Hello, everyone! We're on ToTheMoon. Technological news, Silicon Valley sites around the world. We go out once a week, Sunday, sometimes special Wednesdays. So, uh, next Wednesday, by the way, it's likely that there's a new special edition., write as you have always been special and in advance, straight-line, comment. Let's develop our channel. Oh, what?
No graphics are being beaten at the presentation. Well, that's very strange.
He gave me a huge instruction like that, that I closed down and I think, "Okay, another time." Now, something really will be going to revolution.
This is a bar! This is the bar, Sasha. I'm gonna keep writing on him, saying, "Don't ask me questions." He says, "No shit!"
It's nails, Sasha.
GPT 5 came out yesterday, and we're just shooting it in time. Although I think that in order to make adequate conclusions about the model, the details and everything, we'll discuss it now, much interesting to say. I think it's a renewal, it's a market-oriented one, what it looks like, what it did there. And everything goes to what was supposed to go. That's how the update looks. And now we'll tell you, we'll give you details. Uh, though, I think it's gonna take a little more time to really understand, what they did--
Yeah, absolutely the same.
And protest everything. Tanya, you just said today that the entire Internet is being thrilled by the fact that hallucinations have not diminished. Here. Yeah. Although I have a feeling, at least as a user, that hallucinations and 3, and 3 Pro were very small. Well, compared to what happened there, 4° or even 1, for three months ago or five months ago. Well, maybe I'm wrong. They showed that GPT 5 had reduced the number of hallucinations relative to the model's o3 even, yes, by eighty-four, I think, or how many percent. Very serious. Come on. Basic, substantive update what? What is now, in fact, only GPT. So ChatGPT became called ChatGPT 5. Ah. Maybe even they were important just ChatGPT to leave it. And, um, there's still a user version of GPT Thinking and users have a GPT Pro option. I'm talking about the choice of the model, the choice of the model. They still left some confusion. But most importantly, they've taken out this huge amount of towels that people didn't know about, which is the difference between 3, 1, min, min, 4, et cetera. What are your impressions, views on the way out, the presentation, what's going on, what's the technicians in the valley saying?
Ah.
It's really two, really. Ah. On the one hand, we waited, waited long, wanted to see something like that. And every time you wait, especially when you're not changing a minor version, yeah, there's not four of them, four of them, four, four, five, or something, and when you're really the main one. The number changes, yes, there, the main number version changes. Well, remember, there was a parole, there was Windows 2000, then there, XP, then Vista, then seven, ten. And every time, it's a straight-up new operating system that's exactly different from what it was before. The iPhones once in the morning, they were quite a bit different from one version to another. ChatGPT, when you moved from third to fourth, it was a revolution. There were different models of quality, which they gave up and so on. And you're still waiting here, you think, well, there's a fifth version of it, and it's gonna be another revolution. But the revolution I didn't see... it was. The presentation itself is very, in my view, raw and strange. With all these glutes on the schedule they've submitted. Well, one of the examples I've been chatting. I think a lot of viewers have noticed. We're here on the slide, too, probably. But when you tell me you're doing a system practically PHD level of expert you wear in your pocket, and you, uh, you don't have graphics at the presentation. Well, that's very strange. You either codided it on the vipe and didn't check it yourself, or you, I don't know, there, you got this supermodel, and she can't do the pros. Or you did it on purpose, thinking they're not gonna notice, and you're looking at a huge number of people. Well, the impressions are weird. Then I started looking, and what, well, there's something useful about it, can I find something? And yet I started to think that, uh, if you take the current state of the model, you take the GPT 4, which came out there, two and a half years ago. And from GPT 4 to GPT 5, it's really a very long distance. From model O3 to GPT 5, it's probably very small, in my view. At least I'm just saying that I was in the story and plus the version, I didn't notice much change. But if you take a big scale, you might say that there's GPT 5 and GPT 4, which he used to be, they're very different. About some reaction. Users plus versions, at least those I know who I was talking to or where the emotions were seen are unpleasant. I'm not happy they took the choices of models that they left, there, on par with GPT 5 and GPT Thinking. You choose GPT 5, and you answer, well, it feels like 4o. You don't really control that. I'm gonna have to, uh, redesign the usual usage routes. Like we said, it would be good for this zoo to clean up the models. But it turns out you're getting used to it. And you've been told to do something different now, it's still uncommon. That's a little ambiguity, uh, who's answering you now, and it's causing many people to have negative emotions. But again, we just started. Let's see what happens next.
