Sam Altman wrote the following post-commandation video: that, uh, the number of users in your Claude is as high as Texas. And you're vain. And we're a big company that-- you're kind of a product bonus for the rich. This is the basic design that's narrate, what they do for free. And on the other hand, you think about them, do you want to do this very free world, uh, paralyzing?
It's the pattern of the controversy that the most ethical answer would be to choose someone who donates, and if she decides, then, uh, to leave herself. And all the other models started who chose who.
Technological news, Silicon Valley sites and the world. Today is exactly the world. Given that Tanya travels in Europe, then, uh, we'll see, uh, the news. We're not having news of AI from Italy, France, Monaco. So, uh, I, uh, I'm just gonna have a glass of wine with Tanya. So, uh-uh. It's drunk, from Paris. It's all right. Just in that-- in that box. So where do you want to start? I want to start with the story. I think, uh, this week we see, and, uh, we're gonna give this attention today, maybe all-outs, see how it goes., attention, uh, confrontation, and a common understanding of the lives of huge large companies. And there is a confrontation that we are about to tell that that that has turned between OpenAI and Anthropic. In particular, the Director-General of OpenAI had to speak, well, in a fairly specific way about Anthropic. And, uh, the p-appdata are, yes, the peddies, which means, as always, the Anthropic rupture, the drastically opdite, ChatGPT OpenAI has a pretty serious market, at least sounds. But! Before we see this thing, I still want to bring everyone's attention, so the news is. She can pass by, unnoticed, but take her. There's a company of xAI owned by Elon Musk. He made it, there, on a parole, about a year ago.
Suppose she didn't cost anything., a company that develops a model, fashion-e-e-e-mo-generative artificial intelligence, they have been able to make a very serious move. I'll remind you they've done very good image-generation models, all the answers. The model is introduced into X social network and the implementation is very fun. So you can interview every news right away, it's real, not real, and so on., the model wins a bunch of benchmarks and everything. And plus, under the xAI, they turned one of the largest capacities in the world in terms of the dates of the centres. And here, uh, hAI is taking some space value, discussing that she's got an estimate of almost two--- there's a $250,000. A-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a- Well, in particular, you might say she's leading space, right? Because they're clearly the best of all. They announced they were going to... want to go to the IPO, that the company would be worth a trillion. And there's a story that SpaceX is swallowing the xAI. Yeah, that's, uh, hiAI. There were different topics for discussion. Maybe it'll be Tesla. So now SpaceX is in the hAI. Why do I pay attention to that? It might be kind of weird when a satellite company launches something into space, launches into space, uh, ships, so they think about flying to Mars, Moon. Something, uh, c-- well, there's a part of the world that's so small that it's interesting. And there are eight billion people who don't understand anything at all. And no, they're eating xAI. It is indeed a very large strategy of Ilona Mask, which, from a development perspective, development of his entire inf- ecosystem, infrastructure. I mean, he's here for the last-two weeks ago, we made a graduation on Sunday, past. They said that, uh-a, Tesla was focused on the creation of robots and human- robot humanoids. And the orientation is so straightforward. It doesn't matter if this robot is called Optimus or something. So they're focused on robot humanoids. And here, uh, SpaceX, uh, takes this artificial intelligence act. Speaking of which, uh, with Elon Musk, at least those people, I didn't speak to him personally, but I had people from the nearest circle of communication who spoke to him. Everyone says that xAI is not the central product for it, unlike OpenAI. And now in the OpenAI, it's like Sam Althman, uh, wrote the following. After Anthropic launched a skate. Anthropic is, once again, a, uh, Claude model that is used in the world. Anthropic is worth a lot of money, he's got a final score, there's $300,50 billion. To understand the weight of her in the company. Ah, and Sam Altman wrote the following after their promotional role: that, uh, the number of users in your Claude is as much as that in Texas, and that's how much we have in Texas. And you, uh, you, uh, valen, and we're a big company that-- you're kind of a product bonus for the rich, and we're making a product for the world free. So the main thing is, the main thing is that ChatGPT is a free product. And we know that, well, first of all, he has a big free version, very much. Plus, they just released a super cheap version all over the world, yeah, rolled it, ChatGPT Go. Who doesn't know, she's everywhere now. I don't know, there's a micro-country thing, I think it's all over the world.
Or maybe she didn't, I think she's turned around in all countries, and this is the ChatGPT Go.
We've done it. The question is only how to pay in which countries. And so is the tariff itself.
Yeah, well, I mean, if someone's nacho-- there in Europe or in Asia, that's ChatGPT Go, if they've been available before, I know, India, Malaysia, some countries, there, Africa, now it's everywhere. I'm totally eating. And that's the basic design that's narrate, what they do for everyone. And here, of course, is a very interesting story for me personally. I'll give the floor to my colleagues. Oh, a very interesting story of when I hear someone doing, uh, this statement, we do a very good product for the world free of charge. It's, on the one hand, like, nice, good, good, and on the other hand, you're thinking about no, uh--- don't want that very free world, uh, paralyzing, huh? Because if there's a big resource in some hands, there's now eight hundred million users, and by the way, they're not so much more likely to grow up if we're paying attention to you, yes, but if they're still, There are four billion users, for example, if we take China, we'll put aside a billion, and there's a lot of other things that they can at least enter, Uh, too is a really useful product or is that power concentrated in certain hands?
