Hello, everyone! We're on ToTheMoon. Technological news, Silicon Valley sites, U.S.A. worldwide.
They have therefore achieved such phenomenal results.
I'll use my negative experience. Well, I'm not talking about GPT chatting.
I'm sure it's different. It can't be that different.
I don't want to face a story in person when I'm out of my question.
He's been in touch with my voice from the beginning and knows where I speak and diagnose in my text. I mean...
Two dollars is not a sum to die.
But with negative experience, I didn't.
It's an interesting issue. To discuss Chinese models, including the only data on Tencent. Sasha is very interested in knowing from the Tencent model and Baidu model. Plus, you know, Manus, Manus, yeah, he's Manus, he's real or unreal, and what's going on there. And I'll share, including, because I finally got a long-losted pendant, I've been waiting for him for a year. How he works in AI, at least my conclusions on the skies, the ones I did. And we'll discuss in detail, of course, deep research and, in general, what's happening in models.
Sasha, look, you're wondering, now that we started the graduation, you're a very interesting subject about blocking, not locking, such speeches by American companies and American leaders that Chinese We don't need to let models, do we? You can explain this detail.
Yeah. I mean, there's another Ernie model from Baidu, which looks twice as different as DeepSeek. And it just shows us that a few teams, company teams, Chinese teams are very competitive on the foundation of the level, yes, we also did, we talked Manus. I mean, he's still not declared anywhere, but it's very, very, very much, as it is, nobody says it's a full bench. So people get... people who get access, some write a mix rye, but nobody says it's a bench. Yeah, I'll remind you that when the Cognition Devin was coming out, which, by the way, is already on the estimate of four billion, it was a clear conclusion that it was a bench. Yeah, that's a conclusion that was on, on YouTube, on, on Hacker News, et cetera. I mean, it looks like this Manus didn't use that dirty tactics.
And they do have more advanced resoning and long-term planning capabilities, yes, for the start of agents. It's like a direction they're moving in. Which has very much to do with it, of course, because I'm already having deep research on using what I'm waiting for it in ten, fifteen, twenty minutes, I can actually wait for it. and two, three hours, four hours. I mean, I'm not gonna get any more espionage on my camper. And naturally, I think in the industry people are willing to pay for that, for such requests, tens, hundreds of dollars. So, what I'm, like, upset is that OpenAI and I think Anthropic are trying to regulate Chinese models.
I think it's useless simply because the start-ups are always gonna be. The start-ups will use everything that is available to open source. They'll be a little fine tuning adding some extra filters over the top. And how do you say it's their model, it's their own stap, right? I mean, just Manus, it looks like where he connected and how resoning and the coding of the generation together, right? And they have achieved such phenomenal results. There'll be start-ups that will do more than that, and they'll do less. But somehow, the start-ups will always be able to smell different models, collect workflows. And I don't think the open source is possible to stop it. Yeah? I mean, I think it's gonna be very complicated and unfamiliar, why waste energy on it.
Moreover, as we know, if anyone who represents America in mathematics, programming, it's mostly Chinese. And it is also known that, for example, in San Francisco, Google Chinese, OpenAI Chinese, they are living in, like, maybe not in one place, physically, but, nonetheless, in one room like comunity and one. One way or another, they share all the work together. So it's hard to expect that data don't leak in Mainland China, right? So, we're not talking about the camper data, the data is not leaking because, as much as the securities are working in American companies, very professionally. But the ideas and the kind of Lørnings that cost tens, hundreds of millions of dollars, they certainly are leaking. I-- I'm really hard to believe that's not true.
What do you think about the system spread? How should Chinese models be restricted or not restricted? And you're more focused on your own systems within your country and using the systems you have in your country? Or do you try to use the best solutions you have and find the best solution now? Or used to one thing and used one solution? It's interesting to know in the comments, please write. It's, uh, such an important choice.
I think people are very different here, differently.
Oh, well, it's probably here, yeah, about security a little gesture-- well, tough offtop, but I think it's super important to talk. And that Riplink filed a case with Deal for corporate espionage. Oh, that's, uh, that's really amazing what's going on between American companies. So, yes, it's said that there was some kind of access to internal databases, right? To understand how clients use Riplink, what they have, what degree of satisfaction they have, and so on. Yes, so that Deal could potentially tag these clients, it's a slob-day thing. I'll tell you soon, yes, there's probably a few lines, huh? And it was standardly considered a little gray, sometimes sued. It's when companies use designs and some public stuff, right? I mean, I'm saying, someone's testing something. There are some tricky ways to track who's testing, yes, there, through different spai service, and so on. And then these designs sometimes, so to speak, steal, sometimes change a little bit. And it's like what happens in the industry at both the start-up level and the big companies like TikTok, Instagram, Snapchat. And that, well, that's kind of a legal thing to pay for, but it doesn't usually get to court, does it? I mean, it's usually trying to figure out these things. And if we're talking about, and if we're talking about, uh, this industry, and if we're still talking about... AI, there's no point in stealing anything specifically.
Well, this dile situation, with the rpling, of course, is very disappointing.
