I agree with, uh, what Carpaty describes, that the level implementation is not more than 10 per cent. GPT's chat room is eight hundred million.
So the number of Fortune 500 companies that say directly, "Aha, here we are, implementing AI, we saw such benefita." They're just a little short.
If there's one thing to be done, it's a reasoning model. On January 20, if I'm not mistaken, they put their R1., it's got them to be a top rolling app for a while at AppStore.
Uh, I think the twenty-sixth year will be the year when the baccalaureate is completely devoid.
There's a strange feeling we can get a very strong shift at all. Hello, everybody! Uh, ToTheMoon Channel.
Technological news, Silicon Valley sites around the world. We have a special issue, uh, pre-emptive, new year's day. Today's graduation is about to show what happened two thousand and twenty-fifth year. You, uh, you know, have an interesting thing. Now, Elnar, you've been doing a lot of ana-- these, uh, uh, those, uh, uh, like a picture? I'm a photographer, yeah. On how to gather what you have the opportunity, there, YouTube-Canal, and she's collecting you a picture of what happened. And in general, if we're gonna analyze, there, our YouTube-canal, considering that we've done more than 50 episodes, uh, news, of course, and once--- well, information, everything has been going on. On the one hand, right? Uh, on the other hand, uh, I agree with, uh, what Carpaty describes that implementation at a level not exceeding 10 per cent is not more than 10 per cent, using a model of not more than 10 per cent. And that, it must be in life, um, and in everyday life of people, and in, uh, various tools we use. I thought yesterday, "Where is my unworking Alexa, where, or where is my normal assistant at Amazon?" Uh, yee, and in and in companies, of course, it's clear that it's been introduced, well, very little, yeah. At least I know a lot of business. People are putting things like that in a very simple way. So we're going to have a lot ahead of us. And here I want to see, "A" and "A" and "A" "the "M"" "the "M"" topics" for two thousand twenty-fifth year.
What do you think is the most important in the twenty-fifth year in terms of AI technology?
Look, if you're talking about AI, uh, first of all, it's hard to separate one thing, but, uh, if you're gonna have one thing, it's a reasoning model. Yes, I mean, how much they've changed, uh, how we use it, how we used to use the different, uh, chat-bots we used to use before, yes. Here, uh, there was a DeepSeek in January when, uh, well, before that, the ChatGPT model was, but, accordingly, DeepSeek on 20 January, if I'm not mistaken, they put their R1., it's got them to be a top rolling app for a while at AppStore. Uh, they're available, there, around the world, practically without restrictions. And, indeed, we've been developing this whole year. Reasoning models all have, they all develop. The time that comes out for reasoning, that is the number of tokens, just grows. And the main thing about this year's superb development is, I think it's just something to do with, uh, reasoning models. Deep Research is the same, it's the same way.
But the reasoning didn't come up with DeepSeek anyway, right? And, uh-- but you mean, that gave some, uh, new approaches?
Look, DeepSeek published an article in May, if I'm not mistaken, last year, i.e., not this year, the twenty-fifth, the twenty-fourth year. I'm writing , DeepSeek Math, and it was the GRPO algorithm that was used quite actively. Mm-hmm, other companies didn't publish anything on this subject. First, they put on access to the O1 is OpenAI, yes. But it's like saying, they've come up with it or they haven't come up with it. DeepSeek used, uh, mathematical reasoning. Here. It's hard for me to tell who specifically invented it.
But there's something near DeepSeek that's definitely hanging around.
Look, Deep Research said. Deep Research came in the twenty-fifth year?
February. Yes, February. I mean, it's not even a year. I mean, it's so used to it that it exists.
They're used to it, it seems he's been, like, a long time ago, and not just in the twenty-fifth year.
Yeah, he only showed up in February, yeah.
Yeah, our audience, write down how often you use Deep Research and, uh, where, in which models you're actually starting a different research. But I think the fact is, uh, that's what turned it over for me, and the new world did. Here's a new world, It's clear that when they really dilated the reasoning model and showed up there, the version and so on, I stopped doing, for example, Deep Research, but I'm just doing it with the usual models, Models, there, thinking. At the same time, of course, Deep Research has given us an unbelievablely strong approach. It's such a big deal. Although I wanted to do this yesterday, you know, so, uh, a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-GPT was giving out what kind of products he was producing and what kind of products was left on the shelf somewhere. Uh, well, Deep Research is a cool product. At the same time, he didn't get, uh, I think it could have been incredibly developed and made out of it , a useful, working, easily accessible tool. Because, to date, I think Deep Research is using a very small audience of people, probably interest from the eight hundred million users that ChatGPT has. I'll assume interest because I'm personally close to myself, and I see Deep Research only uses people who are professional users of ChatGPT. Everyone who's a private person like this, even who's routine, they don't go right beyond launching Deep Research. I have this, uh, thugger, that's my wife, yeah. She's such a super user of ChatGPT, but if she's gonna run Deep Research, I don't know what's gonna happen.
This should be a certain type of request. You know? You're not gonna be on some kind of home thing to start Deep Research. Why? So, in what field you use for professional work, there's Elnar, for example, in some scientific field, yeah, that's understandable, Deep Research.
Well, you should be able to compare it to something.