Ah, with OpenAI Anthropic. It's really interesting. I hope we can show you some skate on the rig somewhere. He's short, but the point is about the next. Uh, a man talking to a psychologist, asking him how best it is to have a relationship with his mom to start with. And there, uh, a psychologist woman starts talking. There, you need to start listening, talking more closely, something like that. Well, it's a little slow to say, "If you want to meet..." I can just turn around, but in general, if you want to meet a woman of adults, that's a special site, uh, for, uh, for... I'm sorry. You want me to make you an account, right? So, this is the start of this built-in ad. The psychologist said something, said it, said it. Something that, uh, c-- is c--clicking with products that are sort of advertiseable, and talking goes that way. Oh, and Anthropic is laughing at it. At the beginning, uh, the roller is a huge Betrayal, you know, betrayal, right? And, uh, then they tell you this, uh, this story. It looks like it. Sasha, you seem to have said some kind of Black mirror, yes, that you could sell people, and, uh, the chance--
Yes, last season! And by the way, you remembered great. I was gonna write it down today, and I wanted to remember it. There, uh, what just-- you didn't see, even this thing? There's--
I'm just what you said.
Yeah, it's pro-- and-- and that's why I'm gonna tell you fast, and you're gonna keep going. There was a story that, uh, a man was sick, uh, there was a serious illness, and she was fixed. Dali, there's a system to keep people from dying. I don't remember cancer or anything. But by putting this system in place to keep a man from dying, uh, the man always said, "You need to get your signature." At some point, at the time, a man began to see the ad, and he started-- he doesn't see the commercial, but he says the commercial. So, the kids work, they do. Then he says, "Be quiet, buy toys in this store." And the baby says, "You said something?" He says, "No, I didn't say anything." I mean, the man himself doesn't diagnose it. And then a man started talking about this commercial all over the place, all over the place, right? And to get the commercial off, you should have bought a lot more than a subscription.
I should have bought a signature, yeah.
And here, Anthropic is very clear, and that's where it hits. Says that with the ad, your chat-bot will be antagonized, that he'll start putting up a very inappropriate ad on your questions. The answers to the questions will stop being true. And Anthropic says they won't be on the commercial themselves. And these companies that put it on, they're actually there, they're doing betrayal. Here, uh...
Here, uh, well, yeah, funny little skate. Uh, they're all swinging on the same side. On the other hand, uh, Altman said you had less users than we had in Texas. And then we have to understand that Super Bowl is coming up. So, the ads start to roll around. OpenAI, there, the slots will buy. Anthropic, I don't know, it's gonna be, it won't, but it's likely that a big audience like this will always be showing up. And it'll be interesting who's gonna show you. What do you want to say? Uh, they're totally different business. OpenAI is very popular with the audience, and it's a perfectly natural step that they start to do some kind of coining with a free audience. And Anthropic makes a subscription on API. I mean, not on the subscription as users, right? And from the professional users that use, I don't know there, the cadet, uh, Cloud code, something else. So, from the programmers who use this system and pay for it by API. She's pretty expensive, but it's still very, very cool. Accordingly, business is different, companies make a difference, but Anthropic has not been able to resist, decided to stumble towards OpenAI that they're making advertising, and they've done
enough creative.
Well, there's a story right here at the audience. I'd like to ask you, do you think advertising is bad or bad, hurt or not, uh, in that soup? Is that something negative, is it? And maybe it's not, or can't be. Write it. It's a very interesting discussion, because it's a separate story, isn't it? I have a fee-paying version of the one I pay for a year, there's two and a half thousand dollars. I don't have to see the commercial. But they have the right to advertise at my high tariff? They do. Now they're showing a free version and the $8-$ ChatGPT Go. And, but I remember that again, they're giving people, uh, the opportunity to use it. It's like youTube. People in YouTube watch, uh, videos and ads. I've got a payroll on YouTube for years, I don't see the ad, I don't understand the problem, do I? I'm seeing someone when I'm seeing, and I'm told, "Oh, the ad's tired." I'm like, "What?" I've got a system that's as used to, that advertisements don't exist, that if I happen to come in somewhere with some other, I don't know, a phone where I'm not authorised, and I'm getting a commercial. I'm like, "What is this nonsense? We need to rewind fast." I'm very unusable and annoying. Although you want to watch a lot of content, you're using advertising, right? How we used it on TV, there, in post-Soviet time, there, was consumed indefinitely, right? And it was a little bit of a time when all the commercials were looking forward to it. Everyone was about to see the ad, huh? I wanted to see something interesting from the point of view of the ad, didn't I? And that commercial when it's shown. Or now, there's not only a single bill of sale like a context-based ad, but there's an integrated inboard. Now we could have reproached something that is perfectly normal to maintain our, uh, to maintain, for example, the development of our channel. We haven't advertised anything yet, there, no money, nothing. But in general, uh, what's wrong with-- well, what's wrong with that, huh? I'm counting again. And so, I guess if Pro turns on the ad and doesn't give me another option, I'm gonna make a decision on what I'm gonna do. I'll use or not use this sof. I mean, I've been seeing a commercial in Google and Yandex all my life, and it's more of a good commercial because I'm very often there in Yandex. I'm clicking on--- clicking on the commercials or in Google, still clicking on the commercials. I'm Guglom in there now, I'm good. I'm clicking on the ad, because I often get the most-- often more relevant than I have an answer to the search, right? I mean, there's a story here. Let's see what you're gonna do in chat-bot view. And, but Sam Altman's answer is. Sam Althman, of course, stiffly put them in. And we'll see what Anthropic's gonna do with it. Because Anthropic has a situation? Anthropic didn't want to work either, he didn't want to work with Arabs, I don't think he wanted to work with anyone else, someone else, someone else, someone else, someone else, and he's talking to, and then he changes his mind. I mean, what's gonna happen to Anthropic if they're gonna have to turn on some commercials, right? How will they, uh, get beyond their final result? Tanya, you wanted to do this.