Look, I'll be right back. You were telling me the subject, you said you were ready for deep research requests to wait there for ten, fifteen, twenty minutes or even a few hours. That's a very interesting comment. I'm a deep researcher now, and I'm very often used. And it really realized that I actually changed my model of interaction with LLM. So you've been waiting for an answer instantly, you've asked, you've got the answer right away. When I started using O1 Pro in reasoning, there's a few minutes to wait. ChatGPT allows it to work in the background, you know, like many systems. I just loaded it in the background. I was notified that it was there in three minutes. And with deep research, yes, you're going to be a long distance. And I'd probably be some things, maybe I'd be ready to wait a day. The question is, well, what quality will be given just to make people understand the case. Sasha, you're in deep research, and you're gonna bring some examples of the boxes you're using. I'll give you some examples of my own, my briefcases. I don't know how much Tanya and Ellery use deep research, so if-
I'm using, I'm having negative experiences.
Negative experience.
It's also interesting, by the way, negative experiences when you get started, tell me later.
Sasha, what are your case files?
Yeah. So now, in the generative AI, there is a constant convergence and as many as the media, they are not public at all, are they? And for example, there's a lot of companies behind every big Marvel movie, selling studios, right? And they're already actively using AI. They get it in a very wide eye, it doesn't come on Twitter, but
This is not on Twitter, but it's, for example, the largest, well, one of the largest VFX sales companies DNEG. She bought Metaphysic, which was just making generative AI decisions, you know, bye-bye generative AI decisions for, uh, babing, lipsing, etc. I mean, to do a full-fledged market research, and, uh, very, very, very, very, very, very good.
That's the one-specific thing, yes, that's the example.
Tanya, did you, uh, get wrong?
listened to you about how wonderful you found a coach for your dog, and I decided to find a couple of techs that I've been missing for five years. Oh, and I'd rather be there, uh, a certain area where I don't have too much to go because I have a little baby. And, well, all he found me, although I gave a lot of extra water, I didn't get the whole thing. And the level of search was, uh, about a plus-minus like, that I would be given Google or Yelp, which is-- that's very weird. I watched him this reasoning, and he'd literally check three websites. I don't know, I don't know. I went in and looked at them, I think why he chose them to give me that kind of answer.
Deep Research, yes. Then, uh, uh, about ChatGPT, I'm not talking at all. We were watching, uh, movie, and I was wondering if he got Oscar or not, fast. Here. He told me that there was actually no movie, he didn't go out on the roll. I'm like, "Oh, I get it." I mean, it was about the movie--
No, it's okay. The base is trained.
"Sube station."
Did you connect the search on the Internet or did he not search the Internet?
He just didn't look, probably, yeah. He probably didn't look, he gave me a reply. So I need to re-examine, check online now.
Yeah, and I told him that Oscar was actually going through, and he was a long time ago. He's like, "Oh, yeah, sorry. He got one of those, like that." Well, in fact, returning to Deep Research. Then I decided to compare certain products, and he also made it so superficial, because I-- I decided to do it with those products I know well to assess how much he can. I'm not gonna be useful. And he did it very superficially. I mean, I don't know, maybe there's some extra building, something else, but I'm just, like, a regular user at the household level, just pumping into this Deep Research button. And then when I was there more than anything, uh, he told me I ran out of my free attempts, so that's all.
And I'm a little disappointed in this paradise.
Sas, what's the chance Deep Research, like, has a $200 subscription different from Deep Research for $20,000?
Oh, I'm sure it's different.
That's great, isn't it?
Judging by the expirinence of me, it's very different.
I'm sure it's different. It can't be that different, can it? I mean, OpenAI loses a lot of money. And now I have a request, like a half hour, like this. I'm not sure they're, they can't even ask for one of those, I think, free, which lasts half an hour.
Oh, yeah, and here, all this stuff is saying that, well, I think Tanya's right about something, that there's a lot of criticism about OpenAI that their search is not working well. It's just not the websites that find, it's just the wrong sites. And, well, it's still me-- well, I mean, I'm a lot of positive, positive news about Google, right? That Gemini 2 came out that Gemma was the only source of them, that the reasoning model is already on the rise. But for me, and I don't see any of these practical applications. It's kind of great that Google wakes up. And now they're Vis Security's company now, and they're gonna buy for $32 billion.
I mean, they're moving, of course, you can't say they're not moving, but something's up until I see value.
Finding some information, if it's not just, you know, there's a question-- answer from the series as something, and you want to find a topic globally in a big cut. It could be math and programming. But not just the answer, yeah, there, like, write me a code or something, but you want to get the whole concept straight, so I'm the Deep Research guy. And I like the results in general. I mean, here, well, on those briefcases I tried, I got what I got. I mean, compared to the usual o1 or the base DeepSeek with the discussion, that's the option when a large structured
text comes up, I'm fine.
Here is an interesting one, in fact, that we are now clearly at a time when it is very difficult to make such rapid comparisons, comparisons and conclusions. I mean, I know there's some math baccalaureate, programming. OpenAI officially stated in its blogs a few days ago that AI would be better there by the end of the year for the programmer. So everyone's programming. Y Combinator said how many of them did it? Forty percent of students or how many, what's the number?
Ninety percent of the code base is written with the generator.