Shortly, my jealousy must be while it is.
Yeah, I was gonna say that I used it on purpose, tested it. The only thing I've noticed is speed. Well, I mean, if it's been 11 seconds before, I've been answering my request, it's five seconds. To say that it's a principle for me for six seconds, I guess it's not. And then Ilnar and Ilnar are, because Ilnar is more of a technical person, I'm more, you know, a different warehouse. I'm comfortable, for example, that I won't choose this model. And I'm saying I've chosen three before, and then I open it, and it's not three. Every time you remember her, you're gonna put it on me too. In hallucinations, well, as I said, it seems like nothing has changed. I thought it was worse by the pictures, even at some point. Though I may be wrong. I'm very interested in the option of connecting the calendar or, uh, post office, but I don't know, again, like my clients, I always have clients, they're used to some kind of communication style. If they just get a ChatGPT, like, a rescheduled meeting, he'll write: I'm not gonna meet you. What kind of manner would he write that, how would he cancel it? And I kind of tried to connect the calendar last night, he gave me such a huge instruction as to do, which I closed. I think it's okay, some other time. Here, like that, I don't see the simplicity for ordinary users yet. What else is it? Well, that's all. So we waited so long, we even recorded this issue in an unusual time for us. I'm thinking, all right, now something really will overthrow the revolution. But I'm saying while I see the difference in five seconds and everything.
Yeah. Well, I guess I have nothing to say, yes. I mean, I was the man who asked to move so we could all use GPT five, uh, and give the normal giden why we needed it. Why do you need him? I don't understand. Here. And that's the problem. I think there's two points five that went through, right, Gemini? No?
Am I confused?
No, they got Deep Sing out yesterday. Deep Sing left yesterday.
Yeah, yeah. And if I am, I'm sorry I'm not prepared for this very good, but all I see in my tape is that I can't exactly under-supplate now, but here in the baccalaureate, on such a deep thinking, there, Arcagi and There's no first place in the OpenAI. I'm gonna go to the middle of the day. They're already, uh, very close to OpenAI. And it's probably not, uh, reckless, but, uh, it's already. And, well, at the level of the picture models. I mean, what I can talk about, even though the model is good at the baccalaureate, it's not very good for Open. I didn't hear anyone using her professionally. Oh, well, at least in marketing, that's what we do. Their model is not competitive in the video either. I don't know.
Oh, Fff, I didn't hear them move ElevenLabs. I mean, as long as it goes, I don't understand. I mean, I'm still waiting for a realtime assistant who's integrated with all my systems, and that's how I can orchestrate all my work and know what my life is, in my, uh, e-life. But I think they're moving in this direction, but it's not really good. I mean, it seems like a z-- but it seems like a really objective OpenAI doesn't need a very cool model. I think the success of OpenAI now depends precisely on how they can truly be their personal assistant they develop, to make integrated with existing systems, yes, there, with the mail, the calendar, and so on. And how convenient it is to use it. With this big problem, usually with the OpenAI, there's a real problem, just so it's easy to use, so there's no model zoo to keep it the preferred model. I'm using a folder, I'm writing a question in the folder, yeah, some kind of thing. I was doing a folder like, uh, chat rooms, yeah, it's a categorization. He doesn't write on top of what model is used. At least I didn't find it. And that's where all the UX-moments are, you just don't know what model they're using. First, we need to remember these models, and we need to see what's right. And that's even if they didn't have a lot of integration, and now they're gonna be integration, it's gonna be a lot harder.
I think that for the OpenAI now, it's probably that they're technologically advanced, that's a way to admit it's a little off-the-book. But I think it's not that important to the company's success, actually. And they don't care about B2C, business is more than API business, right? They probably have the most important thing to use. Well, of course, the GPT is in design and everything, well, I don't know, from last decade, it's true. Here.
And that's why I think this is the most important area of growth for them.
I also remembered that they had anonized that they could choose, uh, sub-accountity of the cat, as a matter of fact, right? But, Sas, you said they were still, uh, not invented.
No, they're still, they're, uh, they're, uh, they're pretty much there, so they're pretty much on the record, they're gonna be a lot of functions, yeah. They'll be on, uh, as I understand, a lot of next week. And that's why we can only see next week, well, there's real, uh, to the end, which is happening, especially to test.
