Skip to content
Transcript

Transcript · 201 · GPT-6.2 a Letdown? Claude Advances, the NVIDIA Crisis and a New Tesla | AI News — ToTheMoon

English machine translation of the Russian-language episode. Timecodes open the source video.

Episode overview
00:00:00–00:00:50Episode Highlights
Ilnar Shafigullin✓00:00:00

6.1 got really dumb. I'm thinking, what's this crap you're writing me?

Alexander Volchek✓00:00:04

I don't know about you, I just saw the harshest degradation of OpenAI's models, of the ordinary chat. And Codex has become a terribly slow story for me altogether. What Claude Design does is mind-blowing. Yesterday I ate through a Claude Ultra plan, the tokens in one account, in a day. If there are no problems, then it means all the famous ones, Jensen Huang too, and everyone can be replaced, including Elon Musk.

Mentions: OpenAI
Ilnar Shafigullin✓00:00:32

Will ChatGPT help you?

Mentions: ChatGPT
Alexander Volchek✓00:00:34

Do you understand what this means for the world at all? Buying people a car without a license is a different matter! Are Ilnar and I going to argue every episode about new models coming out or not?

00:00:50–00:01:59Anthropic's IPO: What's Happening in the Artificial Intelligence Market
Alexander Volchek✓00:00:52

Ah, hi everyone! We're on the ToTheMoon channel, tech news and insights from Silicon Valley and all over the world. We're starting our Sunday podcast. Well then, Anthropic went public, today, just now, and the exact final price isn't known yet. And actually we need to wait at least, in our time zone it's usually around one in the afternoon, roughly, to see how they finish the day, and, and, of course, watch tomorrow too. Preliminary reports at the start said one and a half trillion, now they're writing 1.8 trillion. But Tesla went public at, I think, a bit more than two, and then rose to three, then slid back again.

Mentions: Anthropic
Alexander Volchek✓00:01:35

Plus right now the market tied to computing, to technology, to artificial intelligence is stormy. So I think that Anthropic's listing will bring another new storm, in no clear direction, and it's unclear what will happen with it. But there will definitely be more money in the market in terms of artificial intelligence, right? So, we can start with the fact that ChatGPT got an update, 6.2 Astra and 6.2 Sol came out.

00:01:59–00:03:25GPT-6.2 Astra and Sol Are Out: First Impressions and a Comparison with Claude
Alexander Volchek✓00:02:05

And roughly a week after 5.2 Fable and 5.2 Mythos came out at Anthropic. And to my regret, probably, in terms of what I read about the Astra and Sol model, and yesterday I even ran it a bit, uh-um, switched a number of tasks to this model, I'm somehow more and more disappointed in Codex for developing staff-staff code. And these days I even went in and studied which companies, in the US in terms of tech companies, use Codex and Claude Code for programming. Well, and the result is obvious, right? There, if you really look, it's clear everyone says they use Codex too.

Alexander Volchek✓00:02:53

But Claude, of course, is a super-leading company, and they even say that at Google they program not with Gemini, Gemini exists of course, but with Claude Code. That is, they're allowed to use Opus. You can see that Fable there is kind of, it's probably a separate, exclusive story, understandably, because it's very expensive. And although I really like this 5.2, over the last five days I've been really glad about certain things I've done with it. What are your feelings about the release of 6.2 Astra and 6.2 Sol?

Mentions: GPT-6.2
00:03:25–00:05:19Has ChatGPT Gotten Worse? GPT-6.1 vs GPT-6.2 and New Restrictions
Ilnar Shafigullin✓00:03:25

6.1 got really dumb. That's actually how I found out that 6.2 had come out. I started doing my usual tasks that I do with ChatGPT, and I think: "What's this crap you're writing me?" I go to the news-

Mentions: GPT-6.2 · ChatGPT
Alexander Volchek✓00:03:39

What happened?

Ilnar Shafigullin✓00:03:40

...I see 6.2 is out. Yeah-yeah-yeah-yeah-yeah. I switch to 6.2 and think: "Well, my dear one is back, the one that more or less handled tasks normally."

Mentions: GPT-6.2
Alexander Volchek✓00:03:48

Seriously? Is it really like that for you? So you clearly saw it, right in 6.1?

Ilnar Shafigullin✓00:03:51

I did, yes. That is, I give it an ordinary task we've already worked on. Maybe it happened specifically in this area, maybe somehow, I don't know. Well, I mean, I found out about it after I thought: "Well, what's this crap you're doing for me?" That is, 6.1 started to disappoint me on the tasks I worked on with it. And that's how I found out 6.2 was out, and switched. And here it's worth noting right away that 6.2 Astra, on some tasks I calmly did with 6.1, started showing off, saying: "No, I won't do this, this kind of violates my rules", and so on.

Mentions: GPT-6.2
Ilnar Shafigullin✓00:04:28

I hadn't gotten into such situations with ChatGPT before. Well, I mean, you can see that with cybersecurity, with cyber security, they tightened the model even more. And on those tasks where everything was fine for me, on 6.1 and any models before it, because I usually didn't get into such situations, 6.2 began refusing to do quite ordinary tasks for me, in my view. Yes, you can look at them from another angle and decide that maybe I'm a fraudster. But considering I have full access to the project's codebase, that's silly. Am I going to break my own project?

Mentions: ChatGPT · GPT-6.2
Ilnar Shafigullin✓00:05:05

Well, okay, sort of. But one way or another, with 6.2 I, first, stumbled on this, and second, I'll repeat, 6.1 really disappointed me in one place. And it turned out that maybe this is connected with the release of 6.2.

Mentions: GPT-6.2
Alexander Volchek✓00:05:18

Listen, well, in general, I don't know about you, I just saw the harshest degradation of models from OpenAI, of the ordinary chat, of the ordinary chat over the last two or three weeks.

Mentions: OpenAI
00:05:19–00:07:43Why ChatGPT and Codex Are Slowing Down: Speed and Quality Problems
Alexander Volchek✓00:05:32

I think it's connected with the fact that, well, I was saying three years ago that OpenAI gives too many resources and too much capacity, and you can do simply unreal things. Claude never gave the opportunity to do analyses so voluminously, so massively, upload so much data and exploit it and also not spend tokens, right, because the chat didn't spend tokens. Now ChatGPT has become very similar to Work. Maybe one migration has already happened, we don't know, right, we don't fully understand who has which version.

Mentions: OpenAI · ChatGPT
Alexander Volchek✓00:06:02

And Codex has become a terribly slow story for me altogether. I have the Ultra Fast mode by virtue of, well, Pro 500. It works eight times faster. By the way, in 6.2 Astra it's not eight times but four times faster. And in Sol, yes, I think it works four times more. And the ordinary Fast is not two and a half but two. But Ultra Fast eats up tokens, it kind of, it instantly starts eating up tokens. And before, tasks that ChatGPT, specifically about processing a large volume of information, I always liked that, yes.

Mentions: GPT-6.2 · ChatGPT
Alexander Volchek✓00:06:35

That is, a huge number of reports, a large amount of data, large arrays, analysis of various texts, for example, it did quite quickly. That is, in an hour, or two, you could analyze, I don't know, 300, or 400 thousand, I don't know, conversations, for example, right, or some materials. It would go through all the materials. Now it has, so, started spending time almost sequentially. Yesterday it told me it started spending time almost like ElevenLabs on recognition. That is, for its analysis I even wrote: "Do I understand correctly that we're spending time practically like on recognition in ElevenLabs, where separately we..."