Oh, ninety-five percent of the codes, code, yes, the base is written with AI. The question, uh, what a common user now gets is how he really understands in these tools, is capable of correctly putting different questions in there, and he can also understand, is worth paying for it. money or not. I mean, it's like $200 a month. I have, for example, my sister--- when I asked Sasha ten, I still need to buy two hundred dollars or not? Or I insisted on my company directors, on the company, that they should buy two hundred versions at work and start using them. I told you to get right through yourself. Well, there I see it's a different amount. O1 Pro, obviously, it's not O1. Well, that's guaranteed. And especially when you're a lot, a lot of queries, or, like Sasha says, well, Sasha's very technically in there saying, "There's a Deep Research, so they can't even just, just, just, pay-paid. I'm telling you, there's $20,000 worth of it. It's just that they're not, like, the point of this, it's kind of wrong, I don't think it's just that she can't work. Yeah. How do we make a decision? I, for example, my sister, when she bought two hundred dollars, I explained it to her as follows. I said I, for example, personally don't want to face a story at all when I'm done asking. I don't want to face a story when I have some kind of pause, I don't want to face history. I'm ready to go to the servers, there's a few, uh, periodicals there, yeah, ChatGPT there, but I understand these things. But I don't want to face the request when I'm writing like, "You're out of your request limit," and I'm gonna need to keep going. I mean, especially on my specialty, I run a professional day, there are dozens, hundreds of requests for different analysis, and there's a very, uh, very specific volume. In the course of working with the tables, with some conclusions, in checking texts, in finding all the patterns, in examining something. But I wonder if Deep Research is still more on my private personal requests. Why, Sasha, I asked you, are you professional? I'm just curious that I'm still very professional, I'm very much using Deep Research, but my main concentration in my normal life has become. And I really get it. Today, I think you've devoted a very cool thing to this new concept of waiting in time, which is totally adequate in terms of life. Well, I did, I did. I'm running a parallel analysis. Here we were doing the survey. I came in, I looked at the findings of the system, why not, right?
And I've been scrutinizing them.
I'm a user experience, I've got some kind of total garbage. No, not a good answer. And he says, "You've got an end to your attempts. You can buy a bill of $200 a bill of sale, yes, well, a more expensive version." In fact, I can be like two hundred dollars, it's not a sum to die. But with negative experience, I didn't do it because, well, why? How this system needs to be redesigned somehow. That is for the user experience to be more positive.
The user experience is clearly a problem in the system. Yeah, it's definitely big, big pain. Yeah, we're not gonna argue with anyone. It's like a movie. It could have written that the film was out then. This request clearly indicates that he could have had an outdated data. He could have given more detail that the data were outdated. Do you want me to make this request? I was supposed to run it myself, make a request. We'll be there. Well, I'm sure that's what's gonna come to, and the systems will be quiet to handle such requests. There will be no glucs, there will be no such problem as just some of them. And how much does that affect the business of these companies, how much they might lose, not lose their customers?
That's the question. I mean, I still have a business question for me. There's Perplexity. What will they do when everyone develops and--
When everyone develops more and more. So they, uh, someone's gonna buy them or they're gonna be an entirely independent, huge player like OpenAI. A, and-- although it is not clear how they are doing by their base, especially given the vast number of different systems that are emerging, including different Ki--- different Chinese systems. I don't know how this is going to happen, of course.
Well, for Perplexity, that's good. I mean, they're very much behind OpenAI in their absence, they're not foundational models, yes, that's how Sasha says. So they have, the OpenAI has its models that are different from what's on the market. And before the DeepSeek appeared, it was still, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, well, it was a lot higher than the middle market. And Perplexity is clearly dependant on all the others. And when the open source of models emerges, which they can deploy there and their servers, if necessary, and somehow, they are getting closer to such essential players.
They're theirs, they've been doing their own. And you have to choose, yes, where you like more, where you have quality--
How come they have so much money to afford to do these tests, huh, so much pressure? That's an interesting thing. I, uh, Polina, my wife, says her friend uses GPT chat in Russia through the bean. I say, "So, across the belt, it's a different quality of the answers. I mean, you get access through the bot, on a parole, to cheap infrastructure. Yeah, you can pay a part of the stuff, but you still don't get access to the road." You, Elnar, are right, of course, they've got, like, these models, how far they can use them and get this economy all together. That OpenAI still cuts the incredible amount of money there. And this game is, uh, obviously-- again, I've always said that even now I'm paying two hundred dollars, I'm eating these two hundred dollars in their place for a day. Yeah? Well, well, not in a day, but in a couple of days, I'm eating them. Sam Altman has said on several occasions that Pro is a lost version. They're officially talking about Pro's loss.
How do you keep up with this and work? That's a question.
Yeah, it's hard here. What they have, we started talking about Sora a little bit, right? Sora is very hard to use. It's an opportunity to sell Pro to the most professional creators, right? I mean, for the most, most, most. A-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a-
And I'm really curious, really. I mean, in the 20-dollar version, so now Sora's available. I came in, I was stiff, but I couldn't even think of some case, that I'd be interested in being a cenerator, right? There are some samples that are shown as examples. I looked at them and closed them back. And I was wondering if you'd tried it. Uh, did something come out of this? Because I couldn't even make myself a banal text write and sgenerate. And do you know any of the cabs where people use it, right? Or, Sas, you only say in some specific thing with the transitions, there, with one foot on another, someone does it.
Yeah, yeah. That's 100 percent. That's 100 percent, right? I mean, basically, even the theme- theme of social media use case, yeah, that's where the most video-- people are making videos, right? Oh, I mean, it means that people used to pay up to $100-- to $100 for, like, five-six programmes they use, right? There's a Videoleap, FaceTune, CapCut, et cetera. Oh, and, and in my opinion, it's not, uh, uh, well, I think that's how it is now when you have a second of the video, well, round, ten cents, well, six even cents, right? Too won't be able to make the economy.