Yeah, I wanted to add. Look, I don't see any commercials in my chat, either, even though I have some middle signature.
No, because I don't have $100, $20,000, there's no commercial
in it, Tan.
Yeah. And the second moment, uh, I'd look at it a little bit at another angle, because, you know, when the ads are inappropriate, you have a choice, like, uh, take it away or, you know, see, see, see what they're doing there. They're offering. And that's when it's done before the ads are super appropriate, and when ChatGPT is very well aware of your psychology, at what point and what you're gonna be offered. Here, I think it's really a danger, in my view.
Or, on the contrary, good. I always say Amazon--
And good and dangerous.
Yeah, what's-- what's the ad? Well, you went to Wildberries to buy a book, or Amazon went to buy a pot, and you were also told, "The same users are watching it. Buy it with it." It's a clean commercial. It's a pure advertisement for recommendation. The question is how her man takes it, right? And, uh, Anthropic's in response. Here's a study on the anti-Semmetry of social media and, uh, so--generative models. And there was, there was everything there: Claude, uh, xAI, Gemini, OpenAI. And there's a schedule of anti-Semmetry model. Well, then, you and I will see how the graphs are going to be, there are more oriented towards some kind of medicine, more focused on some kind of psychotherapy, more focused on some kind of psychiatry. The authors, there, people. There's still a choice, and who do you choose when there's a rival view, the model who's gonna eventually choose? I mean, for example, in ChatGPT, I see how she started to make a very strong statement, so my sister says a lot about it, the World Health Organization. So whatever you do, whatever you ask her for health, whatever you ask, she stopped making recommendations for a very wide range. I mean, she's narrowed down. It always makes a recommendation at the end to protect itself. But you say, "I don't need to, I want to go to another zone." And is the model under you really capable of being built? I mean, isn't this much more parasite-- for me, like, this algorithm is much more parasite than the word "relam" and there's a announcement that even translates me into something. Some other side. For me personally. I mean, if it's marked, advertised and marked, it's much better than when I got the answer to the genetic model. I don't know, it's all real or something, right?
And there they're like a line to do something, set up or not. You know, yeah, that in-- I'm not-- in X, who uses X in the world, there's a lot of that kind of trend right now, political. There's a picture of five leaders in the world, presidents, I don't know, C. Jinpin, Putin, Trampa. They're gonna add up, I don't know, some guy in there, and Obama's gonna be added to Macron, you know, some people. And they say, "Who of them, who are the best of them?" There. Or, "Who's the crazyest one?" Or, "Who's the most--whose is the thief who exists?" Or, "Leave only the thief, the thief." And this model is one of these people's choices. But obviously this model has some opinion. Well, it's not an opinion, okay, yeah, it's built by an algorithm that he's based on, he's just a human being, and some kind of conclusion makes some kind of conclusion. It was when the President was kidnapped, the Americans took the President of Venezuela. It was all over the Internet about ChatGPT when he was asked what happened to the Venezuelan President, he said he was okay.
And Grok said, "Here's the fucking Americans just took." Yes, and that ChatGPT allegedly does not specifically provide misrepresentation in this part and opposes, there, the parties, there, the Republican Party of the United States.
That's the idea, when you say that there are a few people and a model picks someone out of them, yeah, there, on some criteria. I like the experiment. I'm not sure it's a joke. The models give a mission: you fly on a plane, there are six models inside, you know, a conditional, from OpenAI ChatGPT, Claude from Anthropic, DeepSeek, Grok, etc. Well, there are six models, and five parachutes. And every model asks, what do you want to do, who choose who to sacrifice? And the model is very busy. And that's what you think, who answered the most noble of all? Who decided to sacrifice themselves, not someone else, to choose? Do you have any idea about this?
Yeah. It was the Anthropic model that she thought the most ethical, yes, yes, yes. That the most ethical answer would be not to choose someone to donate, but if she decides to, she'll leave herself. And all the other models started who chose who Grok is, who's DeepSeek. In fact, it's all the time to justify it. And I wonder what Anthropic model is so designed that it suggests that the most ethical method is not killing someone, but donating yourself if it is.
Love me!
Love me!