Alexander Volchek✓00:07:13

It says: "No-no-no-no-no, it just seems to you that this is happening." But the quality of the work too. That is, I see it in terms of quality, specifically on complex tasks. It's clear, again, complex ones. Maybe it's habit, it's nitpicking somewhere, some, well, expectations. Ilnar and I have discussed this many times. Ilnar and I are the harshest fans, so, of expecting, so. And since you've already waited, then leave it as is. There. But it seems to me that a certain dullness has set in.

Alexander Volchek✓00:07:42

But I want to note that recently in Claude, instead of code, I just launched the chat in Claude, and the quality of execution was disgusting.

00:07:43–00:12:20Claude Code vs Codex: Where Anthropic Beats OpenAI
Alexander Volchek✓00:07:51

That is, Claude Code, in terms of executing a task, on the design of certain modules, did simply top-class work for me. And, by the way, Claude Design is closing for them. Let me remind everyone, you need to migrate from it. Now, if you simply write in Claude Code: "Make me a page design", it will launch internal modules. And, of course, it's simply mind-, mind-blowing. So, what Claude Design does is mind-blowing. I don't understand why use other tools for app design development, well, UX/UI, designing such things, instead of Claude Design.

Alexander Volchek✓00:08:28

But Claude, ordinary Claude, to me, of course, it's not ChatGPT, and we'll see how much attention they'll direct there. Anya, do you have any, were there some tests of 6.2, or haven't you looked yet? You have a lot of tasks, you just have a lot of tasks in Codex, right?

Mentions: ChatGPT · GPT-6.2
Anna Volchek✓00:08:45

I have several thoughts here. First, it seems to me that this conversation repeats in roughly every episode of the channel as soon as new models come out. So, it's straight-up déjà vu. You could even make a compilation.

Alexander Volchek✓00:09:00

Well, they're really showing a very cool result. That is, it reached the "holy cow" level.

Ilnar Shafigullin✓00:09:06

Has it gotten worse somewhere? Yes, it has gotten worse. I have exclusively positive impressions of 5.4. I like it.

Alexander Volchek✓00:09:14

Well, I noticed that ChatGPT 5.5 Pro became much cooler than it was. Then Solo went back to 5.5 Pro, because it seems to me that Pro does better. And it put together a hell of a study, totally crazy, right, which I read. ChatGPT, by the way, has been infuriating for the last four weeks. It gives an unreally cool answer. When new models seem to degrade. That ChatGPT started answering very poorly.

Mentions: ChatGPT
Anna Volchek✓00:09:43

So I get déjà vu every time. I just caught myself thinking that in these couple of days I seemingly lived through, over the last week, all these scenarios. On the one hand, it seemed to me that the models I work with had gotten dumber. And it was just some kind of disaster. On repeated, that is, there are regular tasks that I do, for example on a weekly basis, and exactly the same solution to this task gives me a radically different output, which I'm not at all ready to approve and accept.

Anna Volchek✓00:10:15

On the other hand, the latest new Fable or the latest Astra models seemed super, how fast everything became and what a good result it gives. But in one scenario that's so, in another scenario I switched to other models, switched at random from Fable, in Claude, or switched, for example, to Sol from Astra. So everything depended so much on context, on what task I'm solving. And the third point, this is probably the takeaway of this week, it's a rare case here when I destroyed my entire usage altogether, on absolutely all possible platforms, completely, and essentially just sat and waited.

Mentions: GPT-6.2
Anna Volchek✓00:11:00

I ate through even Pro, the mode in ChatGPT, and in Codex.

Mentions: ChatGPT
Alexander Volchek✓00:11:07

And ordinary ChatGPT too? And ordinary ChatGPT too?

Mentions: ChatGPT
Anna Volchek✓00:11:09

Yes, and the reset, I kept thinking: "Damn, well, I have Pro mode in Codex, I usually didn't reset." But here I, so, already reset this mode, because I wrote myself several, more precisely, upgraded a number of my skills regarding more economical use, especially when it comes to versioning, and more economical use of the output of results in general, and so on. Basically, let's do it nicely, work on design only in some approved final version, because I really began to see and feel that all of this can be eaten up in two days.

Alexander Volchek✓00:11:47

Anya, I want to tell you, yesterday, yesterday I ate through Claude, the Ultra plan, the tokens in one account, in a day. So I keep saying: in general, if I sit all day working and doing exactly the tasks that concern, well, automation or development, I would eat through Codex and Claude completely to zero over the day. There are chats, but mine is chat 500, there, I think, there are no, by the way, blocks, there are no chat limits there, but we're waiting for chat limits. So, blocks. Let's move on.

00:12:20–00:16:58NVIDIA After Jensen Huang's Departure: Who Controls the Company Now?
Alexander Volchek✓00:12:20

I want to recall, first, at NVIDIA, why else is the market stormy now? Because they removed Jensen Huang, after all, and NVIDIA's board ruled out Jensen Huang's return, that is, he definitely won't be back. And separately there are showdowns. Congress requested explanations about chip deliveries. In general, we expect all the showdowns in the artificial intelligence market from January. Our episode came out on Friday, I shot an episode about blocks of ChatGPT and Anthropic in countries where it should, where it's prohibited to use it.

Alexander Volchek✓00:12:52

Whoever watched, share your comments, whoever didn't, watch it, a cool episode. I told there about 20 really specific points, details, of what these systems block you for in different countries. So, Congress requested explanations about chip deliveries, regarding NVIDIA. And let me remind you, the US Treasury Secretary declared outright that the head of the company should be hired from outside the company. That's pure politics that's started. It seems to me that politicians don't fully understand how these businesses are run and how they move. This will bring, of course, such discord to the market.

Mentions: United States
Alexander Volchek✓00:13:32

And I think that this, of course, can create certain problems in the market, because Jensen Huang really had a huge number of personal agreements, really very personal agreements. We'll see. If there are no problems, then it means all the famous Jensen Huangs, everyone can be replaced, including Elon Musk.

Anna Volchek✓00:13:47

And what problems could it lead to?

Alexander Volchek✓00:13:50

What problems, what could there be? Well, he negotiated, look, he had a huge number of personal contracts, for example, on allocating certain capacities. And the question is: is this his personal, personal story that he supported and drove, while the board of directors, for example, resisted internally, or the team resisted, or does everyone there really approve it and so on. There's a lot of different movement inside. That is, right now at Meta, a scandal or not a scandal, there's a conflict over the allocation of certain shares.

Alexander Volchek✓00:14:20

That is, some people say: so, certain shares of a certain voting type should be kept, leaving power with Mark Zuckerberg. Others say: no, they shouldn't. And then came lawsuits, disputes, conflicts, details. Well, it's, it's a fairly serious story, who will get power. I think that, of course, at NVIDIA the question now is: who will get power? Who could influence NVIDIA? That is, how could one influence NVIDIA? There was Jensen Huang, who clearly, as head of the company, made fairly serious decisions.

Alexander Volchek✓00:14:49

He had his own policy, he voiced it, how he thinks, whom to sell chips to, how to develop them for whom, what to do, which projects to do, and so on. And, by the way, right away, I think this week, a ruling came out, a clarification about blocks and the ban on using infrastructure and chips in the world. Before, the infrastructure could be used, look, before it could be used, you couldn't, for example, sell certain chips to China, for example, or sell certain ones to Iran, or, I don't know, you couldn't sell certain chips to Russia.