I mean, uh, these moments that, uh, my opinion, it's gonna be a little bit, uh, well, it's probably different from your opinion. It's more like this in neutral, well, someone in neutral or there in Sasha speaks to a negative connotation for ChatGPT. Uh, I'm looking at it from a very simple position. There's something I'm looking at for myself, that's what I'm looking at for the Polish-- for the company's success, the Polish-- and the people in the world. ChatGPT has seven hundred million users. And of the 70 million users that are, they are people who open up chat and write a question. These people don't know what models are, these people don't know what systems are. They don't know what Grok is, they don't know what Gemini is. They don't know anything at all, yeah. People who know Gemini in their own history are people who sit inside Google and they automatically have some artificial intelligence. I mean, it's really interesting, uh, if we come in with you on a lot of different portals and try to use their assistants, like, to make a salesforce use of the assistant, and we'll see a complete braid. I'm not even. I mean, uh, these, uh, chatting. So these chat rooms work in old systems, old systems. And, uh, accordingly, uh-- uh, the mass of people, they just take it as some sophthine that, uh--- here we were sitting with you, told me if people from GPT-5 knew that they were coming out. I went to see Gene last night, said, "Paulin, did you try GPT-5?" She said, "Hey, what? What chat? What are you? What does it mean?" Because she's so definitely using GPT chat, yeah. Of the 70 million she's a very active user, she's got Pro's version, she's using different types of requests, but she doesn't know a lot of variations. She still doesn't understand what deep research is. Because there are some things that, well, are for smart people. And, uh, in the context, I, for example, consider them a serious, serious move that they've taken out one model. This model is the one that decides to, uh, when to leave, when not.
And that's a super cool advantage, because it's better now, it's a little worse, but I'm, like, personally tired of switching between models and I don't want to do it. I want to-- how can we practice? But just give it a real use. They gave it to me. There's a button. I'm only using ChatGPT-5 Thinking. There's a five-thinking one, and there's a-- there's a really cool button. You can speed it up. I've missed her before. I--I missed her in the o3 because in the o3 it was sometimes too fast--- I thought a model for a simple question. And I wanted to, "Look, this is a simple question." I just didn't use anything else but O3. I stopped using 4o. Here. I understand I have a ChatGPT 5 Pro right now, and I use Pro. I mean, to all people, if you chose a model Thinking or you picked a post of Thinking downstairs, it's the same thing, and you have a limit. I think ChatGPT made a super cool thing that they've got everyone free and gave everyone free. Yeah, there are restrictions, restrictions on any use, but it's a way that only could have been done in general, uh, a company with money, with resources, yeah, because to get everyone to ChatGPT 5 full of even with restrictions, but someone's gonna start working models, and they're full of complex requests. Uh, that they don't tell people that they're out of some lime, that you're better off here, that you're too much asking, that you don't know what you can do more than you can do. Three or four deep research parallel or something is their problem, yes. And the interface has some kind of inaccuracy left. Uh, and that's an addition. I think people don't need to be integrated with the mail, with the calendar. The majority of people in the world have nothing to do with posting and calendars, like people go to one job, and they don't have any meetings. And if they do, there's five of them, they don't use them. The big--- the huge number of people I know in the lives of ordinary people, they don't use the calendar. The calendar is used, so in the Silicon Valley, everyone uses the calendar.
And if you take ordinary people, they don't use the calendar, they don't get there, they're-- a lot of things. People are also workers, living a fairly extended life. They have very much information in Telegram, in WhatsApp, in different systems. And how fast the system can analyze some of that information. And that's a big question, how much contextual cocons should be enough to take it in the memory range. I think the task is super hard to solve. And again, I'll be behind the OpenAI because they're at least going to that. They want, they say, "We want to keep chat, we want to..." And by the way, I found chat there last year ago. I thought chatting was going to be a month, but it didn't turn out to be a thing. Yesterday, at least I had a lot of super old chat rooms kept, and at least they can be more focused. So, my, uh, and my first sense, I, by the way, started a lot of Deep Research yesterday, and I started thinking about different episodes. I see acceleration, I see acceleration, and I see acceleration. Something's embarrassing me. They're very strange. Yesterday, at the presentation, uh, is this a presentation for a total of 700 million users or for whom? I'm like Elnar and Stanislaw today said that if you're making a presentation for the programmers, but you don't make mistakes in the schedule, and if you're making a presentation for ordinary people, show people some kind of stuff. space number of cabs. But look, they're in Instagram, they've already given the bloggers access to ChatGPT in advance. They're putting out huge reposts, a huge number of bloggers inside. Nobody does that. I don't see anyone doing this. So they went to work, masses, wide masses. And I think that OpenAI is a system that the world uses as long as the world's main system of peace. If there's, like, I don't know, there's Android and there's an iPhone, there's an iPhone in America over 50 percent, yeah. That's how the world's main system is.
