, on the other hand, I can't say that I don't like it because, for example, there are very, very funny, uh, options that would never really be. Well, the last one, I think, heard, was in there, faced in YouTube or something, Kanye West, I think, was all the songs he could reach. He was there and "Blunt roses" was singing and he didn't sing. He even reached, like, he did, even reached the Tatar song that he sang in English. And it was really fun. I'm a reference to a discount, we can add. I mean, I don't know how familiar our viewers are with the Tartar folklore, but... As long as it never happened. I would never have Kanye West, there, translated this song into English and sing it. It sounds, well, it's fun, I like it. I mean, never would have this thing come up. On the one hand, it's a neuroslop, yes, as if it wasn't living people or that, but, well, whatever. And when you just want to hear this idea, and that idea, you either like it or you don't like it. It's very cheap and fast. That's what the agents are. Look, on the one hand, uh, this is the story of Open Cloud, when everyone started trying to do all this, yeah, there, and that all the big companies are coming this way, it's promising on the one hand. On the other hand, I feel that the iron computing power is very much behind the opportunities that companies are going to be providing. Oh, I'll explain. If you take the same Anthropic, you take the limits they give to the tokens for use at different tariffs, yes, there, at 20, at the stadolare. On the 20-dollar, if I'm not mistaken, Claude Code has already been cut off. They're slowly tearing down, uh, cutting them, getting smaller and smaller. What's the connection? The models are stunning. Tokens need a lot, resources are not that much. The number of users who want to use this is increasing. If we take-- and this is just a demand-driven generation, yeah, when you make a request, and that's where, well, not exactly a question-and-answer system, but you asked, you got an answer. That's not what you're talking about. When you put the request in, and she's starting to call some other agents, create it like that, right? I mean, on the one hand, you have a lot of users growing, on the other hand, your models are getting more stunning. On third hand, you're starting a different model, starting a whole ensemble of models, right? And when we say we have a lot, uh, inside the sub-tasks and steps, you have a lot of computing powers to do. And there are two options, either that will come and be very expensive, yes, well, good quality agents, they'll be worth a lot of money just because the computer's missing. Either we have to wait. And really good quality things will only come when we, there, are moving the next stage, connected to, there, with video cards, right? Now, Vera Rubin is developing, soon to enter the market, the next ones are doing, but we're still working on video cards that have been there for over a year or two. Consequently, the market should move to the next stage of the computing process, and in some way optimize all the work of neuronets. And then, after that, the next irration of agents, I think it might already be. Now, I think it's still so demoversive. Or that's how Claude Mythos was history: they only opened up for 40 companies, for the chosen ones, that's where they're strictly checked, and so on. Also, with these agents, not everyone can afford them. It's either big companies or rich people or something. Now, this is a direct mass development in the current amount of computing capacity, I can't see it, to be honest. I may be too skeptical, but I have a feeling about what's going on with the limits on the different subscriptions.
It's a yes, it's a supplement to yours, including to my question, which, in fact, to create an environment that keeps in touch with everything, it's all interacting, of course, it's a huge number of tokens. And we have to go to the model so that this environment can come, we have to go to a model of completely different pay for the system, yes.
And to give it to ordinary people, something else must happen. I don't like the story right now. If everything is to focus on enterprise now, I think, I think, that there will be a loss in quality creation and development, in the creation and development of AGI itself. If you focus on enterprise, right? Because AGI enterprise is not needed from the start. That's right, yeah. And the enterprise has some fiction that they need. Even if you and I are to see the examples that OpenAI has given, where they have implemented, the Codex. I read these examples. They look like a nurse to me, honestly, Ildar. I mean, uh, I just know what we do in our different companies to analyse data research and their examples from the series that their code has improved, there, it's easier for a sales manager to do, there, no. I know I saved him five hours. It's just something-- some kind of nonsense. Well, because what is it? I mean, we're talking about systems now, not about six hours of the manager getting free. We're talking about getting a very new knowledge. Like I said before, two hundred thousand calls were used to analyze and build a degree from one to ten calls, good or bad, because if you were to do it, you should have. I'd start building a system that would learn. What's the good and bad calls? And she had to do that on millions of phone calls, implicated or learn something. Now you're just putting any calls into the system on any subject. And since she already has a lot of vast, trained, different information, from math, biology to the psychology of ordinary people, she's making you this assessment. But it's not a question of six to eight hours or ten. It's about a whole different knowledge, that's actually a whole different knowledge, which, you know, about what was actually impossible. Just like the API I've been putting on, by the way, that subject, she, I think that all people, someone, know what API is or who's programming, supposedly, well, what difference does it make? That's all bullshit. But in fact, if you're looking at it, it's a very serious move forward. The way the system did it now. And without the guithab, huh? Because I'm saying that I had to because she offered me to build a guithab, I built a guithab, I did, but I went back, I left it because I realized that I, I--- for me, the creation of a guitar, Now, agents are a story without this, uh, programming environment. I mean, I don't want to go on this Wednesday.