So when it's worth a price or less, then yes, it's gonna be possible to count on a wider adoption, and so, um, Sora has a very heavy interface for such use.
Look, from the point of view, Elnar, what you're talking about on video and your use, I think it's important that an important story that... it's very important for yourself to use all systems when you really need this case, not... I'm not sure how much fun I am. I mean, if we use it as a doll, everyone will stay in the GPT chat room in the poem, uh, and, uh, some story there. And remember how much it used to be, and many people have left it. I mean, it's like a prank. You don't use it all the time. To be truly involved, you have to make yourself go straight through a lot of real cases and real things. I mean, if you work with-- you do, there, the covers, uh, you use models, there, and Google is different. There are really cool things right now that Google's last, uh-uh--- Won's got the water stamps removed, yeah, yeah, yeah, um, pictures, images.
Yeah, but, Sas, I'll be honest with you, yes, that's for any use of the case, but, uh, when you need anything else to explain, externally, you know, I mean, any presentation, any pitch, like I, you know, So much value for the ears of all these systems, great value.
What about the video? I mean, with pictures, I understand, but with the video, Sas--
Same thing!
- You're
in the presentations?
Same thing. If, if it's just minimal, Elnar, less of the case. For example, we've been trying to use different boxes in different projects, and there's a micro-stop, you can, for example, generate a few seconds when you have to play that box. Or, for example, we at The Moon could take and make five-second, seven-second videos on our request. Question: Are we sure we need them to do this or not? I mean, if the editer sits, it's a very simple example, and it's gonna be sitting, like our craftsman, and the mounter's gonna add this thing, well, he's gonna add this thing, he's gonna sgenerate. Just like the people who put out the posts, I know, in Facebook, in LinkedIn, they're putting out posts. Like someone's making pictures. I don't, for example, generate pictures. I'm not interested, I don't need that. Well, I mean, I don't see any point in supporting my content. I like to back up the real pictures. And, for example, from the viewpoint of the video, we in the mounting are backing up this real thing. I know companies and I know people who support, generate videos in these systems. That's it. They're also in advertising, someone in advertising. We're in advertising, like, real estate-tested to do things like that. Then we see that, well, quality isn't that we're not interested or what we're looking for. And you should spend some time on it, and you should be doing it separately. It's much easier for us to do differently. That's it. And you stop doing it, you're not on it. I mean, I'm a lot of just connected to teams that are developing different video systems and videos, video stories. It depends on the individual. But there are people who use it. I just see there are people who generate enough microwaves of some kind of video.
If you watch the mass, I sometimes come to the Instagram account of people who are making video production. I remember that, you know, it was Viber 100 thousand years ago, and there were grandmothers, aunts sending such strange videos. I think it looks like a mass-- it looks like a plus-minus like this. I don't know, I think there's some super-tech, but their units.
There's a good thing, there's good things, even now, they put out, there's some micro-sized rolls, shorts and riels that appear in Facebook or in Tictoca, and it's even hard to distinguish, for example, animals. These are real animals or non-real animals. You start seeing the details. I mean, there's still micro things that look pretty good. I don't care about real animals like I don't care about watching. And I...
You're probably not really good at real things, either.
Well, I don't, I do, I do, I do, I'm in some kind of animal supply, and I'm interested in seeing, for example, animal behavior in a wild environment. I may be personally, I can see a video like that, but I don't need a real video. That's why. But these channels have subscriptions. Well, we all know there are subscriptions.
Yeah.
AI has sgenerated images not only, but even videos have a large number of AI subscribers. So people who consume that content, huh?
They like that content. That's it. Let's get to the subject. I want to touch that on, I think, this is a new yuzkeis, a very new yuzkeis, it's not very common, I can't hear it. It's a skiis for this kind of culin, right? I mean, there's a corn, this calon has twenty-four for seven, all of it. What's interesting is that when this penis was released, I didn't know what the use scenario would be with him. He doesn't write audio. I mean, I don't get access to audio. He's been in the middle of my head and he knows where I speak and diagnose in my text. So he's transcriting me, he's showing me the text, and it says in that text that you said it. And there are other Speakers, one, Speaker two, there's a Speaker, three, four, five, fifteen, and so on. And he doesn't detail those syringers. Speakers' voices cannot be added. Well, maybe they'll do it in the future. You can add, I don't know, there's a wife, there or something. Some kind of permit. No audio. I think they're losing some privacy. I think they're not keeping it, though there's a discussion in the audio field, whether it's written or not. Because, in fact, if you don't write audio, you don't have another person's voice, you don't have it as a direct confidential body, that's the structure and direct confidential information.
So, speaking of the text, which is fun, I put it on this pendant.
What's fun? I put it on, I put it on, I've been wearing it for a couple of days, I think I'll give it to you. Although I bought it, by the way, for $600 when they're gonna get it. He's worth four hundred. I think there's a hundred dollars discount, $400. Sold. So, and I think I'll give it to you. And there was a signature, like a signature of $2, $40, $20,000, a regular $40 anlim. I think I'm paying for something again. I've got a hundred signatures that I pay ten, twenty, 100, two hundred, I don't know how much money I've been writing. I don't know, $00,000, I guess some signature's off somewhere. And I went, and I was here for the third day, and I have a lot of meetings, meetings, and at some meeting, I realized how cool I had this pendant. Because all I'm telling you is that all my assignments I give them, I can restore them. I mean, I clearly realized at the meeting that I don't care about the side of the place, people are sitting there, they've recorded or they haven't signed up, I can restore. And I was very important to me as a structure of certain thoughts. I came by, like, saying that I was talking about Google in the morning or that I was talking about a sales strategy yesterday or whatever we were talking about, or I had a meeting at 9:00 in the morning that I had on it.