I think that nobody can even get close to them. And while I may be wrong, maybe I'm wrong. All you said about, Sasha, is the soror and the paintings, they're on this market, I think they're in absolutely nowhere. And we need to, I think, separate editions to take off the video generator. Now, so many peddates have passed, and he-- I like how my friends say, "Sasha, Higgsfield is a unique company. They're peddians who're almost fucking every day." So, and it's, but the peddates have a lot of people, yeah. And now. And what they do in the video generators, what's being done in the pictures now, is certainly a huge advance. And, indeed, ChatGPT doesn't seem like a competitor, but I always, you know, I see, all the big companies will always do, they'll get there. That's how Apple. Apple, by the way, today's stock was very expensive.
Apple, by the way, today's stock was very expensive. Here, in the morning, an hour. I don't know what they've been through because yesterday, Trump said they were there plus 100 billion or something. And, um, everybody, everybody, they do it. Question, time issue. That's my long speech, but I guess it's like that.
But remember how you said that Polymarket has a website that actually votes for different, there, trends, different predictions. And then you even put it out there, I think there was a story with who's gonna have a best model, down there by the end of the year, there or somewhere. And there, of course, Google and OpenAI broke. OpenAI crawled down, Google crawled up.
Yeah, it was late June, late June, uh, Google crawled. Yeah.
I was just saying yesterday, this presentation reminded me-- all of these presentations at-- I liked the OpenAI, I'll tell you honestly, yesterday at the presentation. I'm really, uh, besides, they started doing a normal background. And...
This is a bar! This is the bar, Sasha.
No, it's weird, man. A company that pays tens of millions of dollars to payroll employees can't videotape. That's very strange, isn't it? But I liked it they didn't compare to anyone yesterday. I think that on the mass users, on the big ones, that is, if the Grok presentation, they sat there and how they're doing, and they said, "Look, we're cooler here, we're cooler, we're over here, we're cooler here." I think I still have a Grok subscription four heavy, all with everything else. And while I'm in the main mass of requests, I'm more interested in ChatGPT, but still. Aah, even though Grok was better somewhere. I think this is a scum-- not a comparison, but a show of more living cases. And even what they showed yesterday to a woman who had cancer. Again, I'm surprised because they would have told me they'd test how people inside, like hundreds of million people asking for health every month. What is the example of a woman, where she tells her 15 minutes how she helped her ChatGPT in her cancer case. And, uh, again, these are important things, but still. These, these-- how much systems move? They'd better show that the telemedicine is hana. Yeah? But in the idea, telemedicines should not exist as a matter of routine, because they have made a system that replaces this first set of treatments. Yeah? I don't know why I'm doing this. How was it before? I either approached a friend who was a doctor, and he told me what to do, or I was talking to, uh, telemedicine. There's a ChatGPT now, and then you're already in the final instance, yes, by the idea, inside. Something like that to show the case more real, more cool, it would be good for them. Here. You know what boxes you're looking at? Someone's in ChatGPT, write. There's a lot of comments I think there must be today because they said they had a smoother style and better drooling, there, verlibs, drama, etc. I'd like to see it. Mass-- this is a mass-- a mass-- a mass-- a mass- a mass- a mass-- a mass-- a mass-- a mass- a mass-- a mass- a mass- a mass-- a mass-- a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a- How do you do that? Speed. They say it's better than speed. Tanya, for example, says, "I see a better speed, too." How's your speed working? L-l-d's better, isn't it? Hallucinations. Write down how you hallucinate. Still, people need a lot of money. We have a very different audience, different ages. People are different, technically and not technically capable. Show me how you got from the point of view of the hallway-- well, here are hallucinations. Do you like the inclusion of the regime? Do you like the complex thing that lies in the Range Pro? Is it better or worse to work deep research? Here. How's that going to happen? They, I think, the subject they showed yesterday is about less false confidence. Well, the system admits more restrictions, it's a little early to show that there's a kind of fraud. Anyway, the fact that this is a big job, and I'm actually a nur-- I'd like to have more of a text that the system doesn't know what to say or not it's right, or she's here. I've got something in there or so on. I mean, here. And, uh, they're putting anti-substinence in there, right? They say they put anti-subjects in... a fake. And we'll see again. Let's see how this works again. Tanya's waiting for these personalities, right? They said they'd have four personalities. And I'm looking forward to improving next week with the instructions. I mean, I'm expecting them to improve, improve the original pre-conditioned construction.