Look, just try to try. Just try to try. If you want to find a good, good master, good beauty salon, or you want to find some sort of type, I don't know, some jewelry, some jewelry, like, it's, like, saying... it's doing it. Where to buy something from. That's the best thing. It's gonna start Deep Research. I mean, if you're looking for a place to explore, you want to find out how I used to lead a recent example on the skates, find a place to walk a dog without a lead, you'd better start Deep Research than you did. To ask me, in terms of the fundamentality of the big answer. Very simple things. Though Tanya seems to be right. It seems Deep Research is only for some very specialized, complicated things, right? But by the fact, even if I remember that when I started using Deep Research, I made such requests from the series to study to me how to feed a dog. Yeah, I bought a dog in March and February, Elnar, you say you're in. I've done a lot of Deep Research on this subject because, yet Deep Research is analysing a whole different number, uh, coming through, a whole different number of sites and resources, and documents. Another point is, once again today, we can write so cool that the usual thinking of the model will be 15 minutes to think, you know, unusual, but in general, it'll take 15 minutes to think and walk you, all of which will be there. search, collect, and so on.
But Deep Research has a kind of...
The moment that people are really spoiled right now, they're gonna have a little Internet hanging out there. They're already re-uploading the page 30 times. So, Sas, not everyone like you, is ready to wait 15 minutes to answer the domestic question where to go to the shave. They want it all. That's why I think there's a moment here, too. Typology of the person. Everybody wants it fast and easy.
Look, well, it's like people are just like this and they understand these systems. The question is, what is not available in the twenty-fifth year has not learned to make a good interface, some new one that would allow these questions, which, Tanya, you say, to a man to decide that a man would be safe. I mean, there's a part of the questions that-- that's normal. You wait for days, that's okay, right? If you're thinking about it, you don't change the hairdresser every day, and you need to spend some time. Yeah, not just a little bit of a nut job, yeah, not just a little bit of a bait. And people are really in a hurry. I noticed an interesting thing. I'm in a regular class, and I saw that in Zoom people go to class at zero minutes or plus one or two. People stopped coming in in advance. I often open Zoom in advance, in ten minutes, there's no one. It's really weird. I mean, this is straight, if you think about what's going on. People don't just come in there at fifty-nine minutes, they come in, like, zero one or zero two or zero three. Don't you understand why? Because...
So that the welcome speech is over and over.
Or because people think they're waiting anyway. Yeah, but everyone's waiting five minutes anyway.
Everyone's got time.
Yeah. And you came in like this in ten minutes, and there's only one person there, right? Ten, huh? Or from--
Yeah, tens, yes, or hundreds that should be. It's a very strange story. That's probably the case you're bringing, and, uh, it's gonna be... what's that? I don't know if they're gonna have this job, uh, but it's gonna be cool if they can. So people, uh, interface, yeah, to help the interface, and, uh, we're, like, helping, uh, calmly, like, that some kind of queries can be made for, like, a few days, yeah. Here. But they, I don't think they think much about people, yeah. They think about some kind of space--
We've already said in the past graduation that we're clones.
Spaceships.
No, obviously, yes. This, this is, like, this is like Deep Research. Look, it's a cool product, and it's still sharp, it's just as good as it was. That's where this is the gift, and it was so cool to finish. And once again, these products were very much, uh, and in terms of updating.
Here. Sas, what do you say in terms of, uh, the twenty-fifth, uh, the twenty-fifth year? We understand that now part of the band--- well, let's go through, get it. You think the coolest thing was the twenty-fifth year?
I think it was cool that in the twenty-fifth year, the winners of the twenty-fifth year were the concimers. AI quality improved, aaa, the re-deep Research system became more rapid, especially if Gemini was there to see. Thato, that's what we saw, that's DeepSeek, making everyone get stressed. Google's got everyone to get tense now. Anthropic is still very competitive. And I think that the twenty-fifth year was a year, uh, it was a year of AI consumers.
You mean, uh, uh, well, you mean that a lot of pepgreides have been good for people in terms of the usual system? What's the, uh-- what's the increase, and when did the systems grow, uh, growth, if you watch? Now, we're calling it a number even in GPT chat rooms of eight hundred million, because they, uh, Google, were there, and they're called four hundred. I'm guessing Google might be able to reach 800 million, too. Uh, there's a feeling they had, like, a few hundred or maybe even less than 100 million, aa, users. So the pgreade went through a huge one. I'm very much a huge pgreid, uh, modelled, that Meta disappeared from the radar in the twenty-fifth year. But what's very much...
It seems, too, Sasha. Llama 4 was out in the summer. She's been driving away since, but, like, taking the second half of the year. And the first half of the year, Llama 4 they released.
First, yes.
I thought...
Look...
...mixture of experts, they-- they've made big models.
Yeah. Yeah, yeah. I'm talking to December now, of course, you're absolutely right. I mean, she's out of here now. I mean, if you can see what's going on now, uh, Claude, Gemini, there, ChatGPT. I see there are Chinese models, there are Chinese models. DeepSeek has done a bunch of apegreeds now, yes, just a few weeks ago. Uh, but still Meta, as we discussed it in the first quarter as one of the biggest competitors, then the Grok came up and started talking. And of course, what the HAI and Ilon Mac did again in a year is unthinkable. Here. Bring the company to two hundred and thirty, there's billion dollars, uh, at least potential estimates, I don't know, maybe the deal's already over or over these weeks at the time we're filming, p-p-- We're taking off, uh, our graduation. That's a very serious achievement. I mean, of course, the models were unbelievably stunned by quality. But I have a feeling in person that people have not been used. People have started using models now, as we called for a year ago. I'm talking about the main mass of people, well, that's the big, big, big block. And I have at least among, um, friends and acquaintances-- those people who were supposed to be used a year ago, they kept using. And those people who didn't use it, they didn't start using it. Well, or they use it so indirectly. So they open up a few queries and some microves ask, they don't know what's going on in these systems. Here.