Alexander Volchek✓00:15:25

But companies located there could buy this infrastructure, for example, in Singapore with these same chips, or buy it in the US. And kind of through a number of such chains it worked. Now a very clear explanation has come out, where specifically a ban was issued that you can't, that this middleman side can't sell it to these companies. That is, they're introducing a lot of restrictions on such countries. Although, by the way, I thought more restrictions would be introduced for all countries in general, other than the US.

Mentions: United States
Alexander Volchek✓00:15:57

But here OpenAI has already concluded a fairly serious agreement with NATO countries and with NATO, and, in particular, the deployment of their systems, well, into the military industry and so on, well, very serious ones. Therefore, and will suffer, Anya, well, since they do, I know, well, part, inside teams at NVIDIA, and since there's such a number of projects that Jensen Huang led, and whatever agreements there were, I think some people there may suffer. But someone will win, as always. We'll see. It's hard for me to forecast here, too, too huge a company, right, when your company is worth 5 trillion dollars, it's not all that simple.

Mentions: OpenAI
Anna Volchek✓00:16:34

Well, you always think you don't fully understand, really, how everything worked and what the decision-making chains were, who influenced what, what all these reporting lines are, it's still quite, all the way out, it seems to me, extremely hard to assess.

Alexander Volchek✓00:16:50

Anya, it's immediately clear that you work at an American company. Uh-um, about reporting lines. This, here's what I want to say.

Mentions: United States
00:16:58–00:21:05AI Built Software for AMD: A Threat to NVIDIA's Leadership?
Alexander Volchek✓00:16:58

Look, there's a very interesting topic about chips, and it seems to me this story is altogether explosive. News came out that Anthropic's models fully wrote software for AMD's infrastructure, for AMD's new chips, kind of. And the gist, the practical meaning, is that it turns out what they've done now took three days. Before, it's written very clearly there, this would have taken six months and a whole team of developers. That is, it's a very serious, big, complex task, because there's a huge amount of computation there, it all has to be fully checked and so on.

Mentions: Anthropic
Alexander Volchek✓00:17:42

And Anthropic's statement says very clearly that this was done 100% by artificial intelligence. And what's the applied construction that's now possible at all, the changes that are possible in terms of, again, NVIDIA, what happened now at AMD, and, of course, it's a big question, a huge amount of money is still needed there anyway. That a very fast transition to other hardware is possible. That is, if a transition to other hardware used to require months of expensive engineering work, then it's clear that whoever bought such systems was very strongly tied to, to some particular supplier.

Alexander Volchek✓00:18:22

And it's clear that in this case, if artificial intelligence greatly reduces the work, then choosing AMD now can be much easier. And it's a very interesting case, it seems to me. And I'm sure, of course, that OpenAI and Anthropic are working on writing software for various microcontrollers and for such infrastructure in general. But this open case personally impressed me.

Ilnar Shafigullin✓00:18:47

The same story is just as useful for NVIDIA, because, for example, there's research, that is, a new graphics card comes out, well, say, Vera Rubin or the previous architectures there are. And at the moment they come out, they deliver about 30 percent of the potential built into them. And then over six months to a year, engineers, the libraries used on these graphics cards, well, the same CUDA and so on, they refine them. And only six months to a year after release do these graphics cards reach their design capacity that can be squeezed out of them, out of this hardware.

Ilnar Shafigullin✓00:19:28

And essentially, as far as I understand, the news is precisely that this gap, of six months to a year, between the card's release and the raw software that exists before it's brought to a properly working version, is exactly what's being sharply reduced. For NVIDIA this is also most likely very useful, well, in terms of reduction. For AMD it's a huge plus if they can make a story similar to what NVIDIA has with CUDA. That is, NVIDIA isn't only hardware, it's also all the infrastructure around it. Many frameworks tied to training neural networks are all made for this story.

Ilnar Shafigullin✓00:20:09

And, actually, that's why NVIDIA has such hegemony, right, in the market. Accordingly, if they can pull off a similar story, yes.

Alexander Volchek✓00:20:18

They actually talked about it, in this code there was a comparison with NVIDIA, with CUDA, that NVIDIA has this, right? And AMD's is called ROCm. And it's a software platform for accelerators. Yes, a software platform for accelerators.

Ilnar Shafigullin✓00:20:35

Yes, it's worth recalling that Huang and the head of AMD are relatives. There. There were a bunch of jokes after Huang was fired, like: "I have a friend, lots of experience at a tech company, who can refer him to some company?" And they attach this fake résumé of Huang's. I think that if NVIDIA suddenly doesn't take him back, he'll find somewhere to apply his talents even in that same market, if anything.

00:21:05–00:22:30Gemini and Grok Bot: Artificial Intelligence Starts Running Errands
Alexander Volchek✓00:21:05

Listen, I want to recall another story, and it'll be interesting for our viewers, whoever writes in, whoever has tested such features. Let me remind you, we recently talked about Grok Bot launching the ability to create automatic calls, kind of, so that the service would call through, help, for example, with purchases, or help resolve various complicated issues you have, or call tech support, or, I don't know, book a car in for service. Now Google has had an update, within Gemini. They opened a free errands mode.

Alexander Volchek✓00:21:38

They opened it, and in six countries, and in the US additional ways appeared to ask the assistant for some, so, matter or assignment you need to give it. And in a separate tasks tab you can describe the desired result. But I'm interested in the comment of a Google vice president, who said: "Sometimes a task may remain unfinished, or the result may not be the one you expected." And here, of course, it's very interesting, right? So what do you do, then? Ask it, to call support, well, over the phone to get money back, and for you, and the system will call for you and agree that you'll also, so, connect some new plan there or spend something.

Mentions: United States
Alexander Volchek✓00:22:20

It's, it's a very specific story. And here I want to recall, I don't know if you, you probably don't have this yet, an important topic for our viewers: GPT Wallet came out, the GPT wallet came out.

00:22:30–00:26:47ChatGPT Wallet: How Alexander Made Three Purchases Through AI
Alexander Volchek✓00:22:33

You can add your cards there, and judging by your faces, you can see you're not, not into this topic, right? You can add your cards there, and I added my cards there and let ChatGPT make purchases. I made three purchases. In some, it seems to me, we even had episodes this week, I recalled this story, I made purchases and I'll keep making them, and I'll keep making these purchases. It's unbelievably convenient, of course.

Mentions: ChatGPT
Anna Volchek✓00:22:58

And convenient how? Well, explain.

Alexander Volchek✓00:22:59

ChatGPT itself glitches. Here's a simple example for you. I needed to buy a New Year's present here, and I went into ChatGPT and said: "Find me", so, it found me a present, this and that, chose it and told me the price. Then I tell GPT Wallet: "Make this purchase", because GPT Wallet works in Dots, and you don't launch it from the chat, but for now you launch it from Dots. And Dots tells me: "They didn't tell you everything there. There are also taxes on the purchase, there's also this, and this, and this, and this."

Mentions: ChatGPT
Alexander Volchek✓00:23:27

That is, what does this, this, kind of, Wallet functionality actually do? It completely, Anya, carries the purchase through not just, say, putting it in the cart, it does all the registration there, puts it in the cart, makes all the statements and gets the final, kind of, purchase ticket. Like in Apple Pay, you press Apple Pay to pay... Well, in the Russian-speaking market, maybe many have forgotten how to pay with Apple Pay, but that's humor. So, when they press the "pay" button, it shows what you're buying. That is, the item will be listed there automatically.