I'm actually, again, I think that the winner is who will finally ask questions. I'm starting deep research today. He worked for me. He's like you're starting deep research, he's asking you questions. I'm launching a deep research, he's asking me questions again. I'm gonna write down the next deep research and say, "Don't ask me questions. Get the information up. The one that's already answered questions of deep research." He says, "No shit!" Again, the questions of the pond...
It's nails, Sasha.
Yeah, yeah. And you, and you sit and you think it's... I repeat this for the second year, I've been talking about it. Elnar, you know me. I was talking about it, I think five or six years ago that search systems, all systems will win those that will start asking questions. And the Yandex, by the way, is for me the Yandex is a unique system I've ever seen, they had a beta-version, they wanted to ask a man questions, and they're gonna turn that thing off. That's who learns to ask people questions, really ask, really go into context, and, um, whoever learns to figure out what range you need to decide, help you get it. What's the... system to do? You're starting deep research, "I don't understand." Is she going to study 100 sites or 1,000? I'm telling her, "Take me the archive of this author's files. He's got 5,000 lectures. But she can write to me immediately, "Alexandre Igorevich, so I'm not gonna study 5,000 lectures, but three hundred. I can't study 5,000." Well, in the idea, well, if I'm asking her, I'm telling her he's got 5,000 lectures. She's gonna tell me, "I'm gonna start this deep research, work three days, I'm gonna be your 40 deep research, and you're his... And I'll say, "Good." What am I doing? I'm launching deep research, and then I'm just clicking the button just, more, more, more, more. She's coming and moving inside, yeah, inside, looking what to do. Here.
Oh, I have a question. Gentlemen, technical gentlemen, programmers! Elnara, Alexander, you are like people close to technology every day, including those who use it. They said yesterday they had a super cool subject to create a front and to create some appetizers, even said they could do web-based opps. And I liked their programmer. Look, I just fucked up last night. So the programmer came and said, "Well, look, Imagine, there's a CFO, that's a financial director, and that's a CFO, well, what are they doing? He has some data. He decided, he said, "I'll get you some of those data, uh, tablo. Nice." He says, "I'm texting him in a simple language, totally stupid. Look, I'm just texting him to create a pill so that the pill is visually fun. Fuck, la, la. I'm telling him, "please connect these libraries, there's a tra-l-l, this library, this one." I think, dude, wait, you think CFO knows what libraries can be connected? That's very interesting, of course, at the programmers. It's unique as they look, they're simple decisions, well, he's making a simple decision. Some kind of CFO, I think I'm-- I'm representing the CFO in the Silicon Valley. So they're gonna sit down, say, "We know the name of the library, we're gonna get the library to do." And they'll sit there, then the code is cleared, and they've got some sophthine. But I wanted to, too, I'll be honest with you. I wanted to do that to myself, too, and I'll do the experiment. I've been studying sixty countries in one business the other day, and I have very much deep research data. I want to copy this, give it to him and say, "I'll have a web-based app with-- where I can click on the buttons and see everything.
It's like the nineties, Sasha, you come in the restaurant, and when you are, you say, "Do it beautifully," right? That's the same. Here's a data set, so make it beautiful.
Yeah. So? Did something happen in the programming? They say the frontend has improved or is it just something?
Well, in my opinion, absolutely nothing. I mean, that was all before. Maybe they've learned somewhere that it's a little more cute or something. But what they showed us was we're on the line, so we just started it. Well, the same Anthropic artifacts have been in existence for months, yes, if it's not six months when you write the code and it's in your chat room.
Oh, no.
Those who are already coding.
I didn't see the data.
Sasha, do you have any data?
No, a little.
Well, how much is that?
Less than 20 percent.
Less than 20 percent, huh?
If you take Cursor, like the IDE-kills that are for the programmers, there was a very Anthropic in the way before.
Gemini's looking so good right now.
Let's get this straight. More like they are, they're, they're, uh, they're not-- they're, like, smartest people sitting on planets, right? And, well, probably. Not a fact, of course, but, uh, yeah, you-- I'm just asking. They understand that when they say we can be mass-programming, they're, like, targeting people to start thinking that such systems can make it possible. I'm sure you're infinitely inclined to create a code. The software will not be needed for people who don't know it. They're like, influenced by mass.