Although, of course, er, revolution, uh, revolution is super. Now, about Carpathova, these 10 percent of the introductions are just about to have a moment like this, um, talk or put. Uh, his comment on what's been introduced by 10 per cent is using models. Me, I-- how do you even think this is a topic? Because I have a real feeling that maybe even less than 10 per cent of the opportunities are even being used, and they're still, um, little. I mean, most things--
Well, Carpathians, Sasha, yes, I think he's got a little bit of it. I mean, if we look at this, yes, too, well, whatever the boxes, that is, the number of Fortune 500 companies that say directly, "Yeah, so we've introduced AI, we've seen these benefits," it's just a little, they're not. It's just a little, so, uh, it's not gonna make it up to ten percent. That's why I agree with you. It's, uh, it's, uh, it's, uh, it's, uh, it's, uh, it's probably a ten percent overall introduction to the Conscimers, but it's a little less in the interpranise.
What do you think, uh, Elnar, Elnar, Tanya, about this?
First of all, it's hard to take. Uh, about ten percent of me-- there's not much to say right now. Second, any technology needs time. Uh, we've got literally three years of GPT and a reasoning came up a year ago. Models grow, quality grows. So, well, any adoption takes time. So, so the Enterprise is going to be well-approved, well, another year, I think it's time to wait, and plus or minus it's all going.
And there's new systems and new models, and they won't be able to implement them again.
Yeah, that's what it is.
I think you know, there's always a new one among the ordinary people. There's a stage of denial of such mass hysteria and fear that maybe I'd rather not inject, there, I don't know, electricity , and then it'll take my house. That's what AI is, I think there is. Early followers, well, like you, for example. And there are ordinary people, and they're used to watching with caution from the side that this is going to get out of it and what kind of benthic is, as Sasha says, you can get. Well, returning to the first question you asked, which is how personal these instruments have already been introduced and become part of my life, which I think is, what was the year they came out, the twenty-fifth, Twenty-fourth, when did they even come?
Because it's a part of reality. But if you're gonna get a little bit of a memory, yeah, it's like all these pictures that I use in a professional life, they all showed up in the twenty-fifth. First, ChatGPT, there's a curve, a cosot, but it was a really breakthrough for me, that I might not go to the Photoshop, but that's how he's gonna get some screaming, throw it away, and he's already doing something there. Then Nana Banana was a breakthrough. Now, Google's doing it, too, Gemini. So I think that-- that feels like it's been years. I don't know, I feel it.
Well, Midjourney has been a long time ago, but it's from the point of view of using it to make it work, yes, it's the change of this year, especially Nana Banana, of course.
I've always been a little confused. I suspect there's a way to work with him. I've never given it enough time because it's still more conceptual, and I need precise things, precise dimensions and everything. So, yes, Midjourney is a little different story, but it was also visual and revolutionary in time. So if you ask me now when he showed up, I think he's been for ten years, and he's probably how old? One year, two .
Twenty-third year, probably somewhere.
Oh, yeah. I mean, I think the success of these products for consumers is not a business, how organic they have been implemented and seem to be part of the daily process. You can judge them like that.
And now I'm in the ChatGPT, even in the ChatGPT, I'm in the video test, and I'm in the video. So he's been doing a quiet video now in the normal line, right? And he's got a problem with audio analysis, but at least he's asking-- he's gonna check out the videos where he's located, he's gonna make all the pictures. So, progress in processing the files of the various documents and in generating the various documents, well, enormous.
I am not saying the progress that has been made in programming, but we are not discussing this issue, yes, but the progress that has been made in the development of the application code is simply colosal. I mean, it's a casle-- just a crazy move. What's that gonna get? The reverse side, which will result from the end result, when there are more people who can easily create something. I mean, on the one hand, it's like a lot of people are easy to create, at the same time, well, not that we see it--
Uh, we see more than 10 per cent, there, or even 5 per cent of the enterprise implementation.
That's really-- that's amazing, by the way. Uh, but I remembered, by the way, while we were talking about-- uh, it's just that progress is very fast. And Ilnar personally said it was normal, that it was a cho, uh, that was just a little late. At the same time, uh, the emergence of the model, they have significantly accelerated everything, including even the companies themselves, there are other money, and they have other quality products. But I'm thinking about the market, uh, company automation, that's ERP, XRM, CRM system, and how much, if we even now take two thousand-- leaving 2025, how much of the people in the world are. This is not understood by the end and how we are in the ancient century, for example, in the CRM system. So if you see the CRM system, there's a salesforce leader, and see what they've got to do with their client, that process that came up with, 30 years ago, Excel. Yia, yes, there are tasks, yes, there's some kind of automation. So, uh, take a look at HubSpot. So it's all very simple until now, simple systems. There's no way that a shift-- something revolution-- that's happening, and people are trying to automate. Uh, a lot of people are trying to automate, uh, routine their current processes. But, of course, um, I don't-- not, for example, in my opinion, I don't know if a more significant introduction to enterprise will take place in the 26th year. I think that, uh, people's daily lives, that's exactly the system-- the number of people will grow up using systems. I might even say, maybe, 2 billion people at ChatGPT next year, not in 2030 or where they put their own plan. Uh, I wouldn't be surprised. I mean, I wouldn't be surprised if I was told. But the fact that companies in the world, uh, are gonna start using, like, artificial intelligence models, uh, I'm gonna be doing this for the rest of the time, um, next year, that's the con-- that's very relatively true. At least, given that the economy is young and the financial model for the implementation of such instruments is not clear how to work. Here. Mm-hmm, let's see what happens.