Alexander Volchek✓00:23:59

That is, essentially, the Apple Pay system immediately sees the goods placed inside it. GPT Wallet is a similar thing. That is, at the end they send you this final cart, and you buy it. Obviously, if you, you were made some other purchases or you didn't approve these purchases, then, there will be no problems. Well, clearly they rolled it out in the US. Let me remind you that in the US in general, if someone used your plastic card, credit card, and you didn't make this purchase, the bank will instantly return the money, always, right?

Mentions: United States
Alexander Volchek✓00:24:29

That is, it's kind of a basically protected mechanism. So nobody worries if their card gets stolen. I remember back when I started going to America, constantly, every year, well, 15 years ago, we were amazed, so, how easily everyone hands over their card, in a restaurant and everywhere. And we immediately recalled, so, Minsk or Moscow, that your card isn't taken away from you, and taking it away is wrong. Because if it's taken there, it can be copied, and if it's copied here, you'll always get the money back. It's pointless. That is, it's a pointless story.

Mentions: United States
Alexander Volchek✓00:24:59

It makes no difference to me at all, right? There. The service is really cool, and it really placed the order, and it placed it on very strange sites. Why is that interesting? Before, before, for me personally ChatGPT, when it searched for something-

Mentions: ChatGPT
Anna Volchek✓00:25:11

What does "placed on very strange sites" mean?

Alexander Volchek✓00:25:13

Well, for example, sites I didn't know, what these sites are, how to do things there, place orders. That is, I'm used to buying on sites I trust. Or I like buying, for example, on Amazon, because for me a purchase on Amazon takes, roughly, one click. That is, ChatGPT gave me a link to Amazon, I went in and immediately clicked "add to tomorrow's delivery". Usually there's roughly that kind of button there. You press the button, and that's it, you have the product. Well, you know, Anya, right?

Mentions: ChatGPT
Anna Volchek✓00:25:38

Yes, yes-yes-yes.

Alexander Volchek✓00:25:39

That is, it's instant, it's instant. You, it's an automatic action, essentially, right? And, by the way, I don't understand why Amazon still has no integration with Dots. More precisely, it's clear, because Amazon is trying, kind of, to stay in its own infrastructure. But it seems to me that for them it will be a big loss. And so, when ChatGPT searched other sites for me, I kept saying: "Can you give me it on Amazon?" Because on another site you need to do some registration, you need to log in, enter data. A lot of complicated, well, complicated procedures, right, for me at least.

Mentions: ChatGPT
Alexander Volchek✓00:26:08

I don't want to spend time on that. And now, of course, this is really an incredible solution. And if they don't leave this service as such a pet project and don't leave it, again, some employee somewhere came up with something on the side and made some micro-thing that isn't connected to the overall infrastructure, but for real, let me remind you, by the way, that OpenAI, and maybe ours will find the episode, our editorial team can easily search episodes now, will find the fragment where ChatGPT, Ilnar, remember, we talked about it?

Mentions: OpenAI · ChatGPT
Alexander Volchek✓00:26:38

Released a cart, the ability to make purchases. Remember, there was such a thing? They had a cart, after all.

Ilnar Shafigullin✓00:26:42

A startup there, they also-

Alexander Volchek✓00:26:43

And then they quietly dropped this project. I want to tell you about checkout, about this checkout, what they did, and why I paid close attention to it.

00:26:47–00:30:17Instant Checkout: OpenAI Is Changing the Online Shopping Market
Alexander Volchek✓00:26:55

They integrated directly with several American companies, in particular, well, Shopify, for example, so that a person searching for some product in ChatGPT would buy it right away in ChatGPT. That is, essentially, what they've made now is a single-item cart with one product to one delivery address. But what's cool? Every day I search for something in ChatGPT, where it gives me links mostly to Amazon. I click Amazon, add it to my cart, right? And for now this interface, I'm used to it anyway. That's money, large volumes and so on.

Mentions: United States · ChatGPT
Alexander Volchek✓00:27:29

But overall, probably, in terms of ChatGPT, if its UX/UI is cool for working with carts, I'd buy right away through ChatGPT from my Amazon account. That is, it made the integration. We'll show the diagram now. Here's the diagram, I think it's shown now. There's literally the path, the path where the user talks in ChatGPT, ChatGPT talks with the store and fully places the order, with a credit card, or Apple Pay, or something else, right? That is, it fully makes the purchase. What does the topic say? That, of course, OpenAI definitely wants to get into the marketplace market and create its own marketplaces.

Mentions: ChatGPT · OpenAI
Alexander Volchek✓00:28:12

That's obvious. And from the stores' side it's always a double-edged story here. On the one hand, it's cool for you to integrate with them. Shopify's shares rose instantly after the announcement of this Instant Checkout. Their shares rose on NASDAQ too. At the same time, a person who gets used to using it, to making purchases in ChatGPT, stops caring where they make the purchase in the background, right? If everyone's integrated with them, their payment system is stored in OpenAI, then what difference does it make to me, which company will deliver to me, if there's, well, there is a difference: quality, service and everything else. So, orders appeared for them, they had, well, there were orders.

Mentions: ChatGPT · OpenAI
Alexander Volchek✓00:28:55

That is, their original logic was like that. And there the questions arose precisely: who will bear responsibility for these orders? Because the main story here is that OpenAI doesn't bear responsibility for orders. Here responsibility, essentially, for the order lies with the seller. OpenAI is merely an assistant for making a purchase. And it directly, look, to add cards, I got authorization, for example, through the card, directly through American Express. I don't remember how I authorized it, maybe the American Express interface even opened for me, I don't remember. Chase definitely opened for me, the Chase interface. That's important, right?

Mentions: OpenAI
Alexander Volchek✓00:29:32

That is, it's not just that I entered some card number. That is, OpenAI double-checked with the issuer of these cards, essentially, that this card can be used. And I know that a number of my friends tried cards from-...different countries, some cards didn't work. Well, that is, no-no-no-no, it won't work for everyone. That is, there clearly must be a check by the bank. What's more, Chase officially says that only the owner can add it to ChatGPT, that is, essentially, whoever owns the account must also own this card, right?

Mentions: OpenAI · ChatGPT
Alexander Volchek✓00:30:04

There. But for me this integration is extremely interesting.

Anna Volchek✓00:30:07

Very cool. I'd try it. Unfortunately, there's no such option, but I'd definitely try it.

Alexander Volchek✓00:30:11

But it seems to me this thing will appear in Europe, Anya. Well, I think it's about to appear.

Anna Volchek✓00:30:15

And I think, on the contrary, it's unlikely.

00:30:17–00:35:38Safety of Shopping Through ChatGPT: Refunds, Cards and Liability
Ilnar Shafigullin✓00:30:17

Historically my anxiety is much higher. Especially when it comes to buying on obscure, as you put it, sites. Well, I mean, if the product arrives, well, I don't know, not the quality you expected and so on, will ChatGPT help you with returns and so on?