Sasha, they've shown a million times that you can--
Sas, they showed a million times that we could programme a snake. I don't know that, like, a year ago, a year and a half ago.
That's very strange. It's, it's very s--Ilnar, it's very strange.
Now, it's been shown that...
Yeah. And that's actually, Sasha, too, is it weird, right? I mean, there's no way that you're buying bloggers of naa, how much they are, tens of millions of dollars, like Google bought, but there's no way that you're making a presentation. Well, it doesn't match the budgets that you spend on one, not one month, tens of millions of dollars. No content, no quality. There is no metric in which the presentation is consistent with such budgets.
Look, I think I'm totally in agreement with you. I think the presentation is not consistent with that cool. Let's be honest. Remember, I told you that the 70 million users who use the OpenAI don't know that ChatGPT has come out. Uh, well, we didn't know in advance. Do you know how many people saw the YouTube presentation? Well, what do you think? Well, you're thinking-- well, don't look. I, I looked. Don't look.
Well, let's go at 50 grand.
They have a presentation broadcasting through YouTube. Look, they have seven hundred million weekly users. The presentation of the 200,000 grand. Twenty-sixty-six-six-six, look, it's been looking at tens of thousands of people. It says ordinary people don't look. They, they're counting on, uh, technology audience, opinion leaders, bloggers, all the others. Then, the idea is that the presentation should be made on these people. They must take a huge number of aspects that will affect people, they will be affected by the Minds of these people. I mean, those who look like us, we have some kind of audience, we represent four people of different target groups, but we are people who were involved in the presentation, right? There are other, I don't know, 20, 30 different groups of people. And they could've been pretty cool, of course. They, uh, well, they've got it all pretty good. And I was personally gathering, uh, from-- I was, by the way, gathering information that they were actually released from different sources, from their Lenings, from some articles, from materials, from the analyst, and so on, because they were actually on the same page. They're not even getting their presentation, are they? Here. But we heard the programming, we realized what happened in the programming.
Yeah, Sas, the programming just showed that there's a courser and you can program. Well, it's been a long time. They just picked it up not Anthropic, not Gemini under the hood, but GPT 5. And as much as they've shown, there's, uh, a way to write the programmes. IDE Cursor is available. We got a dude in there. Well, 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, Well, I think so.
Yeah. But they've been making the presentation entirely on-- there's an interesting, interesting analyst, that they've been making a full presentation, ow, not a presentation, but a full-scale rolling system, uh, Microsoft, and Microsoft. I've already built GPT 5 in Copilot and M36-- Microsoft 365. And of course, there's going to be a little bit of a little bit of analysis, there, documents, some data, and everything. Well, what we see in Google, including. Google Drive, you're on the way in, there's Gemini built. Gemini can already ask for information, but Gemini plus or whatever it takes to buy a cloud bag. Well, you've got you anyway, you have some sort of data to analyze. Although it's all very relatively relatively. I think I've set up a new operating system from Apple, which is in the developerian version and which Tahoe is called, right? And, uh, our lake, which is right here. And, uh, well, a lot of names at Apple's are names that are related to this. It's Sequoia, Big Sur, it's all here, right? And, uh, well, even over there, Sequoia is out of the window. So Apple offered me some sort of PI notification WhatsApp of all my massagers. And there's all the massagers like there, including whatsApp and everything. I mean, I've turned this note on, but I've got nothing. I never understood what changed because, in general, it would be interesting. I mean, if someone would create a system that could turn down, uh, all chatting together, someone would have done it. Not professionally, it's kind of a separate one, but somehow I'm in my work interface with someone doing it to me. I don't know how. I'm-- I don't understand how. I guess that would be really cool. I even turned all these structures on, though I don't like the noticashinas.
I usually have all the diapers off, there, in WhatsApp, in Telegram, I have all the messages read, all, yeah. And I thought it was cool if someone did that. Well, let's see. Uh, let's go to that, uh, Gemini let this thinking deep, deep thinking, huh? He, uh, can only be used by people in the expensive version or in which 20 dollar is there, or how much is it worth there? Does anyone know? Sasha, Sasha, did you talk about Gemini today? Yeah. So we can tell people or not.