There's one big block that, uh, we've started since the beginning of the year. The Trump came to power and there were big statements that there would be a lot of money to throw into AI. And remember those numbers when you started-- there were hundreds of billions of dollars. That is the difference between the business awareness and the econo-consciousness of the economy of these things. It's, of course, that this year's hundreds of billions of numbers have become very calm for the rule, for the sake of consciousness, right? So we see the Anthropic, uh, estimate of more than a few hundred billion dollars. We see the OpenAI evaluation. There's someone saying billion dollars, $750 billion. Not a-- obviously it's not a billion, trillion dollars, trillion, $1,000, right? Or $750 billion. We see, uh, hei's crazy grades. Well, we understand that Google has been able to capitalize $4 trillion. And, uh, these projects that are under discussion, there's tens of billions of dollars, that's kind of, uh, someone's doing something there. I'm on the news, if I can see that someone gave someone $5 billion or $10 billion, you're not gonna stop being so sensitive. Yeah, there's a problem that says that often these moneys are kind of short, there, Oracle's missing, something else that there's no way ChatGPT needs more than $3illion to move on. Next year, I don't know, they're gonna need, um, $10 trillion to develop. But the point is, the numbers are different and the infusions are different. And for me, my point of view was confirmed, which I had originally been, three years ago, that there is virtually no one in this race... well, the countries that won. I mean, in the way that there is a US, there's China.
We see that China has made tremendous progress in the 25th year, and we have even discussed it since the beginning of the year, told us how much it has moved. And he has progress, of course, huge, and we don't even feel it. If you look at the introduction, they're not in the process of some kind of Haip, not at the level of the baccalaureate, but at the level of, for example, how many tractors in the country without the driver. Or how many of them-- how much of their logistics is highly automated in the country, or how many AI services have been introduced into the public administration system, and how they are implemented and all scan and analyse. Here, uh, again, I think America brags about benchem marks and big money, and China, uh, uh, um, here it comes, by the number of engineers they have, by the number of patents they have, by the number of patents they have, The number of technology companies that work there, in integration, ine-e, in life, people's lives, from government to people's lives and from integrating technology into different sectors of the economy, I think they're, We're out. And, uh, there's a , there's a strange feeling that we can actually get a very strong shift in terms of, uh, well, that--that maybe China's gonna start doing something technological, super-high. Well, that's when we say, I don't know, AGI or ASI on our side, and it's not gonna be done in America.
Here. Given their speed, and other countries, I just want to add, we even had a special probability, right? About countries, different countries, I think. We've been filming, Lenar, about education this year, right? The releases were made in a separate manner, as in different countries, and, uh, the introduction of artificial intelligence within education. I'm saying, again, the largest number of things is, in-- again, in some sense, a visual, because too much is done in China. And, for example, there were many statements in the States of some incredible number of projects and everything. And by the fact, uh, such serious direct implementations and serious decisions, we don't see. But we only see that coursera decided to buy Udemy, so they decided to merge them. Here. Here you are, here you are, the technology of America, right? Here. And I don't see any changes like school. I mean, if, because I still have my kids covering, uh, elementary and middle, and high school in the U.S., I don't see a change in my face that would affect, for example, the educational system in the U.S. schools. In particular, we read about it in China. Again, yeah, I don't have kids in China. We can only judge on the news. And Chinese news is not the end, m, the truth. But other countries, it feels like they're completely back.
Other technologies like Deep Research started and similar discussions. On 23 January, OpenAI presented operators, yes, that is, a version in which, uh, the system controls, there, your mouse, clicks like this. In fact, before that, if I remember correctly, Anthropics started doing it last year. But on 23 January, at the beginning of the year, it appears that OpenAI presented it, and gradually it has resulted in a large number of browsers. Well, in fact, everyone has those, which, uh, have inside themselves an intrudered AI, an agent built by some kind of, uh, some kind of department can do for you. It's the same thing Perplexity, and it's the same thing that OpenAI has and the rest of the companies have. And this branch, that you said that there were some technologies that didn't get that spread, right? What Deep Research, I think is a proliferation, but...
Well, there's no browser!
Perplexity, yes.
Yeah.
So atlas is so stupid that you, like Ctrl C, Ctrl V, work on your English dressing, you don't work on the Russian dress. And a bunch of these are bagos, underwork, and I don't use any other stuff. I ended up closing him up and returning to my standard browsers. Do you use this kind of thing and do we use our viewers? Interesting. I think the technology is still not ready.
Look, from the point of view of the browsers, it's interesting, of course. Because these huge statements and some discussion are endless. You're sitting, you don't understand, and why do you need a side-broiser, some sort of auxiliary stuff like a series of translators or summit, some sort of article or something. It's kind of useful.
At the same time, you're doing an extra analysis with all these materials, not just the summit reading somewhere. For example, in Apple Intelligence, I think it's a year's failure, yes, the twenty-fifth year. I'll remind you, by the way, that this is Apple's failure. Because Apple promised to clean up everything in March. That, by the way, has nothing to do with their continued capitalization and adding hundreds of billions of dollars to them, so, yeah? And not even a few more.
And they didn't even add a few, I think, interest on the twenty, uh, they've been jacked for a year. Uh, and, uh, there, like, when Apple Intelligence is using, you're integrated with GPT chat, and you're in GPT chat, uh, you're doing some sort of question through Apple Intelligence, you don't have GPT chat history. You're coming, and you're on it, you're not gonna work. You don't have infrastructure. Or, like Elnar says, there's no simple things in Atlas you're used to. And I-- that's a little example, right? There's still a lot of stuff there, which is not there, yeah, you need to be standard for work at the browser, not just the summit, right?
And as I'm not, it's a problem when I have to analyze some article, just sharing, yeah, pushing it together, and everything. And here's the question of agents. Look, a huge number of companies this year have started agents, automated agents or agents that can microcod it, well, a little. And they're supposed to be doing a lot of things. It feels like it's not an alpha or beta in the twenty-fifth year, it's research, right? Well, we're in the research zone of this whole thing. And ChatGPT, by the way, released his agents, too, and they, uh, I remember, a little bit of time I tried to do a lot of things to them. I--we've even brought a lot of cabs about it to The Moon, and there were working certain cases, but I finally gave it away. I mean, I'm not using an agent at the moment, and I'm not using ChatGPT different reminders because they've been glucchising, doing something wrong, and the products haven't developed. ChatGPT is a fucking product that doesn't develop anything but moving these models between itself, right? Well, the model is developing itself, okay. But they're also going to end up there all the time. I mean, you might say they're kind of developing it. That's what's weird.