I mean, I want to stay in the future and bend in the future rather than under the past. Yeah. That's why I had this conclusion. And Claude Design's out, Claude Design's out. You saw, Ildar, no?
Yeah, I'm just wondering if I could see it again, and I'll see it from the side. Then we wait for Tanya. She's gone for different reasons today. I wonder if it's Claude Design that this system, which, uh, actually describes it as a condition, is not just a function to make a beautiful slide, which Anthropic describes it. As a product of Anthropic Labs, which allows Claude to design, interactive prototypes, there, slides, etc. And is it not the same hip, there, or just some kind of thing that OpenAI does, or is it some sort of editing to get into this world. Who makes one-page puppets, who makes beautiful prejudges? I hear a lot of people saying, "We don't do GPTs, we make predicaments, we do them in Claude because Claude Prezes make it more like beautiful and delicious." Yeah. But I'm not interested in that at all. I'm interested in the quality inside. What's the data on it? Plus, I don't like the presentation, but I understand that we're living in a corporate world today where people like all these beautiful predicaments, beautiful pageants, which are exciting, so these single-page super cool, It's all so super-funny and super cool, right? And that's how if people look at the world's resume now, I think, if you look, everyone should have a super perfect summary of the right terms, with very smart words, with the numbers written, some kind of... - Yes, it's not. Well, you're looking, and everyone's fucking crazy like organi--
It's fucking organized, right?
Well, look, with regard to the resume, it's always a little more, uh, working on two sides. So the researchers make their resume visible. On the other hand, actually, er, hiring, uh, adding filters that remove the extra summaries. And you're always fighting. Perhaps at some point, the employers just started to look at the well-established summary, where the numbers, achievements and other things were better. And the market, on the other hand, is clear, adding.
So the cause of the investigation here, by the invoking, may be different. But yes, the resume is the same.
We've got a comunity on YouTube. It's a separate function that's at YouTube level. You can support our channel now, connecting this comunity. There's a very small fee, a few dollars is a month's payment, and you'll even have your special badge. Especially those people who often leave comments or make it clear. You'll be identified. Actually, it's for, uh, the development of our channel. We have over 600,000 subscribers and tens of thousands of people watching our videos. Yeah, uh, you keep us, you keep developing together, and we'll, uh, be reinforced, including all the interesting materials, right?
Aah, I want to tell you something now. Aaah, you did, uh, research. Gallup's doing a research. I'm not sure I'm not the type of research, but I'm still. This study was a story that said that half of the working Americans already used artificial intelligence. Yeah? And when I read it yesterday, I was thinking. Well, at first, I thought it might be some kind of faeque, because, uh, they were probably asking people who could have answered them. And those people who could have answered them, uh, they don't represent the whole society, because there are people in society and in Americans, including those who, uh, never answered, wouldn't have gotten into that sample, right? And they don't think so, but they took them to this sample and immediately set it up. There wouldn't be a half in there, would there? Ia, but still among the people they interviewed, half said they were using. And there, there's a decent percentage of people there who use this system-- well, using artificial intelligence professionally and on their side. I'm all thinking it's all unprofessional, right? Because a comrade came to me yesterday, I opened it. First thing I did, I chose a model of thinking and she chose a regime one-second. So, all the people I take a phone with GPT chat, all the defaults keep standing instant. Illar, this is amazing, right? Well, I'm not talking about people I don't know, I met, like, a man from-- and talking to a man who starts up here, right? Or who works in a corporation, some big, serious, and he's dealing, and he's connected to artificial intelligence, it's the artificial intelligence that connects, yeah. All those who are not connected are, of course, they-- they think they're using, by the way, professionally. So they're asking at the GPT chate, he's answering them. Yesterday, a man is coming with me in the car, asking him a question that's not easy enough. He's answering him in two seconds. I understand what the answer was? From the model's memory. But the memory of the model on this block could have been updated a year and a half ago. Here.