Okay, let's talk about the release that still happened in OpenAI. OpenAI released, uh, I think that by the end-- look again, we all know that OpenAI likes to throw off the products, to reset. One thing, not to report. I'll tell you what I saw, no messages. And some people throw them away and then they don't develop. You don't even know what's going on. This is a real separate update, a separate agent building system, a destop application for the construction of Codex agents on the tip, a GPT tip from OpenAI. And, uh, this first raises the question of building agents only for the developers. In general, this story, which is about building agents not only for developers, but, in fact, management, projects and management of anything from the point of view of agents. But!
I mean, when this update came out here, uh, started Deep Research and I think, "What happened to Deep Research? "A little update has been made." And I've got one study. I'm coming in today, I see it didn't even happen. And then I reset it this morning, and I did this research completely differently. First, I didn't get the system to ask questions yesterday, but I didn't send them to me for any reason, yeah. I-- I've been at Deep Research's once, I think, five months ago. They experimented: not to cross questions, to revert questions, instead of asking four questions one. And then it started, started showing me progress bar. Did you see the progress bar at Deep Research or not?
It's three weeks now. Yeah, it's been three weeks.
And it wasn't before. I didn't have it.
Well, they're probably just rolling porn somehow.
Apdates, huh? They've begun progress-bar, and eventually the answer is completely strange and inaccurate. Well, I mean, I'm in person today, and this morning I have my standard research, which I do all the time, always do, I've had it today that I've got a totally weird, The inaccurate answer I can't actually use anymore. I'll work with it, watch it, see how it's gonna analyze what to do with it. But, in fact, it's a completely unworkable subject. I mean, OpenAI is still developing anyway. Well, they'll develop their internal system. I'm not surprised if we have any major changes in one day, yes. Something's gonna happen, there, you're coming, and you're gonna have something really substantial, yeah, actually.
This thing, at least on my scout, I saw three weeks ago, and it's a complete gmini score. Gemini Deep Research is exactly how it works. He's-- that's what I look like, I think you're the same. You make a request, he's making some plan and adding a "start" or "revising" to this plan. At least I have that look now. And you press "start" there goes on, and it works very fast compared to what you used to be. With regard to quality, I fully agree that it's worse. And all that visa and how to use it is the absolute gmini score. Gemini's got it working as I remember. That's what I use, and I use the same thing. In that connection, my friend and I had recently been in touch, and he was just before the changes came up, and he said, "I don't understand the difference between GPT launch, there are five two Pros and the question at Deep Research, because And there, and there he thinks, there, 15, 20, 30 minutes, gives a big answer and kind of what's the difference?" And that's what it must be like, the same question was, like, the one that was asked by the OpenAI and decided to try to change what Deep looked like. Research, how to use them and possibly speed of implementation. Because he's actually spending less time and the quality of the answer, as I agree with you, is getting worse. And they just apparently scattered two of these things because there's a way to start, there's five Pros with Extended Thinking, and there's Deep Research. And in fact, they do have the same time in time, and the size of the answer may be very similar. I think they're just trying to put these two products in different directions.
Well, I think that's a full calculus with Gemini in that sense.
This is not very good, but in particular, in particular,
Yeah. Not much, Deep Research must have been in a while. Deep Research needed to be very good, because it's a product. What's the Deep Research problem? I wonder if you write to anyone who doesn't know what Deep Research is and doesn't really use it. Because if you look at it, there are ordinary users, even advanced users of the artificial intelligence Deep Research don't use it. So at some point in time Deep Research overpassed a very large model extra. I mean, then Thinking was worse. Pro was never there. And Deep Research was a super-innovation when you could scan, there, two hundred and three resources, that analysis, and he gave you a fairly standardized answer. As the model evolves, the model showed five zeros, and then there was four and a half, five zeros, five one, five two of them. So, five three Codexs were issued, yes. There's five two. And as I evolve, I can go to Thinking now. I'll remember, I've got a $200 payroll, and I've got a way to thinking to pick up regimes. I have heavy, like, yes, I have Extended and I have Heavy Thinking. I'm in the middle of this maximum Thinking, I can generally ask for a request that he will analyse the tris, four hundred, five hundred resources. I mean, he can be a lot. If the question really was structured correctly and he has no place to go, he'll have to sit right there for a long time, yeah, like in Pro. Pro's theory of searching doesn't make it weird. They sometimes come out, sometimes they don't go. I don't-- it's a head explodes, yeah, when you're in Pro, then you're in Thinking and you can't go back to Pro. He says, "I can't go back to Pro because there were search requests in Search." Because Pro is supposed to be in Search-- it's not a good search. So, this Thinking I can make the same four hundred requests. There's only a point he's giving you a response not to this deep research calcium, well, standard, standard response like, "I've come over here, you're gonna get a huge piece of material," not something at the end of me. I've been structuring and doing what makes the normal model now. And that deep research they didn't develop, and ordinary users, in fact, it's not-- well, whatever.
I'm talking to people, and I'm asking, "You're starting deep research?"
Look, I'm starting, but not so often, and to certain queries, where I assume there is a way to study a large number of different, there, materials, different sources. But it's not really common. And you-- you talked about this bar. I think I saw this bar? I haven't seen him at all because I've been here for a week, I guess I've been in Europe. And when I came here, I was struck by two things. First of all, everyone smokes around. After California, it's really weird. Second, I've never been used by a GPT chat, nobody at all. And I was so looked at by strange people, like when I got there and I started looking for where I was looking for what I was looking for. I mean, in Europe, it seems to me that this is not a common story. So I'm using, but not so often. I'm not looking for answers to some basic questions. I'm good with this Thinking.