And by the agents, remember the first time they came as bright as it was. It was March of this year that Manus showed up, and we even discussed that this is a deep sick moment, that business analysts can no longer learn that this will be done. And it looked really cool. Noo's again, the presentations were a little better than
reality.
Yeah, it's an interesting thing that Chrome has seventy-one percent of the market, uh, browsers, and you know it's Google, and Google has Gemini 3.0 right now. And in fact, Nana Banana and Veo, and a lot of things. And, obviously, in them, their browser will be built--- first, they're already there. Let's do this. There's already a huge amount of AI chips in Chrome, and they'll still be unbelievably big. And they also have mail, calendars, uh, and all together are integrated, all with this browser is. But can this browser somehow lose popularity? He might lose the popularity that I'm currently on GPT chat room about the browser spending ninety-five percent of the time and five percent in the browser. What if ChatGPT creates something not a browser, it'll increase, then I'll stop being necessary. That's how you and I, remember, a few of the episodes back led by an example, Elnar, you gave an example that when the car cheeks were changing, you'd opened the video. So, the idea is, ChatGPT should think about that case, right? Like the boxes where people still need a browser or still need YouTube, they'll be in the usual GPT chat, and it'll look cool. They won't have to open these systems for any reason, I think, not browsers. I didn't believe or believe in the browser personally. Here.
In keeping with your thought, this research Elnar dropped suggests that people are very bad at all, especially these super-long texts. And for people, it's much more understandable, uh, pictures, infographics, video is perfect. I mean, if we move in this direction, they'll be more users if we, like we did this time, we've been discussing, in response to a request for a way to do something, to generate video instruction, Well, or at least a picture of it, like a 20-page text. Nobody reads instructions. It's like... knowing you're not really smart. If you've asked for instructions to collect the IKEA closet or something.
Yeah, the question is, if they can do it now. Because they can see that they're not going up there until they're in this direction. Even Deep Research is being produced in the same verb as it was.
I mean, she's not really changing. You even ask Deep Research to add another vervain, it's like he's not listening to you, right? And it was, remember, there was a canvas, yes, in a GPT chat where you have... I used a moment when you had a document. It was really fun. And then you work with that document inside and you're with him. The idea itself is very good, as usual in the skies, even there are things I'm always discussing. Like, in one chat, I have, in there, I don't know, stock chat or some kind of house chat, I don't know, there, uh, a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a- It would be cool if they saw you in the same chat, that whole information was collected in some way, maybe some graphic or table or some kind of infographic, right?
Gathering these data together. But what do we see? We see that there are five versions of the story now, and five versions of the same chat can give you the fourth message and write to you, uh, that she doesn't know where you live or what you're asking, even though you're in the first one. She was totally in the line of her message, and she reads memory, and then in the fourth message in the same chat, yeah, the same chat, it includes an absolute stupid chat. I mean, this migration, model migration. Here's the question: uh, they still have a strong model improvement or as long as it's just a drink, and, uh, a resoning, and then, uh, these cars, some kind of auto-coupling tracks. The system ' s auto-request, not the real development, uh, real development, some kind of cool development of systems that could lead to what they're like. There's KGI and Psyche, the rest of it.
I was just, uh, asking, there's people's opinion about Carpathova's post, yeah, where's it going? Because the current reasoning, he... It shows that it is true that inference can continue to be scaled up by computer. We saw there were big breaks in the training too. It's all very cool, but it's very interesting if new concepts are gonna be. Carpathian wrote it pretty well. And when I look at what Carpathian wrote, yes, I mean, he gave it to me six, yeah, that's the six biggest ones that need to be expected.
I think one of the most, uh, one of the most interesting topics, yes, is that, and I think, the twenty-sixth year will be the year when the baccalaureates are completely dead, right? We still have every model release, it is accompanied by baccalaureate. And I think the main thing that's gonna happen is that in the twenty-sixth year, the baccalaureate will be killed.
But when they're dying, Sasha, it's that these companies will start to focus more on real skies and tell people about the real serious stuff, not how to order sushi. So they put the baccalaureate on the one hand that they made space--- a cosmise-- a cosmonaut with eight hundred brains, and then they tell you about the case, how they help the system to help the sushi in Instacarta. I'll order it. So when they're gonna kill the baccalaureate, what's going to happen?
I don't know. It's one thing I think, I think it's a super big question I'm reading about right now, there's comments, and I-- I think he's not answered yet. Oh, I don't think anyone has a response to it. What do you mean? So what's the paradigm? I mean, when we see a couple-- a change of paradigm, until I think I don't know.
Look, come on, let's fake it. Imagine, so OpenAI comes out in January, sets out a new model and doesn't show any benchmarks. Well, just puts a new model on the website and writes a new model, and writes a new number. The usual, ordinary users, ordinary consumers, the usual users, this billion people, do they have a difference in the baccalaureate? No, they didn't know what the baccalaureate was. They just had a new number. They don't even know what it is. They got a car or five. They don't understand it. They don't even know how this switch is pressed, so it's inside, right? Here. So they don't have any difference at all, they'd like-- the usual user would be fun if they'd at least report it, what's inside. At least, yeah? What happens differently? Now they've released a model, well, people are gonna start testing these models, something about them, something to tell them. So they're gonna start hiring themselves, like, the engineers? Or if this year, by the way, this year, I think this is a huge market problem, by the way, the crazy market problem is that at one point, someone started to pay too much money to engineers. I think they're starting to play this game very much, Meta. At least this hat caught metu if I'm not wrong, right? When Meta first started talking about... It was spring. First, you-- millions of dollars and tens of millions of dollars-- millions of dollars that pay people.