It was happening. And he's got me a consolidated information, so he collects and gives me a whole lot of good stuff. But there's one nuance. Turns out, by collecting information for two weeks, the war-- paying anlim, because I was there seven minutes because I was talking all day and, well, at least ten hours a day, and, uh, So, there's too much talk, talk. I'm very much of a content-- talking questions- answers to people, and I'm very much in the middle of a different debate. I mean, there's a lot of cat-catastrophic content. And I saw the next problem. Turns out he doesn't do all my data. And I think it's one of the fundamental problems of systems in general. That there's no simple system for a normal person to keep him up with his data and deal with them. That's how much I've been trying to find these systems, I haven't found it yet. Everyone says they are. But then, once you're dealing with this real thing, so she's really there, so she's really going to be looking for it to really try to get in and find, really study. Turns out I was writing to him what I said yesterday. He answers me sometimes, sometimes he doesn't answer, and he answers it, if I give him, like, yesterday from 11 to 1:00, and says, "In this segment, do me an analysis." And that's kind of a thing, because I can't, like, take it and make it up. Tell me, seven days ago at 1350, I said something, didn't I? I have too many, um, meetings, too many details, and how much they're gonna do this cool search, it's gonna be a result of course. They're probably in there, they're writing a dysfunction and a clot. I don't know how much it's technically possible, but the skies were for me, I can honestly say, fundamental. Even to share with my wife, do I know what I've learned? I realized that I could now easily refer to any completely phrase. I got it all written down. Terrible. Terrible. Yeah, it's, uh, look, on the one hand, it's a black mirror and it's a servitude, there, man, and everything. I'm very glad I got into this case because-- and I don't think it's legal. Because, uh, no-- that's a big question about the rule of law. Because when you go with hours or go with the phone that listens to you or turns on-- listens, yeah, yeah, he does. Or Alexa's listening too. Yeah, let's-- nobody's saying that publicly that they're listening. Let's talk about it, yeah, come on. You know. Come on, come on. Now I suggest that we just get to the law, talk, uh, and they wouldn't sell, America wouldn't sell this company, burn it, burn it down, burn it down, eat it with the guts, would it? And they'll make billions of dollars just and eat the faunders and everyone else, right? We need to see what happens next. I have a question. Because it's... Tanya, what am I suggesting we discuss? I'm suggesting we discuss the law, but we discuss... I always...
It's like blocking, and Europe is blocking something like that, and they lose their starts. For me personally. I saw this super, I wanted to live this skirt. That's how the black mirror shows when a man walks, he keeps all the information in his eye. There's a video like that. It's interesting, yes, when he can bleach and re--- he's being vouched and said, "That's what you said four days ago." To survive this, to see what happens to a person if he's in this context, in such a way, in such a way as to use information that's what it is, right? And that doesn't stop any man I'm talking to, talking to me, totally. It's, uh, it's a well-known, famous subject that if a person writes, he still keeps talking, especially if they write everywhere and all the time. I had a question, uh, for me, like me with those data, where I get some plus or something, or I get, like, minus, a little bit of abundance. And I must have shared exactly what I've learned from my point of view, and I've realized the exact new skies for myself, and I'm very interested if they can. They've got a rizoning, they've got a rizoning inside, but it's not working the entire data set yet, I think. And I want to see what happens to this data set, with this data set. By the way, Tanya, a simple case, very nice. They're talking about him, well, they're just cabins. You sit there, you stay home, like, and you ask me what to buy. And they say you're gonna buy it. That's a pretty cool case, by the way. And you're kind of going to the store, opening up the app, and you're saying you should buy it. I mean, there's so much of it. So we've discussed, what have we just discussed? I mean, if you really see it, they're positioning it as a cool record book, such a unreal record book, if you don't need it, you've taken off and you've put it down. I mean, if you see that this subject, like, closed, you have no right to write. There's a law to know, yes, where you can write, and where you can't. And I had a case of interest in the States. We're in, I've got the phone and started writing. The man got the phone, started writing. I say, "I can write, and you can't. There you go. Where's the question? What's that? I really like that there's no voice. What do you think about these systems anyway?
A little, precise question. Tell me, please, the Internet is needed all the time to make transcribation happen?
No, not at all, not at all. So you're connected...
I mean,
the device is happening.
He's recording himself inside, and he's, uh, there's a memory inside. I don't know what the volume is. Well, look, it's a micro-memorial thing, it's not that small, you know? There he is. And the batteries are 24 hours, two. Well, I've been talking for a little more than 24 hours, and he's, uh, then synced up with the phone. So you open the phone, it syncs up and everything. I mean, you don't need the Internet. It's not much of a synchronized with an app, and then it's already doing it separately, separate analysis.
I'm trying to see where the transcribation is going. She's...
And of course, transcribation is happening on the Internet. Then the mobile application is tied up, transcribated, reset the data. These data, 100 per cent of them go to, outside and outside the world, are analysed, returned. It's guaranteed.
Yeah, returning to what you said, audio writing and audio is leaving.