Mentions: ChatGPT
Alexander Volchek✓00:30:36

Look, in America there's no problem, once again, right? That is, if you ordered a product and got the wrong product, and they don't want to refund your money at the click of one button, then the bank will simply refund your money. Well, you know this thing, right? You open the transaction in the bank, press "refund", they refund your money, and then they'll sort it out. It's an absolutely well-known case. I agree with you that, here, I wanted to order one product from Europe, and you know what ChatGPT told me?

Mentions: United States · ChatGPT
Alexander Volchek✓00:30:59

"Use PayPal to buy this product, because the seller is questionable." And I tell it: "Why PayPal, if I have a Chase card?" — "Ah, well, then order without any problem." That is, it suggested it to me itself.

Ilnar Shafigullin✓00:31:11

And then, yes, you were right, it should have been done differently and you shouldn't have ordered from this seller.

Alexander Volchek✓00:31:15

Look, you really need to understand the system. That is, you're right that if you, for example, pay with a non-American card, once again, pay in dollars, you'll have questions. I don't know how it works in Europe, how much, Anya, Anya, how quickly a European bank refunds money if the product didn't arrive, or arrived in the wrong quality, or the service wasn't provided to you.

Mentions: United States
Anna Volchek✓00:31:35

I don't know, it seems to me there's nothing like that here at all. I might, to be honest, I might well not know, because-

Alexander Volchek✓00:31:39

You most likely haven't tried. But listen, this was-

Anna Volchek✓00:31:41

I haven't tried-

Alexander Volchek✓00:31:42

...this was even, this was even in Russia, and it worked pretty well. It seems to me it still works in domestic banks, in domestic systems.

Anna Volchek✓00:31:48

Well, in general I just haven't had a situation where a product didn't arrive. And the return system is so light and simple that I also haven't had a situation where I resolved something through the bank rather than through, for example-

Alexander Volchek✓00:32:01

Rather than through the operator. Yes, here you can-

Anna Volchek✓00:32:04

So, to be honest, I've simply never had such a case at all.

Ilnar Shafigullin✓00:32:08

Look here, with Amazon it's absolutely clear to me, because there's Amazon as the integrator, sellers on one side, the buyer on the other, and it takes on itself, well, picking up the product you didn't like. Remember, we bought you a pull-up bar, it turned out to be lousy, and in the end, there, even with a torn box, they came, picked it up and took it away.

Alexander Volchek✓00:32:26

Yes, yes, yes.

Ilnar Shafigullin✓00:32:27

But if you buy not from Amazon but from some, I don't know, specialist from the north of America, from Alaska or somewhere else, they bring you some parcel. What happens with that? Well, ChatGPT found this seller, they sent some product, and in the end you didn't like it.

Mentions: United States · ChatGPT
Alexander Volchek✓00:32:43

It's all a very simple story. The wrong product arrived, you write to this seller that the wrong product arrived. If he doesn't answer you or doesn't refund the money, or doesn't say I'll send you another product, you go into the transaction in the bank and click: "I received a low-quality product". The money appears in your account right away. That's it. That's the end of it, look, look, that's the end of it. The probability that this seller will prove to the bank that he sent the right product is zero, first, first, practically zero.

Alexander Volchek✓00:33:05

Of course, you need to photograph it. I've had such cases. I've actually had a whole series of such cases, because I do order things, there are some exclusive things that Amazon actually doesn't have a lot of. And mostly, in the US the buyer's market is very strongly protected from all this, right? And especially now, everything regarding the market of cancellations, returns or low-quality products has fairly strong protection. I even, I order for the dog, my dog is on a raw diet, she's lying there, that sleeping one.

Mentions: United States
Alexander Volchek✓00:33:36

So, and I order from a state, it seems to me, several thousand kilometers away, from some, I don't even remember which. I order him a certain by-product, and they ship them in special packaging, frozen, right? And once a lot of packages arrived, and one of the packages, I opened it, and it had thawed for me, and I didn't like how the product looked. I wrote to him, he told me: "Nothing terrible at all." And instantly refunded the money right away. It's a standard case here. Here it's a standard case.

Alexander Volchek✓00:34:06

If someone starts bickering with you, you write: "Fine", and write to the bank, and they refund your money right away, right? That is, or they resolve your story right away, because otherwise this store will have problems with the bank later too. Another point is that there really can be a situation, of course, where you'll have to argue, where you'll have to sort things out. In Europe, for example, Polina's mom recently flew to us, and they charged her extra money for suitcases. A Polish airline, by the way.

Alexander Volchek✓00:34:32

So, when my wife wrote to this Polish airline, it said: "We know nothing, not our business." There. And I told them: "Next time pay with American cards." And when mom flew back, this Polish airline charged her money again. Only there's one nuance: my wife paid with an American card, and she says: "Sasha, write there again, sort it out?" I say: "No, just go into the bank and cancel the transaction. Write that they charged you money illegally." That's it. This is really, I remember very clearly, the Polish company replied: "Well, those are our rules."

Mentions: United States
Anna Volchek✓00:35:02

Well, by the way, I just checked that in Poland you really can request a refund through the bank that issued the card, say, if the product didn't arrive. But the bank first reviews the dispute, and only if the dispute is decided in your favor, only then will they return this money to you.

Alexander Volchek✓00:35:21

That's a standard thing, yes, it's standard world practice.

Anna Volchek✓00:35:23

Yes.

Alexander Volchek✓00:35:24

Yes, it's standard world practice. But you know why it's done this way? Because America owns the dollar, it's their system, they can take the dollar back. There. It's strange, by the way, that in Poland they can't take back the euro, but overall, overall, so, it's surprising. Yes, we'll see.

Mentions: United States
00:35:38–00:40:28Tesla Cybercab: A Car with No Steering Wheel and No Driver — Who's Ready to Switch?
Alexander Volchek✓00:35:38

What, by the way, do you think about the Cybercab that came out, and they released it on the market? I shot an episode on Monday, for those who haven't seen it. But would you buy yourselves such a car or not?

Mentions: Cybercab
Anna Volchek✓00:35:48

I would buy it. I thought: "Cool! I'd definitely buy it."

Ilnar Shafigullin✓00:35:54

I, of course, don't like driving, but with a live driver I'd somehow feel more comfortable for now.

Alexander Volchek✓00:36:00

Yes, you rode in San Francisco yourself.

Ilnar Shafigullin✓00:36:01

It's connected only, yes, I rode. But look, why would I feel more comfortable with a live driver for now? Because there are still a bunch of live drivers all around. This is, probably, still a somewhat irrational fear of mine, because, well, statistics say, as far as I understand, that, uh-um, autonomous vehicles get into accidents less often than live drivers who, well, drive at the wheel, ordinary drivers. But still, when there's a huge number of people around driving at the wheel, I somehow feel calmer with a human. Yes, in the back seat, but with a human.

Anna Volchek✓00:36:40

And it seems to me it's safer, more convenient, you don't have to talk to anyone, you can do things yourself, save a ton of time. Although in general I love being at the wheel, but due to certain circumstances lately I don't spend that much time at the wheel. It seems there are a lot of pluses here, and it all sounds so coherent, safe, logical, ergonomic and so on, that it really wins you over.

Alexander Volchek✓00:37:08

You, by the way, say you don't have to talk to anyone, but here it's really still a personal car. In a personal car you don't have to talk to anyone either, right? You just recalled that for yourself.