And, uh, he doesn't even understand that. I mean, how does he start to treat... Ilnar, by the way, is that interesting thing, huh? I don't know if you noticed or not. People start, start to do it, start to treat it as reliable information. I mean, I'm thinking about GPT Pro reasoning, Pro or there's GPT thinking reasoning with a high level of heavy, that it's reliable information because he went on the Internet, looked, looked, rechecked. I'm a GPT chat, I like it. He writes, I don't-- I like it. He's saying, "I'm not taking the data I already have from the model, but I'm going to re-examine, maybe something's been up." I don't know if you ask him, there's a pay-for-pay parking or a free-of-charge parking on this street. He says, "I'm not taking it, I'm going to study." Because he understands you chose a model that needs to recheck the data. So these people who ask, they're sure they're answering, uh, as updated information, yes, like they have, uh, checked and fresh. I mean, like Google, and then I thought, "Well, Google's already, uh, lost this fundamental quality of knowledge a long time ago, right? There's a story you can get out of there that's 15 years old, she's describing, I don't know how to borrow, right?
Sasha, that's what I was gonna say. It's a little sad. I mean, you know, uh, I don't know. Okay, you and I don't remember that much, but there was a television show that always the truth was on TV. Then there's a place in the nineties where they stopped believing the TV. So, at the beginning of the zero, it was, uh, zero that, uh, there's no lie on the Internet, right? And on the Internet, like, right, right, right, right, right, right, right, free space, and so on. Now, we're actually saying that Google references and Internet information can be very different. Now we're going to see that there are GPT, Claude, Gemini and others, they're not generally very different, and most questions of their answers can be believed. But I'm afraid that whatever happens to them is the same thing that happened, uh, there with the radio, the television, the newspapers, the Internet and so on. So the list is just gonna be going to be going on with those chat-bots. At some point, just those who run them, those who do the training, those who do the next fainting and the other, with so much, uh, leverage on people, they're gonna be just gonna wake them up.
use it.
There'll be 100 percent of it, of course. And it's important for people to understand and understand, uh, understand it. I was sleeping yesterday, uh, considering what happened here, it's an API connection, and I've put my question back in terms of the future, future programmers, because, yet, the programmers. They say. But the point is not about the programmers, but the point is, people who can, actually, um- everybody says that the programmers don't need them. But the question is, the problem is not about the software. The problem is, what? What kind of programmers is this or what analysts are they? So what is this analyst that gives me six numbers now, not giving me a picture of what I'm doing? Or what kind of programmer does wrong and doesn't check the data? Or what is this programmer who doesn't use artificial intelligence today to analyse his quality code or code quality architecture or to restructure it? I see a lot of that. The question here is, what are these people gonna be? And programmers, analysts, designers, and salesmen.
What are the people who are supposed to be able to come into this real wide, fundamentally take it and rightly set the challenge? And you and I know that Elnar was always the problem, and the problem is always gonna be. And I'm not that I fell asleep with a calm thought, I'm not really worried about the whole thing, but I'm sure it's just a little redistribution. Who gets more money or who has more advantages or more opportunities. Or if there was a time when you were a programmer, you'd make a very high salary anyway. Well, now there's a time when it's just this wage, it's probably going to be moved to some zone.
But there's no question about the programmer, it's probably about all the specialties, any job, profession, there, human life, money, anything. I mean, it's gonna move just the other way. It's like Google released the ends of this Google in Google Finance. So they had a Google finance update, and they're, like, writing it separately, Google's updated, planning to update the finances. Now they'll be in the marketplace not just saying what position he has, but collecting some new data. I opened up, looked at it, and I thought it was as irrelevant as it was all dead. Because if you're really working in this system, you're in a state of Gemini, you can do things that didn't dream of Google Finance with a regular chat and that's a lot.
You need more. You don't need this visual interface. And I will still proclaim my thought that the vast amount of sophth that is written in the world will disappear and the vast amount of it will be redistributed, it will be different. And the vast number of systems we know, for example, we know, the system that CRM is, I think in ten years CRM, CRM will not exist, that there will be any other name, some other name, other names or other approaches. It's not gonna be these systems, not these business processes, and not these standard concepts as sales plugs. And that the manager is counting the number of calls, the call there or his just proceeds. Thank you very much, everybody. I'll see you in a week.