Oh, no, it's written, I mean, it doesn't show up to me. That's what hit me. It's probably getting wiped out, right? I don't know. Look, I'm having a hard time. I don't know if they're keeping all the audio. Let them, okay, keep them. But what I read, what I read or what I don't have. Look, Ildar, it's fun, I don't have access to audio. I mean, this is really interesting for me. The fact that they gave the a priori access to audio and, in general, they gave you as a spyker, and no other people. That was interesting. He also understands in different languages, different designs. Well, that's obvious, right?
Mm-hmm. No, well, for the sake of justice, the challenge is to unblock the Speaker dialogue, especially if a few participants, it's not trivial, is it?
I'm not.
And maybe they just don't decide it because they can't, right? One specific syker they take, and the others can't. But what you said was interesting, and it could, well, be used like that or not as good as a skill. That's what we actually started, right? That we used to get used to having our messages very quickly. Then, in fact, there was a reasoning. And that's what we're all using. And there was a lot of other updates that we don't use. So they're pretty-- reasoning, uh, a pretty natural story, there, a time-consuming response, and so on. That's what you said, it's actually the same natural way that you can use, possibly in new applications, new devices. Because you showed some kind of interesting skies, in my opinion.
I was thinking, I thought it should be like a ring. Tanya has an Ora, right? Or I just don't wear it. I have it, too, there, I diagnose sleep last night. Or there are people with hours, like, walking. Clearly, the watch, they're closed, you don't see it. It's still here. From the point of view of the conversation, it's simple. I see she's probably gonna be made by some little girl. I mean, she's just like a micro device that writes itself internally and does it. Well, you're like some... And it's obviously not a phone, is it? Because your phone isn't always with you, it's not that accessible, it can be in your pocket or somewhere else. So, from the point of view of keeping data, it's not-- here, the key design was that you're always in it. They say there are two skies, there's a skill skid. They even have a tariff like that: data for work and data for life, right? So when you write like that everywhere, you write all your talk to the kids, you write your talk to your friends, you write all the talking to the people, you're doing it all the way to analyze it. I have a new number of communications because I can communicate on a day, there are thirty, forty, fifty people of different kinds, absolutely, and I'm a good keeper, like--
I'm, uh, lucky to keep the content, for example, right? I can even generate content, I can tell you a simple case to create a content. Making a morittelling every night is a basic task. Imagine I can write today what I did, do it this afternoon. Or write me six conclusions I made today. Or tell me, please, the four phrases I made today to scale, there, and human development. Or tell me what the main advice you gave to business today. He'll give me all this. Or tell me, there, make me the summit what we discussed today in the tusum. He'll give me the summit, I'll put it in there, in Telegram or in the dentalling. It's all yours. I mean, I don't need to go in there somewhere separate, send someone audio, watch it separately. I mean, it's a central assembly. I saw it very clearly on the content generation. I mean, for people like blogs, it's just an amazing tool. Well, just have a lot of different teams, and there's someone else out there who's helping out. But in general, if a man is involved all day, it's a incredible tool, just, uh, to create a content of your own. You can tell him, "Gener me, please, content of my words, not changing my vocabulary, and keeping all the details and so on."
It's a little like Ray Ban's other side of the glasses, yeah, with which many people are going to go...
Yeah, yeah.
- with built-in cameras.
This is where the sound starts. I don't care, and even though we agreed not to talk about it, the question of security, well, it's not much of a secret to me, because you still have audio flowing the river online, and you're gonna start breaking up, Listen, there, you don't have to go on like this, re-sell some data. But the direction seems interesting.
I have, I have a-- yeah, I don't-- I don't say, by the way, that I'm not talking, I'm open to talking about it. I'm saying that I always have a fact that I have. I'll give you an example that it's easier to hack this meeting, a recording from Riverside or hack, eh, Limitless? Which is easier to hack into my post office in Google, where there are a bunch of records, or Limitless? What's easier to hack my iCloud into Apple or something? I mean, I use so much different services, I keep so much data in notes, different designs, everywhere. How many of us use different CRM systems, banking systems, a lot of things. There's so much data everywhere. So any system can be hacked. What's to break in? I have cameras at home, standing. Can cameras access and access systems? I guess I could. And you can access the server to the house or cloud stores to connect inside the line. Or the TV and the camera on it. Samsung has his own, Sony has his own, LG has its own. I mean, here, there's a mom-- easier my car, so I can get my car on, and, uh, one in the other, there, and then I can get in there, see that we're in the car talking to date. Or turn off the car or get a schedule. I have all my ways, travels write because I have a remote opening, everything is done. Get my whole track where I went, where I looked. God, how many applications do that! So from a system perspective, I'm treating this system as just one of two hundred, maybe three hundred, or four hundred systems that I'm writing forever. I see reports every day, you and I see the re-swept of different systems. They always break someone up, they leak some personal data. I'm always getting some systems. Your passwords, uh, are compromised because some base is regular, where you once left passwords a year and a half ago, someone scattered and swept.
But we live where we live.
Well, for the sake of justice, the start-ups are much easier to break. Especially the news we said about Y Combinator having ninety-five percent of the code base being generosized, it's clearly--
No questions.
- it's a little less quality. Yeah, that's a lot of talking about, too.