But then Zuckerberg
But then Zuckerberg beat the nail to the end when one man engineer promised five years more than a billion dollars, didn't he? He's still out of his mind. Here. One and two or how many billions of dollars were there. And what does it have to do with it, so these baccalaureates are now? The way people migrate between themselves means people between these companies, someone's firing someone. We're just hearing, so we got fired somewhere, hired. If we open up most of the company's employees, like OpenAI, their profiles, we'll see their previous work in their profiles. They're in circles, so they're walking. And we'll see a lot of new staff, like from Meta, and in Meta we'll see a lot of staff, out of Google. Here. XAI or X- or , Ilona Mask. Just what, I wonder, is it gonna happen when-- what's the problem with getting the baccalaureate? I don't think there's any problem at all. Actually. I'd have been there for Ilona Mask, I'd have taken it, just clean the baccalaureate. But they're still small. The game is very small to do it, yes.
Look, while the baccalaureates give customers, investors or someone else's sense that it's upwards, they're being used. It's like, you know, a long time ago, and still, I think it exists, but nobody's looking at it. Or I don't look at it. When the mobile phone comes out and he's getting a new processor, those who use the Android, everyone looked at the AnTuTu bench. Like how many parrots you're getting, there's a processor on the new phone, right? And you're watching, you think, "Oh, I have the coolest phone. Or I'll buy that phone because he's got very high numbers in there." I think everyone stopped watching this for a while, because the plus-minus is the same number on the processor. You never start, there, I don't know, LLM modeling on the phone, and for domestic use, as long as he's on AnTu, nobody's interested. The same story will gradually happen to the baccalaureate because the plus-minus they're recruiting for Humanity Last Exam, who has one percent more, who has one percent less, and, in fact, nobody
cares.
The problem is, ninety-two percent is almost 100, because the baccalaureate is not clean. There's always a printout, there's always something there. And the model can give the right answer, but the test, there, is a print--
The test itself, yes.
And she's making a mistake.
Yeah.
Yeah. Well, as much as you can, there are hundreds of thousands of questions to do exactly exact, checked, somewhere there's gonna be a human factor, a fingerprint, something like that. So, ninety plus is, you know, a brand-new baccalaureate, and you can't look at it anymore.
And the question is, what's the right or wrong baccalaureate, right, when are we talking about something? I mean, compared to the simple way that a person responds to it, people are also responding differently. We see models give a very different information on the same answer. And you don't understand, what's the right answer? That's the right answer, short, long, superficial, uncompleted, with a pill or not, with smart words, with simple words. You don't know what that is anymore. Here. I wonder if you'd write as an audience ours, as you're aware, and how much trouble you have to make some small hallucinations or mistakes. I'm just hallucinating, there's no super-signation errors right now. I mean, I'm just calm about what might be a mistake, right? And that was a good talk, by the way. Sam Althman said that, and other people in the market, that you don't take the models as if the device you're guaranteed by 100 percent would give you a straight superficial answer.
About-- the subject that I want to raise. A very important subject for viewers is a profession. Yeah, we talked about a lot this year, and there was a discussion on the market that would take a huge number of jobs. Well, so far, these discussions are still going, so people don't have to work, and they're just gonna make money. Or that people will be fired and people will become poor, that more rich people will be born through the use of artificial intelligence, and so on. Sign who's got a lot of skill this year, who's got artificial intelligence, too. What professions have you been able to earn more money on this? Who lost his job because of the intellect of artificial intelligence? Here. How much is that going on? I think people lost their jobs during the cowid. More changes have occurred than artificial intelligence is now. And that's a very interesting aspect, by the way, because the covid came up unprognosed and I think it brought a lot more chaos and much more change. I understand chaos not necessarily something bad, but the fact of uncertainty that has brought artificial intelligence models. Well, we've been watching for two years. Although, of course, again, the trend is incredibly cool, and we'll continue to film this subject for the twenty-sixth year so that you can increasingly understand how to use and use in your life. artificial intelligence.
What else do you see in the year that passed, what is the point of drawing attention to-- even to forming this for the twenty-sixth year?
Look, I'd like to point out the following in general. People have this, uh, cognitive distortion, yes, when something's done, it seems that it's very easy. Well, in a rude way, remember, they used to say that, uh, "I know computers are just computers, they don't know anything. Here's the chess, this is the direction where you need to think. That's when the computer starts playing chess, that's when the computer's smart." Well, as long as people can't do anything against the computer. Then they started saying, "The chess isn't that complicated. Oh, here we go! There's a lot of options there. That's not real. People will never get down there." Well, actually, we have a famous story with AlphaGo, right? And the go was done, too. And since then, it's like, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, since then, it's been like, like, like, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, since then, well, well, well, well, well, well, well, well, well, well, well, since then, well, well, well, well, Well, that's what's the problem? And the same thing happens to models. Now, remember the very first, the first ones that were, the first words that were, the ends were agreed, you think, "This is my respect!" And now there's one little fingerprint or something, and You're like, "What kind of nonsense did you get out of? Is that what this is all about?"
And here, I just want to say that there's a lot of things that happened in the last year.
Mm-hmm.
Wait, but the one's up, the one's up this year.
No, the one came out last year.
Oh, oh, past one.
Yeah. And the O3 came out this year in spring. Then they disappeared, respectively. It's there, GPT 5, then 5.1, then 5.2, right? All this, there with Pro, with Extended Thinking, that's all. Gemini has the same story. 2.5. It was spring this year, then there's 2.5 Pro Flash, then 3 Pro, which has all been blown up. Grok, if I remember correctly, you're Grok 3-Heavy's still buying this year.