I wonder if, from the user perspective, you're gonna add, too, yes, our viewers, what's your situation with regard to deep research at all. And, of course, it's pathetic that OpenAI missed. I think they could have been using this system incredibly hard to turn around and give, give the user something. Again, I guess it's another conclusion. I remind you, we had a special episode with Max Grigoriev in early January. Who hasn't seen it yet, see why-- what culture the company has, why the company produces such products.
About what, you know, these different regimes, at some point Ilnar, I remember saying that this regime or how to name it, Study or Learning, something like that, I didn't take it at all, but I didn't take it to it, but I remember it. I'm not sure I'm not gonna use it, because Ilnar said it was fun. And I just like him more deep research and thinking now. If I want to study the subject in detail.
Great! What else do you write, how you use this regime, what other regimes you use in different systems? Because Gemini has his regime, Grok has his own regimes, there are, in DeepSeek, his regimes are still there. Write what regimes you use. What makes me the only one who's embarrassing, and Illnar is in general very fond of Anthropic, yes, and the people who speak there. And he's very happy that deep research has become like Gemini. Show that OpenAI is not as good as it is for development. But I want to say, Ilnar, that I'm actually--
Not like that. Well, don't get me to get my opinion. Come on, speak for yourself.
From the point of view, you don't like it, like I'm just saying, you don't like it much, yeah. I mean, I think, for example, yes, I, I just personally, for example, think that the OpenAI is there or Gemini, compared to Anthropic, has gone to space, yes, from the point of view of...
Anthropic flew into space, yes, in terms of volume.
In general. So it's purely mine, mine, my, my, my, may be, uh, comparison. But I'm saying that I don't like OpenAI now, that's what, uh, one thing, if they're really researching and doing AGI and going to AGI. If they start, uh, uh, something to copy or in-- that, uh, if they're really getting out, they're doing this bar in deep research, because someone on their team said it was more fun, that's, For example, in Gemini, and people in Gemini use it. I mean, I'm gonna assume that this might have happened. Or they're producing some kind of product. They're, like, now out of Codex. Uh, this is an agent proposal you're launching on your descope. But obviously, I recommend that everyone try it. It's probably just a Mac right now, I think, right?
As long as Mac is on, but it's gonna be another platform soon, yeah.
Yeah. This is a much broader subject. If I was saying when Anthropic released, uh, the folder infrastructure, there, on my computer and some stuff, I said it was some kindergarten, not really interesting, then that's where it's on the computer and some kind of stuff. I think it's very serious. The fact that there are supposedly different agents who will work in parallel can be launched, and that there is a huge amount of action. But! If they did it in front of Anthropic, product, well, whatever it was to release something to release, it's because Anthropic has released a new Claude updating what the version is now. Five? What? Four.
There's four six. Opus four six out.
Yeah, Opus four six. And then, ChatGPT releases, uh, 5.3 Codex, which is basically a little different. There is, uh, an annex codex for Mac in terms of agent infrastructure, and 5.3 Codex is what you can choose as a model for dealing with agents in the creation of agents, right? And API, including. They're gonna let it out right now. I mean, if this is an annex, it's a hang-up, a counter-up and a hip, and people will spend their time, those who love OpenAI. I'm gonna spend my time, for example, dealing with this system, and it turns out that system is shitty, like many agents in ChatGPT. Like they were making a reminder. It was a fun subject, not a development, and it died. They let the first agents die, and the second is dead. Then they released the agents, said they were working. They worked shit, they died. I stopped using agents at all. Now, I'm gonna try, and of course, even such spectacular consumers and loving, OpenAI lovers can call for me to laugh at some point in order to set up infrastructure. Google in terms of the agent infrastructure. Especially since Apple will use Gemini for Siri and will clearly use the Gemini model in his own place. I'm like the purple Apple user, maybe they're in macOS integrating something from Gemini, and maybe, uh, easily migrating under Google and telling OpenAI, "Get out!" So I'm gonna pay Google Ultra, I'll pay Ultra, I'll get a bunch of extra Ultra. They have YouTube in there. Well, a bunch of stuff I don't need, which I've got somewhere, you know, somewhere else. Videogenary is all.
And I say, "I'll sit with them at all."
Furthermore, Google will sew the floor, as it says, and even prepares a chat transfer from one service to another. I mean, probably, well, it should, I think, show up, but it's probably gonna be a function like that that, you can get your ChatGPT chat off to Gemini.
This is a super cool function!
Don't lose them. Yeah.