Anna Volchek✓00:37:14

Well, we were just talking, in this case Ilnar, like, made a comparison with a live driver. Like if we're talking with a live driver or without a driver. If you, of course, drive yourself alone at the wheel and without a wheel, then here you can just find some other pluses. It seems to me it's a great option, but at least in our world, well, at least for me, it would be a good second option. And there are definitely some scenarios, wishes, moods and trips where you want to drive, but just the same-

Alexander Volchek✓00:37:47

You think there are such ones, right?

Anna Volchek✓00:37:48

...there can be options where it's much more convenient, I don't know, standing in a traffic jam, I'd rather read, study, write, learn, than stand, for example, in a traffic jam holding the wheel. Let a specially trained car do it, changing lanes competently, in complete safety, take me where I need to go. Imagine how much time people could simply save, and not be in such a state of tension, irritation, during the time it takes them to get, for example, to work.

Alexander Volchek✓00:38:16

It seems to me it's an illusion that people love driving.

Ilnar Shafigullin✓00:38:18

I think so too, because I myself don't like driving. Here I'm a bit biased, as they say, my perception is skewed anyway. Another story is that the scope of autonomous driving is still quite limited anyway. Remember-

Alexander Volchek✓00:38:33

Of course, of course

Ilnar Shafigullin✓00:38:34

...how a year ago, probably, we said that Tesla drives much better from Elon Musk's house to the office, SpaceX-

Anna Volchek✓00:38:46

Yes-yes-yes

Ilnar Shafigullin✓00:38:47

...than in other places. That's because it was simply trained better on this route, right? And so, kind of, the people who travel along this route feel that the autopilot works, well, at full power and really well. If we take some back alleys where the car that collects data hasn't driven, on which cars are then trained and so on, right, it will drive too, but the quality and the safety it provides will be lower. If we take some completely atypical weather conditions, say New York, Canada or something else, where the seasons change a lot.

Ilnar Shafigullin✓00:39:21

Yes, probably now they'll write in the comments that such cars already drive and so on, but still they drive much more in California and in Texas, where after all-

Alexander Volchek✓00:39:29

Well, have you even seen how self-driving works in Europe, in difficult countries, on some narrow roads? You know, self-driving is very widespread not in California at Tesla, right? And by the way, I'll remind everyone, now again they'll put out the figures, the number of Tesla sales in China. Well, everyone just loves to talk about the Chinese there. Look at the sales volumes in their categories and how Tesla is winning. By the way, they just had September, they won the sales volume in the category there altogether, right? That is, they hold a fairly strong position, and they're seriously opening up Europe too.

Alexander Volchek✓00:40:01

And I, probably, I don't know, what I see, once again, it's hard for me to say, I don't have a Tesla, right? Here we've had some persuading, the Cybertruck is closing, they won't produce them anymore. And we, after all, probably decided to get one, so that we'd have a Cybertruck, but also this full self-driving will appear, right? And they say it's very cool in terms of full self-driving. There was one comment, in my Monday video, whoever wants to look in more detail at the Cybercab, have a look, because the car is interesting.

Mentions: Cybercab
00:40:28–00:44:47Cybercab Without a Driver's License: New Possibilities and Limitations
Alexander Volchek✓00:40:28

There was a comment that because this car is also interesting, and you know, right, that this car can be bought without a license.

Anna Volchek✓00:40:34

Mm-hm.

Alexander Volchek✓00:40:35

Do you understand what this means for the world at all?

Anna Volchek✓00:40:36

Yes-yes.

Alexander Volchek✓00:40:37

Buying people a car without a license is a different matter. Of course. And later they'll also make it possible to just send your child alone. For now you can't until 18, but in general. Then the laughter started, so, because I was looking into whether you can ride drunk at the wheel, and with a child, with a child next to you. So, then the comments came: "And can you ride at the wheel and with a drunk child next to you?" Well, there it all went completely off the rails. So, someone wrote: "You can ride drunk at the wheel", right? Just so you know, in the Cybercab. Well, clearly, there's no wheel. Once again, there's no wheel.

Mentions: Cybercab
Alexander Volchek✓00:41:08

And ChatGPT at first, can you imagine, it so didn't understand how to answer me when I was researching the Cybercab to make the episode, that it tells me: "Well, if you're really not driving, you have the mode fully on, then you can be drunk." I'm like: "Wait, but in the Cybercab you can't drive, you can't drive, you can't get the wheel there", right. And someone, by the way, gave a case and says: "Well, I think it's not very convenient, because imagine, I came to go trekking somewhere, so.

Mentions: Cybercab · ChatGPT
Alexander Volchek✓00:41:34

So I arrived, and the whole parking lot is full. Well, to walk, to hike, to walk. So, the whole parking lot is full. And usually I as a human will get my bearings much faster, drive somewhere past the bridge, find parking and that's it." I sit and think: "Wait, do you think you can't tell a Cybercab or a Tesla: 'Drive ahead and find a new parking spot?'" And this topic, of course, people are very divided on it. I know all the people who drive a Tesla, I don't know, I personally don't know any such people, not a single person with full self-driving.

Mentions: Cybercab
Alexander Volchek✓00:42:03

Everyone says: "We won't buy other cars at all anymore." Well, maybe because, people basically don't have some huge number of cars, I don't know, right? Maybe they like, well, the quality inside this car in terms of the cabin, design, interiors, but okay. That the quality of full self-driving is astonishing. And really, I, even in the Cybertruck, a big plus, a huge, big checkmark for me, they really have an unbelievably cool full self-driving. It's not my Cadillac full self-driving.

Alexander Volchek✓00:42:33

We've now driven 400 km north of my home, and the road is so winding, Highway One goes there, and you drive, I turn on the autopilot, and it periodically tells me: "Oops, I'm not working." And you, so, on a winding road suddenly start grabbing the wheel, right? It's such a special effect, yes, with Cadillac. But that's Cadillac. It'll die out somewhere, somewhere. Although, in my class, there, it's super, well, a beautiful car, there, because it's a modern Escalade, QL, it goes very far, right, it goes 700 km, even more, 440 or 450 miles of full range.

Alexander Volchek✓00:43:12

But you can come here and go back without charging, right? But still we'd gladly, we're going back home today with Dima, we'd gladly get into the autopilot with him and just ride for four hours. And especially when it's raining, some thunderstorms. Here, Ilnar, by the way, you said: "Rain, thunderstorms — no", but I'd gladly, in this weather, because the weather here is varied right now, right? Now eight-meter waves, now pouring rain, now the sun shining, nothing is clear at all. I'd gladly ride on autopilot.

Alexander Volchek✓00:43:42

And it seems to me that the quality of these cars in terms of camera recognition, a human can't work like that. Well, they can't, a human can't see 360, in vision, instantly. Well, I mean, I don't believe in that.

Anna Volchek✓00:43:55

Well, it also seems that with roads one hundred percent filled exclusively with such self-driving cars-

Alexander Volchek✓00:44:03

There won't be traffic jams.

Anna Volchek✓00:44:05

...the issues related to traffic jams, safety and so on will decrease quite seriously. Probably, of course, it's-

Alexander Volchek✓00:44:11

There won't be any. Anyuta, there won't be any.

Anna Volchek✓00:44:13

I don't know how long-term that prospect is, right, if we're talking about one hundred percent filling, but the problems we face today when driving, yes. And, by the way, this point, that you can operate this car without having a license, that was also one of my pluses. I have a license, but I also thought how many people who for various reasons, I don't know, simply out of fear, out of anxiety-

Alexander Volchek✓00:44:39

Yes.