What, look at where you have a vault, again, where you have a data repository that securities use. If they use your private code, it's a separate subject, how they get it. Well, that's a lot of things. You know what history is? It's always been a story in Telegram that people think Telegram is safe. Well, Telegram is safe if you're making private chat, right? How many people know that Telegram is safe and the data don't scramble when they're making private chat? Well, one of the thousand, I guess. And ordinary people don't know what it is. What I'm thinking. What's so secure and whatsApp you have in default like a sippro-crypt, and it's kind of a default code here. Here, here, I think that an important subject for any man is a little more to figure out and see where you're at risk. For example, for me, I'm giving my full voices, my own. You want to show me all the conversations I have in the day, I'm ready to open them and keep watching where they can be used. First, I know there's a law, there's jurisdiction that can't use my data if they were stolen, and if I didn't-- I didn't agree, and accordingly, I can't show or produce anything there. I'll do something. I'm willing to sacrifice for, for example, a certain technology test. I know people who don't get home that Alex doesn't get to be installed, and they don't go out with the iPhone, they go with the button phone and turn the phone on, just when they call, and they change it every week, right? And Tanya met them, too. I've met a whole series of people who have nothing to do with them, and I don't have to be present at any conversation or communication system. Well, here, I think it's a question of choosing everyone. And it's gonna be great if you write as an audience, what do you think about security? You, uh, sceptical about everything and you're putting everything in your head, and you're trying to figure it out? You don't understand anything at all? Well, you don't even know where you can write, how to do some construction, how to turn on, turn off those constructions on your current phones. Why does he listen to you, you're showing the same commercial? Or you, uh, understand that risk balance, right? You choose where you risk personal data, yes, and where personal data don't run any risk. I mean, that's what you do in terms of risk systems, that's very, very good to know. I'll tell you soon, yes, that it's possible to start with the OpenAI, and defolt is a setup that is used to improve the model for all. Oh, and that's a way to turn it off. How many people are off? Look, there's a big moment here, how many people are shutting down.
That's news.
Sure. The only place where the data is not default--
Tell me where to
turn it off.
Well, there's a set in there, they're default data. So Semenar and I had a workspace, and in workspace they say we're not on this, we're not learning, but that we're not learning, it doesn't mean that no one can access these data, can't have access to them. I'll see. And Sasha just gave the example of a company there, Deal, which I use. Sasha, this is Deal, which reports, yes, sells and allows payments to do anything at that time.
Deal is a pay-as-you-go.
Yeah, I have, I have two companies, two corporations on this system or three, and three, three corporations. And we use that data. I mean, I know perfectly well, well, maybe in theory they can do. I was the CEO of Megaplan. We've had a hundred thousand legal persons in there, and all their systems of the Task Force, the Task Manager or the CRM system, have been retained, but obviously the Salesforce sees the sale just around the world, yes, here the world. Not only sales, conversations, profits, receipts, correspondence, legal and illegal sales, legal and illegal contracts, black accounts, white accounts, sees everything. Well, let's just, honestly, see everything. Or a 1C system. Well, people had one C, so they were in the 1C, and they were all accounting and one and that, which means they're doing the day. And here, as you know, I've always said, when about cloud service, how much easier it is to hack your own service at your house or to sellforce. I'm, like, using a lot of Apple vaults. I think, for example, that hacking Apple is more complicated than hacking my server. I'd be here if I had, put a server for a million dollars, he'd have a metal wall, it would be easier to hack than hack Apple, right? Well, in mine, in my mind again. No, well, then you must be paranoid and not write any data, not write any data.
Yeah, I mean, yeah, that's what matters. It is important for itself to clearly define the criteria for which this decision is to be made. Because the next generation of companies that have to build business will be forced to-- they will be more in-depth. So you imagine yourself, that's how the sales-- from the 1C to the Salesforce, you're gonna come in, you never get out, you know, just impossible. And that's gonna be more. And the next generation of AI-companies, they'll be more busy, more on the date-management, and they'll switch the cost. That's what I was gonna say. Now switching cost zero is practical, and in two years, AI will be everywhere and switching cost will be huge. So we know that, yes, for example, for one of the simplest business processes, you know, comparatively, sales, yes, there are a lot of different systems, and as far as we can't say, Salesforce is all there. Right, right?
And then, and when we see AI, using AI for more complex business processes or personal efficiency processes, it will be even more important to choose a proper, right system.
So, Sasha, can you explain a little bit what switching cost means to make sure everyone understands?
Yeah.
Switching cost, that's, like, when all the data came in, it's all on Google Cloud Platform, so we can switch the cost, but it's two, I don't know, well, it's a month, a man-month, right? I mean, you can appreciate it. Then, uh, that's how much it takes to move on to the best alternative. But sometimes it's worth a lot, sometimes it's worth a lot, and sometimes it's just gonna be very expensive, yeah. Because if everyone on the team has learned to pump some...
It's easier to pump certain buttons in some workflow, then it's gonna be harder.
Look, it's saying that in a normal life, those who will long be planting a man for some kind of diversity, a number of briefcases will win. Because it's time to migrate to the user now, if a person uses an iPhone, well, it's critical, right? Because when you use a lot of different services, you, you, uh, really, ten times, you'll think about how much you can turn it off. I mean, it's all about you, you're already on it, you're on it, you know, you're on it. Or I've got, like, ten or 15 Alex. I need to invest in new columns to get Alexa to change the system. I'll be, I'll be sure they're so cooler to put so much money in there because you throw those trash columns out, and you don't use them anymore.