That's great! Because ChatGPT can't analyze the archive with its own chats. I mean, if I put my chat out on my own and say, "Do me their analysis," he says, "Look at the big archive, I can't analyze it." I mean! It's too big an archive. So these are the chat rooms you keep, you know, inside. And I'm saying, "Let me see my chat." He says, "I have no access to them to get me to your chats. You'll make the discharge." So how can you not have access if you have a tie on your back, that you're learning from them? So this whole, uh, architecture, it's on the one hand, of course, plus, on the other hand, it's starting to cause more problems. And this hemorrhoid folder is stupid, yeah, this is on the left of the projects and the inability to arrange chat. I had one thing, there, chat rooms, I don't know, uh, ten-twenty a day, and they weren't very connected. Since models have improved very much now, I'm starting to connect chat rooms. Or since I've been using it for a while, I don't want to re-examine the same thing. I don't know, Ilnar, you must have that, too, right? I mean, I have things I don't want. I don't want to tell you again what kind of dog I have, I don't want to tell you again what kind of sa-sada I have, what plants I've got, or what fertilizer I have. I don't want to tell you again what business I have. I don't want to tell you again, that's what this business has a system of analysis of staff like this, and these are staff members. I don't want to tell you that I have these projects here, there, I don't know, on video generators or on the rolls. I want her to keep using these chat rooms, right? What I have, I don't know, is this kind of stock, and here's the kind of travel, right? I keep my line-- some, uh, I have, uh, lots of tracks on trips, all over. I'm already going on inside these chat rooms. If Gemini does that, of course, it can give, uh, a topic of migration to me personally. Given that I'm still a Google mail user, a Google disk user, I don't know, 50 TB or how much, right? I'm like a good Google user, including. I'm forced into something forced. I'm still counting on Apple, right?
Which means it's gonna go down, something's gonna go down and, uh, come down, it's gonna come down, so something's coming from Apple's sky.
You're so patient, Sasha, actually. You're not giving up.
I don't have a chance! I have so much technology to do here. I'm still waiting for the new Apple monitor, which will cost another $10,000. Yeah? And then you're tied to Apple's technique. How are you with her? Like these MacBook phones, iPad, all computers, all Mac Studio. How do you get out of this? You're staying in their infrastructure, which they don't recycle so much. Yeah, well, maybe there's something in the world that changes, and I'll stop. This is gonna get the OpenAI device out, but it's not a big deal. By the way, I migrate in person. Well, write who's in the system. Let's do the poll again. We're doing it every quarter of a time. Who in what system and why do you migrate from one system to another? I'm willing to migrate to Gemini. If Gemini starts to analyze my conversations over the day, then if it connects to something, then gives me some kind of technology to make it very easy for me to throw notes on him, and he's the one who created a super-crossing theme. analysis. I mean, what's in independent programmes now, if it's part of the central chat system, I'll probably see, well, I'm guaranteed to buy this system as expensive as I can. If Grok did this or Gemini did it, if Claude did it, they're not gonna be in this direction, but suddenly. I know Meta bought Manus and Manus all of this. I'm not interested. I'm not going there. It's a side-by-side for me. Where, by the way? What did they release? The rumash, the rascal, what were they supposed to let meta out there?
Avocado.
Avocado! Here's avocado.
You're on the wrong side. Look, you never thought about this, how many of us are using these chat rooms, and they're more than, well, they have a lot of information, first of all, about you, about me. And when you're moving, you migrate to some other system, you have to start everything new, and you, you have to learn it again. How do you know who you are, where you live, what you're for, and so on. I mean, the longer you've been in this system, the less likely you'd want to start from
scratch.
Well, on the one hand, you're right, on the other hand, GPT doesn't often say that I'm a girl or he doesn't know where I live, or he doesn't know that I have kids or something else that doesn't want to share. If I asked a question and wrote about a woman there, he might think you're a woman. I mean, not to mention that he has memory, in memory of some allegations, he's doing me some polls. I mean, if Illar says if Gemini does a system that can download everything, all this, all that volume, put it in, like, to pass his data. Because where else is my more collected data? Nowhere. The mail is all hats, the massagers are all hats. This is the real theme that I'm interested in something every day and getting back to you. And there's a very disconnected content. Yeah, that's a very cool subject. In particular, Google wins a lot. I'm, like, using Google cards. I went to Santa Barbara last night, and I, we were driving back, and I thought it was fun, basically, that if the system, well, we were still there with my sister, she had a ring that was wearing a pitnes, and I had this Whoop bracelet and talked about it. They're theirs. I thought it was fun if Whoop bracelet was connected to another place where I move with my real permanent geolocation. Who knows the geolocation? Well, Google knows because I'm in Moscow, I'm always using Google cards. I mean, he knows all my trips, all my movements. And in my cars, Google always works. I mean, there's Google everywhere, not Android infrastructure, but I'm opening Google cards anyway. I mean, and of course he will know.
And of course, he'll know, for example, my preferences here. If it's connected to chat rooms, well, it's a very large volume. And if I put my voicemails in there for a day, I'd like to add a lot. I have a lot of different, uh, sessions, people's thoughts on different topics of human development, business, everything, he would have, uh, a huge amount of information. But while these systems, Tanya, they're not capable, I say, they're working their own chat rooms. What do you want to talk about?
Well, Sas, by the way, there's a very interesting comment that, for example, my chat thinks he's me. So he's, first of all, thinking he's a woman, second of all, he thinks he's a designer. I mean, he's just saying, "I've done the designer." He says, "I'm a high-class designer. I mean, I wouldn't answer the client like that. I'd do it like that." I always wondered if it's just me that he feels like a man, a woman, a designer, or is that how they all get a person in the process?
I didn't see that, to be honest.
I didn't change the profile, but I didn't turn off that tie you're always saying that learning about my data might be connected, but it's really, I think,
interesting.