Anna Volchek✓00:44:40

...also because of some indicators-

Alexander Volchek✓00:44:41

Our mom and yours, our mom and yours.

Anna Volchek✓00:44:42

...can't, for example, pass the driving test, right? And how many problems this would solve.

00:44:47–00:49:05Tesla Subscription, Self-Charging and the Future of Roads Without Traffic Jams
Alexander Volchek✓00:44:47

The car, by the way, still costs money, it has a subscription, and, of course, you'll have a lifetime subscription, you, it's not entirely clear to me why they set this price. Well, maybe it's like that at the very beginning, most likely it'll be canceled, then there'll just be some monthly leasing payment, because otherwise it makes no sense, right? That is, if you're not on a subscription-

Anna Volchek✓00:45:03

I didn't quite understand-

Alexander Volchek✓00:45:04

Yes.

Anna Volchek✓00:45:05

...this binding to each specific person, because, as far as I understood, it's impossible, I don't know, if this car is mine, Ilnar can't drive it. I can't hand it over to him.

Alexander Volchek✓00:45:14

A family member can drive it. For now a family member can drive it.

Anna Volchek✓00:45:18

Yes, but that is, even a family member is like other subscriptions, different levels of access to use.

Alexander Volchek✓00:45:25

Look, they'll develop this. It's just a story launched right now.

Anna Volchek✓00:45:27

Well, obviously, of course.

Alexander Volchek✓00:45:28

There's even such a clause, look, there's even such a clause: can this car be used as a taxi? It's written there that you have to get a separate agreement. Or, for example, the car is now declared able to go by itself to a charger to charge. Here, though, Dima and I came to some village, Fort Bragg, in the north, to a charger, and we thought: "And who, who will plug it in, who will plug in this cable, right?" So, there's a person driving there. Well, probably in Austin, in Dallas, in Texas, well, in the places where it's being launched, there's a huge number of charging stations, probably, where an operator works, right, and an operator will work.

Alexander Volchek✓00:46:02

The car can go to service by itself, it will deliver itself too. You know, right, that in Texas in general Teslas deliver themselves periodically, right? They have this thing there: you go out onto the street, and your car has arrived, you ordered it online, and it came to you. That's super convenient, of course, in-in-incredible. So, I asked the question, studied the rules there: can the car, if it took me to work, go home? There. That's not fully defined yet, and we'll see how it will really work. It seems to be written that with certain approvals it can be done.

Alexander Volchek✓00:46:34

And it's, of course, a really great thing.

Anna Volchek✓00:46:36

That's also the solution for all the parking: they can drop you off and then pick you up.

Alexander Volchek✓00:46:40

So that's, Anya, also the solution for traffic jams. Look, traffic jams won't exist, they can't be. Traffic jams, well, only if it's some city where everyone came for some festival, and even then, even for a festival, they'll distribute themselves correctly, they'll schedule the time correctly, they'll tell the person in advance that we have a queue, bought, I don't know, or not bought, distributed, at what time you should leave. Today we'll be driving, we need to get through all of San Francisco from the north. There are a bunch of bridges, people going in different directions to work, home, everywhere.

Alexander Volchek✓00:47:08

I'd be glad if they told me that you'll get there without traffic jams, you should be there between one and two, so you need to leave here at such and such a time. Then we'd gladly plan our time and drive that way, right? Actually, people try to guess and do that with traffic jams, but they don't always guess right, right? By the way, we also have road closures on the road here, today, by the way, we have two one-way closures where we can still, can get stuck and can stand. We'll see what happens, we'll see what comes of it. So, Ilnar, I'm waiting for you.

Alexander Volchek✓00:47:37

When the Cybertruck arrives for you, if after all I, I'm standing in the queue there, right? But it should, should, in theory, be approved any moment now, and it'll arrive, come over, we'll test it in full self-driving. It's scary. My wife and I rode the Cybertruck in full self-driving, the two of us. It also has a big window in front. At first, when you're not used to it, it's a bit... Well, you feel uneasy. And, by the way, one of Polina's friends here got into an accident in a Tesla, because full self-driving was driving and changing lanes sharply between cars on the highway.

Alexander Volchek✓00:48:09

And she thought it was doing it wrong and wouldn't make it, and grabbed the wheel. As soon as she grabbed the wheel, she got into an accident. There. And so, of course, very calm people need to ride, not grandmas and people who have never sat in one in their lives. Here mom, Anya, by the way, is coming in a month, and she hasn't ridden, I think, in a Waymo, right? And she needs to-

Anna Volchek✓00:48:29

Yes, no, we definitely need to put her in a Waymo.

Alexander Volchek✓00:48:30

She needs to be put alone in San Francisco, put alone in a Waymo, close the door and say: "See you in 20 minutes." Especially on these hills, right? Ilnar and I, it seems to me, also rode over the hills, right, with you in a Waymo. And yes, I'm very calm in a Waymo, by the way. Altogether, for me it's such a simple thing there. You, Anya, rode with Danila just now, right?

Anna Volchek✓00:48:50

I also love it. I really like it, yes. Well, I like it, this clean car without a driver and...

Alexander Volchek✓00:49:01

Well, you can be yourself, make it dirty, yes. Can, I want to mention one more topic, that, so, Instagram is testing a standalone messenger, Direct, and a social network account will be optional.

00:49:05–00:52:05Instagram Tests a Standalone Messenger: Why Does Meta Need a New Direct?
Alexander Volchek✓00:49:11

They want to enter the market, of messengers. I don't fully understand whether they can succeed in today's reality in entering the messenger market. Especially, I personally expect, how you, your opinion is interesting, how you, do you expect the launch of messengers from Anthropic and from OpenAI? I'm still not counting Google, because these are companies, still, this company is a corporation, a mastodon. I'm not counting Grok, because there's xAI, and inside xAI there's a messenger, right?

Mentions: Anthropic · OpenAI
Alexander Volchek✓00:49:39

And what do you think about the topic of messengers from OpenAI and Anthropic?

Mentions: Anthropic · OpenAI
Ilnar Shafigullin✓00:49:45

Well, let's talk about Google. Google did have its own messenger. In their mail there was the ability to chat, and you could also use this messenger outside, kind of, not only inside the mail. But, like many Google products, as we said almost in the first episode, Google has its own graveyard of projects that they started and didn't finish. That is, they entered this market even before Telegram appeared, well, back when WhatsApp was around. But somehow they didn't go actively in that direction.

Mentions: Telegram
Ilnar Shafigullin✓00:50:10

Maybe they didn't see monetization, maybe, I don't know, the team that worked on it somehow couldn't prove itself, something else like that. And as for OpenAI and Anthropic, when you said it, I realized I'd never heard this news and never even thought they could go there. But it seems as though, I mean, look, it would definitely be interesting. Clearly, AI would be built in there natively by the creators of that AI and so on. But it seems as though for me, at least, it's not their market. Better they race to AGI, and other companies will add AI to the chats.

Ilnar Shafigullin✓00:50:48

That is, I wouldn't really want the companies to spread themselves thin on this. But again, maybe this is as rational as the desire to ride with a driver. The only thing I want to add is that Anthropic now, after the IPO, still depends heavily on the market, on what announcements it will make, on what products it will make, on how it will present them, its share price will depend. So, accordingly, maybe irrationality, well, from the point of view of achieving AGI, will appear in its behavior too.