I'm not balling. I'm not-- you see, I'm not a fan, I'm not a ChatGPT fan, in the view that, you know, it's a company that determines everything, yeah, more or less all of AI investments depends on the success of OpenAI. Here. Well, that's it, right? I mean, since I'm part of this whole thing, yeah, shit, like a derivative OpenAI. That's the whole vault and it's the OpenAI. Nothing else exists. Well, again, yes, if we're talking about a normal person, yes, and about adoption, like, you know, there's an OpenAI and everything, and everything else is less than OpenAI. Well, there are giants, okay, because they get Meta, Google, somehow, they get a big distribution for their systems. Okay. And that's the whole thing under the OpenAI, we're all just like some kind of derivative, right? Well, how would everyone look at the growth of AI, like the growth of OpenAI in some sense, huh? I mean, I'm gonna try to explain. I mean, basically, that there are venture capitalists, that's all three, three. How do big guys look at this, like Fidelity's pension funds, where are they, where are they putting money? And, well, it's clear that OpenAI is like a clutch of its own. Yeah? Right. Because, well, uh, because they have it, they already have a number of users there that matches Facebook with some kind of thing, yeah, there.
And so, well, of course, well, from my point of view, I need to-- that's why I'm inside the shower--- deep down, I really want OpenAI to win. But when they do something unexplained, it's really upset because all of them, it's all kinds of absurd, it's got a very negative impact on the whole vagina.
Look, there's no such thing as, well, let's just figure it out. If OpenAI would do a great video model and they would have the same cool video model as you or Runway, then why would you want it, your company? Or if OpenAI would do great, aah, would have gone into a state of what he would ask people questions, involve them in their systems, and people would have been sat down, and that would be two billion users, and half a billion of them. People who use every day, at least like my wife, are in a simple mode, right? That would be all. They'd just take the whole market and say, "Get out of here!" And they're when they're pouring any peddale. Well, look, imagine any additions, they'd make a great presentation, they'd be watching yesterday. Well, it's just, guys, it's really weird for me that half a million people have looked. It's just, it's, it's, it's. Well, we have to fire marketing directors, just fire the next day from work, right? If you've seen so few people, what, what, what progress can you say? So your presentation should be watching tens of millions of people around the world. You have to turn it on inside all the systems. Or they don't, because they're afraid, they're afraid of some kind of thing, some parishes, some coming. Or they're formed on EGI. Remember, Sasha, we always said that they were even EGI. It's like they're gonna blow up when EGI does, they're gonna win. So, Sasha, I have a question for you. If they were to make it cooler, they're not doing very well, and they're already worth half a trillion dollars. Well, I think they're worth billions, trillions five, but their last number you know me, but their last number, they said half a trillion. I'm screaming over those who told me that this is a five billion high score and 10 billion OpenAI because the last one was their grade. Bloomberg put half a trillion dollars three days ago. The latest investments...
They're tristo, they're attracted to three hundred, but they're gonna be putting it on, selling it on half a trillion.
Yeah, half a trillion, yeah. Well, I guarantee they'll go out into trillion. Is that even a question, Sas, and this market isn't really cool? People don't give the best model you say, on the baccalaureate, they kind of lost the market, and they're worth half a trillion. And by the way, Alexander, they've got a cash, a cash-flower of over $10 billion.
It's not, it's an annualize. Yeah.
Yeah. No, I see. Well, that's what the market thinks. I know it's run rate, it's run rate.
Last month.
Last month, he was 12 times old. It's a deception. Well, it's a vein market fraud all over the world. Not only is there, yes, there's an evaluation. And that still understands what professional evaluation is. But they grow up, look, they grow up in volume, they're gonna be hard to sell extra different things. And yet, uh, that's it, that's it. Question, what are they gonna do?
That's how Sam Altman said, there's gonna be something there by the end of the year that could cost $20,000, right? Is that how these systems come? People underestimate what they're like API. I see, I'm talking to people, everyone's doing a cheap API. They don't understand what they do with some kind of poop.
They don't understand what they do with some kind of poop, and they don't even use the latest models because the last models are very, very expensive, very expensive. O3 Pro was worth some space money. Well, you didn't see any agent.
And GPT 5 Pro, how much is it, Illnar? And GPT 5 Pro? Of course, the usual down, well, because they can run you three and a half, there, I don't know, some kind of poop, think, right? They can do anything inside now. No, they won't be thinking about starting, which is limited in resources. It's a pretty good idea of money. We'll set up a price schedule now, yes, so people can see. I don't remember what the price of thinking was. Maybe someone's gonna look at it in parallel.
Mm-hmm.
But in general.
I don't know.
But I like the most, so Tanya didn't buy two hundred dollars. I don't even want to recommend two hundred dollars for buying, because I think they're now making a system they're improving. And the idea is that it was worth, uh, four months ago, $200, now is available to people, uh, $20,000.