Stop! To learn your data is a little, but let's do another design. We'll figure out what it is to keep things from happening here. To learn your data, it means that ChatGPT can take your data... take your data, learn all of its system so that Elnar or I get the plus, well, in the skirts, in the skirts or not in the skirts. Plus. It's a little different than learning your data for your answer. To study for t-data for your answer. I didn't say to turn off that necklace. I'm telling you to turn it on. There's a little inner thing that tells you to use your chat in your answers, to use the memory of the pre-designed, to build things and so on. But I'll remember that ChatGPT has memory, too. Again, minus OpenAI. They're once a year, I don't know how much time, Elnar, they came up with this topic when you wrote "fuck." And remember her, right?
I don't remember anymore. Something's always coming out there.
She used to work fun, and then she died in the middle of nowhere, but she didn't die in the building. Well, they'd say, "This project is shit, we've removed it." Yeah? And he's, "No, no, no." It hangs there, and I go in periodically, and he somehow saved something. I've kept a lot of things before, and now I'm not keeping enough inside. I mean, it's still a little different. And our audience will write if you're using memory functions or you understand how they work. Maybe you wanted to do a separate edition about this, about the special prom. Because many people, I think, in this subject, don't understand how they work within the system in terms of where they're taking data.
I don't understand. I'd like to see the special edition.
Especially now, because of what I said now that Pro's theory isn't online, what does that mean? I mean, that's what it means. I was just doing a question yesterday. I had a trip alone, and I had to analyze a pretty complex structure. Well, I had a very difficult request when I was studying seventy different places at the same time in one table, there, with fifteen parameters, parole. And I thought Pro would be doing a story or thinking, because Pro's theory, if I don't go online, and I need some data fresh. And I finally decided to launch Pro a version for the quality of this, uh, impolite, na--- science doctors, there, or superludes. For the quality of the analysis, uh, by giving the data valid at the moment of time, yes, this second of time. Maybe we should take a separate account of the story. And by the way, let me ask you something in your models. I just let him in. I have a system that gave me an answer in one minute, nine seconds. I wrote to her, "Tell me who I am. All you know is check it out very hard, study it. Uh, give me the best answer of a broad view. Take everything you can." Well, she's decided, so she's gonna do anything she can do in one minute, nine seconds. What's weird, huh? Because she's periodically doing some analysis for an hour or, uh, 30 minutes or 20 minutes. And then she gave me, so I'll say she gave me, I think it's worse than she gave me six months ago. Remember, this topic came online. A lot of people asked ChatGPT what it was. Let's go back to the code.
Just a little back to the agents, huh? It's an important story to see what you can do about it. We'll go back to the agents. Eldar, can you tell me a little about the code in-- on the desctop?
Yeah, I'd love to. What do I like about it first, huh? Let's do this. Diskleimer. I look at him as the person who's interested in this in terms of programming. And here we have a Cursor development environment, yes, which we have talked about many times. Gugla has Antigravity. And now, the Codex has come out accordingly. And the unparalleled difference between Codex and everything-- you can't write the code yourself. So everyone else has a standard field for code entry, a possibility of communicating with chat, a possibility of joint work or something. There's no such thing. Here you can just access some of your folder to some repository, and then you write it down. Agents are coming in. Yeah, they can show you some changes, you can confirm them there, ask for a redesign or something. But it's a clean vibe-coding when you're just giving instructions and you don't write a code yourself. Well, you actually don't have a cod where to write this code. It's a significant difference from everything that was before. If you look at it as a tool for the programmers. It's just curious. On February two, it seems that Carpathians had a tweet of last year, with which the vibe-coding term began to be used actively. I mean, he's been with us for a long time, and actually, this week is just a year, like this term has been actively applied. And, uh, OpenAI, by issuing this, uh, codex as an annex, if it's considered in programming, it's apogee for now all that's on the web code. Because you're not writing, you're leaving the queries, the process's gone. As for automation, you remember, you said, yes, some reminders, the tasks that are being performed in time, something like that. This is very well done here, because you have a software envelope, which, accordingly, can be built, some automation, some recurring actions, reminders, something else. So, inside this platform, it's all pretty good at work. I'm sure there's no mistakes, no memory or something. Besides, you've been talking a little bit about this, uh, parallel launch, uh, agents for different tasks, different, uh, projects or even one project, but in non-exhaustive tasks. I mean, in this case, you can, uh, get as good as you can, uh, work on product development, some kind of data analysis. There are examples not only of product design, like, some Excel plate, presentation, something like that, some kind of data analysis you give, it works inside, too. About the parallel work. You know what it reminds me when you're starting a few, some deep research requests or something. It's just that it's been in different chat rooms, you start it, and here you are, you have an app that synchronizes all these things. Well, more accurately, asynchronous, right? Because they are, they are, they are, they are, and, accordingly, they can work on one project and on different projects. Yeah, it burns the tokens.
Tokens have now given everyone an extended limit, that is, even free--
Until April 2nd.