That's it. That's 100%.
Then Grok 4. Yeah, then Grok 4 came out with all those--
Grok is actually the story of this year, yeah.
Yeah. Anthropic, uh, Claude 3.7 is also this year. We had four very good models now, then 4.5 showed up. Here. Same with coding models. The anthropics released the Claude Code. There was a Codex later. In parallel with this, Cursor, Windsurf and others are also taking the market. And it ended up still with the Antigravity that we, uh, had Google. So, look, uh, at the moment, it seems like the models aren't that cool and so on. And if you look at it in the dynamic, there's a lot going on. The models were getting really cooler than they were. Yeah, in some ways, it seems like there's something missing in the moment, right? But if you look at the dynamics, the changes are very significant.
It's a fact. It's a fact. And here, just in parallel, and you wonder how much people don't feel or see it. And many people even belong to-- and many, they never get used to many, not even a lot of them.
Sas, I was just saying that I think I was just thinking about this and I came to the conclusion that again about Meta, what you said, I got knocked out. But they actually looked like a player like that in June. Then there was-- then there was this negative pyara spiral that took them down there.
But-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and-and- actually, yeah. I mean, AI companies probably have to show some sort of breakthroughs in every quarter in the year 20, not in the baccalaureate, but in some real skies. I mean, like, a case-stadi, you know, a whole-of-the-art ad about some kind of thing, you know, like, like, a little bit more, like, a-a-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b-b New York, in L.A. They're gonna have to make commercial commercial advertising to show how their model solves the skies. Because, well, you're just gonna get out in six months. I mean, six months...
- You do. I mean, six months, you, uh, like Meta's already in the ra-- the market is just gone. I was just a little bit late last night, having a chat with my friends, yeah, everybody's got a negative yelling about Metu, just like I did. And that's why I think that next year I'm gonna be a little bit of a year, a year, a year of skies, and in the vocabulary, and all of these systems will be really going to be in this broad skill. And, uh, OpenAI is definitely gonna double.
I looked at the traffic. They seem to have a little traffic from South-East Asia. There's a full-blown growth potential.
Bilbordas are unexpected.
Well, no, wait--
I thought they were already dead.
You're still coming. And as I hear, uh, your husband, we were going to New Mexico with him now, and, uh, and I'm gonna tell you. We were driving to New Mexico, and he asked me, "A billboards working?" There were billboards hanging. And I say, "Well, the billboards are clearly working on a mass market, like Apple, but they don't work in San Francisco for AI products that hang. There's a Nabius Billboard hanging, something else.
Yeah, because it's all about AI.
And he says, "Oh, my wife bought my phone from the billboard the new Apple." So, that's why... About you, Tanya.
I was the one who manipulated. Ehal's next to-
That's why--
And I think how do I tell him about the phone?
Oh, one, and the billboard.
It's a billboard. I said, "Ilnar told me on the sub-cate that he was out."
Here. So...
Orange. So it's a very cleverly built communication. But actually, a funny story, because, uh, uh, you know, now, I think it's gonna be a little more valuable than analogism. I mean, it was a rabid growth, it's all online, and now that his vast amount of billboards are probably gonna come back in life.
I'm just wondering if the bubbards in San Francisco are hanging, uh, often different starttaps. There are serious, really, companies hanging, yeah, I don't know, like Figma hangs, you know, more serious in terms of audience. But a lot, uh, little shit hanging. They're hanging...
Yeah, and I don't know what they meant.
They hang big players, yeah, when the Apple hangs the billboard and says, "This picture is made with the iPhone." Nice commercial, good commercial. When you see it forever, see, see, you see, you see, you clearly show that, uh, quality, quality of camera, right? Why doesn't there be a publicity in the OpenAI? Why don't you have a Gemini commercial like that? Why doesn't Claude have that commercial, huh? Why is there no such ad at X? That's a big question. That's a big question. And that's the entrance to the real-- because, as I said, uh, my friend a few weeks ago, because they're research, you know, in OpenAI, because she's a research organization, and research.
business is not needed, and they don't need billboards.
You got an elf from Santa Claus.
Yeah.
But Meta, by the way, Sasha said the interesting theme about Meta. Meta wasn't just, Sasha, summer, and they were there at the end of spring, they were super cool at the beginning of summer. And when they've all hired these guys, yeah, and they've started to get people out of their way, and... They did the trend, they made the trend of expensive staff. That's Zuckerberg talking in the summer and said I'd be okay with spending $250 billion, uh, not even finding a solution, right? Just spend and lose. I mean, well, that's his subject. And you, Sasha, you think they're gonna get out or out? I asked you that question. I'll ask you again.
I mean, of course, they'll let her out sometime, uh, but they'll be much easier. And you know, they'll be a lot easier for them to release a closed source mo-- that's, uh, they're obviously in big trouble, they're losing a lot of bad shit. Patience ends, too. To release a top-down open source model, you know, time-miming is very necessary. They're not in the market anymore, they're not in the market anymore, are they? I mean, by the way, those companies that have been in the last six months... So what's my point? That companies that haven't been in the market for the last six months will be very hard for them to return to the market where--
Mm-hmm.
You need to release some other big-time campaign every quarter, besides the baccalaureate, yeah, big promotion campaigns, besides the baccalaureate. And Meta will be much easier to build on what they do to consumer AI, that it's a closed model that delivers some kind of safe experience there, interactive experience, playful, appropriate. Metadata, et cetera. I mean, I think there's no way they're gonna get back in the open source. That's it, the train's gone.