I mean, it's such a hard-on design. But in some systems, it's easy to move. For example, while I feel like there's no LLM models, there's no company that would be so attached to it today that it would not be possible to move from it. I'm talking about physics. For the law-- for the company, I think so. Well, I mean, it's a condition that we forget GPT chatting tomorrow and migrate to something else? Or can we forget Grok and migrate to something else? You can, uh, there, that's all that's all. I mean, for now, whatever I am, I'm a very good user of GPT tea, right? But I know that if someone does something that is fundamental, I'm gonna see revolutionary in the movement and have, I don't know, more use case integration, I'm gonna move in at a moment, I'm gonna move from the GPT to this. system. I mean, I'm, like, migrating easily.
Yeah, yeah, yeah, yeah, yeah. And that's very interesting who can... So I asked about Perplexity. What do you think about migration, by the way? Are you linked to a system you can't migrate with? For example, it may be in terms of the creation of pictures or the generation of votes or, in terms of, I don't know, the transcrimination of a text, the generation, you know, video, in terms of the creation of something, maybe, Screaming or something.
I mean, do you have any connection to a particular model? How much can you or can't go to something else? And of course, it's gonna be a great guy who's gonna come up with a new interface. Apple, by the way, I think we should remember Apple today. Apple promised in the summer, and then promised in the fall that they would end in March, so turn their Apple Intelligence and how it's gonna work really hard. But I think they're failing. Here. I'm a Apple amateur of technology and everything, but from the point of view of this new use, a new button case when you migrate with this chat dialogue window, I think Apple is completely out of hand. Someone knows the case of use through this button, so...
No, no one needs it, no one needs it. Everyone's already forgotten about her.
Absolutely, yes, absolutely. And, of course, for Apple, I think it's just a phenomenal failure. And the question is, that's the thing they're dealing with, that they must have been scared, right? They were as if they were so scared that the system was so moving that they said they would be integrated with GPT chat, they would be integrated with all the others. They should have stayed in their standard mode. We're gonna decide ourselves when we start something, but we're gonna start something cool. Still, it's probably one of the first failures of Apple, so big in the past, because I'm a big user of different types of technology, and they're, of course--
Well, actually, don't know why I couldn't do it, well, how could you... why couldn't you, like, fully integrate the normal Grammarly? That's what I don't understand. Like Grammarly is a work product. Many people need to write better. Why couldn't Grammarly be integrated, like Grammarly do? And it's kind of obvious that there's, like, a, uh, quality thing, a model, a billion parameters. Obviously, you can do that. And, well, English is exactly the same, right? And why can't they at least roll out for the pyara that they have, like, a Grammarly, that's kind of a kind of a kind of thing they've got, yeah, built everywhere, right? Which knows exactly about you. Why would these stupid boxes with pictures I can't even understand? I mean, for example, well, Grammarly, you can do a little better, too. So Grammarly, who even better understands context and chaos-- well, what kind of tone you need, right? So, basically, Grammarly can categorize it in six different tones, right? But understand what kind of f-- well, how do you understand a little context, but not in terms of what it is, but just what use the case, huh? I mean, naturally, if you're in there talking to investors, you need to be as sure as possible, as you can, right? If in the family, it's just that you're clear, right? And so on. I mean, well, why, like, why, and for investors, who, for people who are strangers, need more formal tone, probably less formal tone for people with whom communication has already been done. I mean, it's like these, that's, like, a choice of tone, it can be built as simple as, like, some very simple rules, right? I mean, even then, just like that smart, just need a more advanced Grammarly for everyone, for everyone, for each tone. That's it. I don't understand why they couldn't do it.
I can't understand that at all.
Look, they let it go inside, and that Proofread and Rewrite, and it's a tone that can be chosen, by the way. But the question is, they could've gotten into this case and made this case great. He glutes, doesn't always work, not everywhere on the device, not all the programs and systems.
It must be accessible everywhere, like Grammarly made integration.
Yeah, they could have done it absolutely like the top of everything, so you could go up, everything about speech and creation, and take it away. And in the idea, Apple could have translated all the scenarios of speech from all the systems in six months. And all the quieter ones would stop using ChatGPT, Grok, DeepSeek, anything. Because you're a ud-- I'm more comfortable. I can honestly say that I'm still a messengers of an iPhone, yes, or from, there, mail writing or notes, I'm not leaving. I'm often checking the text, yeah, when you're texting--
Yeah, well, he just needs to, so if they were still making different deposits in the iMessage, just so that the worker and the non-worker aren't in the same place, it would be super.
This, this, this is...
That's not real, I know it's not real. It's not a real development, but, well, yeah, that's how it is. And of course they could have, they've released a journaling app, but journaling annex should be made to voice AI, obviously, it's time. Something like that is talking to you, so you don't know. You were interested in the subject of self-development. There, like, what you learned in a week of positive, some kind of positive affirmations you're grateful to, something like that. Well, shorter, I understand that you can develop these skies in your CI. Such a journaling, so super base, huh? But no matter how much it's done. Journaling app is a bit ungrateful.
Yeah, I don't think they'll give it that fast, yeah.
Yeah, and--
No, the iPhones don't give up on--
And there's no need to roll it out now, even if they can do it technically. Here's the 17th iPhone, new sales will say that we did this and will raise the next sales.
Yeah, but you see, Apple's still trying to avoid that thing, too, because there's more to do-- first, by the way, they didn't have a great sales, they didn't have a great deal of sales. And there's still a subject that when the 17th iPhone comes out, and they'll say we have another thing, and it's gonna be so specific. See you later. I'll see you on the To Moon Channel in exactly a week.