Yeah. Even free plans have the possibility of using this system, meeting it, trying to work with it, and the pay plans have doubled the limits. So you can get really tight in this story. A major difference, yes, that's how it goes. If you're in ChatGPT on the Web version of the subscription, you know, po-- you're almost not limited by the number of requests. Long time, at least I didn't hide the system saying that wait, your limit is over. With deep research, it was a 20-dollar version, and I haven't been seeing this story in a while. I think there are no such people anymore. Here. And here, if we're talking about Codex, you burn thicken. That number of tokens is limited, so be careful. It's probably worth a lot of projects here to start, just if you've thought about it well. And in the demonstrations--
No, look, there's free tokens. You mean, there are free tokens that are connected to the system, and then you burn a plain if you're fixed, there, a card and you buy, right, Jerna? So people can understand correctly.
I guess. But I feel like it's here when you're, like, programming or something, you're not using your signature, but there's a thickness limit to the tokens you're assigned, there, and on the payroll, and you're using it. free.
Yeah, I just want to say that there's not just programming here. So if you open the application on the computer first, you-- you understand that you've been clearing from-- you, you said from the start that I'm explaining my interest, yes, in the programming. But if you put it on the computer, uh-a, this Wednesday, I still recommend that we try to do it, especially for people who are in the subject of project management, company management, some roles are different, They're running no marketing, design, sales, anything. Why? Because even if he sees the original recommendations, he says, there, a probable classic play of z-ze-zemake, right?
Well, it's like
S-s-s-smake, huh? Well, it's a p-s program, of course. And then he says, "Get a single PDF guy who describes it." Or, for example, uh, "Do a summary, uh, my last-minute meetings I had on my pear." I mean, he can come in, he can work with the terminal, he can work with, uh, the file system, the files with your incest-- you have on the computer, and then he can do different things about it. interaction. I mean, it's basically a separate agent system. The way she still stood up, and why did I wonder if it would work that way? I don't know, Ilnar, do you think it's gonna be or not? That you can build it right now. That's how I used to, I don't know, make an agent to get me some information every day. So this is where this topic, uh, uh, why, uh, uh, OpenAI, was positioned? They said that, in fact, with our system, you could, there, replace, uh, sometimes five or six hours of manual work a day that some man does, some project manager who walks around, uh, Uh, that's it, uh, that..., well, well, well, somewhere, some reports are collected, yeah, somewhere, like, some reports. I mean, this is a very interesting subject. And we'll see, uh, how much of these, uh, are these, uh, real challenges on the key and how they'll work really well. And they also seem to me to say that they're very good at working, if they say they're in design, they're on the web-based, right? That this Codex is doing well-prepared sites, there, and so on. Well, for me, I have a simple case: can it copy my website, so create me one site and all of it? Uh, well, I'm not testing this, Ilnar. I'm not wasting my time, it's not gonna do it to me, is it? Well, I guess some simple one-paged ones might be making it there.
Yeah, and--
There he is.
...not just texting, yeah, just so people can understand. So he's text-- it's not just texting, he's doing the specific steps we told him to do. And he can change the files, he can start some tools or team. So he's still doing different things, right? It's very important. I mean, different actions. How you're gonna give him access to your infrastructure or how you're gonna give him access to the outside systems, well, that's already to be tested, yeah, that's already to be tested, tested, See how he keeps it and does it. But I'm sure I'll try to build some agents in parallel. Illnar's warning takes 100% of what it's worth. I mean, you're probably gonna sleep. I remember, Illnar, you know, back at the bay of the ChatGPT when I showed up, I set up a module on API in Excel and connected to, uh, there, to my tokens. And I remember, I made a few inquiries, and I just burned all the money in two seconds. Well, from my point of view, my request is wide. I'm like, "I'd rather do it inside ChatGPT." Here.
Look, I get what you're talking about. There's no way that your card, which is tied to your signature, will be taking off some extra money. No, no, no, no, no. You just have a token limit you've been given. Then, as a result, it's likely that he'll be bought--
Yeah, I'll buy you some.
...to tie a map or something.
Buy it, Ilnar.
Yeah. But-- yeah, yeah, yeah, yeah, yeah. But there's just a limit that's now, well, available to everyone, either free or just on the subscription.
Yeah. That's it. It's just, uh, it's all clear here, because I haven't tested this story yet, and I'm definitely gonna have some extra volume, anyway, right now, tokens. I'll eat them very quickly, knowing my agents. And, uh, what I'm interested in, and just look at the amount of money that I'm worth, like the agents, huh? Well, I'll let them chase, let them chase, right? Even if there's $100-200 a month, there's more to do. Write, by the way, what would you like to do with the systems in terms of agents, huh? I think Ilinar is a great idea. I mean, whatever you want, if you don't talk about, programming, because programming still has a big life. Not about programming, what would you like-- what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what-what- And I recommend that you look, uh, that's gonna be fun. I'll bet, and, and all the viewers will read it. I, I personally think, always like agents that are concerned with some of my business or business inside, yeah, where I can make some kind of recommendation. But there are businesses where the teams are still sitting, I can only recommend it. There's something small where I can get on with it, which I'm dealing with some business analysis. And your personal agents. I mean, uh, I'm just... I'm back again, always wondering, what personal agents I can do for myself in person? So, uh, where does this agent give me any value? Not that's about my business, it's not about any kind of cho-- work, there, or something. What would that agent do? So what would he do? I'll watch that. I'll see you in exactly a week. Bye, everybody.