Well, not open source, they can make a cool model? We saw examples when someone comes out and says, "I have a cool model." Well, uh, as far as, well, everyone's gonna smile now and say Apple won't, but in the... in theory, they could do it, come out in the 26th year, like, "Ha-ha! We're having a model, so it turns out to be born, and a model to show. Because they lost. Apple didn't just lose the model, he lost his integration, uh, with the integration of different models into his own, his own phone. They couldn't, I don't know, I think it's the first time they made a button that... Well, here's all Apple people, Apple, Apple sitting. You use Apple Intelligence button?
No, it's awful, it's a disgusting decision.
No, it's not serious, it's very bad they did that, because really need it, real time like the assistant who's been integrated there. Trillion three rubles of Apple could do it, but like zero terations, well, literally zero irrations a year after the button, it's disgusting.
mail, documents, photos, all infrastructure. Well, that's how you say the assistant could have done it, right? They didn't make A OpenAI. Here, uh, last minutes, so, the one that's leaving our New Year. OpenAI, so he started to develop the deposits there, so it's endless. There are Pulse, Images, Apps, there, Codex, Atlas, GPTs. He-- was that store-- was that a dead cemetery or something? He's in there, in the warehouse-- in the inside. No, there are serious things there, there's Codex, so I have nothing against it. Here. But still, the question. I still have a question, by the way, why not Sora? So Images are there, and Sora's gone. Oh, and I think she was. I think she was at some point in time and she was taken away.
Yeah, she was there.
Yeah. Or I have this in my interface, which means there's some kind of a missing person in the interface, and someone in the interface has Sora. So. And with the only ChatGPT and GPTs in the dextop app, see the dextop app. There's no other deposits. Well, OpenAI is a straight giant UX/UI. But, uh, I think if they don't start running this design, well, they don't, uh, start making products for users, they'll have problems. Or they'll have to create, uh... Well, unless they create a super-smoky model there, huh? Unbelievable new model.
Otherwise, people in Google, given their infrastructure, will migrate quietly. I wonder if there's any loss now, if the loss in the last quarter of the OpenAI auditor's auditor's migration to Gemini. Because if it happened, these people migrate without, well, irresponsible. Elnar, you've been migrating a lot on Gemini, by the way. You migrated to him after all? Or is ChatGPT still left?
Gemini, I like it more, but I haven't let OpenAI go yet. I like both, and I like it, but Gemini is very nice to me. Especially with regard to the infographics. I don't know, there's no way we're gonna use it, but how he does it, I really like it.
Sasha, what did you mean?
About Gemini, well, that's actually it. Actually, about Gemini, I'm all right, so you know what's interesting? I've had a very long dialogue with him recently, very good. And, and it's 5.2 Pro. I'm actually feeling that I'm about to start using a cadet, and some orchestra, even for these routine tasks, business like this, is quite simple, yes, let's just say. Ah, because models, because, well, these are 5.2. 5.2 Pro is more difficult to use, she's so unpleasant, I'd say. If you can say that, it's unpleasant, it's unpleasant. Oh, you know, that's how a staff member who's predictable, you know what he did. This is Gemini 3.0 Pro. That's good. You're really gonna want to share any problem with him. 5.2 - He's such a more problematic Pro, but the value is obviously very important. Especially if he's thinking long. And I think you know there's a-- there's a s-- that, you know, the possibility that the same courseor would go to many other verticals, as well.
I think I'm gonna change-- I think I'm gonna be, uh, I think I'm gonna be on, on the courser, move more, yeah. I mean, I think we should be going to have to go too...
Yeah, but it's professional, Sasha. Sasha, it's professional.
Sasha, Sasha, it's professional. And that's all about business requests, personal inquiries, the general use of the system--
Grok's gone. Well, just to get some Twitter to analyze. But this is one, twice a month.
Yeah, but Twitter's gonna
analyze it as simple as a plagin, you can say, to the X, right? Unwanted to--
Yes
...un-- for you, yes, for you, what you're saying now. I mean, it's not a separate use. What do you think Grok will still be able to build an audience outside this X-ray audience, or do they have no chance, uh, except X, uh, to get out? And they'll use Grok in their infrastructure, all right, there in Tesla, Starlink, and so, yeah.
That's a very interesting question, isn't it? I mean, that's a very interesting question. I mean, of course, if Elon learned to sell cars, then the assistant would learn to sell, yeah, people. So I can't bet against Ilon, but there's no way that X AI would do that, right? But Ilon can do it. All he did was launch a mass product, yes, Starlink became a mass, and so on.
Well, as we know, he's not just putting his bet on this system, he's putting the whole infrastructure on a big one, yeah, Tesla's link, Starlink, robots, and everything. The use of these systems is important. In the technical panel, Illnar called the Moon Channel to be the girls and said that the girls should end and we should go to the Ironia of Destiny. You didn't see it because it was a technical pause, but so good look at the irony of fate. With the coming New Year or New Year, because someone's already looking into our graduation. Watch us every Sunday, special episodes of the dreams, Wednesdays, special subjects on Wednesdays. I, I-- I have one person out there who's always been killed by these media. Write it, you'll have to write it. This is where we were about the twenty-fifth year. Your conclusions, they're very interesting. It's not even the twenty-sixth year trend, that's the sense of artificial intelligence in your life in the twenty-fifth year. We'll still be doing some special episodes and, after the twenty-fifth year, something extra, and what happens in the twenty-sixth year. We've had a graduation early in the year, and we've been telling the trend for the twenty-sixth year that we've been putting.
Twenty-fifth.
We'll make a comparison. Twenty-fifth, yes, twenty-fifth. I'm already, I'm in the twenty-sixth. Here. We'll do it. We'll make the twenty-sixth year trends and tell us what happens in the twenty-sixth year. Oh, well, I'll see you in New Year. Happy New Year! Bye, everybody.