Mentions: Anthropic
Ilnar Shafigullin✓00:51:23

So, actually, we'll be waiting for various interesting news that will swing the market cap in different directions.

Alexander Volchek✓00:51:29

Listen, here I wanted to explain a bit what I also meant by a messenger. I agree with you that if you create it as an independent standalone product, it's unclear what will come of it and work out. And it seems to me this push from Meta is very strange in the current market, moving people, they had this Direct anyway. And, well, so you released a separate app? What's the point of it? Probably there's some benefit in it. Like when Telegram once appeared, for example, when you're in some country where you have blocks, I don't know, of WhatsApp, or you don't trust WhatsApp, you, for example, start using it.

Mentions: Telegram
Alexander Volchek✓00:52:00

Such things used to work once. Now it's unclear how to take this market at all. But what I wanted to say is that with the appearance of artificial intelligence, Dots appeared, for example, right?

00:52:05–00:59:12The Future of Messengers: AI in Telegram, WhatsApp and Personal Chats
Alexander Volchek✓00:52:09

And I gave this case, that now you can write to Dots from a message. They made this feature. I consider it absolutely pointless, because why? I opened the chat and wrote to Dots the same way. Maybe it works somewhere. For example, if the ChatGPT app isn't available to me right now, I can write to my Dots, but again, only from my own phone, only from my own authorized number. Well, anyway. But what's interesting is that if I could, we tried with Dima to add it to our chat together with Dots, or to add my wife there.

Mentions: ChatGPT
Alexander Volchek✓00:52:43

That is, or if this infrastructure starts working for them, where there'll be a messenger in which you'll be able to simply chat with a person and at the same time artificial intelligence will work in conjunction with other messengers, in conjunction with various agents, bots, overall with such an artificial intelligence module. Because I'd like to move to some system that will control, essentially, all my chats, right, and understand all the chats inside: WhatsApp, and Telegram, and email.

Alexander Volchek✓00:53:13

Because, for example, with email there's a great integration now, I really like it, I really enjoy how my Dots works with email, how it reads them, how it sends messages there to us about the schools, with my wife, how it answers all sorts of corporate letters. I really like this thing. But although there are various glitches and some questions there, I'd be glad to connect Dots to my messengers too, well, that is, to WhatsApp and to Telegram. I'd connect it right now. Just how do you connect it? You don't have a normal healthy connector, you don't have normal healthy restrictions. For example, connect it only as read-only.

Mentions: Telegram
Alexander Volchek✓00:53:49

Yes, and by the way, you can connect mail as read-only. You can connect mail, for example, only in that mode. And, and why not? And when you connect through the computer, WhatsApp, it, by the way, Codex many times, for you too, probably, right, asked me on the Mac, for example, to give access, I don't know, to Telegram or to something. And you're like: "I won't give it to you, because I don't understand at all what you'll do with it. Maybe you'll erase all my chats there, or send some mass mailing to people, or do something else."

Mentions: Telegram
Alexander Volchek✓00:54:17

That is, there are no written standards, written connectors. And I don't know how interested WhatsApp and Telegram are in this. Because, of course, they'll lose audience, just as Google Gemini now, they had an update in the sense that in mail they can do all sorts of tricks with artificial intelligence. So I don't open the mail, I haven't opened Google Gemini for a long time, because I'm such an Apple person, right? And I have Mail, with my mail connected. And here I don't even open Mail anymore, right? My Dots reads everything and looks through everything completely.

Ilnar Shafigullin✓00:54:54

Well, about the messenger, the case you're describing, it reminds me, you know of what? Anthropic has an integration with Slack, and Anthropic themselves said that in practically all their work chats they've added Claude Bot, which carries out assignments given to it within the chat, can facilitate the communication process, and does some other things. That is, most likely, rather than gathering everyone in some single messenger, such integrations will develop further, so that you'll be able to connect, well, that same Dots, I don't know, there, to your conversation with, there, friends, to the family chat, somewhere else, to, well, Telegram, WhatsApp and some other messenger.

Mentions: Anthropic · Telegram
Ilnar Shafigullin✓00:55:41

These integrations, yes, they're probably being held back now by the companies themselves that own these messengers. But with that same Slack it all turned out very convenient to do. And there, kind of, among developers in companies, it's more of a corporate story, it's actively used. There. I think this story will also come to the B2C market.

Alexander Volchek✓00:56:04

And I think that Slack too may have very serious problems because of this. In theory they should, right, because you should have a slightly different infrastructure with artificial intelligence. Although Slack, of course, is fully integrated now, right? Now there's no problem for artificial intelligence from different systems to write there. And here's the question: will there be this transition or, maybe, for example, Slack will stay as it is for corporate use? Maybe it'll stay, or maybe there'll be this transition. I, for example, had moments when, in companies connected with me or where, well, I own or have work on some project, there was Slack.

Alexander Volchek✓00:56:38

I never went in there, they tried to add me there so many times, I always said: "I won't work in this chat. That is, you have to forward me information separately", right? Clearly I had the ability to do that given my roles, but for me having yet another chat somewhere or yet another place somewhere where something is posted uniquely is a certain problem, right? And I understand that basically, if a person works at only one company, on only one project, he comes to Slack, and he may not understand my worry, right, or my problems, or my reasoning.

Alexander Volchek✓00:57:11

Really, a person may not even understand. Damn, may not even understand, why would I chat with my wife in Dots, right? And have Dots send us certain messages. But here, for example, well, I see now, I checked, and every day it keeps, I'd even forgotten about it, but every day a message goes out to me with her in messages, in ordinary iMessage, yesterday it went out: "Important from the GAN school." That's where my elder daughter studies. There: "On Thursday the last lesson by the general schedule ends at a different time, and on Friday it's a shortened day."

Alexander Volchek✓00:57:41

That is, because next week there's Thanksgiving there, a week of holidays for the kids. That is, very clear, useful, good information, right? It reminds me too: "Oh, so the kids have holidays." And we, we on this topic, that is, we, it stopped for me, because I didn't read this mail, that's it, it kept reminding me, informing me about different things, it stopped reminding me about them.

Anna Volchek✓00:58:04

Here, here, what a separate messenger is for, Instagram, I don't really understand. Well, and really the difference from the direct that's there now, chats, messengers inside models altogether, it would be really interesting to me. And I'm probably one of those people for whom it's hard to have many messengers, everywhere these different, certain, especially if there are different links, documents, important information that needs to be saved, it's a constant problem for me to bring it all together and save it. And I also wanted to toss out this thought, maybe.

Anna Volchek✓00:58:41

Perhaps the future of all messengers lies precisely in this integration. That is, on the one hand, nobody wants it, because they understand that it's like a loss of such real usage, and there's no longer a need to develop the product so deeply for the user to come there. But, on the other hand, if sooner or later everyone comes to the models through integration with them, then it seems whoever does it first will keep a large audience for themselves.

00:59:12–00:59:44What's Next for ChatGPT, Anthropic and the AI Market?
Alexander Volchek✓00:59:12

We'll see what happens. It's very interesting how the story will unfold in terms of Anthropic and in general with the artificial intelligence market, infrastructure and the new, including new, US rules and legislation. Because, of course, it will depend on this whether Ilnar and I will argue every episode about new models coming out or not. We're waiting for your comments and see you in our next episodes. Our episodes now come out practically every day. Bye, everyone.