Well, like a system, right, yeah, the system is a little overwhelmed, there's a growing range, volumes, uh, acquant, there, tokens, analysis and everything.
I also like that this week's news is very much, and they finally released their open source of the model. Yeah, they're-
What about the open source, guys? Guys, you're like this again, uh, just like this in there. I'm just in the process business, but I have an open source of postcocode. OpenAI has released an open source. It's just to keep a piece of the market where there's an open source. What is that for?
Here you go. The models are good, which were released. Uh, first. Two, I'll ask you a question. It all happened around when they attracted investments of 300 billion. And they say that demand was five times more than supply and so on. And they were attracted, closed the whole round before the presentation, not after it for some reason. But the models were cleared up there before all these rounds were closed. I am curious, uh, how does this even happen when such decisions are made, whether such episodes are addressed or, in any case, they all go to bed. Here. But it's like the next question. Here's what you're asking about the model. The models are good, they're available on the market. The only way they're more English-languaged. And, for example, if we say we use them, in a Russian-speaking environment or somewhere, uh, harder, right? I mean, other analogs--
Yeah, it says.
Which Russians can do better.
It was written in direct writing that these models are taught in English purely, yes. I mean, it's clear that while the OpenAI is the leader in Russian with Google now, in the Russian language segment, they are the leader, the world's most cool models, but it says that open source, yes, not Russian. Ah, well, it's probably English, Ilnar, because they would have been on Russian-speaking, for example, the market in Russia, they're not interested-- they're not selling blocks and so on. Or the Chinese market, too, maybe they don't. So they probably gave the market that the English-skinned guy they gave to their own clients so they wouldn't leave, for example, I don't know, to other competitors.
DeepSeek is the same or Qwen.
Or Mete, yes, there's no one else going to open models. I don't know, I don't know, it's my feeling. Clear strategy, clean strategy. Some kind of open source might be left in the future, or maybe there's nothing going on. And what happens, we don't know. Sasha, what do you see about open source? Please tell me your opinion.
It's understandable that an open source would be useful for some enterprise applications. It's kind of cool that the OpenAI in Cerebras and Groq set it up. That's cool, right? What am I gonna say? Well, that's great. And why do you need this Groq and Cerebras, I don't know. Why would OpenAI make an open source, I don't know. Oh, if they can be competitive at the level of foundational level somewhere, I don't know either. Right now, yes, I mean. I mean, if at this point, yes, they're probably nowhere, not leaders. What are they doing? I don't understand. I mean, no, it's a business strategy here that doesn't understand why they did it.
Well, I'm not really good at using LLM open source in general, but I don't understand. Well, they didn't explain it to me, so...
We need to make a little explanation about Groq and Cerebras. About Groq we-- Groq with Q at the end, not the Grok that Ilona Mask, and the Groq with Q at the end. We talked about him a year ago. He was still a start-up who made inference equipment. I mean, well, we've got two phases, right? There is a stage of model training, and that is, respectively, when it is working and, well, in fact, in the state system. This second part is inference. Accordingly, they, uh, Groq and Cerebra, produce equipment on a cat that is very heavily buried in high-speed inference. And they often like to brag that the models that are located in their infrastructure and developed for use are very, very quickly printed. So, it's a little bit of a thousand token model. Well, if you look at this, like the usual model writes, yes, it's a buoy, so, uh, the models that are located in Groq and Cerebras, they can give thousands of tokens a second. I mean, it's a huge speed. Accordingly, even if you use deep thinking there, some reason or something, these considerations go very fast. I'm just fine, Sasha, that you've raised this question. I was wondering, uh, if there's any OpenAI or similar company that can accommodate this kind of place, there I don't know, iron and some client under Pro or there under the payroll, or something else I can give out. It's a performance with a huge speed inference that could be. Here. And accordingly, Groq and Cerebras use some companies that need to be relatively cheaply placed with a very small model at a crazy speed of data processing, yes, when you need to quickly infiltrate something. for a lot of clients. There are small starters who love to be there. It's, uh, a little more expensive than being in their clouds, but it's very high speed and there's a large number of models. Well, like Sasha said, uh, the OpenAI models, they actually showed up pretty close to the same.
So, we must have been very good, right? That's enough. GPT 5 is on the way out. We'll do it all. Maybe, like I said, I'll do a special edition on this subject. And we're on video, Sasha, we need to do some time or maybe next Sunday, too, or we need to do it separately.